Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[cascardo/linux.git] / fs / fat / inode.c
1 /*
2  *  linux/fs/fat/inode.c
3  *
4  *  Written 1992,1993 by Werner Almesberger
5  *  VFAT extensions by Gordon Chaffee, merged with msdos fs by Henrik Storner
6  *  Rewritten for the constant inumbers support by Al Viro
7  *
8  *  Fixes:
9  *
10  *      Max Cohan: Fixed invalid FSINFO offset when info_sector is 0
11  */
12
13 #include <linux/module.h>
14 #include <linux/init.h>
15 #include <linux/time.h>
16 #include <linux/slab.h>
17 #include <linux/seq_file.h>
18 #include <linux/pagemap.h>
19 #include <linux/mpage.h>
20 #include <linux/buffer_head.h>
21 #include <linux/mount.h>
22 #include <linux/aio.h>
23 #include <linux/vfs.h>
24 #include <linux/parser.h>
25 #include <linux/uio.h>
26 #include <linux/writeback.h>
27 #include <linux/log2.h>
28 #include <linux/hash.h>
29 #include <linux/blkdev.h>
30 #include <asm/unaligned.h>
31 #include "fat.h"
32
33 #ifndef CONFIG_FAT_DEFAULT_IOCHARSET
34 /* if user don't select VFAT, this is undefined. */
35 #define CONFIG_FAT_DEFAULT_IOCHARSET    ""
36 #endif
37
38 #define KB_IN_SECTORS 2
39
40 /*
41  * A deserialized copy of the on-disk structure laid out in struct
42  * fat_boot_sector.
43  */
44 struct fat_bios_param_block {
45         u16     fat_sector_size;
46         u8      fat_sec_per_clus;
47         u16     fat_reserved;
48         u8      fat_fats;
49         u16     fat_dir_entries;
50         u16     fat_sectors;
51         u16     fat_fat_length;
52         u32     fat_total_sect;
53
54         u8      fat16_state;
55         u32     fat16_vol_id;
56
57         u32     fat32_length;
58         u32     fat32_root_cluster;
59         u16     fat32_info_sector;
60         u8      fat32_state;
61         u32     fat32_vol_id;
62 };
63
64 static int fat_default_codepage = CONFIG_FAT_DEFAULT_CODEPAGE;
65 static char fat_default_iocharset[] = CONFIG_FAT_DEFAULT_IOCHARSET;
66
67 static struct fat_floppy_defaults {
68         unsigned nr_sectors;
69         unsigned sec_per_clus;
70         unsigned dir_entries;
71         unsigned media;
72         unsigned fat_length;
73 } floppy_defaults[] = {
74 {
75         .nr_sectors = 160 * KB_IN_SECTORS,
76         .sec_per_clus = 1,
77         .dir_entries = 64,
78         .media = 0xFE,
79         .fat_length = 1,
80 },
81 {
82         .nr_sectors = 180 * KB_IN_SECTORS,
83         .sec_per_clus = 1,
84         .dir_entries = 64,
85         .media = 0xFC,
86         .fat_length = 2,
87 },
88 {
89         .nr_sectors = 320 * KB_IN_SECTORS,
90         .sec_per_clus = 2,
91         .dir_entries = 112,
92         .media = 0xFF,
93         .fat_length = 1,
94 },
95 {
96         .nr_sectors = 360 * KB_IN_SECTORS,
97         .sec_per_clus = 2,
98         .dir_entries = 112,
99         .media = 0xFD,
100         .fat_length = 2,
101 },
102 };
103
104 static int fat_add_cluster(struct inode *inode)
105 {
106         int err, cluster;
107
108         err = fat_alloc_clusters(inode, &cluster, 1);
109         if (err)
110                 return err;
111         /* FIXME: this cluster should be added after data of this
112          * cluster is writed */
113         err = fat_chain_add(inode, cluster, 1);
114         if (err)
115                 fat_free_clusters(inode, cluster);
116         return err;
117 }
118
119 static inline int __fat_get_block(struct inode *inode, sector_t iblock,
120                                   unsigned long *max_blocks,
121                                   struct buffer_head *bh_result, int create)
122 {
123         struct super_block *sb = inode->i_sb;
124         struct msdos_sb_info *sbi = MSDOS_SB(sb);
125         unsigned long mapped_blocks;
126         sector_t phys;
127         int err, offset;
128
129         err = fat_bmap(inode, iblock, &phys, &mapped_blocks, create);
130         if (err)
131                 return err;
132         if (phys) {
133                 map_bh(bh_result, sb, phys);
134                 *max_blocks = min(mapped_blocks, *max_blocks);
135                 return 0;
136         }
137         if (!create)
138                 return 0;
139
140         if (iblock != MSDOS_I(inode)->mmu_private >> sb->s_blocksize_bits) {
141                 fat_fs_error(sb, "corrupted file size (i_pos %lld, %lld)",
142                         MSDOS_I(inode)->i_pos, MSDOS_I(inode)->mmu_private);
143                 return -EIO;
144         }
145
146         offset = (unsigned long)iblock & (sbi->sec_per_clus - 1);
147         if (!offset) {
148                 /* TODO: multiple cluster allocation would be desirable. */
149                 err = fat_add_cluster(inode);
150                 if (err)
151                         return err;
152         }
153         /* available blocks on this cluster */
154         mapped_blocks = sbi->sec_per_clus - offset;
155
156         *max_blocks = min(mapped_blocks, *max_blocks);
157         MSDOS_I(inode)->mmu_private += *max_blocks << sb->s_blocksize_bits;
158
159         err = fat_bmap(inode, iblock, &phys, &mapped_blocks, create);
160         if (err)
161                 return err;
162
163         BUG_ON(!phys);
164         BUG_ON(*max_blocks != mapped_blocks);
165         set_buffer_new(bh_result);
166         map_bh(bh_result, sb, phys);
167
168         return 0;
169 }
170
171 static int fat_get_block(struct inode *inode, sector_t iblock,
172                          struct buffer_head *bh_result, int create)
173 {
174         struct super_block *sb = inode->i_sb;
175         unsigned long max_blocks = bh_result->b_size >> inode->i_blkbits;
176         int err;
177
178         err = __fat_get_block(inode, iblock, &max_blocks, bh_result, create);
179         if (err)
180                 return err;
181         bh_result->b_size = max_blocks << sb->s_blocksize_bits;
182         return 0;
183 }
184
185 static int fat_writepage(struct page *page, struct writeback_control *wbc)
186 {
187         return block_write_full_page(page, fat_get_block, wbc);
188 }
189
190 static int fat_writepages(struct address_space *mapping,
191                           struct writeback_control *wbc)
192 {
