ext4: sanity check the block and cluster size at mount time
[pandora-kernel.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/version.h>
30 #include <linux/wait.h>
31 #include <linux/blockgroup_lock.h>
32 #include <linux/percpu_counter.h>
33 #ifdef __KERNEL__
34 #include <linux/compat.h>
35 #endif
36
37 /*
38  * The fourth extended filesystem constants/structures
39  */
40
41 /*
42  * Define EXT4FS_DEBUG to produce debug messages
43  */
44 #undef EXT4FS_DEBUG
45
46 /*
47  * Debug code
48  */
49 #ifdef EXT4FS_DEBUG
50 #define ext4_debug(f, a...)                                             \
51         do {                                                            \
52                 printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:",        \
53                         __FILE__, __LINE__, __func__);                  \
54                 printk(KERN_DEBUG f, ## a);                             \
55         } while (0)
56 #else
57 #define ext4_debug(f, a...)     do {} while (0)
58 #endif
59
60 #define EXT4_ERROR_INODE(inode, fmt, a...) \
61         ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
62
63 #define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...)                 \
64         ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
65
66 #define EXT4_ERROR_FILE(file, block, fmt, a...)                         \
67         ext4_error_file((file), __func__, __LINE__, (block), (fmt), ## a)
68
69 /* data type for block offset of block group */
70 typedef int ext4_grpblk_t;
71
72 /* data type for filesystem-wide blocks number */
73 typedef unsigned long long ext4_fsblk_t;
74
75 /* data type for file logical block number */
76 typedef __u32 ext4_lblk_t;
77
78 /* data type for block group number */
79 typedef unsigned int ext4_group_t;
80
81 /*
82  * Flags used in mballoc's allocation_context flags field.
83  *
84  * Also used to show what's going on for debugging purposes when the
85  * flag field is exported via the traceport interface
86  */
87
88 /* prefer goal again. length */
89 #define EXT4_MB_HINT_MERGE              0x0001
90 /* blocks already reserved */
91 #define EXT4_MB_HINT_RESERVED           0x0002
92 /* metadata is being allocated */
93 #define EXT4_MB_HINT_METADATA           0x0004
94 /* first blocks in the file */
95 #define EXT4_MB_HINT_FIRST              0x0008
96 /* search for the best chunk */
97 #define EXT4_MB_HINT_BEST               0x0010
98 /* data is being allocated */
99 #define EXT4_MB_HINT_DATA               0x0020
100 /* don't preallocate (for tails) */
101 #define EXT4_MB_HINT_NOPREALLOC         0x0040
102 /* allocate for locality group */
103 #define EXT4_MB_HINT_GROUP_ALLOC        0x0080
104 /* allocate goal blocks or none */
105 #define EXT4_MB_HINT_GOAL_ONLY          0x0100
106 /* goal is meaningful */
107 #define EXT4_MB_HINT_TRY_GOAL           0x0200
108 /* blocks already pre-reserved by delayed allocation */
109 #define EXT4_MB_DELALLOC_RESERVED       0x0400
110 /* We are doing stream allocation */
111 #define EXT4_MB_STREAM_ALLOC            0x0800
112 /* Use reserved root blocks if needed */
113 #define EXT4_MB_USE_ROOT_BLOCKS         0x1000
114
115 struct ext4_allocation_request {
116         /* target inode for block we're allocating */
117         struct inode *inode;
118         /* how many blocks we want to allocate */
119         unsigned int len;
120         /* logical block in target inode */
121         ext4_lblk_t logical;
122         /* the closest logical allocated block to the left */
123         ext4_lblk_t lleft;
124         /* the closest logical allocated block to the right */
125         ext4_lblk_t lright;
126         /* phys. target (a hint) */
127         ext4_fsblk_t goal;
128         /* phys. block for the closest logical allocated block to the left */
129         ext4_fsblk_t pleft;
130         /* phys. block for the closest logical allocated block to the right */
131         ext4_fsblk_t pright;
132         /* flags. see above EXT4_MB_HINT_* */
133         unsigned int flags;
134 };
135
136 /*
137  * Logical to physical block mapping, used by ext4_map_blocks()
138  *
139  * This structure is used to pass requests into ext4_map_blocks() as
140  * well as to store the information returned by ext4_map_blocks().  It
141  * takes less room on the stack than a struct buffer_head.
142  */
143 #define EXT4_MAP_NEW            (1 << BH_New)
144 #define EXT4_MAP_MAPPED         (1 << BH_Mapped)
145 #define EXT4_MAP_UNWRITTEN      (1 << BH_Unwritten)
146 #define EXT4_MAP_BOUNDARY       (1 << BH_Boundary)
147 #define EXT4_MAP_UNINIT         (1 << BH_Uninit)
148 /* Sometimes (in the bigalloc case, from ext4_da_get_block_prep) the caller of
149  * ext4_map_blocks wants to know whether or not the underlying cluster has
150  * already been accounted for. EXT4_MAP_FROM_CLUSTER conveys to the caller that
151  * the requested mapping was from previously mapped (or delayed allocated)
152  * cluster. We use BH_AllocFromCluster only for this flag. BH_AllocFromCluster
153  * should never appear on buffer_head's state flags.
154  */
155 #define EXT4_MAP_FROM_CLUSTER   (1 << BH_AllocFromCluster)
156 #define EXT4_MAP_FLAGS          (EXT4_MAP_NEW | EXT4_MAP_MAPPED |\
157                                  EXT4_MAP_UNWRITTEN | EXT4_MAP_BOUNDARY |\
158                                  EXT4_MAP_UNINIT | EXT4_MAP_FROM_CLUSTER)
159
160 struct ext4_map_blocks {
161         ext4_fsblk_t m_pblk;
162         ext4_lblk_t m_lblk;
163         unsigned int m_len;
164         unsigned int m_flags;
165 };
166
167 /*
168  * For delayed allocation tracking
169  */
170 struct mpage_da_data {
171         struct inode *inode;
172         sector_t b_blocknr;             /* start block number of extent */
173         size_t b_size;                  /* size of extent */
174         unsigned long b_state;          /* state of the extent */
175         unsigned long first_page, next_page;    /* extent of pages */
176         struct writeback_control *wbc;
177         int io_done;
178         int pages_written;
179         int retval;
180 };
181
182 /*
183  * Flags for ext4_io_end->flags
184  */
185 #define EXT4_IO_END_UNWRITTEN   0x0001
186 #define EXT4_IO_END_ERROR       0x0002
187 #define EXT4_IO_END_QUEUED      0x0004
188 #define EXT4_IO_END_DIRECT      0x0008
189 #define EXT4_IO_END_IN_FSYNC    0x0010
190
191 struct ext4_io_page {
192         struct page     *p_page;
193         atomic_t        p_count;
194 };
195
196 #define MAX_IO_PAGES 128
197
198 typedef struct ext4_io_end {
199         struct list_head        list;           /* per-file finished IO list */
200         struct inode            *inode;         /* file being written to */
201         unsigned int            flag;           /* unwritten or not */
202         struct page             *page;          /* page struct for buffer write */
203         loff_t                  offset;         /* offset in the file */
204         ssize_t                 size;           /* size of the extent */
205         struct work_struct      work;           /* data work queue */
206         struct kiocb            *iocb;          /* iocb struct for AIO */
207         int                     result;         /* error value for AIO */
208         int                     num_io_pages;
209         struct ext4_io_page     *pages[MAX_IO_PAGES];
210 } ext4_io_end_t;
211
212 struct ext4_io_submit {
213         int                     io_op;
214         struct bio              *io_bio;
215         ext4_io_end_t           *io_end;
216         struct ext4_io_page     *io_page;
217         sector_t                io_next_block;
218 };
219
220 /*
221  * Special inodes numbers
222  */
223 #define EXT4_BAD_INO             1      /* Bad blocks inode */
224 #define EXT4_ROOT_INO            2      /* Root inode */
225 #define EXT4_USR_QUOTA_INO       3      /* User quota inode */
226 #define EXT4_GRP_QUOTA_INO       4      /* Group quota inode */
227 #define EXT4_BOOT_LOADER_INO     5      /* Boot loader inode */
228 #define EXT4_UNDEL_DIR_INO       6      /* Undelete directory inode */
229 #define EXT4_RESIZE_INO          7      /* Reserved group descriptors inode */
230 #define EXT4_JOURNAL_INO         8      /* Journal inode */
231
232 /* First non-reserved inode for old ext4 filesystems */
233 #define EXT4_GOOD_OLD_FIRST_INO 11
234
235 /*
236  * Maximal count of links to a file
237  */
238 #define EXT4_LINK_MAX           65000
239
240 /*
241  * Macro-instructions used to manage several block sizes
242  */
243 #define EXT4_MIN_BLOCK_SIZE             1024
244 #define EXT4_MAX_BLOCK_SIZE             65536
245 #define EXT4_MIN_BLOCK_LOG_SIZE         10
246 #define EXT4_MAX_BLOCK_LOG_SIZE         16
247 #define EXT4_MAX_CLUSTER_LOG_SIZE       30
248 #ifdef __KERNEL__
249 # define EXT4_BLOCK_SIZE(s)             ((s)->s_blocksize)
250 #else
251 # define EXT4_BLOCK_SIZE(s)             (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
252 #endif
253 #define EXT4_ADDR_PER_BLOCK(s)          (EXT4_BLOCK_SIZE(s) / sizeof(__u32))
254 #define EXT4_CLUSTER_SIZE(s)            (EXT4_BLOCK_SIZE(s) << \
255                                          EXT4_SB(s)->s_cluster_bits)
256 #ifdef __KERNEL__
257 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_blocksize_bits)
258 # define EXT4_CLUSTER_BITS(s)           (EXT4_SB(s)->s_cluster_bits)
259 #else
260 # define EXT4_BLOCK_SIZE_BITS(s)        ((s)->s_log_block_size + 10)
261 #endif
262 #ifdef __KERNEL__
263 #define EXT4_ADDR_PER_BLOCK_BITS(s)     (EXT4_SB(s)->s_addr_per_block_bits)
264 #define EXT4_INODE_SIZE(s)              (EXT4_SB(s)->s_inode_size)
265 #define EXT4_FIRST_INO(s)               (EXT4_SB(s)->s_first_ino)
266 #else
267 #define EXT4_INODE_SIZE(s)      (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
268                                  EXT4_GOOD_OLD_INODE_SIZE : \
269                                  (s)->s_inode_size)
270 #define EXT4_FIRST_INO(s)       (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
271                                  EXT4_GOOD_OLD_FIRST_INO : \
272                                  (s)->s_first_ino)
273 #endif
274 #define EXT4_BLOCK_ALIGN(size, blkbits)         ALIGN((size), (1 << (blkbits)))
275
276 /* Translate a block number to a cluster number */
277 #define EXT4_B2C(sbi, blk)      ((blk) >> (sbi)->s_cluster_bits)
278 /* Translate a cluster number to a block number */
279 #define EXT4_C2B(sbi, cluster)  ((cluster) << (sbi)->s_cluster_bits)
280 /* Translate # of blks to # of clusters */
281 #define EXT4_NUM_B2C(sbi, blks) (((blks) + (sbi)->s_cluster_ratio - 1) >> \
282                                  (sbi)->s_cluster_bits)
283 /* Mask out the low bits to get the starting block of the cluster */
284 #define EXT4_PBLK_CMASK(s, pblk) ((pblk) &                              \
285                                   ~((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
286 #define EXT4_LBLK_CMASK(s, lblk) ((lblk) &                              \
287                                   ~((ext4_lblk_t) (s)->s_cluster_ratio - 1))
288 /* Get the cluster offset */
289 #define EXT4_PBLK_COFF(s, pblk) ((pblk) &                               \
290                                  ((ext4_fsblk_t) (s)->s_cluster_ratio - 1))
291 #define EXT4_LBLK_COFF(s, lblk) ((lblk) &                               \
292                                  ((ext4_lblk_t) (s)->s_cluster_ratio - 1))
293
294 /*
295  * Structure of a blocks group descriptor
296  */
297 struct ext4_group_desc
298 {
299         __le32  bg_block_bitmap_lo;     /* Blocks bitmap block */
300         __le32  bg_inode_bitmap_lo;     /* Inodes bitmap block */
301         __le32  bg_inode_table_lo;      /* Inodes table block */
302         __le16  bg_free_blocks_count_lo;/* Free blocks count */
303         __le16  bg_free_inodes_count_lo;/* Free inodes count */
304         __le16  bg_used_dirs_count_lo;  /* Directories count */
305         __le16  bg_flags;               /* EXT4_BG_flags (INODE_UNINIT, etc) */
306         __u32   bg_reserved[2];         /* Likely block/inode bitmap checksum */
307         __le16  bg_itable_unused_lo;    /* Unused inodes count */
308         __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */
309         __le32  bg_block_bitmap_hi;     /* Blocks bitmap block MSB */
310         __le32  bg_inode_bitmap_hi;     /* Inodes bitmap block MSB */
311         __le32  bg_inode_table_hi;      /* Inodes table block MSB */
312         __le16  bg_free_blocks_count_hi;/* Free blocks count MSB */
313         __le16  bg_free_inodes_count_hi;/* Free inodes count MSB */
314         __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
315         __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
316         __u32   bg_reserved2[3];
317 };
318
319 /*
320  * Structure of a flex block group info
321  */
322
323 struct flex_groups {
324         atomic64_t      free_clusters;
325         atomic_t        free_inodes;
326         atomic_t        used_dirs;
327 };
328
329 #define EXT4_BG_INODE_UNINIT    0x0001 /* Inode table/bitmap not in use */
330 #define EXT4_BG_BLOCK_UNINIT    0x0002 /* Block bitmap not in use */