193         return mpage_writepages(mapping, wbc, fat_get_block);
194 }
195
196 static int fat_readpage(struct file *file, struct page *page)
197 {
198         return mpage_readpage(page, fat_get_block);
199 }
200
201 static int fat_readpages(struct file *file, struct address_space *mapping,
202                          struct list_head *pages, unsigned nr_pages)
203 {
204         return mpage_readpages(mapping, pages, nr_pages, fat_get_block);
205 }
206
207 static void fat_write_failed(struct address_space *mapping, loff_t to)
208 {
209         struct inode *inode = mapping->host;
210
211         if (to > inode->i_size) {
212                 truncate_pagecache(inode, inode->i_size);
213                 fat_truncate_blocks(inode, inode->i_size);
214         }
215 }
216
217 static int fat_write_begin(struct file *file, struct address_space *mapping,
218                         loff_t pos, unsigned len, unsigned flags,
219                         struct page **pagep, void **fsdata)
220 {
221         int err;
222
223         *pagep = NULL;
224         err = cont_write_begin(file, mapping, pos, len, flags,
225                                 pagep, fsdata, fat_get_block,
226                                 &MSDOS_I(mapping->host)->mmu_private);
227         if (err < 0)
228                 fat_write_failed(mapping, pos + len);
229         return err;
230 }
231
232 static int fat_write_end(struct file *file, struct address_space *mapping,
233                         loff_t pos, unsigned len, unsigned copied,
234                         struct page *pagep, void *fsdata)
235 {
236         struct inode *inode = mapping->host;
237         int err;
238         err = generic_write_end(file, mapping, pos, len, copied, pagep, fsdata);
239         if (err < len)
240                 fat_write_failed(mapping, pos + len);
241         if (!(err < 0) && !(MSDOS_I(inode)->i_attrs & ATTR_ARCH)) {
242                 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
243                 MSDOS_I(inode)->i_attrs |= ATTR_ARCH;
244                 mark_inode_dirty(inode);
245         }
246         return err;
247 }
248
249 static ssize_t fat_direct_IO(int rw, struct kiocb *iocb,
250                              struct iov_iter *iter,
251                              loff_t offset)
252 {
253         struct file *file = iocb->ki_filp;
254         struct address_space *mapping = file->f_mapping;
255         struct inode *inode = mapping->host;
256         size_t count = iov_iter_count(iter);
257         ssize_t ret;
258
259         if (rw == WRITE) {
260                 /*
261                  * FIXME: blockdev_direct_IO() doesn't use ->write_begin(),
262                  * so we need to update the ->mmu_private to block boundary.
263                  *
264                  * But we must fill the remaining area or hole by nul for
265                  * updating ->mmu_private.
266                  *
267                  * Return 0, and fallback to normal buffered write.
268                  */
269                 loff_t size = offset + count;
270                 if (MSDOS_I(inode)->mmu_private < size)
271                         return 0;
272         }
273
274         /*
275          * FAT need to use the DIO_LOCKING for avoiding the race
276          * condition of fat_get_block() and ->truncate().
277          */
278         ret = blockdev_direct_IO(rw, iocb, inode, iter, offset, fat_get_block);
279         if (ret < 0 && (rw & WRITE))
280                 fat_write_failed(mapping, offset + count);
281
282         return ret;
283 }
284
285 static sector_t _fat_bmap(struct address_space *mapping, sector_t block)
286 {
287         sector_t blocknr;
288
289         /* fat_get_cluster() assumes the requested blocknr isn't truncated. */
290         down_read(&MSDOS_I(mapping->host)->truncate_lock);
291         blocknr = generic_block_bmap(mapping, block, fat_get_block);
292         up_read(&MSDOS_I(mapping->host)->truncate_lock);
293
294         return blocknr;
295 }
296
297 /*
298  * fat_block_truncate_page() zeroes out a mapping from file offset `from'
299  * up to the end of the block which corresponds to `from'.
300  * This is required during truncate to physically zeroout the tail end
301  * of that block so it doesn't yield old data if the file is later grown.
302  * Also, avoid causing failure from fsx for cases of "data past EOF"
303  */
304 int fat_block_truncate_page(struct inode *inode, loff_t from)
305 {
306         return block_truncate_page(inode->i_mapping, from, fat_get_block);
307 }
308
309 static const struct address_space_operations fat_aops = {
310         .readpage       = fat_readpage,
311         .readpages      = fat_readpages,
312         .writepage      = fat_writepage,
313         .writepages     = fat_writepages,
314         .write_begin    = fat_write_begin,
315         .write_end      = fat_write_end,
316         .direct_IO      = fat_direct_IO,
317         .bmap           = _fat_bmap
318 };
319
320 /*
321  * New FAT inode stuff. We do the following:
322  *      a) i_ino is constant and has nothing with on-disk location.
323  *      b) FAT manages its own cache of directory entries.
324  *      c) *This* cache is indexed by on-disk location.
325  *      d) inode has an associated directory entry, all right, but
326  *              it may be unhashed.
327  *      e) currently entries are stored within struct inode. That should
328  *              change.
329  *      f) we deal with races in the following way:
330  *              1. readdir() and lookup() do FAT-dir-cache lookup.
331  *              2. rename() unhashes the F-d-c entry and rehashes it in
332  *                      a new place.
333  *              3. unlink() and rmdir() unhash F-d-c entry.
334  *              4. fat_write_inode() checks whether the thing is unhashed.
335  *                      If it is we silently return. If it isn't we do bread(),
336  *                      check if the location is still valid and retry if it
337  *                      isn't. Otherwise we do changes.
338  *              5. Spinlock is used to protect hash/unhash/location check/lookup
339  *              6. fat_evict_inode() unhashes the F-d-c entry.
340  *              7. lookup() and readdir() do igrab() if they find a F-d-c entry
341  *                      and consider negative result as cache miss.
342  */
343
344 static void fat_hash_init(struct super_block *sb)
345 {
346         struct msdos_sb_info *sbi = MSDOS_SB(sb);
347         int i;
348
349         spin_lock_init(&sbi->inode_hash_lock);
350         for (i = 0; i < FAT_HASH_SIZE; i++)
351                 INIT_HLIST_HEAD(&sbi->inode_hashtable[i]);
352 }
353
354 static inline unsigned long fat_hash(loff_t i_pos)
355 {
356         return hash_32(i_pos, FAT_HASH_BITS);
357 }
358
359 static void dir_hash_init(struct super_block *sb)
360 {
361         struct msdos_sb_info *sbi = MSDOS_SB(sb);
362         int i;
363
364         spin_lock_init(&sbi->dir_hash_lock);
365         for (i = 0; i < FAT_HASH_SIZE; i++)
366                 INIT_HLIST_HEAD(&sbi->dir_hashtable[i]);
367 }
368
369 void fat_attach(struct inode *inode, loff_t i_pos)
370 {
371         struct msdos_sb_info *sbi = MSDOS_SB(inode->i_sb);
372
373         if (inode->i_ino != MSDOS_ROOT_INO) {
374                 struct hlist_head *head =   sbi->inode_hashtable
375                                           + fat_hash(i_pos);
376
377                 spin_lock(&sbi->inode_hash_lock);
378                 MSDOS_I(inode)->i_pos = i_pos;
379                 hlist_add_head(&MSDOS_I(inode)->i_fat_hash, head);
380                 spin_unlock(&sbi->inode_hash_lock);
381         }
382
383         /* If NFS support is enabled, cache the mapping of start cluster
384          * to directory inode. This is used during reconnection of
385          * dentries to the filesystem root.
386          */
387         if (S_ISDIR(inode->i_mode) && sbi->options.nfs) {
388                 struct hlist_head *d_head = sbi->dir_hashtable;
389                 d_head += fat_dir_hash(MSDOS_I(inode)->i_logstart);
390
391                 spin_lock(&sbi->dir_hash_lock);
392                 hlist_add_head(&MSDOS_I(inode)->i_dir_hash, d_head);
393                 spin_unlock(&sbi->dir_hash_lock);
394         }
395 }
396 EXPORT_SYMBOL_GPL(fat_attach);
397
398 void fat_detach(struct inode *inode)
399 {
400         struct msdos_sb_info *sbi = MSDOS_SB(inode->i_sb);
401         spin_lock(&sbi->inode_hash_lock);
402         MSDOS_I(inode)->i_pos = 0;
403         hlist_del_init(&MSDOS_I(inode)->i_fat_hash);
404         spin_unlock(&sbi->inode_hash_lock);
405
406         if (S_ISDIR(inode->i_mode) && sbi->options.nfs) {
407                 spin_lock(&sbi->dir_hash_lock);
408                 hlist_del_init(&MSDOS_I(inode)->i_dir_hash);
409                 spin_unlock(&sbi->dir_hash_lock);
410         }
411 }
412 EXPORT_SYMBOL_GPL(fat_detach);
413
414 struct inode *fat_iget(struct super_block *sb, loff_t i_pos)
415 {
416         struct msdos_sb_info *sbi = MSDOS_SB(sb);
417         struct hlist_head *head = sbi->inode_hashtable + fat_hash(i_pos);
418         struct msdos_inode_info *i;
419         struct inode *inode = NULL;
420
421         spin_lock(&sbi->inode_hash_lock);
422         hlist_for_each_entry(i, head, i_fat_hash) {
423                 BUG_ON(i->vfs_inode.i_sb != sb);
424                 if (i->i_pos != i_pos)
425                         continue;
426                 inode = igrab(&i->vfs_inode);
427                 if (inode)
428                         break;
429         }
430         spin_unlock(&sbi->inode_hash_lock);
431         return inode;
432 }
433
434 static int is_exec(unsigned char *extension)
435 {
436         unsigned char exe_extensions[] = "EXECOMBAT", *walk;
437
438         for (walk = exe_extensions; *walk; walk += 3)
439                 if (!strncmp(extension, walk, 3))
440                         return 1;
441         return 0;
442 }
443
444 static int fat_calc_dir_size(struct inode *inode)
445 {
446         struct msdos_sb_info *sbi = MSDOS_SB(inode->i_sb);
447         int ret, fclus, dclus;
448
449         inode->i_size = 0;
450         if (MSDOS_I(inode)->i_start == 0)
451                 return 0;
452
453         ret = fat_get_cluster(inode, FAT_ENT_EOF, &fclus, &dclus);
454         if (ret < 0)
455                 return ret;
456         inode->i_size = (fclus + 1) << sbi->cluster_bits;
457
458         return 0;
459 }
460
461 /* doesn't deal with root inode */
462 int fat_fill_inode(struct inode *inode, struct msdos_dir_entry *de)
463 {
464         struct msdos_sb_info *sbi = MSDOS_SB(inode->i_sb);
465         int error;
466
467         MSDOS_I(inode)->i_pos = 0;
468         inode->i_uid = sbi->options.fs_uid;
469         inode->i_gid = sbi->options.fs_gid;
470         inode->i_version++;
471         inode->i_generation = get_seconds();
472
473         if ((de->attr & ATTR_DIR) && !IS_FREE(de->name)) {
474                 inode->i_generation &= ~1;
475                 inode->i_mode = fat_make_mode(sbi, de->attr, S_IRWXUGO);
476                 inode->i_op = sbi->dir_ops;
477                 inode->i_fop = &fat_dir_operations;
478
479                 MSDOS_I(inode)->i_start = fat_get_start(sbi, de);
480                 MSDOS_I(inode)->i_logstart = MSDOS_I(inode)->i_start;
481                 error = fat_calc_dir_size(inode);
482                 if (error < 0)
483                         return error;
484                 MSDOS_I(inode)->mmu_private = inode->i_size;
485
486                 set_nlink(inode, fat_subdirs(inode));
487         } else { /* not a directory */
488                 inode->i_generation |= 1;
489                 inode->i_mode = fat_make_mode(sbi, de->attr,
490                         ((sbi->options.showexec && !is_exec(de->name + 8))
491                          ? S_IRUGO|S_IWUGO : S_IRWXUGO));
492                 MSDOS_I(inode)->i_start = fat_get_start(sbi, de);
493
494                 MSDOS_I(inode)->i_logstart = MSDOS_I(inode)->i_start;
495                 inode->i_size = le32_to_cpu(de->size);
496                 inode->i_op = &fat_file_inode_operations;
497                 inode->i_fop = &fat_file_operations;
498                 inode->i_mapping->a_ops = &fat_aops;
499                 MSDOS_I(inode)->mmu_private = inode->i_size;
500         }
501         if (de->attr & ATTR_SYS) {
502                 if (sbi->options.sys_immutable)
503                         inode->i_flags |= S_IMMUTABLE;
504         }
505         fat_save_attrs(inode, de->attr);
506
507         inode->i_blocks = ((inode->i_size + (sbi->cluster_size - 1))
508                            & ~((loff_t)sbi->cluster_size - 1)) >> 9;