331 #define EXT4_BG_INODE_ZEROED    0x0004 /* On-disk itable initialized to zero */
332
333 /*
334  * Macro-instructions used to manage group descriptors
335  */
336 #define EXT4_MIN_DESC_SIZE              32
337 #define EXT4_MIN_DESC_SIZE_64BIT        64
338 #define EXT4_MAX_DESC_SIZE              EXT4_MIN_BLOCK_SIZE
339 #define EXT4_DESC_SIZE(s)               (EXT4_SB(s)->s_desc_size)
340 #ifdef __KERNEL__
341 # define EXT4_BLOCKS_PER_GROUP(s)       (EXT4_SB(s)->s_blocks_per_group)
342 # define EXT4_CLUSTERS_PER_GROUP(s)     (EXT4_SB(s)->s_clusters_per_group)
343 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_SB(s)->s_desc_per_block)
344 # define EXT4_INODES_PER_GROUP(s)       (EXT4_SB(s)->s_inodes_per_group)
345 # define EXT4_DESC_PER_BLOCK_BITS(s)    (EXT4_SB(s)->s_desc_per_block_bits)
346 #else
347 # define EXT4_BLOCKS_PER_GROUP(s)       ((s)->s_blocks_per_group)
348 # define EXT4_DESC_PER_BLOCK(s)         (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
349 # define EXT4_INODES_PER_GROUP(s)       ((s)->s_inodes_per_group)
350 #endif
351
352 /*
353  * Constants relative to the data blocks
354  */
355 #define EXT4_NDIR_BLOCKS                12
356 #define EXT4_IND_BLOCK                  EXT4_NDIR_BLOCKS
357 #define EXT4_DIND_BLOCK                 (EXT4_IND_BLOCK + 1)
358 #define EXT4_TIND_BLOCK                 (EXT4_DIND_BLOCK + 1)
359 #define EXT4_N_BLOCKS                   (EXT4_TIND_BLOCK + 1)
360
361 /*
362  * Inode flags
363  */
364 #define EXT4_SECRM_FL                   0x00000001 /* Secure deletion */
365 #define EXT4_UNRM_FL                    0x00000002 /* Undelete */
366 #define EXT4_COMPR_FL                   0x00000004 /* Compress file */
367 #define EXT4_SYNC_FL                    0x00000008 /* Synchronous updates */
368 #define EXT4_IMMUTABLE_FL               0x00000010 /* Immutable file */
369 #define EXT4_APPEND_FL                  0x00000020 /* writes to file may only append */
370 #define EXT4_NODUMP_FL                  0x00000040 /* do not dump file */
371 #define EXT4_NOATIME_FL                 0x00000080 /* do not update atime */
372 /* Reserved for compression usage... */
373 #define EXT4_DIRTY_FL                   0x00000100
374 #define EXT4_COMPRBLK_FL                0x00000200 /* One or more compressed clusters */
375 #define EXT4_NOCOMPR_FL                 0x00000400 /* Don't compress */
376 #define EXT4_ECOMPR_FL                  0x00000800 /* Compression error */
377 /* End compression flags --- maybe not all used */
378 #define EXT4_INDEX_FL                   0x00001000 /* hash-indexed directory */
379 #define EXT4_IMAGIC_FL                  0x00002000 /* AFS directory */
380 #define EXT4_JOURNAL_DATA_FL            0x00004000 /* file data should be journaled */
381 #define EXT4_NOTAIL_FL                  0x00008000 /* file tail should not be merged */
382 #define EXT4_DIRSYNC_FL                 0x00010000 /* dirsync behaviour (directories only) */
383 #define EXT4_TOPDIR_FL                  0x00020000 /* Top of directory hierarchies*/
384 #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
385 #define EXT4_EXTENTS_FL                 0x00080000 /* Inode uses extents */
386 #define EXT4_EA_INODE_FL                0x00200000 /* Inode used for large EA */
387 #define EXT4_EOFBLOCKS_FL               0x00400000 /* Blocks allocated beyond EOF */
388 #define EXT4_RESERVED_FL                0x80000000 /* reserved for ext4 lib */
389
390 #define EXT4_FL_USER_VISIBLE            0x004BDFFF /* User visible flags */
391 #define EXT4_FL_USER_MODIFIABLE         0x004B80FF /* User modifiable flags */
392
393 /* Flags that should be inherited by new inodes from their parent. */
394 #define EXT4_FL_INHERITED (EXT4_SECRM_FL | EXT4_UNRM_FL | EXT4_COMPR_FL |\
395                            EXT4_SYNC_FL | EXT4_NODUMP_FL | EXT4_NOATIME_FL |\
396                            EXT4_NOCOMPR_FL | EXT4_JOURNAL_DATA_FL |\
397                            EXT4_NOTAIL_FL | EXT4_DIRSYNC_FL)
398
399 /* Flags that are appropriate for regular files (all but dir-specific ones). */
400 #define EXT4_REG_FLMASK (~(EXT4_DIRSYNC_FL | EXT4_TOPDIR_FL))
401
402 /* Flags that are appropriate for non-directories/regular files. */
403 #define EXT4_OTHER_FLMASK (EXT4_NODUMP_FL | EXT4_NOATIME_FL)
404
405 /* Mask out flags that are inappropriate for the given type of inode. */
406 static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags)
407 {
408         if (S_ISDIR(mode))
409                 return flags;
410         else if (S_ISREG(mode))
411                 return flags & EXT4_REG_FLMASK;
412         else
413                 return flags & EXT4_OTHER_FLMASK;
414 }
415
416 /*
417  * Inode flags used for atomic set/get
418  */
419 enum {
420         EXT4_INODE_SECRM        = 0,    /* Secure deletion */
421         EXT4_INODE_UNRM         = 1,    /* Undelete */
422         EXT4_INODE_COMPR        = 2,    /* Compress file */
423         EXT4_INODE_SYNC         = 3,    /* Synchronous updates */
424         EXT4_INODE_IMMUTABLE    = 4,    /* Immutable file */
425         EXT4_INODE_APPEND       = 5,    /* writes to file may only append */
426         EXT4_INODE_NODUMP       = 6,    /* do not dump file */
427         EXT4_INODE_NOATIME      = 7,    /* do not update atime */
428 /* Reserved for compression usage... */
429         EXT4_INODE_DIRTY        = 8,
430         EXT4_INODE_COMPRBLK     = 9,    /* One or more compressed clusters */
431         EXT4_INODE_NOCOMPR      = 10,   /* Don't compress */
432         EXT4_INODE_ECOMPR       = 11,   /* Compression error */
433 /* End compression flags --- maybe not all used */
434         EXT4_INODE_INDEX        = 12,   /* hash-indexed directory */
435         EXT4_INODE_IMAGIC       = 13,   /* AFS directory */
436         EXT4_INODE_JOURNAL_DATA = 14,   /* file data should be journaled */
437         EXT4_INODE_NOTAIL       = 15,   /* file tail should not be merged */
438         EXT4_INODE_DIRSYNC      = 16,   /* dirsync behaviour (directories only) */
439         EXT4_INODE_TOPDIR       = 17,   /* Top of directory hierarchies*/
440         EXT4_INODE_HUGE_FILE    = 18,   /* Set to each huge file */
441         EXT4_INODE_EXTENTS      = 19,   /* Inode uses extents */
442         EXT4_INODE_EA_INODE     = 21,   /* Inode used for large EA */
443         EXT4_INODE_EOFBLOCKS    = 22,   /* Blocks allocated beyond EOF */
444         EXT4_INODE_RESERVED     = 31,   /* reserved for ext4 lib */
445 };
446
447 #define TEST_FLAG_VALUE(FLAG) (EXT4_##FLAG##_FL == (1 << EXT4_INODE_##FLAG))
448 #define CHECK_FLAG_VALUE(FLAG) if (!TEST_FLAG_VALUE(FLAG)) { \
449         printk(KERN_EMERG "EXT4 flag fail: " #FLAG ": %d %d\n", \
450                 EXT4_##FLAG##_FL, EXT4_INODE_##FLAG); BUG_ON(1); }
451
452 /*
453  * Since it's pretty easy to mix up bit numbers and hex values, and we
454  * can't do a compile-time test for ENUM values, we use a run-time
455  * test to make sure that EXT4_XXX_FL is consistent with respect to
456  * EXT4_INODE_XXX.  If all is well the printk and BUG_ON will all drop
457  * out so it won't cost any extra space in the compiled kernel image.
458  * But it's important that these values are the same, since we are
459  * using EXT4_INODE_XXX to test for the flag values, but EXT4_XX_FL
460  * must be consistent with the values of FS_XXX_FL defined in
461  * include/linux/fs.h and the on-disk values found in ext2, ext3, and
462  * ext4 filesystems, and of course the values defined in e2fsprogs.
463  *
464  * It's not paranoia if the Murphy's Law really *is* out to get you.  :-)
465  */
466 static inline void ext4_check_flag_values(void)
467 {
468         CHECK_FLAG_VALUE(SECRM);
469         CHECK_FLAG_VALUE(UNRM);
470         CHECK_FLAG_VALUE(COMPR);
471         CHECK_FLAG_VALUE(SYNC);
472         CHECK_FLAG_VALUE(IMMUTABLE);
473         CHECK_FLAG_VALUE(APPEND);
474         CHECK_FLAG_VALUE(NODUMP);
475         CHECK_FLAG_VALUE(NOATIME);
476         CHECK_FLAG_VALUE(DIRTY);
477         CHECK_FLAG_VALUE(COMPRBLK);
478         CHECK_FLAG_VALUE(NOCOMPR);
479         CHECK_FLAG_VALUE(ECOMPR);
480         CHECK_FLAG_VALUE(INDEX);
481         CHECK_FLAG_VALUE(IMAGIC);
482         CHECK_FLAG_VALUE(JOURNAL_DATA);
483         CHECK_FLAG_VALUE(NOTAIL);
484         CHECK_FLAG_VALUE(DIRSYNC);
485         CHECK_FLAG_VALUE(TOPDIR);
486         CHECK_FLAG_VALUE(HUGE_FILE);
487         CHECK_FLAG_VALUE(EXTENTS);
488         CHECK_FLAG_VALUE(EA_INODE);
489         CHECK_FLAG_VALUE(EOFBLOCKS);
490         CHECK_FLAG_VALUE(RESERVED);
491 }
492
493 /* Used to pass group descriptor data when online resize is done */
494 struct ext4_new_group_input {
495         __u32 group;            /* Group number for this data */
496         __u64 block_bitmap;     /* Absolute block number of block bitmap */
497         __u64 inode_bitmap;     /* Absolute block number of inode bitmap */
498         __u64 inode_table;      /* Absolute block number of inode table start */
499         __u32 blocks_count;     /* Total number of blocks in this group */
500         __u16 reserved_blocks;  /* Number of reserved blocks in this group */
501         __u16 unused;
502 };
503
504 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
505 struct compat_ext4_new_group_input {
506         u32 group;
507         compat_u64 block_bitmap;
508         compat_u64 inode_bitmap;
509         compat_u64 inode_table;
510         u32 blocks_count;
511         u16 reserved_blocks;
512         u16 unused;
513 };
514 #endif
515
516 /* The struct ext4_new_group_input in kernel space, with free_blocks_count */
517 struct ext4_new_group_data {
518         __u32 group;
519         __u64 block_bitmap;
520         __u64 inode_bitmap;
521         __u64 inode_table;
522         __u32 blocks_count;
523         __u16 reserved_blocks;
524         __u16 unused;
525         __u32 free_blocks_count;
526 };
527
528 /*
529  * Flags used by ext4_map_blocks()
530  */
531         /* Allocate any needed blocks and/or convert an unitialized
532            extent to be an initialized ext4 */
533 #define EXT4_GET_BLOCKS_CREATE                  0x0001
534         /* Request the creation of an unitialized extent */
535 #define EXT4_GET_BLOCKS_UNINIT_EXT              0x0002
536 #define EXT4_GET_BLOCKS_CREATE_UNINIT_EXT       (EXT4_GET_BLOCKS_UNINIT_EXT|\
537                                                  EXT4_GET_BLOCKS_CREATE)
538         /* Caller is from the delayed allocation writeout path,
539            so set the magic i_delalloc_reserve_flag after taking the
540            inode allocation semaphore for */
541 #define EXT4_GET_BLOCKS_DELALLOC_RESERVE        0x0004
542         /* caller is from the direct IO path, request to creation of an
543         unitialized extents if not allocated, split the uninitialized
544         extent if blocks has been preallocated already*/
545 #define EXT4_GET_BLOCKS_PRE_IO                  0x0008
546 #define EXT4_GET_BLOCKS_CONVERT                 0x0010
547 #define EXT4_GET_BLOCKS_IO_CREATE_EXT           (EXT4_GET_BLOCKS_PRE_IO|\
548                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
549         /* Convert extent to initialized after IO complete */
550 #define EXT4_GET_BLOCKS_IO_CONVERT_EXT          (EXT4_GET_BLOCKS_CONVERT|\
551                                          EXT4_GET_BLOCKS_CREATE_UNINIT_EXT)
552         /* Punch out blocks of an extent */
553 #define EXT4_GET_BLOCKS_PUNCH_OUT_EXT           0x0020
554         /* Don't normalize allocation size (used for fallocate) */
555 #define EXT4_GET_BLOCKS_NO_NORMALIZE            0x0040
556         /* Request will not result in inode size update (user for fallocate) */
557 #define EXT4_GET_BLOCKS_KEEP_SIZE               0x0080
558
559 /*
560  * Flags used by ext4_free_blocks
561  */
562 #define EXT4_FREE_BLOCKS_METADATA       0x0001
563 #define EXT4_FREE_BLOCKS_FORGET         0x0002
564 #define EXT4_FREE_BLOCKS_VALIDATED      0x0004
565 #define EXT4_FREE_BLOCKS_NO_QUOT_UPDATE 0x0008
566 #define EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER   0x0010
567 #define EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER    0x0020
568
569 /*
570  * Flags used by ext4_discard_partial_page_buffers
571  */
572 #define EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED   0x0001
573
574 /*
575  * ioctl commands
576  */
577 #define EXT4_IOC_GETFLAGS               FS_IOC_GETFLAGS
578 #define EXT4_IOC_SETFLAGS               FS_IOC_SETFLAGS
579 #define EXT4_IOC_GETVERSION             _IOR('f', 3, long)
580 #define EXT4_IOC_SETVERSION             _IOW('f', 4, long)
581 #define EXT4_IOC_GETVERSION_OLD         FS_IOC_GETVERSION
582 #define EXT4_IOC_SETVERSION_OLD         FS_IOC_SETVERSION
583 #define EXT4_IOC_GETRSVSZ               _IOR('f', 5, long)
584 #define EXT4_IOC_SETRSVSZ               _IOW('f', 6, long)
585 #define EXT4_IOC_GROUP_EXTEND           _IOW('f', 7, unsigned long)
586 #define EXT4_IOC_GROUP_ADD              _IOW('f', 8, struct ext4_new_group_input)
587 #define EXT4_IOC_MIGRATE                _IO('f', 9)
588  /* note ioctl 10 reserved for an early version of the FIEMAP ioctl */
589  /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