509
510         fat_time_fat2unix(sbi, &inode->i_mtime, de->time, de->date, 0);
511         if (sbi->options.isvfat) {
512                 fat_time_fat2unix(sbi, &inode->i_ctime, de->ctime,
513                                   de->cdate, de->ctime_cs);
514                 fat_time_fat2unix(sbi, &inode->i_atime, 0, de->adate, 0);
515         } else
516                 inode->i_ctime = inode->i_atime = inode->i_mtime;
517
518         return 0;
519 }
520
521 static inline void fat_lock_build_inode(struct msdos_sb_info *sbi)
522 {
523         if (sbi->options.nfs == FAT_NFS_NOSTALE_RO)
524                 mutex_lock(&sbi->nfs_build_inode_lock);
525 }
526
527 static inline void fat_unlock_build_inode(struct msdos_sb_info *sbi)
528 {
529         if (sbi->options.nfs == FAT_NFS_NOSTALE_RO)
530                 mutex_unlock(&sbi->nfs_build_inode_lock);
531 }
532
533 struct inode *fat_build_inode(struct super_block *sb,
534                         struct msdos_dir_entry *de, loff_t i_pos)
535 {
536         struct inode *inode;
537         int err;
538
539         fat_lock_build_inode(MSDOS_SB(sb));
540         inode = fat_iget(sb, i_pos);
541         if (inode)
542                 goto out;
543         inode = new_inode(sb);
544         if (!inode) {
545                 inode = ERR_PTR(-ENOMEM);
546                 goto out;
547         }
548         inode->i_ino = iunique(sb, MSDOS_ROOT_INO);
549         inode->i_version = 1;
550         err = fat_fill_inode(inode, de);
551         if (err) {
552                 iput(inode);
553                 inode = ERR_PTR(err);
554                 goto out;
555         }
556         fat_attach(inode, i_pos);
557         insert_inode_hash(inode);
558 out:
559         fat_unlock_build_inode(MSDOS_SB(sb));
560         return inode;
561 }
562
563 EXPORT_SYMBOL_GPL(fat_build_inode);
564
565 static void fat_evict_inode(struct inode *inode)
566 {
567         truncate_inode_pages_final(&inode->i_data);
568         if (!inode->i_nlink) {
569                 inode->i_size = 0;
570                 fat_truncate_blocks(inode, 0);
571         }
572         invalidate_inode_buffers(inode);
573         clear_inode(inode);
574         fat_cache_inval_inode(inode);
575         fat_detach(inode);
576 }
577
578 static void fat_set_state(struct super_block *sb,
579                         unsigned int set, unsigned int force)
580 {
581         struct buffer_head *bh;
582         struct fat_boot_sector *b;
583         struct msdos_sb_info *sbi = MSDOS_SB(sb);
584
585         /* do not change any thing if mounted read only */
586         if ((sb->s_flags & MS_RDONLY) && !force)
587                 return;
588
589         /* do not change state if fs was dirty */
590         if (sbi->dirty) {
591                 /* warn only on set (mount). */
592                 if (set)
593                         fat_msg(sb, KERN_WARNING, "Volume was not properly "
594                                 "unmounted. Some data may be corrupt. "
595                                 "Please run fsck.");
596                 return;
597         }
598
599         bh = sb_bread(sb, 0);
600         if (bh == NULL) {
601                 fat_msg(sb, KERN_ERR, "unable to read boot sector "
602                         "to mark fs as dirty");
603                 return;
604         }
605
606         b = (struct fat_boot_sector *) bh->b_data;
607
608         if (sbi->fat_bits == 32) {
609                 if (set)
610                         b->fat32.state |= FAT_STATE_DIRTY;
611                 else
612                         b->fat32.state &= ~FAT_STATE_DIRTY;
613         } else /* fat 16 and 12 */ {
614                 if (set)
615                         b->fat16.state |= FAT_STATE_DIRTY;
616                 else
617                         b->fat16.state &= ~FAT_STATE_DIRTY;
618         }
619
620         mark_buffer_dirty(bh);
621         sync_dirty_buffer(bh);
622         brelse(bh);
623 }
624
625 static void delayed_free(struct rcu_head *p)
626 {
627         struct msdos_sb_info *sbi = container_of(p, struct msdos_sb_info, rcu);
628         unload_nls(sbi->nls_disk);
629         unload_nls(sbi->nls_io);
630         if (sbi->options.iocharset != fat_default_iocharset)
631                 kfree(sbi->options.iocharset);
632         kfree(sbi);
633 }
634
635 static void fat_put_super(struct super_block *sb)
636 {
637         struct msdos_sb_info *sbi = MSDOS_SB(sb);
638
639         fat_set_state(sb, 0, 0);
640
641         iput(sbi->fsinfo_inode);
642         iput(sbi->fat_inode);
643
644         call_rcu(&sbi->rcu, delayed_free);
645 }
646
647 static struct kmem_cache *fat_inode_cachep;
648
649 static struct inode *fat_alloc_inode(struct super_block *sb)
650 {
651         struct msdos_inode_info *ei;
652         ei = kmem_cache_alloc(fat_inode_cachep, GFP_NOFS);
653         if (!ei)
654                 return NULL;
655
656         init_rwsem(&ei->truncate_lock);
657         return &ei->vfs_inode;
658 }
659
660 static void fat_i_callback(struct rcu_head *head)
661 {
662         struct inode *inode = container_of(head, struct inode, i_rcu);
663         kmem_cache_free(fat_inode_cachep, MSDOS_I(inode));
664 }
665
666 static void fat_destroy_inode(struct inode *inode)
667 {
668         call_rcu(&inode->i_rcu, fat_i_callback);
669 }
670
671 static void init_once(void *foo)
672 {
673         struct msdos_inode_info *ei = (struct msdos_inode_info *)foo;
674
675         spin_lock_init(&ei->cache_lru_lock);
676         ei->nr_caches = 0;
677         ei->cache_valid_id = FAT_CACHE_VALID + 1;
678         INIT_LIST_HEAD(&ei->cache_lru);
679         INIT_HLIST_NODE(&ei->i_fat_hash);
680         INIT_HLIST_NODE(&ei->i_dir_hash);
681         inode_init_once(&ei->vfs_inode);
682 }
683
684 static int __init fat_init_inodecache(void)
685 {
686         fat_inode_cachep = kmem_cache_create("fat_inode_cache",
687                                              sizeof(struct msdos_inode_info),
688                                              0, (SLAB_RECLAIM_ACCOUNT|
689                                                 SLAB_MEM_SPREAD),
690                                              init_once);
691         if (fat_inode_cachep == NULL)
692                 return -ENOMEM;
693         return 0;
694 }
695
696 static void __exit fat_destroy_inodecache(void)
697 {
698         /*
699          * Make sure all delayed rcu free inodes are flushed before we
700          * destroy cache.
701          */
702         rcu_barrier();
703         kmem_cache_destroy(fat_inode_cachep);
704 }
705
706 static int fat_remount(struct super_block *sb, int *flags, char *data)
707 {
708         int new_rdonly;
709         struct msdos_sb_info *sbi = MSDOS_SB(sb);
710         *flags |= MS_NODIRATIME | (sbi->options.isvfat ? 0 : MS_NOATIME);
711
712         sync_filesystem(sb);
713
714         /* make sure we update state on remount. */
715         new_rdonly = *flags & MS_RDONLY;
716         if (new_rdonly != (sb->s_flags & MS_RDONLY)) {
717                 if (new_rdonly)
718                         fat_set_state(sb, 0, 0);
719                 else
720                         fat_set_state(sb, 1, 1);
721         }
722         return 0;
723 }
724
725 static int fat_statfs(struct dentry *dentry, struct kstatfs *buf)
726 {
727         struct super_block *sb = dentry->d_sb;
728         struct msdos_sb_info *sbi = MSDOS_SB(sb);
729         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
730
731         /* If the count of free cluster is still unknown, counts it here. */
732         if (sbi->free_clusters == -1 || !sbi->free_clus_valid) {
733                 int err = fat_count_free_clusters(dentry->d_sb);
734                 if (err)
735                         return err;
736         }
737
738         buf->f_type = dentry->d_sb->s_magic;
739         buf->f_bsize = sbi->cluster_size;
740         buf->f_blocks = sbi->max_cluster - FAT_START_ENT;
741         buf->f_bfree = sbi->free_clusters;
742         buf->f_bavail = sbi->free_clusters;
743         buf->f_fsid.val[0] = (u32)id;
744         buf->f_fsid.val[1] = (u32)(id >> 32);
745         buf->f_namelen =
746                 (sbi->options.isvfat ? FAT_LFN_LEN : 12) * NLS_MAX_CHARSET_SIZE;
747
748         return 0;
749 }
750
751 static int __fat_write_inode(struct inode *inode, int wait)
752 {
753         struct super_block *sb = inode->i_sb;
754         struct msdos_sb_info *sbi = MSDOS_SB(sb);
755         struct buffer_head *bh;
756         struct msdos_dir_entry *raw_entry;
757         loff_t i_pos;
758         sector_t blocknr;
759         int err, offset;
760
761         if (inode->i_ino == MSDOS_ROOT_INO)
762                 return 0;
763
764 retry:
765         i_pos = fat_i_pos_read(sbi, inode);
766         if (!i_pos)
767                 return 0;
768
769         fat_get_blknr_offset(sbi, i_pos, &blocknr, &offset);
770         bh = sb_bread(sb, blocknr);
771         if (!bh) {
772                 fat_msg(sb, KERN_ERR, "unable to read inode block "
773                        "for updating (i_pos %lld)", i_pos);
774                 return -EIO;
775         }
776         spin_lock(&sbi->inode_hash_lock);
777         if (i_pos != MSDOS_I(inode)->i_pos) {
778                 spin_unlock(&sbi->inode_hash_lock);
779                 brelse(bh);
780                 goto retry;
781         }
782
783         raw_entry = &((struct msdos_dir_entry *) (bh->b_data))[offset];
784         if (S_ISDIR(inode->i_mode))
785                 raw_entry->size = 0;
786         else
787                 raw_entry->size = cpu_to_le32(inode->i_size);
788         raw_entry->attr = fat_make_attrs(inode);
789         fat_set_start(raw_entry, MSDOS_I(inode)->i_logstart);
790         fat_time_unix2fat(sbi, &inode->i_mtime, &raw_entry->time,
791                           &raw_entry->date, NULL);
792         if (sbi->options.isvfat) {
793                 __le16 atime;
794                 fat_time_unix2fat(sbi, &inode->i_ctime, &raw_entry->ctime,
795                                   &raw_entry->cdate, &raw_entry->ctime_cs);
796                 fat_time_unix2fat(sbi, &inode->i_atime, &atime,
797                                   &raw_entry->adate, NULL);
798         }
799         spin_unlock(&sbi->inode_hash_lock);
800         mark_buffer_dirty(bh);
801         err = 0;
802         if (wait)
803                 err = sync_dirty_buffer(bh);
804         brelse(bh);
805         return err;
806 }
807
808 static int fat_write_inode(struct inode *inode, struct writeback_control *wbc)
809 {
810         int err;
811
812         if (inode->i_ino == MSDOS_FSINFO_INO) {
813                 struct super_block *sb = inode->i_sb;
814
815                 mutex_lock(&MSDOS_SB(sb)->s_lock);
816                 err = fat_clusters_flush(sb);
817                 mutex_unlock(&MSDOS_SB(sb)->s_lock);
818         } else
819                 err = __fat_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
820
821         return err;
822 }
823
824 int fat_sync_inode(struct inode *inode)
825 {
826         return __fat_write_inode(inode, 1);
827 }
828
829 EXPORT_SYMBOL_GPL(fat_sync_inode);
830
831 static int fat_show_options(struct seq_file *m, struct dentry *root);
832 static const struct super_operations fat_sops = {
833         .alloc_inode    = fat_alloc_inode,
834         .destroy_inode  = fat_destroy_inode,
835         .write_inode    = fat_write_inode,
836         .evict_inode    = fat_evict_inode,
837         .put_super      = fat_put_super,
838         .statfs         = fat_statfs,
839         .remount_fs     = fat_remount,
840
841         .show_options   = fat_show_options,
842 };
843
844 static int fat_show_options(struct seq_file *m, struct dentry *root)
845 {
846         struct msdos_sb_info *sbi = MSDOS_SB(root->d_sb);
847         struct fat_mount_options *opts = &sbi->options;
848         int isvfat = opts->isvfat;
849
850         if (!uid_eq(opts->fs_uid, GLOBAL_ROOT_UID))
851                 seq_printf(m, ",uid=%u",
852                                 from_kuid_munged(&init_user_ns, opts->fs_uid));
853         if (!gid_eq(opts->fs_gid, GLOBAL_ROOT_GID))
854                 seq_printf(m, ",gid=%u",