590 #define EXT4_IOC_ALLOC_DA_BLKS          _IO('f', 12)
591 #define EXT4_IOC_MOVE_EXT               _IOWR('f', 15, struct move_extent)
592
593 #if defined(__KERNEL__) && defined(CONFIG_COMPAT)
594 /*
595  * ioctl commands in 32 bit emulation
596  */
597 #define EXT4_IOC32_GETFLAGS             FS_IOC32_GETFLAGS
598 #define EXT4_IOC32_SETFLAGS             FS_IOC32_SETFLAGS
599 #define EXT4_IOC32_GETVERSION           _IOR('f', 3, int)
600 #define EXT4_IOC32_SETVERSION           _IOW('f', 4, int)
601 #define EXT4_IOC32_GETRSVSZ             _IOR('f', 5, int)
602 #define EXT4_IOC32_SETRSVSZ             _IOW('f', 6, int)
603 #define EXT4_IOC32_GROUP_EXTEND         _IOW('f', 7, unsigned int)
604 #define EXT4_IOC32_GROUP_ADD            _IOW('f', 8, struct compat_ext4_new_group_input)
605 #define EXT4_IOC32_GETVERSION_OLD       FS_IOC32_GETVERSION
606 #define EXT4_IOC32_SETVERSION_OLD       FS_IOC32_SETVERSION
607 #endif
608
609 /* Max physical block we can address w/o extents */
610 #define EXT4_MAX_BLOCK_FILE_PHYS        0xFFFFFFFF
611
612 /*
613  * Structure of an inode on the disk
614  */
615 struct ext4_inode {
616         __le16  i_mode;         /* File mode */
617         __le16  i_uid;          /* Low 16 bits of Owner Uid */
618         __le32  i_size_lo;      /* Size in bytes */
619         __le32  i_atime;        /* Access time */
620         __le32  i_ctime;        /* Inode Change time */
621         __le32  i_mtime;        /* Modification time */
622         __le32  i_dtime;        /* Deletion Time */
623         __le16  i_gid;          /* Low 16 bits of Group Id */
624         __le16  i_links_count;  /* Links count */
625         __le32  i_blocks_lo;    /* Blocks count */
626         __le32  i_flags;        /* File flags */
627         union {
628                 struct {
629                         __le32  l_i_version;
630                 } linux1;
631                 struct {
632                         __u32  h_i_translator;
633                 } hurd1;
634                 struct {
635                         __u32  m_i_reserved1;
636                 } masix1;
637         } osd1;                         /* OS dependent 1 */
638         __le32  i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
639         __le32  i_generation;   /* File version (for NFS) */
640         __le32  i_file_acl_lo;  /* File ACL */
641         __le32  i_size_high;
642         __le32  i_obso_faddr;   /* Obsoleted fragment address */
643         union {
644                 struct {
645                         __le16  l_i_blocks_high; /* were l_i_reserved1 */
646                         __le16  l_i_file_acl_high;
647                         __le16  l_i_uid_high;   /* these 2 fields */
648                         __le16  l_i_gid_high;   /* were reserved2[0] */
649                         __u32   l_i_reserved2;
650                 } linux2;
651                 struct {
652                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
653                         __u16   h_i_mode_high;
654                         __u16   h_i_uid_high;
655                         __u16   h_i_gid_high;
656                         __u32   h_i_author;
657                 } hurd2;
658                 struct {
659                         __le16  h_i_reserved1;  /* Obsoleted fragment number/size which are removed in ext4 */
660                         __le16  m_i_file_acl_high;
661                         __u32   m_i_reserved2[2];
662                 } masix2;
663         } osd2;                         /* OS dependent 2 */
664         __le16  i_extra_isize;
665         __le16  i_pad1;
666         __le32  i_ctime_extra;  /* extra Change time      (nsec << 2 | epoch) */
667         __le32  i_mtime_extra;  /* extra Modification time(nsec << 2 | epoch) */
668         __le32  i_atime_extra;  /* extra Access time      (nsec << 2 | epoch) */
669         __le32  i_crtime;       /* File Creation time */
670         __le32  i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
671         __le32  i_version_hi;   /* high 32 bits for 64-bit version */
672 };
673
674 struct move_extent {
675         __u32 reserved;         /* should be zero */
676         __u32 donor_fd;         /* donor file descriptor */
677         __u64 orig_start;       /* logical start offset in block for orig */
678         __u64 donor_start;      /* logical start offset in block for donor */
679         __u64 len;              /* block length to be moved */
680         __u64 moved_len;        /* moved block length */
681 };
682
683 #define EXT4_EPOCH_BITS 2
684 #define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
685 #define EXT4_NSEC_MASK  (~0UL << EXT4_EPOCH_BITS)
686
687 /*
688  * Extended fields will fit into an inode if the filesystem was formatted
689  * with large inodes (-I 256 or larger) and there are not currently any EAs
690  * consuming all of the available space. For new inodes we always reserve
691  * enough space for the kernel's known extended fields, but for inodes
692  * created with an old kernel this might not have been the case. None of
693  * the extended inode fields is critical for correct filesystem operation.
694  * This macro checks if a certain field fits in the inode. Note that
695  * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
696  */
697 #define EXT4_FITS_IN_INODE(ext4_inode, einode, field)   \
698         ((offsetof(typeof(*ext4_inode), field) +        \
699           sizeof((ext4_inode)->field))                  \
700         <= (EXT4_GOOD_OLD_INODE_SIZE +                  \
701             (einode)->i_extra_isize))                   \
702
703 /*
704  * We use an encoding that preserves the times for extra epoch "00":
705  *
706  * extra  msb of                         adjust for signed
707  * epoch  32-bit                         32-bit tv_sec to
708  * bits   time    decoded 64-bit tv_sec  64-bit tv_sec      valid time range
709  * 0 0    1    -0x80000000..-0x00000001  0x000000000 1901-12-13..1969-12-31
710  * 0 0    0    0x000000000..0x07fffffff  0x000000000 1970-01-01..2038-01-19
711  * 0 1    1    0x080000000..0x0ffffffff  0x100000000 2038-01-19..2106-02-07
712  * 0 1    0    0x100000000..0x17fffffff  0x100000000 2106-02-07..2174-02-25
713  * 1 0    1    0x180000000..0x1ffffffff  0x200000000 2174-02-25..2242-03-16
714  * 1 0    0    0x200000000..0x27fffffff  0x200000000 2242-03-16..2310-04-04
715  * 1 1    1    0x280000000..0x2ffffffff  0x300000000 2310-04-04..2378-04-22
716  * 1 1    0    0x300000000..0x37fffffff  0x300000000 2378-04-22..2446-05-10
717  *
718  * Note that previous versions of the kernel on 64-bit systems would
719  * incorrectly use extra epoch bits 1,1 for dates between 1901 and
720  * 1970.  e2fsck will correct this, assuming that it is run on the
721  * affected filesystem before 2242.
722  */
723
724 static inline __le32 ext4_encode_extra_time(struct timespec *time)
725 {
726         u32 extra = sizeof(time->tv_sec) > 4 ?
727                 ((time->tv_sec - (s32)time->tv_sec) >> 32) & EXT4_EPOCH_MASK : 0;
728         return cpu_to_le32(extra | (time->tv_nsec << EXT4_EPOCH_BITS));
729 }
730
731 static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
732 {
733         if (unlikely(sizeof(time->tv_sec) > 4 &&
734                         (extra & cpu_to_le32(EXT4_EPOCH_MASK)))) {
735 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,20,0)
736                 /* Handle legacy encoding of pre-1970 dates with epoch
737                  * bits 1,1.  We assume that by kernel version 4.20,
738                  * everyone will have run fsck over the affected
739                  * filesystems to correct the problem.  (This
740                  * backwards compatibility may be removed before this
741                  * time, at the discretion of the ext4 developers.)
742                  */
743                 u64 extra_bits = le32_to_cpu(extra) & EXT4_EPOCH_MASK;
744                 if (extra_bits == 3 && ((time->tv_sec) & 0x80000000) != 0)
745                         extra_bits = 0;
746                 time->tv_sec += extra_bits << 32;
747 #else
748                 time->tv_sec += (u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK) << 32;
749 #endif
750         }
751         time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> EXT4_EPOCH_BITS;
752 }
753
754 #define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode)                          \
755 do {                                                                           \
756         (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec);               \
757         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
758                 (raw_inode)->xtime ## _extra =                                 \
759                                 ext4_encode_extra_time(&(inode)->xtime);       \
760 } while (0)
761
762 #define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode)                        \
763 do {                                                                           \
764         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
765                 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec);      \
766         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
767                 (raw_inode)->xtime ## _extra =                                 \
768                                 ext4_encode_extra_time(&(einode)->xtime);      \
769 } while (0)
770
771 #define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode)                          \
772 do {                                                                           \
773         (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime);       \
774         if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra))     \
775                 ext4_decode_extra_time(&(inode)->xtime,                        \
776                                        raw_inode->xtime ## _extra);            \
777         else                                                                   \
778                 (inode)->xtime.tv_nsec = 0;                                    \
779 } while (0)
780
781 #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode)                        \
782 do {                                                                           \
783         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime))                      \
784                 (einode)->xtime.tv_sec =                                       \
785                         (signed)le32_to_cpu((raw_inode)->xtime);               \
786         else                                                                   \
787                 (einode)->xtime.tv_sec = 0;                                    \
788         if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra))            \
789                 ext4_decode_extra_time(&(einode)->xtime,                       \
790                                        raw_inode->xtime ## _extra);            \
791         else                                                                   \
792                 (einode)->xtime.tv_nsec = 0;                                   \
793 } while (0)
794
795 #define i_disk_version osd1.linux1.l_i_version
796
797 #if defined(__KERNEL__) || defined(__linux__)
798 #define i_reserved1     osd1.linux1.l_i_reserved1
799 #define i_file_acl_high osd2.linux2.l_i_file_acl_high
800 #define i_blocks_high   osd2.linux2.l_i_blocks_high
801 #define i_uid_low       i_uid
802 #define i_gid_low       i_gid
803 #define i_uid_high      osd2.linux2.l_i_uid_high
804 #define i_gid_high      osd2.linux2.l_i_gid_high
805 #define i_reserved2     osd2.linux2.l_i_reserved2
806
807 #elif defined(__GNU__)
808
809 #define i_translator    osd1.hurd1.h_i_translator
810 #define i_uid_high      osd2.hurd2.h_i_uid_high
811 #define i_gid_high      osd2.hurd2.h_i_gid_high
812 #define i_author        osd2.hurd2.h_i_author
813
814 #elif defined(__masix__)
815
816 #define i_reserved1     osd1.masix1.m_i_reserved1
817 #define i_file_acl_high osd2.masix2.m_i_file_acl_high
818 #define i_reserved2     osd2.masix2.m_i_reserved2
819
820 #endif /* defined(__KERNEL__) || defined(__linux__) */
821
822 /*
823  * storage for cached extent
824  * If ec_len == 0, then the cache is invalid.
825  * If ec_start == 0, then the cache represents a gap (null mapping)
826  */
827 struct ext4_ext_cache {
828         ext4_fsblk_t    ec_start;
829         ext4_lblk_t     ec_block;
830         __u32           ec_len; /* must be 32bit to return holes */
831 };
832
833 /*
834  * fourth extended file system inode data in memory
835  */
836 struct ext4_inode_info {
837         __le32  i_data[15];     /* unconverted */
838         __u32   i_dtime;
839         ext4_fsblk_t    i_file_acl;
840
841         /*
842          * i_block_group is the number of the block group which contains
843          * this file's inode.  Constant across the lifetime of the inode,
844          * it is ued for making block allocation decisions - we try to
845          * place a file's data blocks near its inode block, and new inodes
846          * near to their parent directory's inode.