855                                 from_kgid_munged(&init_user_ns, opts->fs_gid));
856         seq_printf(m, ",fmask=%04o", opts->fs_fmask);
857         seq_printf(m, ",dmask=%04o", opts->fs_dmask);
858         if (opts->allow_utime)
859                 seq_printf(m, ",allow_utime=%04o", opts->allow_utime);
860         if (sbi->nls_disk)
861                 /* strip "cp" prefix from displayed option */
862                 seq_printf(m, ",codepage=%s", &sbi->nls_disk->charset[2]);
863         if (isvfat) {
864                 if (sbi->nls_io)
865                         seq_printf(m, ",iocharset=%s", sbi->nls_io->charset);
866
867                 switch (opts->shortname) {
868                 case VFAT_SFN_DISPLAY_WIN95 | VFAT_SFN_CREATE_WIN95:
869                         seq_puts(m, ",shortname=win95");
870                         break;
871                 case VFAT_SFN_DISPLAY_WINNT | VFAT_SFN_CREATE_WINNT:
872                         seq_puts(m, ",shortname=winnt");
873                         break;
874                 case VFAT_SFN_DISPLAY_WINNT | VFAT_SFN_CREATE_WIN95:
875                         seq_puts(m, ",shortname=mixed");
876                         break;
877                 case VFAT_SFN_DISPLAY_LOWER | VFAT_SFN_CREATE_WIN95:
878                         seq_puts(m, ",shortname=lower");
879                         break;
880                 default:
881                         seq_puts(m, ",shortname=unknown");
882                         break;
883                 }
884         }
885         if (opts->name_check != 'n')
886                 seq_printf(m, ",check=%c", opts->name_check);
887         if (opts->usefree)
888                 seq_puts(m, ",usefree");
889         if (opts->quiet)
890                 seq_puts(m, ",quiet");
891         if (opts->showexec)
892                 seq_puts(m, ",showexec");
893         if (opts->sys_immutable)
894                 seq_puts(m, ",sys_immutable");
895         if (!isvfat) {
896                 if (opts->dotsOK)
897                         seq_puts(m, ",dotsOK=yes");
898                 if (opts->nocase)
899                         seq_puts(m, ",nocase");
900         } else {
901                 if (opts->utf8)
902                         seq_puts(m, ",utf8");
903                 if (opts->unicode_xlate)
904                         seq_puts(m, ",uni_xlate");
905                 if (!opts->numtail)
906                         seq_puts(m, ",nonumtail");
907                 if (opts->rodir)
908                         seq_puts(m, ",rodir");
909         }
910         if (opts->flush)
911                 seq_puts(m, ",flush");
912         if (opts->tz_set) {
913                 if (opts->time_offset)
914                         seq_printf(m, ",time_offset=%d", opts->time_offset);
915                 else
916                         seq_puts(m, ",tz=UTC");
917         }
918         if (opts->errors == FAT_ERRORS_CONT)
919                 seq_puts(m, ",errors=continue");
920         else if (opts->errors == FAT_ERRORS_PANIC)
921                 seq_puts(m, ",errors=panic");
922         else
923                 seq_puts(m, ",errors=remount-ro");
924         if (opts->nfs == FAT_NFS_NOSTALE_RO)
925                 seq_puts(m, ",nfs=nostale_ro");
926         else if (opts->nfs)
927                 seq_puts(m, ",nfs=stale_rw");
928         if (opts->discard)
929                 seq_puts(m, ",discard");
930         if (opts->dos1xfloppy)
931                 seq_puts(m, ",dos1xfloppy");
932
933         return 0;
934 }
935
936 enum {
937         Opt_check_n, Opt_check_r, Opt_check_s, Opt_uid, Opt_gid,
938         Opt_umask, Opt_dmask, Opt_fmask, Opt_allow_utime, Opt_codepage,
939         Opt_usefree, Opt_nocase, Opt_quiet, Opt_showexec, Opt_debug,
940         Opt_immutable, Opt_dots, Opt_nodots,
941         Opt_charset, Opt_shortname_lower, Opt_shortname_win95,
942         Opt_shortname_winnt, Opt_shortname_mixed, Opt_utf8_no, Opt_utf8_yes,
943         Opt_uni_xl_no, Opt_uni_xl_yes, Opt_nonumtail_no, Opt_nonumtail_yes,
944         Opt_obsolete, Opt_flush, Opt_tz_utc, Opt_rodir, Opt_err_cont,
945         Opt_err_panic, Opt_err_ro, Opt_discard, Opt_nfs, Opt_time_offset,
946         Opt_nfs_stale_rw, Opt_nfs_nostale_ro, Opt_err, Opt_dos1xfloppy,
947 };
948
949 static const match_table_t fat_tokens = {
950         {Opt_check_r, "check=relaxed"},
951         {Opt_check_s, "check=strict"},
952         {Opt_check_n, "check=normal"},
953         {Opt_check_r, "check=r"},
954         {Opt_check_s, "check=s"},
955         {Opt_check_n, "check=n"},
956         {Opt_uid, "uid=%u"},
957         {Opt_gid, "gid=%u"},
958         {Opt_umask, "umask=%o"},
959         {Opt_dmask, "dmask=%o"},
960         {Opt_fmask, "fmask=%o"},
961         {Opt_allow_utime, "allow_utime=%o"},
962         {Opt_codepage, "codepage=%u"},
963         {Opt_usefree, "usefree"},
964         {Opt_nocase, "nocase"},
965         {Opt_quiet, "quiet"},
966         {Opt_showexec, "showexec"},
967         {Opt_debug, "debug"},
968         {Opt_immutable, "sys_immutable"},
969         {Opt_flush, "flush"},
970         {Opt_tz_utc, "tz=UTC"},
971         {Opt_time_offset, "time_offset=%d"},
972         {Opt_err_cont, "errors=continue"},
973         {Opt_err_panic, "errors=panic"},
974         {Opt_err_ro, "errors=remount-ro"},
975         {Opt_discard, "discard"},
976         {Opt_nfs_stale_rw, "nfs"},
977         {Opt_nfs_stale_rw, "nfs=stale_rw"},
978         {Opt_nfs_nostale_ro, "nfs=nostale_ro"},
979         {Opt_dos1xfloppy, "dos1xfloppy"},
980         {Opt_obsolete, "conv=binary"},
981         {Opt_obsolete, "conv=text"},
982         {Opt_obsolete, "conv=auto"},
983         {Opt_obsolete, "conv=b"},
984         {Opt_obsolete, "conv=t"},
985         {Opt_obsolete, "conv=a"},
986         {Opt_obsolete, "fat=%u"},
987         {Opt_obsolete, "blocksize=%u"},
988         {Opt_obsolete, "cvf_format=%20s"},
989         {Opt_obsolete, "cvf_options=%100s"},
990         {Opt_obsolete, "posix"},
991         {Opt_err, NULL},
992 };
993 static const match_table_t msdos_tokens = {
994         {Opt_nodots, "nodots"},
995         {Opt_nodots, "dotsOK=no"},
996         {Opt_dots, "dots"},
997         {Opt_dots, "dotsOK=yes"},
998         {Opt_err, NULL}
999 };
1000 static const match_table_t vfat_tokens = {
1001         {Opt_charset, "iocharset=%s"},
1002         {Opt_shortname_lower, "shortname=lower"},
1003         {Opt_shortname_win95, "shortname=win95"},
1004         {Opt_shortname_winnt, "shortname=winnt"},
1005         {Opt_shortname_mixed, "shortname=mixed"},
1006         {Opt_utf8_no, "utf8=0"},                /* 0 or no or false */
1007         {Opt_utf8_no, "utf8=no"},
1008         {Opt_utf8_no, "utf8=false"},
1009         {Opt_utf8_yes, "utf8=1"},               /* empty or 1 or yes or true */
1010         {Opt_utf8_yes, "utf8=yes"},
1011         {Opt_utf8_yes, "utf8=true"},
1012         {Opt_utf8_yes, "utf8"},
1013         {Opt_uni_xl_no, "uni_xlate=0"},         /* 0 or no or false */
1014         {Opt_uni_xl_no, "uni_xlate=no"},
1015         {Opt_uni_xl_no, "uni_xlate=false"},
1016         {Opt_uni_xl_yes, "uni_xlate=1"},        /* empty or 1 or yes or true */
1017         {Opt_uni_xl_yes, "uni_xlate=yes"},
1018         {Opt_uni_xl_yes, "uni_xlate=true"},
1019         {Opt_uni_xl_yes, "uni_xlate"},
1020         {Opt_nonumtail_no, "nonumtail=0"},      /* 0 or no or false */
1021         {Opt_nonumtail_no, "nonumtail=no"},
1022         {Opt_nonumtail_no, "nonumtail=false"},
1023         {Opt_nonumtail_yes, "nonumtail=1"},     /* empty or 1 or yes or true */
1024         {Opt_nonumtail_yes, "nonumtail=yes"},
1025         {Opt_nonumtail_yes, "nonumtail=true"},
1026         {Opt_nonumtail_yes, "nonumtail"},
1027         {Opt_rodir, "rodir"},
1028         {Opt_err, NULL}
1029 };
1030
1031 static int parse_options(struct super_block *sb, char *options, int is_vfat,
1032                          int silent, int *debug, struct fat_mount_options *opts)
1033 {
1034         char *p;
1035         substring_t args[MAX_OPT_ARGS];
1036         int option;
1037         char *iocharset;
1038
1039         opts->isvfat = is_vfat;
1040
1041         opts->fs_uid = current_uid();
1042         opts->fs_gid = current_gid();
1043         opts->fs_fmask = opts->fs_dmask = current_umask();
1044         opts->allow_utime = -1;
1045         opts->codepage = fat_default_codepage;
1046         opts->iocharset = fat_default_iocharset;
1047         if (is_vfat) {
1048                 opts->shortname = VFAT_SFN_DISPLAY_WINNT|VFAT_SFN_CREATE_WIN95;
1049                 opts->rodir = 0;
1050         } else {
1051                 opts->shortname = 0;
1052                 opts->rodir = 1;
1053         }
1054         opts->name_check = 'n';
1055         opts->quiet = opts->showexec = opts->sys_immutable = opts->dotsOK =  0;
1056         opts->utf8 = opts->unicode_xlate = 0;
1057         opts->numtail = 1;
1058         opts->usefree = opts->nocase = 0;
1059         opts->tz_set = 0;
1060         opts->nfs = 0;
1061         opts->errors = FAT_ERRORS_RO;
1062         *debug = 0;
1063
1064         if (!options)
1065                 goto out;
1066
1067         while ((p = strsep(&options, ",")) != NULL) {
1068                 int token;
1069                 if (!*p)
1070                         continue;
1071
1072                 token = match_token(p, fat_tokens, args);
1073                 if (token == Opt_err) {
1074                         if (is_vfat)
1075                                 token = match_token(p, vfat_tokens, args);
1076                         else
1077                                 token = match_token(p, msdos_tokens, args);
1078                 }
1079                 switch (token) {
1080                 case Opt_check_s:
1081                         opts->name_check = 's';
1082                         break;
1083                 case Opt_check_r:
1084                         opts->name_check = 'r';
1085                         break;
1086                 case Opt_check_n:
1087                         opts->name_check = 'n';
1088                         break;
1089                 case Opt_usefree:
1090                         opts->usefree = 1;
1091                         break;
1092                 case Opt_nocase:
1093                         if (!is_vfat)
1094                                 opts->nocase = 1;
1095                         else {
1096                                 /* for backward compatibility */
1097                                 opts->shortname = VFAT_SFN_DISPLAY_WIN95
1098                                         | VFAT_SFN_CREATE_WIN95;
1099                         }
1100                         break;
1101                 case Opt_quiet:
1102                         opts->quiet = 1;
1103                         break;
1104                 case Opt_showexec:
1105                         opts->showexec = 1;
1106                         break;
1107                 case Opt_debug:
1108                         *debug = 1;
1109                         break;
1110                 case Opt_immutable:
1111                         opts->sys_immutable = 1;
1112                         break;
1113                 case Opt_uid:
1114                         if (match_int(&args[0], &option))
1115                                 return -EINVAL;
1116                         opts->fs_uid = make_kuid(current_user_ns(), option);
1117                         if (!uid_valid(opts->fs_uid))