847          */
848         ext4_group_t    i_block_group;
849         ext4_lblk_t     i_dir_start_lookup;
850 #if (BITS_PER_LONG < 64)
851         unsigned long   i_state_flags;          /* Dynamic state flags */
852 #endif
853         unsigned long   i_flags;
854
855 #ifdef CONFIG_EXT4_FS_XATTR
856         /*
857          * Extended attributes can be read independently of the main file
858          * data. Taking i_mutex even when reading would cause contention
859          * between readers of EAs and writers of regular file data, so
860          * instead we synchronize on xattr_sem when reading or changing
861          * EAs.
862          */
863         struct rw_semaphore xattr_sem;
864 #endif
865
866         struct list_head i_orphan;      /* unlinked but open inodes */
867
868         /*
869          * i_disksize keeps track of what the inode size is ON DISK, not
870          * in memory.  During truncate, i_size is set to the new size by
871          * the VFS prior to calling ext4_truncate(), but the filesystem won't
872          * set i_disksize to 0 until the truncate is actually under way.
873          *
874          * The intent is that i_disksize always represents the blocks which
875          * are used by this file.  This allows recovery to restart truncate
876          * on orphans if we crash during truncate.  We actually write i_disksize
877          * into the on-disk inode when writing inodes out, instead of i_size.
878          *
879          * The only time when i_disksize and i_size may be different is when
880          * a truncate is in progress.  The only things which change i_disksize
881          * are ext4_get_block (growth) and ext4_truncate (shrinkth).
882          */
883         loff_t  i_disksize;
884
885         /*
886          * i_data_sem is for serialising ext4_truncate() against
887          * ext4_getblock().  In the 2.4 ext2 design, great chunks of inode's
888          * data tree are chopped off during truncate. We can't do that in
889          * ext4 because whenever we perform intermediate commits during
890          * truncate, the inode and all the metadata blocks *must* be in a
891          * consistent state which allows truncation of the orphans to restart
892          * during recovery.  Hence we must fix the get_block-vs-truncate race
893          * by other means, so we have i_data_sem.
894          */
895         struct rw_semaphore i_data_sem;
896         struct inode vfs_inode;
897         struct jbd2_inode *jinode;
898
899         struct ext4_ext_cache i_cached_extent;
900         /*
901          * File creation time. Its function is same as that of
902          * struct timespec i_{a,c,m}time in the generic inode.
903          */
904         struct timespec i_crtime;
905
906         /* mballoc */
907         struct list_head i_prealloc_list;
908         spinlock_t i_prealloc_lock;
909
910         /* ialloc */
911         ext4_group_t    i_last_alloc_group;
912
913         /* allocation reservation info for delalloc */
914         /* In case of bigalloc, these refer to clusters rather than blocks */
915         unsigned int i_reserved_data_blocks;
916         unsigned int i_reserved_meta_blocks;
917         unsigned int i_allocated_meta_blocks;
918         ext4_lblk_t i_da_metadata_calc_last_lblock;
919         int i_da_metadata_calc_len;
920
921         /* on-disk additional length */
922         __u16 i_extra_isize;
923
924 #ifdef CONFIG_QUOTA
925         /* quota space reservation, managed internally by quota code */
926         qsize_t i_reserved_quota;
927 #endif
928
929         /* completed IOs that might need unwritten extents handling */
930         struct list_head i_completed_io_list;
931         spinlock_t i_completed_io_lock;
932         atomic_t i_ioend_count; /* Number of outstanding io_end structs */
933         /* current io_end structure for async DIO write*/
934         ext4_io_end_t *cur_aio_dio;
935         atomic_t i_aiodio_unwritten; /* Nr. of inflight conversions pending */
936
937         spinlock_t i_block_reservation_lock;
938
939         /*
940          * Transactions that contain inode's metadata needed to complete
941          * fsync and fdatasync, respectively.
942          */
943         tid_t i_sync_tid;
944         tid_t i_datasync_tid;
945 };
946
947 /*
948  * File system states
949  */
950 #define EXT4_VALID_FS                   0x0001  /* Unmounted cleanly */
951 #define EXT4_ERROR_FS                   0x0002  /* Errors detected */
952 #define EXT4_ORPHAN_FS                  0x0004  /* Orphans being recovered */
953
954 /*
955  * Misc. filesystem flags
956  */
957 #define EXT2_FLAGS_SIGNED_HASH          0x0001  /* Signed dirhash in use */
958 #define EXT2_FLAGS_UNSIGNED_HASH        0x0002  /* Unsigned dirhash in use */
959 #define EXT2_FLAGS_TEST_FILESYS         0x0004  /* to test development code */
960
961 /*
962  * Mount flags
963  */
964 #define EXT4_MOUNT_GRPID                0x00004 /* Create files with directory's group */
965 #define EXT4_MOUNT_DEBUG                0x00008 /* Some debugging messages */
966 #define EXT4_MOUNT_ERRORS_CONT          0x00010 /* Continue on errors */
967 #define EXT4_MOUNT_ERRORS_RO            0x00020 /* Remount fs ro on errors */
968 #define EXT4_MOUNT_ERRORS_PANIC         0x00040 /* Panic on errors */
969 #define EXT4_MOUNT_MINIX_DF             0x00080 /* Mimics the Minix statfs */
970 #define EXT4_MOUNT_NOLOAD               0x00100 /* Don't use existing journal*/
971 #define EXT4_MOUNT_DATA_FLAGS           0x00C00 /* Mode for data writes: */
972 #define EXT4_MOUNT_JOURNAL_DATA         0x00400 /* Write data to journal */
973 #define EXT4_MOUNT_ORDERED_DATA         0x00800 /* Flush data before commit */
974 #define EXT4_MOUNT_WRITEBACK_DATA       0x00C00 /* No data ordering */
975 #define EXT4_MOUNT_UPDATE_JOURNAL       0x01000 /* Update the journal format */
976 #define EXT4_MOUNT_NO_UID32             0x02000  /* Disable 32-bit UIDs */
977 #define EXT4_MOUNT_XATTR_USER           0x04000 /* Extended user attributes */
978 #define EXT4_MOUNT_POSIX_ACL            0x08000 /* POSIX Access Control Lists */
979 #define EXT4_MOUNT_NO_AUTO_DA_ALLOC     0x10000 /* No auto delalloc mapping */
980 #define EXT4_MOUNT_BARRIER              0x20000 /* Use block barriers */
981 #define EXT4_MOUNT_QUOTA                0x80000 /* Some quota option set */
982 #define EXT4_MOUNT_USRQUOTA             0x100000 /* "old" user quota */
983 #define EXT4_MOUNT_GRPQUOTA             0x200000 /* "old" group quota */
984 #define EXT4_MOUNT_DIOREAD_NOLOCK       0x400000 /* Enable support for dio read nolocking */
985 #define EXT4_MOUNT_JOURNAL_CHECKSUM     0x800000 /* Journal checksums */
986 #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
987 #define EXT4_MOUNT_I_VERSION            0x2000000 /* i_version support */
988 #define EXT4_MOUNT_MBLK_IO_SUBMIT       0x4000000 /* multi-block io submits */
989 #define EXT4_MOUNT_DELALLOC             0x8000000 /* Delalloc support */
990 #define EXT4_MOUNT_DATA_ERR_ABORT       0x10000000 /* Abort on file data write */
991 #define EXT4_MOUNT_BLOCK_VALIDITY       0x20000000 /* Block validity checking */
992 #define EXT4_MOUNT_DISCARD              0x40000000 /* Issue DISCARD requests */
993 #define EXT4_MOUNT_INIT_INODE_TABLE     0x80000000 /* Initialize uninitialized itables */
994
995 #define EXT4_MOUNT2_EXPLICIT_DELALLOC   0x00000001 /* User explicitly
996                                                       specified delalloc */
997
998 #define clear_opt(sb, opt)              EXT4_SB(sb)->s_mount_opt &= \
999                                                 ~EXT4_MOUNT_##opt
1000 #define set_opt(sb, opt)                EXT4_SB(sb)->s_mount_opt |= \
1001                                                 EXT4_MOUNT_##opt
1002 #define test_opt(sb, opt)               (EXT4_SB(sb)->s_mount_opt & \
1003                                          EXT4_MOUNT_##opt)
1004
1005 #define clear_opt2(sb, opt)             EXT4_SB(sb)->s_mount_opt2 &= \
1006                                                 ~EXT4_MOUNT2_##opt
1007 #define set_opt2(sb, opt)               EXT4_SB(sb)->s_mount_opt2 |= \
1008                                                 EXT4_MOUNT2_##opt
1009 #define test_opt2(sb, opt)              (EXT4_SB(sb)->s_mount_opt2 & \
1010                                          EXT4_MOUNT2_##opt)
1011
1012 #define ext4_set_bit                    __test_and_set_bit_le
1013 #define ext4_set_bit_atomic             ext2_set_bit_atomic
1014 #define ext4_clear_bit                  __test_and_clear_bit_le
1015 #define ext4_clear_bit_atomic           ext2_clear_bit_atomic
1016 #define ext4_test_bit                   test_bit_le
1017 #define ext4_find_first_zero_bit        find_first_zero_bit_le
1018 #define ext4_find_next_zero_bit         find_next_zero_bit_le
1019 #define ext4_find_next_bit              find_next_bit_le
1020
1021 extern void ext4_set_bits(void *bm, int cur, int len);
1022
1023 /*
1024  * Maximal mount counts between two filesystem checks
1025  */
1026 #define EXT4_DFL_MAX_MNT_COUNT          20      /* Allow 20 mounts */
1027 #define EXT4_DFL_CHECKINTERVAL          0       /* Don't use interval check */
1028
1029 /*
1030  * Behaviour when detecting errors
1031  */
1032 #define EXT4_ERRORS_CONTINUE            1       /* Continue execution */
1033 #define EXT4_ERRORS_RO                  2       /* Remount fs read-only */
1034 #define EXT4_ERRORS_PANIC               3       /* Panic */
1035 #define EXT4_ERRORS_DEFAULT             EXT4_ERRORS_CONTINUE
1036
1037 /*
1038  * Structure of the super block
1039  */
1040 struct ext4_super_block {
1041 /*00*/  __le32  s_inodes_count;         /* Inodes count */
1042         __le32  s_blocks_count_lo;      /* Blocks count */
1043         __le32  s_r_blocks_count_lo;    /* Reserved blocks count */
1044         __le32  s_free_blocks_count_lo; /* Free blocks count */
1045 /*10*/  __le32  s_free_inodes_count;    /* Free inodes count */
1046         __le32  s_first_data_block;     /* First Data Block */
1047         __le32  s_log_block_size;       /* Block size */
1048         __le32  s_log_cluster_size;     /* Allocation cluster size */
1049 /*20*/  __le32  s_blocks_per_group;     /* # Blocks per group */
1050         __le32  s_clusters_per_group;   /* # Clusters per group */
1051         __le32  s_inodes_per_group;     /* # Inodes per group */
1052         __le32  s_mtime;                /* Mount time */
1053 /*30*/  __le32  s_wtime;                /* Write time */
1054         __le16  s_mnt_count;            /* Mount count */
1055         __le16  s_max_mnt_count;        /* Maximal mount count */
1056         __le16  s_magic;                /* Magic signature */
1057         __le16  s_state;                /* File system state */
1058         __le16  s_errors;               /* Behaviour when detecting errors */
1059         __le16  s_minor_rev_level;      /* minor revision level */
1060 /*40*/  __le32  s_lastcheck;            /* time of last check */
1061         __le32  s_checkinterval;        /* max. time between checks */
1062         __le32  s_creator_os;           /* OS */
1063         __le32  s_rev_level;            /* Revision level */
1064 /*50*/  __le16  s_def_resuid;           /* Default uid for reserved blocks */
1065         __le16  s_def_resgid;           /* Default gid for reserved blocks */
1066         /*
1067          * These fields are for EXT4_DYNAMIC_REV superblocks only.
1068          *
1069          * Note: the difference between the compatible feature set and
1070          * the incompatible feature set is that if there is a bit set
1071          * in the incompatible feature set that the kernel doesn't
1072          * know about, it should refuse to mount the filesystem.
1073          *
1074          * e2fsck's requirements are more strict; if it doesn't know
1075          * about a feature in either the compatible or incompatible
1076          * feature set, it must abort and not try to meddle with
1077          * things it doesn't understand...
1078          */
1079         __le32  s_first_ino;            /* First non-reserved inode */
1080         __le16  s_inode_size;           /* size of inode structure */
1081         __le16  s_block_group_nr;       /* block group # of this superblock */
1082         __le32  s_feature_compat;       /* compatible feature set */
1083 /*60*/  __le32  s_feature_incompat;     /* incompatible feature set */
1084         __le32  s_feature_ro_compat;    /* readonly-compatible feature set */
1085 /*68*/  __u8    s_uuid[16];             /* 128-bit uuid for volume */
1086 /*78*/  char    s_volume_name[16];      /* volume name */
1087 /*88*/  char    s_last_mounted[64];     /* directory where last mounted */
1088 /*C8*/  __le32  s_algorithm_usage_bitmap; /* For compression */
1089         /*
1090          * Performance hints.  Directory preallocation should only
1091          * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
1092          */
1093         __u8    s_prealloc_blocks;      /* Nr of blocks to try to preallocate*/
1094         __u8    s_prealloc_dir_blocks;  /* Nr to preallocate for dirs */
1095         __le16  s_reserved_gdt_blocks;  /* Per group desc for online growth */
1096         /*
1097          * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
1098          */
1099 /*D0*/  __u8    s_journal_uuid[16];     /* uuid of journal superblock */
1100 /*E0*/  __le32  s_journal_inum;         /* inode number of journal file */
1101         __le32  s_journal_dev;          /* device number of journal file */
1102         __le32  s_last_orphan;          /* start of list of inodes to delete */
1103         __le32  s_hash_seed[4];         /* HTREE hash seed */
1104         __u8    s_def_hash_version;     /* Default hash version to use */
1105         __u8    s_jnl_backup_type;
1106         __le16  s_desc_size;            /* size of group descriptor */
1107 /*100*/ __le32  s_default_mount_opts;
1108         __le32  s_first_meta_bg;        /* First metablock block group */
1109         __le32  s_mkfs_time;            /* When the filesystem was created */
1110         __le32  s_jnl_blocks[17];       /* Backup of the journal inode */
1111         /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
1112 /*150*/ __le32  s_blocks_count_hi;      /* Blocks count */
1113         __le32  s_r_blocks_count_hi;    /* Reserved blocks count */
1114         __le32  s_free_blocks_count_hi; /* Free blocks count */
1115         __le16  s_min_extra_isize;      /* All inodes have at least # bytes */
1116         __le16  s_want_extra_isize;     /* New inodes should reserve # bytes */
1117         __le32  s_flags;                /* Miscellaneous flags */
1118         __le16  s_raid_stride;          /* RAID stride */
1119         __le16  s_mmp_update_interval;  /* # seconds to wait in MMP checking */
1120         __le64  s_mmp_block;            /* Block for multi-mount protection */
1121         __le32  s_raid_stripe_width;    /* blocks on all data disks (N*stride)*/
1122         __u8    s_log_groups_per_flex;  /* FLEX_BG group size */
1123         __u8    s_reserved_char_pad;
1124         __le16  s_reserved_pad;
1125         __le64  s_kbytes_written;       /* nr of lifetime kilobytes written */
1126         __le32  s_snapshot_inum;        /* Inode number of active snapshot */
1127         __le32  s_snapshot_id;          /* sequential ID of active snapshot */
1128         __le64  s_snapshot_r_blocks_count; /* reserved blocks for active
1129                                               snapshot's future use */
1130         __le32  s_snapshot_list;        /* inode number of the head of the
1131                                            on-disk snapshot list */
1132 #define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1133         __le32  s_error_count;          /* number of fs errors */
1134         __le32  s_first_error_time;     /* first time an error happened */
1135         __le32  s_first_error_ino;      /* inode involved in first error */
1136         __le64  s_first_error_block;    /* block involved of first error */
1137         __u8    s_first_error_func[32]; /* function where the error happened */
1138         __le32  s_first_error_line;     /* line number where error happened */
1139         __le32  s_last_error_time;      /* most recent time of an error */
1140         __le32  s_last_error_ino;       /* inode involved in last error */
1141         __le32  s_last_error_line;      /* line number where error happened */
1142         __le64  s_last_error_block;     /* block involved of last error */
1143         __u8    s_last_error_func[32];  /* function where the error happened */
1144 #define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1145         __u8    s_mount_opts[64];
1146         __le32  s_usr_quota_inum;       /* inode for tracking user quota */
1147         __le32  s_grp_quota_inum;       /* inode for tracking group quota */
1148         __le32  s_overhead_clusters;    /* overhead blocks/clusters in fs */
1149         __le32  s_reserved[109];        /* Padding to the end of the block */
1150 };
1151
1152 #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1153
1154 #ifdef __KERNEL__
1155
1156 /*
1157  * run-time mount flags
1158  */
1159 #define EXT4_MF_MNTDIR_SAMPLED  0x0001
1160 #define EXT4_MF_FS_ABORTED      0x0002  /* Fatal error detected */
1161
1162 /*
1163  * fourth extended-fs super-block data in memory
1164  */
1165 struct ext4_sb_info {
1166         unsigned long s_desc_size;      /* Size of a group descriptor in bytes */
1167         unsigned long s_inodes_per_block;/* Number of inodes per block */
1168         unsigned long s_blocks_per_group;/* Number of blocks in a group */
1169         unsigned long s_clusters_per_group; /* Number of clusters in a group */
1170         unsigned long s_inodes_per_group;/* Number of inodes in a group */
1171         unsigned long s_itb_per_group;  /* Number of inode table blocks per group */
1172         unsigned long s_gdb_count;      /* Number of group descriptor blocks */
1173         unsigned long s_desc_per_block; /* Number of group descriptors per block */
1174         ext4_group_t s_groups_count;    /* Number of groups in the fs */
1175         ext4_group_t s_blockfile_groups;/* Groups acceptable for non-extent files */
1176         unsigned long s_overhead;  /* # of fs overhead clusters */
1177         unsigned int s_cluster_ratio;   /* Number of blocks per cluster */
1178         unsigned int s_cluster_bits;    /* log2 of s_cluster_ratio */
1179         loff_t s_bitmap_maxbytes;       /* max bytes for bitmap files */
1180         struct buffer_head * s_sbh;     /* Buffer containing the super block */
1181         struct ext4_super_block *s_es;  /* Pointer to the super block in the buffer */
1182         struct buffer_head **s_group_desc;
1183         unsigned int s_mount_opt;
1184         unsigned int s_mount_opt2;
1185         unsigned int s_mount_flags;
1186         ext4_fsblk_t s_sb_block;
1187         uid_t s_resuid;
1188         gid_t s_resgid;
1189         unsigned short s_mount_state;
1190         unsigned short s_pad;
1191         int s_addr_per_block_bits;
1192         int s_desc_per_block_bits;
1193         int s_inode_size;
1194         int s_first_ino;
1195         unsigned int s_inode_readahead_blks;
1196         unsigned int s_inode_goal;
1197         spinlock_t s_next_gen_lock;
1198         u32 s_next_generation;
1199         u32 s_hash_seed[4];
1200         int s_def_hash_version;
1201         int s_hash_unsigned;    /* 3 if hash should be signed, 0 if not */
1202         struct percpu_counter s_freeclusters_counter;
1203         struct percpu_counter s_freeinodes_counter;
1204         struct percpu_counter s_dirs_counter;
1205         struct percpu_counter s_dirtyclusters_counter;
1206         struct blockgroup_lock *s_blockgroup_lock;
1207         struct proc_dir_entry *s_proc;
1208         struct kobject s_kobj;
1209         struct completion s_kobj_unregister;
1210
1211         /* Journaling */
1212         struct journal_s *s_journal;
1213         struct list_head s_orphan;
1214         struct mutex s_orphan_lock;
1215         unsigned long s_resize_flags;           /* Flags indicating if there
1216                                                    is a resizer */
1217         unsigned long s_commit_interval;
1218         u32 s_max_batch_time;
1219         u32 s_min_batch_time;
1220         struct block_device *journal_bdev;
1221 #ifdef CONFIG_QUOTA
1222         char *s_qf_names[MAXQUOTAS];            /* Names of quota files with journalled quota */
1223         int s_jquota_fmt;                       /* Format of quota to use */
1224 #endif
1225         unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
1226         struct rb_root system_blks;
1227
1228 #ifdef EXTENTS_STATS
1229         /* ext4 extents stats */
1230         unsigned long s_ext_min;
1231         unsigned long s_ext_max;
1232         unsigned long s_depth_max;
1233         spinlock_t s_ext_stats_lock;
1234         unsigned long s_ext_blocks;
1235         unsigned long s_ext_extents;
1236 #endif
1237
1238         /* for buddy allocator */
1239         struct ext4_group_info ***s_group_info;
1240         struct inode *s_buddy_cache;
1241         spinlock_t s_md_lock;
1242         unsigned short *s_mb_offsets;
1243         unsigned int *s_mb_maxs;
1244
1245         /* tunables */
1246         unsigned long s_stripe;
1247         unsigned int s_mb_stream_request;
1248         unsigned int s_mb_max_to_scan;
1249         unsigned int s_mb_min_to_scan;
1250         unsigned int s_mb_stats;
1251         unsigned int s_mb_order2_reqs;
1252         unsigned int s_mb_group_prealloc;
1253         unsigned int s_max_writeback_mb_bump;
1254         /* where last allocation was done - for stream allocation */
1255         unsigned long s_mb_last_group;
1256         unsigned long s_mb_last_start;
1257
1258         /* stats for buddy allocator */
1259         atomic_t s_bal_reqs;    /* number of reqs with len > 1 */
1260         atomic_t s_bal_success; /* we found long enough chunks */
1261         atomic_t s_bal_allocated;       /* in blocks */
1262         atomic_t s_bal_ex_scanned;      /* total extents scanned */
1263         atomic_t s_bal_goals;   /* goal hits */
1264         atomic_t s_bal_breaks;  /* too long searches */
1265         atomic_t s_bal_2orders; /* 2^order hits */
1266         spinlock_t s_bal_lock;
1267         unsigned long s_mb_buddies_generated;
1268         unsigned long long s_mb_generation_time;
1269         atomic_t s_mb_lost_chunks;
1270         atomic_t s_mb_preallocated;
1271         atomic_t s_mb_discarded;
1272         atomic_t s_lock_busy;
1273
1274         /* locality groups */
1275         struct ext4_locality_group __percpu *s_locality_groups;
1276
1277         /* for write statistics */
1278         unsigned long s_sectors_written_start;
1279         u64 s_kbytes_written;
1280
1281         unsigned int s_log_groups_per_flex;
1282         struct flex_groups *s_flex_groups;
1283
1284         /* workqueue for dio unwritten */
1285         struct workqueue_struct *dio_unwritten_wq;
1286
1287         /* timer for periodic error stats printing */
1288         struct timer_list s_err_report;
1289
1290         /* Lazy inode table initialization info */
1291         struct ext4_li_request *s_li_request;
1292         /* Wait multiplier for lazy initialization thread */
1293         unsigned int s_li_wait_mult;
1294
1295         /* Kernel thread for multiple mount protection */
1296         struct task_struct *s_mmp_tsk;
1297
1298         /* record the last minlen when FITRIM is called. */
1299         atomic_t s_last_trim_minblks;
1300 };
1301
1302 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
1303 {
1304         return sb->s_fs_info;
1305 }
1306 static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
1307 {
1308         return container_of(inode, struct ext4_inode_info, vfs_inode);
1309 }
1310
1311 static inline struct timespec ext4_current_time(struct inode *inode)
1312 {
1313         return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
1314                 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
1315 }
1316
1317 static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
1318 {
1319         return ino == EXT4_ROOT_INO ||
1320                 ino == EXT4_JOURNAL_INO ||
1321                 ino == EXT4_RESIZE_INO ||
1322                 (ino >= EXT4_FIRST_INO(sb) &&
1323                  ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
1324 }
1325
1326 static inline void ext4_set_io_unwritten_flag(struct inode *inode,
1327                                               struct ext4_io_end *io_end)
1328 {
1329         if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
1330                 io_end->flag |= EXT4_IO_END_UNWRITTEN;
1331                 atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
1332         }
1333 }
1334
1335 /*
1336  * Inode dynamic state flags
1337  */
1338 enum {
1339         EXT4_STATE_JDATA,               /* journaled data exists */
1340         EXT4_STATE_NEW,                 /* inode is newly created */
1341         EXT4_STATE_XATTR,               /* has in-inode xattrs */
1342         EXT4_STATE_NO_EXPAND,           /* No space for expansion */
1343         EXT4_STATE_DA_ALLOC_CLOSE,      /* Alloc DA blks on close */
1344         EXT4_STATE_EXT_MIGRATE,         /* Inode is migrating */
1345         EXT4_STATE_DIO_UNWRITTEN,       /* need convert on dio done*/
1346         EXT4_STATE_NEWENTRY,            /* File just added to dir */
1347         EXT4_STATE_DELALLOC_RESERVED,   /* blks already reserved for delalloc */
1348 };
1349
1350 #define EXT4_INODE_BIT_FNS(name, field, offset)                         \
1351 static inline int ext4_test_inode_##name(struct inode *inode, int bit)  \
1352 {                                                                       \
1353         return test_bit(bit + (offset), &EXT4_I(inode)->i_##field);     \
1354 }                                                                       \
1355 static inline void ext4_set_inode_##name(struct inode *inode, int bit)  \
1356 {                                                                       \
1357         set_bit(bit + (offset), &EXT4_I(inode)->i_##field);             \
1358 }                                                                       \
1359 static inline void ext4_clear_inode_##name(struct inode *inode, int bit) \
1360 {                                                                       \
1361         clear_bit(bit + (offset), &EXT4_I(inode)->i_##field);           \
1362 }
1363
1364 EXT4_INODE_BIT_FNS(flag, flags, 0)
1365 #if (BITS_PER_LONG < 64)
1366 EXT4_INODE_BIT_FNS(state, state_flags, 0)
1367
1368 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1369 {
1370         (ei)->i_state_flags = 0;
1371 }
1372 #else
1373 EXT4_INODE_BIT_FNS(state, flags, 32)
1374
1375 static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
1376 {
1377         /* We depend on the fact that callers will set i_flags */
1378 }
1379 #endif
1380 #else
1381 /* Assume that user mode programs are passing in an ext4fs superblock, not
1382  * a kernel struct super_block.  This will allow us to call the feature-test
1383  * macros from user land. */
1384 #define EXT4_SB(sb)     (sb)
1385 #endif
1386
1387 #define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
1388
1389 /*
1390  * Codes for operating systems
1391  */
1392 #define EXT4_OS_LINUX           0
1393 #define EXT4_OS_HURD            1
1394 #define EXT4_OS_MASIX           2
1395 #define EXT4_OS_FREEBSD         3
1396 #define EXT4_OS_LITES           4
1397
1398 /*
1399  * Revision levels
1400  */
1401 #define EXT4_GOOD_OLD_REV       0       /* The good old (original) format */
1402 #define EXT4_DYNAMIC_REV        1       /* V2 format w/ dynamic inode sizes */
1403
1404 #define EXT4_CURRENT_REV        EXT4_GOOD_OLD_REV
1405 #define EXT4_MAX_SUPP_REV       EXT4_DYNAMIC_REV
1406
1407 #define EXT4_GOOD_OLD_INODE_SIZE 128
1408
1409 /*
1410  * Feature set definitions
1411  */
1412
1413 #define EXT4_HAS_COMPAT_FEATURE(sb,mask)                        \
1414         ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0)
1415 #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask)                     \
1416         ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0)
1417 #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask)                      \
1418         ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0)
1419 #define EXT4_SET_COMPAT_FEATURE(sb,mask)                        \
1420         EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
1421 #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask)                     \
1422         EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
1423 #define EXT4_SET_INCOMPAT_FEATURE(sb,mask)                      \
1424         EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
1425 #define EXT4_CLEAR_COMPAT_FEATURE(sb,mask)                      \
1426         EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
1427 #define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask)                   \
1428         EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
1429 #define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask)                    \
1430         EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
1431
1432 #define EXT4_FEATURE_COMPAT_DIR_PREALLOC        0x0001
1433 #define EXT4_FEATURE_COMPAT_IMAGIC_INODES       0x0002
1434 #define EXT4_FEATURE_COMPAT_HAS_JOURNAL         0x0004
1435 #define EXT4_FEATURE_COMPAT_EXT_ATTR            0x0008
1436 #define EXT4_FEATURE_COMPAT_RESIZE_INODE        0x0010
1437 #define EXT4_FEATURE_COMPAT_DIR_INDEX           0x0020
1438
1439 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER     0x0001
1440 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE       0x0002
1441 #define EXT4_FEATURE_RO_COMPAT_BTREE_DIR        0x0004
1442 #define EXT4_FEATURE_RO_COMPAT_HUGE_FILE        0x0008
1443 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM         0x0010
1444 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK        0x0020
1445 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE      0x0040
1446 #define EXT4_FEATURE_RO_COMPAT_QUOTA            0x0100
1447 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC         0x0200
1448
1449 #define EXT4_FEATURE_INCOMPAT_COMPRESSION       0x0001
1450 #define EXT4_FEATURE_INCOMPAT_FILETYPE          0x0002
1451 #define EXT4_FEATURE_INCOMPAT_RECOVER           0x0004 /* Needs recovery */
1452 #define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV       0x0008 /* Journal device */
1453 #define EXT4_FEATURE_INCOMPAT_META_BG           0x0010
1454 #define EXT4_FEATURE_INCOMPAT_EXTENTS           0x0040 /* extents support */
1455 #define EXT4_FEATURE_INCOMPAT_64BIT             0x0080
1456 #define EXT4_FEATURE_INCOMPAT_MMP               0x0100
1457 #define EXT4_FEATURE_INCOMPAT_FLEX_BG           0x0200
1458 #define EXT4_FEATURE_INCOMPAT_EA_INODE          0x0400 /* EA in inode */
1459 #define EXT4_FEATURE_INCOMPAT_DIRDATA           0x1000 /* data in dirent */
1460
1461 #define EXT2_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1462 #define EXT2_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1463                                          EXT4_FEATURE_INCOMPAT_META_BG)
1464 #define EXT2_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1465                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1466                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1467
1468 #define EXT3_FEATURE_COMPAT_SUPP        EXT4_FEATURE_COMPAT_EXT_ATTR
1469 #define EXT3_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1470                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1471                                          EXT4_FEATURE_INCOMPAT_META_BG)
1472 #define EXT3_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1473                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1474                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
1475
1476 #define EXT4_FEATURE_COMPAT_SUPP        EXT2_FEATURE_COMPAT_EXT_ATTR
1477 #define EXT4_FEATURE_INCOMPAT_SUPP      (EXT4_FEATURE_INCOMPAT_FILETYPE| \
1478                                          EXT4_FEATURE_INCOMPAT_RECOVER| \
1479                                          EXT4_FEATURE_INCOMPAT_META_BG| \
1480                                          EXT4_FEATURE_INCOMPAT_EXTENTS| \
1481                                          EXT4_FEATURE_INCOMPAT_64BIT| \
1482                                          EXT4_FEATURE_INCOMPAT_FLEX_BG| \
1483                                          EXT4_FEATURE_INCOMPAT_MMP)
1484 #define EXT4_FEATURE_RO_COMPAT_SUPP     (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
1485                                          EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
1486                                          EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
1487                                          EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
1488                                          EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
1489                                          EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
1490                                          EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
1491                                          EXT4_FEATURE_RO_COMPAT_BIGALLOC)
1492
1493 /*
1494  * Default values for user and/or group using reserved blocks
1495  */
1496 #define EXT4_DEF_RESUID         0
1497 #define EXT4_DEF_RESGID         0
1498
1499 #define EXT4_DEF_INODE_READAHEAD_BLKS   32
1500
1501 /*
1502  * Default mount options
1503  */
1504 #define EXT4_DEFM_DEBUG         0x0001
1505 #define EXT4_DEFM_BSDGROUPS     0x0002
1506 #define EXT4_DEFM_XATTR_USER    0x0004
1507 #define EXT4_DEFM_ACL           0x0008
1508 #define EXT4_DEFM_UID16         0x0010
1509 #define EXT4_DEFM_JMODE         0x0060
1510 #define EXT4_DEFM_JMODE_DATA    0x0020
1511 #define EXT4_DEFM_JMODE_ORDERED 0x0040
1512 #define EXT4_DEFM_JMODE_WBACK   0x0060
1513 #define EXT4_DEFM_NOBARRIER     0x0100
1514 #define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1515 #define EXT4_DEFM_DISCARD       0x0400
1516 #define EXT4_DEFM_NODELALLOC    0x0800
1517
1518 /*
1519  * Default journal batch times
1520  */
1521 #define EXT4_DEF_MIN_BATCH_TIME 0
1522 #define EXT4_DEF_MAX_BATCH_TIME 15000 /* 15ms */
1523
1524 /*
1525  * Minimum number of groups in a flexgroup before we separate out
1526  * directories into the first block group of a flexgroup
1527  */
1528 #define EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME 4
1529
1530 /*
1531  * Structure of a directory entry
1532  */
1533 #define EXT4_NAME_LEN 255
1534
1535 struct ext4_dir_entry {
1536         __le32  inode;                  /* Inode number */
1537         __le16  rec_len;                /* Directory entry length */
1538         __le16  name_len;               /* Name length */
1539         char    name[EXT4_NAME_LEN];    /* File name */
1540 };
1541
1542 /*
1543  * The new version of the directory entry.  Since EXT4 structures are
1544  * stored in intel byte order, and the name_len field could never be
1545  * bigger than 255 chars, it's safe to reclaim the extra byte for the
1546  * file_type field.
1547  */
1548 struct ext4_dir_entry_2 {
1549         __le32  inode;                  /* Inode number */
1550         __le16  rec_len;                /* Directory entry length */
1551         __u8    name_len;               /* Name length */
1552         __u8    file_type;
1553         char    name[EXT4_NAME_LEN];    /* File name */
1554 };
1555
1556 /*
1557  * Ext4 directory file types.  Only the low 3 bits are used.  The
1558  * other bits are reserved for now.
1559  */
1560 #define EXT4_FT_UNKNOWN         0
1561 #define EXT4_FT_REG_FILE        1
1562 #define EXT4_FT_DIR             2
1563 #define EXT4_FT_CHRDEV          3
1564 #define EXT4_FT_BLKDEV          4
1565 #define EXT4_FT_FIFO            5
1566 #define EXT4_FT_SOCK            6
1567 #define EXT4_FT_SYMLINK         7
1568
1569 #define EXT4_FT_MAX             8
1570
1571 /*
1572  * EXT4_DIR_PAD defines the directory entries boundaries
1573  *
1574  * NOTE: It must be a multiple of 4
1575  */
1576 #define EXT4_DIR_PAD                    4
1577 #define EXT4_DIR_ROUND                  (EXT4_DIR_PAD - 1)
1578 #define EXT4_DIR_REC_LEN(name_len)      (((name_len) + 8 + EXT4_DIR_ROUND) & \
1579                                          ~EXT4_DIR_ROUND)
1580 #define EXT4_MAX_REC_LEN                ((1<<16)-1)
1581
1582 /*
1583  * If we ever get support for fs block sizes > page_size, we'll need
1584  * to remove the #if statements in the next two functions...
1585  */
1586 static inline unsigned int
1587 ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1588 {
1589         unsigned len = le16_to_cpu(dlen);
1590
1591 #if (PAGE_CACHE_SIZE >= 65536)
1592         if (len == EXT4_MAX_REC_LEN || len == 0)
1593                 return blocksize;
1594         return (len & 65532) | ((len & 3) << 16);
1595 #else
1596         return len;
1597 #endif
1598 }
1599
1600 static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1601 {
1602         if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1603                 BUG();
1604 #if (PAGE_CACHE_SIZE >= 65536)
1605         if (len < 65536)
1606                 return cpu_to_le16(len);
1607         if (len == blocksize) {
1608                 if (blocksize == 65536)
1609                         return cpu_to_le16(EXT4_MAX_REC_LEN);
1610                 else
1611                         return cpu_to_le16(0);
1612         }
1613         return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1614 #else
1615         return cpu_to_le16(len);
1616 #endif
1617 }
1618
1619 /*
1620  * Hash Tree Directory indexing
1621  * (c) Daniel Phillips, 2001
1622  */
1623
1624 #define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
1625                                       EXT4_FEATURE_COMPAT_DIR_INDEX) && \
1626                     ext4_test_inode_flag((dir), EXT4_INODE_INDEX))
1627 #define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
1628 #define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
1629
1630 /* Legal values for the dx_root hash_version field: */
1631
1632 #define DX_HASH_LEGACY          0
1633 #define DX_HASH_HALF_MD4        1
1634 #define DX_HASH_TEA             2
1635 #define DX_HASH_LEGACY_UNSIGNED 3
1636 #define DX_HASH_HALF_MD4_UNSIGNED       4
1637 #define DX_HASH_TEA_UNSIGNED            5
1638
1639 #ifdef __KERNEL__
1640
1641 /* hash info structure used by the directory hash */
1642 struct dx_hash_info
1643 {
1644         u32             hash;
1645         u32             minor_hash;
1646         int             hash_version;
1647         u32             *seed;
1648 };
1649
1650
1651 /* 32 and 64 bit signed EOF for dx directories */
1652 #define EXT4_HTREE_EOF_32BIT   ((1UL  << (32 - 1)) - 1)
1653 #define EXT4_HTREE_EOF_64BIT   ((1ULL << (64 - 1)) - 1)
1654
1655
1656 /*
1657  * Control parameters used by ext4_htree_next_block
1658  */
1659 #define HASH_NB_ALWAYS          1
1660
1661
1662 /*
1663  * Describe an inode's exact location on disk and in memory
1664  */
1665 struct ext4_iloc
1666 {
1667         struct buffer_head *bh;
1668         unsigned long offset;
1669         ext4_group_t block_group;
1670 };
1671
1672 static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
1673 {
1674         return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
1675 }
1676
1677 /*
1678  * This structure is stuffed into the struct file's private_data field
1679  * for directories.  It is where we put information so that we can do
1680  * readdir operations in hash tree order.
1681  */
1682 struct dir_private_info {
1683         struct rb_root  root;
1684         struct rb_node  *curr_node;
1685         struct fname    *extra_fname;
1686         loff_t          last_pos;
1687         __u32           curr_hash;
1688         __u32           curr_minor_hash;
1689         __u32           next_hash;
1690 };
1691
1692 /* calculate the first block number of the group */
1693 static inline ext4_fsblk_t
1694 ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1695 {
1696         return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
1697                 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1698 }
1699
1700 /*
1701  * Special error return code only used by dx_probe() and its callers.
1702  */
1703 #define ERR_BAD_DX_DIR  -75000
1704
1705 void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1706                         ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1707
1708 /*
1709  * Timeout and state flag for lazy initialization inode thread.
1710  */
1711 #define EXT4_DEF_LI_WAIT_MULT                   10
1712 #define EXT4_DEF_LI_MAX_START_DELAY             5
1713 #define EXT4_LAZYINIT_QUIT                      0x0001
1714 #define EXT4_LAZYINIT_RUNNING                   0x0002
1715
1716 /*
1717  * Lazy inode table initialization info
1718  */
1719 struct ext4_lazy_init {
1720         unsigned long           li_state;
1721         struct list_head        li_request_list;
1722         struct mutex            li_list_mtx;
1723 };
1724
1725 struct ext4_li_request {
1726         struct super_block      *lr_super;
1727         struct ext4_sb_info     *lr_sbi;
1728         ext4_group_t            lr_next_group;
1729         struct list_head        lr_request;
1730         unsigned long           lr_next_sched;
1731         unsigned long           lr_timeout;
1732 };
1733
1734 struct ext4_features {
1735         struct kobject f_kobj;
1736         struct completion f_kobj_unregister;
1737 };
1738
1739 /*
1740  * This structure will be used for multiple mount protection. It will be
1741  * written into the block number saved in the s_mmp_block field in the
1742  * superblock. Programs that check MMP should assume that if
1743  * SEQ_FSCK (or any unknown code above SEQ_MAX) is present then it is NOT safe
1744  * to use the filesystem, regardless of how old the timestamp is.
1745  */
1746 #define EXT4_MMP_MAGIC     0x004D4D50U /* ASCII for MMP */
1747 #define EXT4_MMP_SEQ_CLEAN 0xFF4D4D50U /* mmp_seq value for clean unmount */
1748 #define EXT4_MMP_SEQ_FSCK  0xE24D4D50U /* mmp_seq value when being fscked */
1749 #define EXT4_MMP_SEQ_MAX   0xE24D4D4FU /* maximum valid mmp_seq value */
1750
1751 struct mmp_struct {
1752         __le32  mmp_magic;              /* Magic number for MMP */
1753         __le32  mmp_seq;                /* Sequence no. updated periodically */
1754
1755         /*
1756          * mmp_time, mmp_nodename & mmp_bdevname are only used for information
1757          * purposes and do not affect the correctness of the algorithm
1758          */
1759         __le64  mmp_time;               /* Time last updated */
1760         char    mmp_nodename[64];       /* Node which last updated MMP block */
1761         char    mmp_bdevname[32];       /* Bdev which last updated MMP block */
1762
1763         /*
1764          * mmp_check_interval is used to verify if the MMP block has been
1765          * updated on the block device. The value is updated based on the
1766          * maximum time to write the MMP block during an update cycle.
1767          */
1768         __le16  mmp_check_interval;
1769
1770         __le16  mmp_pad1;
1771         __le32  mmp_pad2[227];
1772 };
1773
1774 /* arguments passed to the mmp thread */
1775 struct mmpd_data {
1776         struct buffer_head *bh; /* bh from initial read_mmp_block() */
1777         struct super_block *sb;  /* super block of the fs */
1778 };
1779
1780 /*
1781  * Check interval multiplier
1782  * The MMP block is written every update interval and initially checked every
1783  * update interval x the multiplier (the value is then adapted based on the
1784  * write latency). The reason is that writes can be delayed under load and we
1785  * don't want readers to incorrectly assume that the filesystem is no longer
1786  * in use.
1787  */
1788 #define EXT4_MMP_CHECK_MULT             2UL
1789
1790 /*
1791  * Minimum interval for MMP checking in seconds.
1792  */
1793 #define EXT4_MMP_MIN_CHECK_INTERVAL     5UL
1794
1795 /*
1796  * Maximum interval for MMP checking in seconds.
1797  */
1798 #define EXT4_MMP_MAX_CHECK_INTERVAL     300UL
1799
1800 /*
1801  * Function prototypes
1802  */
1803
1804 /*
1805  * Ok, these declarations are also in <linux/kernel.h> but none of the
1806  * ext4 source programs needs to include it so they are duplicated here.
1807  */
1808 # define NORET_TYPE     /**/
1809 # define ATTRIB_NORET   __attribute__((noreturn))
1810 # define NORET_AND      noreturn,
1811
1812 /* bitmap.c */
1813 extern unsigned int ext4_count_free(char *bitmap, unsigned numchars);
1814
1815 /* balloc.c */
1816 extern unsigned int ext4_block_group(struct super_block *sb,
1817                         ext4_fsblk_t blocknr);
1818 extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
1819                         ext4_fsblk_t blocknr);
1820 extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
1821 extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1822                         ext4_group_t group);
1823 extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1824                                          ext4_fsblk_t goal,
1825                                          unsigned int flags,
1826                                          unsigned long *count,
1827                                          int *errp);
1828 extern int ext4_claim_free_clusters(struct ext4_sb_info *sbi,
1829                                     s64 nclusters, unsigned int flags);
1830 extern ext4_fsblk_t ext4_count_free_clusters(struct super_block *);
1831 extern void ext4_check_blocks_bitmap(struct super_block *);
1832 extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
1833                                                     ext4_group_t block_group,
1834                                                     struct buffer_head ** bh);
1835 extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
1836 struct buffer_head *ext4_read_block_bitmap(struct super_block *sb,
1837                                       ext4_group_t block_group);
1838 extern void ext4_init_block_bitmap(struct super_block *sb,
1839                                    struct buffer_head *bh,
1840                                    ext4_group_t group,
1841                                    struct ext4_group_desc *desc);
1842 extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
1843                                               ext4_group_t block_group,
1844                                               struct ext4_group_desc *gdp);
1845 extern unsigned ext4_num_base_meta_clusters(struct super_block *sb,
1846                                             ext4_group_t block_group);
1847 extern unsigned ext4_num_overhead_clusters(struct super_block *sb,
1848                                            ext4_group_t block_group,
1849                                            struct ext4_group_desc *gdp);
1850 ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
1851
1852 /* dir.c */
1853 extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1854                                   struct file *,
1855                                   struct ext4_dir_entry_2 *,
1856                                   struct buffer_head *, unsigned int);
1857 #define ext4_check_dir_entry(dir, filp, de, bh, offset)                 \
1858         unlikely(__ext4_check_dir_entry(__func__, __LINE__, (dir), (filp), \
1859                                         (de), (bh), (offset)))
1860 extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1861                                     __u32 minor_hash,
1862                                     struct ext4_dir_entry_2 *dirent);
1863 extern void ext4_htree_free_dir_info(struct dir_private_info *p);
1864
1865 /* fsync.c */
1866 extern int ext4_sync_file(struct file *, loff_t, loff_t, int);
1867 extern int ext4_flush_completed_IO(struct inode *);
1868
1869 /* hash.c */
1870 extern int ext4fs_dirhash(const char *name, int len, struct
1871                           dx_hash_info *hinfo);
1872
1873 /* ialloc.c */
1874 extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
1875                                     const struct qstr *qstr, __u32 goal,
1876                                     uid_t *owner);
1877 extern void ext4_free_inode(handle_t *, struct inode *);
1878 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1879 extern unsigned long ext4_count_free_inodes(struct super_block *);
1880 extern unsigned long ext4_count_dirs(struct super_block *);
1881 extern void ext4_check_inodes_bitmap(struct super_block *);
1882 extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1883 extern int ext4_init_inode_table(struct super_block *sb,
1884                                  ext4_group_t group, int barrier);
1885
1886 /* mballoc.c */
1887 extern long ext4_mb_stats;
1888 extern long ext4_mb_max_to_scan;
1889 extern int ext4_mb_init(struct super_block *, int);
1890 extern int ext4_mb_release(struct super_block *);
1891 extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1892                                 struct ext4_allocation_request *, int *);
1893 extern int ext4_mb_reserve_blocks(struct super_block *, int);
1894 extern void ext4_discard_preallocations(struct inode *);
1895 extern int __init ext4_init_mballoc(void);
1896 extern void ext4_exit_mballoc(void);
1897 extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1898                              struct buffer_head *bh, ext4_fsblk_t block,
1899                              unsigned long count, int flags);
1900 extern int ext4_mb_add_groupinfo(struct super_block *sb,
1901                 ext4_group_t i, struct ext4_group_desc *desc);
1902 extern int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
1903                                 ext4_fsblk_t block, unsigned long count);
1904 extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1905
1906 /* inode.c */
1907 struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1908                                                 ext4_lblk_t, int, int *);
1909 struct buffer_head *ext4_bread(handle_t *, struct inode *,
1910                                                 ext4_lblk_t, int, int *);
1911 int ext4_get_block(struct inode *inode, sector_t iblock,
1912                                 struct buffer_head *bh_result, int create);
1913
1914 extern struct inode *ext4_iget(struct super_block *, unsigned long);
1915 extern struct inode *ext4_iget_normal(struct super_block *, unsigned long);
1916 extern int  ext4_write_inode(struct inode *, struct writeback_control *);
1917 extern int  ext4_setattr(struct dentry *, struct iattr *);
1918 extern int  ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1919                                 struct kstat *stat);
1920 extern void ext4_evict_inode(struct inode *);
1921 extern void ext4_clear_inode(struct inode *);
1922 extern int  ext4_sync_inode(handle_t *, struct inode *);
1923 extern void ext4_dirty_inode(struct inode *, int);
1924 extern int ext4_change_inode_journal_flag(struct inode *, int);
1925 extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1926 extern int ext4_can_truncate(struct inode *inode);
1927 extern void ext4_truncate(struct inode *);
1928 extern int ext4_punch_hole(struct file *file, loff_t offset, loff_t length);
1929 extern int ext4_truncate_restart_trans(handle_t *, struct inode *, int nblocks);
1930 extern void ext4_set_inode_flags(struct inode *);
1931 extern void ext4_get_inode_flags(struct ext4_inode_info *);
1932 extern int ext4_alloc_da_blocks(struct inode *inode);
1933 extern void ext4_set_aops(struct inode *inode);
1934 extern int ext4_writepage_trans_blocks(struct inode *);
1935 extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1936 extern int ext4_block_truncate_page(handle_t *handle,
1937                 struct address_space *mapping, loff_t from);
1938 extern int ext4_block_zero_page_range(handle_t *handle,
1939                 struct address_space *mapping, loff_t from, loff_t length);
1940 extern int ext4_discard_partial_page_buffers(handle_t *handle,
1941                 struct address_space *mapping, loff_t from,
1942                 loff_t length, int flags);
1943 extern int ext4_discard_partial_page_buffers_no_lock(handle_t *handle,
1944                 struct inode *inode, struct page *page, loff_t from,
1945                 loff_t length, int flags);
1946 extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1947 extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1948 extern void ext4_da_update_reserve_space(struct inode *inode,
1949                                         int used, int quota_claim);
1950
1951 /* indirect.c */
1952 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
1953                                 struct ext4_map_blocks *map, int flags);
1954 extern ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
1955                                 const struct iovec *iov, loff_t offset,
1956                                 unsigned long nr_segs);
1957 extern int ext4_ind_calc_metadata_amount(struct inode *inode, sector_t lblock);
1958 extern int ext4_ind_trans_blocks(struct inode *inode, int nrblocks, int chunk);
1959 extern void ext4_ind_truncate(struct inode *inode);
1960
1961 /* ioctl.c */
1962 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1963 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1964
1965 /* migrate.c */
1966 extern int ext4_ext_migrate(struct inode *);
1967
1968 /* namei.c */
1969 extern int ext4_orphan_add(handle_t *, struct inode *);
1970 extern int ext4_orphan_del(handle_t *, struct inode *);
1971 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1972                                 __u32 start_minor_hash, __u32 *next_hash);
1973
1974 /* resize.c */
1975 extern int ext4_group_add(struct super_block *sb,
1976                                 struct ext4_new_group_data *input);
1977 extern int ext4_group_extend(struct super_block *sb,
1978                                 struct ext4_super_block *es,
1979                                 ext4_fsblk_t n_blocks_count);
1980
1981 /* super.c */
1982 extern int ext4_calculate_overhead(struct super_block *sb);
1983 extern void *ext4_kvmalloc(size_t size, gfp_t flags);
1984 extern void *ext4_kvzalloc(size_t size, gfp_t flags);
1985 extern void ext4_kvfree(void *ptr);
1986 extern __printf(4, 5)
1987 void __ext4_error(struct super_block *, const char *, unsigned int,
1988                   const char *, ...);
1989 #define ext4_error(sb, message...)      __ext4_error(sb, __func__,      \
1990                                                      __LINE__, ## message)
1991 extern __printf(5, 6)
1992 void ext4_error_inode(struct inode *, const char *, unsigned int, ext4_fsblk_t,
1993                       const char *, ...);
1994 extern __printf(5, 6)
1995 void ext4_error_file(struct file *, const char *, unsigned int, ext4_fsblk_t,
1996                      const char *, ...);
1997 extern void __ext4_std_error(struct super_block *, const char *,
1998                              unsigned int, int);
1999 extern __printf(4, 5)
2000 void __ext4_abort(struct super_block *, const char *, unsigned int,
2001                   const char *, ...);
2002 #define ext4_abort(sb, message...)      __ext4_abort(sb, __func__, \
2003                                                        __LINE__, ## message)
2004 extern __printf(4, 5)
2005 void __ext4_warning(struct super_block *, const char *, unsigned int,
2006                     const char *, ...);
2007 #define ext4_warning(sb, message...)    __ext4_warning(sb, __func__, \
2008                                                        __LINE__, ## message)
2009 extern __printf(3, 4)
2010 void ext4_msg(struct super_block *, const char *, const char *, ...);
2011 extern void __dump_mmp_msg(struct super_block *, struct mmp_struct *mmp,
2012                            const char *, unsigned int, const char *);
2013 #define dump_mmp_msg(sb, mmp, msg)      __dump_mmp_msg(sb, mmp, __func__, \
2014                                                        __LINE__, msg)
2015 extern __printf(7, 8)
2016 void __ext4_grp_locked_error(const char *, unsigned int,
2017                              struct super_block *, ext4_group_t,
2018                              unsigned long, ext4_fsblk_t,
2019                              const char *, ...);
2020 #define ext4_grp_locked_error(sb, grp, message...) \
2021         __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
2022 extern void ext4_update_dynamic_rev(struct super_block *sb);
2023 extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
2024                                         __u32 compat);
2025 extern int ext4_update_rocompat_feature(handle_t *handle,
2026                                         struct super_block *sb, __u32 rocompat);
2027 extern int ext4_update_incompat_feature(handle_t *handle,
2028                                         struct super_block *sb, __u32 incompat);
2029 extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
2030                                       struct ext4_group_desc *bg);
2031 extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
2032                                       struct ext4_group_desc *bg);
2033 extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
2034                                      struct ext4_group_desc *bg);
2035 extern __u32 ext4_free_group_clusters(struct super_block *sb,
2036                                       struct ext4_group_desc *bg);
2037 extern __u32 ext4_free_inodes_count(struct super_block *sb,
2038                                  struct ext4_group_desc *bg);
2039 extern __u32 ext4_used_dirs_count(struct super_block *sb,
2040                                 struct ext4_group_desc *bg);
2041 extern __u32 ext4_itable_unused_count(struct super_block *sb,
2042                                    struct ext4_group_desc *bg);
2043 extern void ext4_block_bitmap_set(struct super_block *sb,
2044                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2045 extern void ext4_inode_bitmap_set(struct super_block *sb,
2046                                   struct ext4_group_desc *bg, ext4_fsblk_t blk);
2047 extern void ext4_inode_table_set(struct super_block *sb,
2048                                  struct ext4_group_desc *bg, ext4_fsblk_t blk);
2049 extern void ext4_free_group_clusters_set(struct super_block *sb,
2050                                          struct ext4_group_desc *bg,
2051                                          __u32 count);
2052 extern void ext4_free_inodes_set(struct super_block *sb,
2053                                 struct ext4_group_desc *bg, __u32 count);
2054 extern void ext4_used_dirs_set(struct super_block *sb,
2055                                 struct ext4_group_desc *bg, __u32 count);
2056 extern void ext4_itable_unused_set(struct super_block *sb,
2057                                    struct ext4_group_desc *bg, __u32 count);
2058 extern __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 group,
2059                                    struct ext4_group_desc *gdp);
2060 extern int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 group,
2061                                        struct ext4_group_desc *gdp);
2062
2063 static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
2064 {
2065         return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
2066                 le32_to_cpu(es->s_blocks_count_lo);
2067 }
2068
2069 static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
2070 {
2071         return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
2072                 le32_to_cpu(es->s_r_blocks_count_lo);
2073 }
2074
2075 static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
2076 {
2077         return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
2078                 le32_to_cpu(es->s_free_blocks_count_lo);
2079 }
2080
2081 static inline void ext4_blocks_count_set(struct ext4_super_block *es,
2082                                          ext4_fsblk_t blk)
2083 {
2084         es->s_blocks_count_lo = cpu_to_le32((u32)blk);
2085         es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
2086 }
2087
2088 static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
2089                                               ext4_fsblk_t blk)
2090 {
2091         es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
2092         es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
2093 }
2094
2095 static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
2096                                            ext4_fsblk_t blk)
2097 {
2098         es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
2099         es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
2100 }
2101
2102 static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
2103 {
2104         if (S_ISREG(le16_to_cpu(raw_inode->i_mode)))
2105                 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
2106                         le32_to_cpu(raw_inode->i_size_lo);
2107         else
2108                 return (loff_t) le32_to_cpu(raw_inode->i_size_lo);
2109 }
2110
2111 static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
2112 {
2113         raw_inode->i_size_lo = cpu_to_le32(i_size);
2114         raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
2115 }
2116
2117 static inline
2118 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
2119                                             ext4_group_t group)
2120 {
2121          struct ext4_group_info ***grp_info;
2122          long indexv, indexh;
2123          grp_info = EXT4_SB(sb)->s_group_info;
2124          indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
2125          indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
2126          return grp_info[indexv][indexh];
2127 }
2128
2129 /*
2130  * Reading s_groups_count requires using smp_rmb() afterwards.  See
2131  * the locking protocol documented in the comments of ext4_group_add()
2132  * in resize.c
2133  */
2134 static inline ext4_group_t ext4_get_groups_count(struct super_block *sb)
2135 {
2136         ext4_group_t    ngroups = EXT4_SB(sb)->s_groups_count;
2137
2138         smp_rmb();
2139         return ngroups;
2140 }
2141
2142 static inline ext4_group_t ext4_flex_group(struct ext4_sb_info *sbi,
2143                                              ext4_group_t block_group)
2144 {
2145         return block_group >> sbi->s_log_groups_per_flex;
2146 }
2147
2148 static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
2149 {
2150         return 1 << sbi->s_log_groups_per_flex;
2151 }
2152
2153 #define ext4_std_error(sb, errno)                               \
2154 do {                                                            \
2155         if ((errno))                                            \
2156                 __ext4_std_error((sb), __func__, __LINE__, (errno));    \
2157 } while (0)
2158
2159 #ifdef CONFIG_SMP
2160 /* Each CPU can accumulate percpu_counter_batch clusters in their local
2161  * counters. So we need to make sure we have free clusters more
2162  * than percpu_counter_batch  * nr_cpu_ids. Also add a window of 4 times.
2163  */
2164 #define EXT4_FREECLUSTERS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
2165 #else
2166 #define EXT4_FREECLUSTERS_WATERMARK 0
2167 #endif
2168
2169 static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize)
2170 {
2171         /*
2172          * XXX: replace with spinlock if seen contended -bzzz
2173          */
2174         down_write(&EXT4_I(inode)->i_data_sem);
2175         if (newsize > EXT4_I(inode)->i_disksize)
2176                 EXT4_I(inode)->i_disksize = newsize;
2177         up_write(&EXT4_I(inode)->i_data_sem);
2178         return ;
2179 }
2180
2181 struct ext4_group_info {
2182         unsigned long   bb_state;
2183         struct rb_root  bb_free_root;
2184         ext4_grpblk_t   bb_first_free;  /* first free block */
2185         ext4_grpblk_t   bb_free;        /* total free blocks */
2186         ext4_grpblk_t   bb_fragments;   /* nr of freespace fragments */
2187         ext4_grpblk_t   bb_largest_free_order;/* order of largest frag in BG */
2188         struct          list_head bb_prealloc_list;
2189 #ifdef DOUBLE_CHECK
2190         void            *bb_bitmap;
2191 #endif
2192         struct rw_semaphore alloc_sem;
2193         ext4_grpblk_t   bb_counters[];  /* Nr of free power-of-two-block
2194                                          * regions, index is order.
2195                                          * bb_counters[3] = 5 means
2196                                          * 5 free 8-block regions. */
2197 };
2198
2199 #define EXT4_GROUP_INFO_NEED_INIT_BIT           0
2200 #define EXT4_GROUP_INFO_WAS_TRIMMED_BIT         1
2201
2202 #define EXT4_MB_GRP_NEED_INIT(grp)      \
2203         (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
2204
2205 #define EXT4_MB_GRP_WAS_TRIMMED(grp)    \
2206         (test_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2207 #define EXT4_MB_GRP_SET_TRIMMED(grp)    \
2208         (set_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2209 #define EXT4_MB_GRP_CLEAR_TRIMMED(grp)  \
2210         (clear_bit(EXT4_GROUP_INFO_WAS_TRIMMED_BIT, &((grp)->bb_state)))
2211
2212 #define EXT4_MAX_CONTENTION             8
2213 #define EXT4_CONTENTION_THRESHOLD       2
2214
2215 static inline spinlock_t *ext4_group_lock_ptr(struct super_block *sb,
2216                                               ext4_group_t group)
2217 {
2218         return bgl_lock_ptr(EXT4_SB(sb)->s_blockgroup_lock, group);
2219 }
2220
2221 /*
2222  * Returns true if the filesystem is busy enough that attempts to
2223  * access the block group locks has run into contention.
2224  */
2225 static inline int ext4_fs_is_busy(struct ext4_sb_info *sbi)
2226 {
2227         return (atomic_read(&sbi->s_lock_busy) > EXT4_CONTENTION_THRESHOLD);
2228 }
2229
2230 static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
2231 {
2232         spinlock_t *lock = ext4_group_lock_ptr(sb, group);
2233         if (spin_trylock(lock))
2234                 /*
2235                  * We're able to grab the lock right away, so drop the
2236                  * lock contention counter.
2237                  */
2238                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, -1, 0);
2239         else {
2240                 /*
2241                  * The lock is busy, so bump the contention counter,
2242                  * and then wait on the spin lock.
2243                  */
2244                 atomic_add_unless(&EXT4_SB(sb)->s_lock_busy, 1,
2245                                   EXT4_MAX_CONTENTION);
2246                 spin_lock(lock);
2247         }
2248 }
2249
2250 static inline void ext4_unlock_group(struct super_block *sb,
2251                                         ext4_group_t group)
2252 {
2253         spin_unlock(ext4_group_lock_ptr(sb, group));
2254 }
2255
2256 static inline void ext4_mark_super_dirty(struct super_block *sb)
2257 {
2258         if (EXT4_SB(sb)->s_journal == NULL)
2259                 sb->s_dirt =1;
2260 }
2261
2262 /*
2263  * Block validity checking
2264  */
2265 #define ext4_check_indirect_blockref(inode, bh)                         \
2266         ext4_check_blockref(__func__, __LINE__, inode,                  \
2267                             (__le32 *)(bh)->b_data,                     \
2268                             EXT4_ADDR_PER_BLOCK((inode)->i_sb))
2269
2270 #define ext4_ind_check_inode(inode)                                     \
2271         ext4_check_blockref(__func__, __LINE__, inode,                  \
2272                             EXT4_I(inode)->i_data,                      \
2273                             EXT4_NDIR_BLOCKS)
2274
2275 /*
2276  * Inodes and files operations
2277  */
2278
2279 /* dir.c */
2280 extern const struct file_operations ext4_dir_operations;
2281
2282 /* file.c */
2283 extern const struct inode_operations ext4_file_inode_operations;
2284 extern const struct file_operations ext4_file_operations;
2285 extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
2286
2287 /* namei.c */
2288 extern const struct inode_operations ext4_dir_inode_operations;
2289 extern const struct inode_operations ext4_special_inode_operations;
2290 extern struct dentry *ext4_get_parent(struct dentry *child);
2291
2292 /* symlink.c */
2293 extern const struct inode_operations ext4_symlink_inode_operations;
2294 extern const struct inode_operations ext4_fast_symlink_inode_operations;
2295
2296 /* block_validity */
2297 extern void ext4_release_system_zone(struct super_block *sb);
2298 extern int ext4_setup_system_zone(struct super_block *sb);
2299 extern int __init ext4_init_system_zone(void);
2300 extern void ext4_exit_system_zone(void);
2301 extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
2302                                  ext4_fsblk_t start_blk,
2303                                  unsigned int count);
2304 extern int ext4_check_blockref(const char *, unsigned int,
2305                                struct inode *, __le32 *, unsigned int);
2306
2307 /* extents.c */
2308 extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
2309 extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
2310 extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
2311                                        int chunk);
2312 extern int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
2313                                struct ext4_map_blocks *map, int flags);
2314 extern void ext4_ext_truncate(struct inode *);
2315 extern int ext4_ext_punch_hole(struct file *file, loff_t offset,
2316                                 loff_t length);
2317 extern void ext4_ext_init(struct super_block *);
2318 extern void ext4_ext_release(struct super_block *);
2319 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
2320                           loff_t len);
2321 extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
2322                           ssize_t len);
2323 extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
2324                            struct ext4_map_blocks *map, int flags);
2325 extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2326                         __u64 start, __u64 len);
2327 /* move_extent.c */
2328 extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
2329                              __u64 start_orig, __u64 start_donor,
2330                              __u64 len, __u64 *moved_len);
2331
2332 /* page-io.c */
2333 extern int __init ext4_init_pageio(void);
2334 extern void ext4_exit_pageio(void);
2335 extern void ext4_ioend_wait(struct inode *);
2336 extern void ext4_free_io_end(ext4_io_end_t *io);
2337 extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2338 extern int ext4_end_io_nolock(ext4_io_end_t *io);
2339 extern void ext4_io_submit(struct ext4_io_submit *io);
2340 extern int ext4_bio_write_page(struct ext4_io_submit *io,
2341                                struct page *page,
2342                                int len,
2343                                struct writeback_control *wbc);
2344
2345 /* mmp.c */
2346 extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
2347
2348 /* BH_Uninit flag: blocks are allocated but uninitialized on disk */
2349 enum ext4_state_bits {
2350         BH_Uninit       /* blocks are allocated but uninitialized on disk */
2351           = BH_JBDPrivateStart,
2352         BH_AllocFromCluster,    /* allocated blocks were part of already
2353                                  * allocated cluster. Note that this flag will
2354                                  * never, ever appear in a buffer_head's state
2355                                  * flag. See EXT4_MAP_FROM_CLUSTER to see where
2356                                  * this is used. */
2357         BH_Da_Mapped,   /* Delayed allocated block that now has a mapping. This
2358                          * flag is set when ext4_map_blocks is called on a
2359                          * delayed allocated block to get its real mapping. */
2360 };
2361
2362 BUFFER_FNS(Uninit, uninit)
2363 TAS_BUFFER_FNS(Uninit, uninit)
2364 BUFFER_FNS(Da_Mapped, da_mapped)
2365
2366 /*
2367  * Add new method to test wether block and inode bitmaps are properly
2368  * initialized. With uninit_bg reading the block from disk is not enough
2369  * to mark the bitmap uptodate. We need to also zero-out the bitmap
2370  */
2371 #define BH_BITMAP_UPTODATE BH_JBDPrivateStart
2372
2373 static inline int bitmap_uptodate(struct buffer_head *bh)
2374 {
2375         return (buffer_uptodate(bh) &&
2376                         test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state));
2377 }
2378 static inline void set_bitmap_uptodate(struct buffer_head *bh)
2379 {
2380         set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state);
2381 }
2382
2383 #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
2384
2385 /* For ioend & aio unwritten conversion wait queues */
2386 #define EXT4_WQ_HASH_SZ         37
2387 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
2388                                             EXT4_WQ_HASH_SZ])
2389 #define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
2390                                              EXT4_WQ_HASH_SZ])
2391 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
2392 extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
2393
2394 #define EXT4_RESIZING   0
2395 extern int ext4_resize_begin(struct super_block *sb);
2396 extern void ext4_resize_end(struct super_block *sb);
2397
2398 #endif  /* __KERNEL__ */
2399
2400 #include "ext4_extents.h"
2401
2402 #endif  /* _EXT4_H */