1118                                 return -EINVAL;
1119                         break;
1120                 case Opt_gid:
1121                         if (match_int(&args[0], &option))
1122                                 return -EINVAL;
1123                         opts->fs_gid = make_kgid(current_user_ns(), option);
1124                         if (!gid_valid(opts->fs_gid))
1125                                 return -EINVAL;
1126                         break;
1127                 case Opt_umask:
1128                         if (match_octal(&args[0], &option))
1129                                 return -EINVAL;
1130                         opts->fs_fmask = opts->fs_dmask = option;
1131                         break;
1132                 case Opt_dmask:
1133                         if (match_octal(&args[0], &option))
1134                                 return -EINVAL;
1135                         opts->fs_dmask = option;
1136                         break;
1137                 case Opt_fmask:
1138                         if (match_octal(&args[0], &option))
1139                                 return -EINVAL;
1140                         opts->fs_fmask = option;
1141                         break;
1142                 case Opt_allow_utime:
1143                         if (match_octal(&args[0], &option))
1144                                 return -EINVAL;
1145                         opts->allow_utime = option & (S_IWGRP | S_IWOTH);
1146                         break;
1147                 case Opt_codepage:
1148                         if (match_int(&args[0], &option))
1149                                 return -EINVAL;
1150                         opts->codepage = option;
1151                         break;
1152                 case Opt_flush:
1153                         opts->flush = 1;
1154                         break;
1155                 case Opt_time_offset:
1156                         if (match_int(&args[0], &option))
1157                                 return -EINVAL;
1158                         if (option < -12 * 60 || option > 12 * 60)
1159                                 return -EINVAL;
1160                         opts->tz_set = 1;
1161                         opts->time_offset = option;
1162                         break;
1163                 case Opt_tz_utc:
1164                         opts->tz_set = 1;
1165                         opts->time_offset = 0;
1166                         break;
1167                 case Opt_err_cont:
1168                         opts->errors = FAT_ERRORS_CONT;
1169                         break;
1170                 case Opt_err_panic:
1171                         opts->errors = FAT_ERRORS_PANIC;
1172                         break;
1173                 case Opt_err_ro:
1174                         opts->errors = FAT_ERRORS_RO;
1175                         break;
1176                 case Opt_nfs_stale_rw:
1177                         opts->nfs = FAT_NFS_STALE_RW;
1178                         break;
1179                 case Opt_nfs_nostale_ro:
1180                         opts->nfs = FAT_NFS_NOSTALE_RO;
1181                         break;
1182                 case Opt_dos1xfloppy:
1183                         opts->dos1xfloppy = 1;
1184                         break;
1185
1186                 /* msdos specific */
1187                 case Opt_dots:
1188                         opts->dotsOK = 1;
1189                         break;
1190                 case Opt_nodots:
1191                         opts->dotsOK = 0;
1192                         break;
1193
1194                 /* vfat specific */
1195                 case Opt_charset:
1196                         if (opts->iocharset != fat_default_iocharset)
1197                                 kfree(opts->iocharset);
1198                         iocharset = match_strdup(&args[0]);
1199                         if (!iocharset)
1200                                 return -ENOMEM;
1201                         opts->iocharset = iocharset;
1202                         break;
1203                 case Opt_shortname_lower:
1204                         opts->shortname = VFAT_SFN_DISPLAY_LOWER
1205                                         | VFAT_SFN_CREATE_WIN95;
1206                         break;
1207                 case Opt_shortname_win95:
1208                         opts->shortname = VFAT_SFN_DISPLAY_WIN95
1209                                         | VFAT_SFN_CREATE_WIN95;
1210                         break;
1211                 case Opt_shortname_winnt:
1212                         opts->shortname = VFAT_SFN_DISPLAY_WINNT
1213                                         | VFAT_SFN_CREATE_WINNT;
1214                         break;
1215                 case Opt_shortname_mixed:
1216                         opts->shortname = VFAT_SFN_DISPLAY_WINNT
1217                                         | VFAT_SFN_CREATE_WIN95;
1218                         break;
1219                 case Opt_utf8_no:               /* 0 or no or false */
1220                         opts->utf8 = 0;
1221                         break;
1222                 case Opt_utf8_yes:              /* empty or 1 or yes or true */
1223                         opts->utf8 = 1;
1224                         break;
1225                 case Opt_uni_xl_no:             /* 0 or no or false */
1226                         opts->unicode_xlate = 0;
1227                         break;
1228                 case Opt_uni_xl_yes:            /* empty or 1 or yes or true */
1229                         opts->unicode_xlate = 1;
1230                         break;
1231                 case Opt_nonumtail_no:          /* 0 or no or false */
1232                         opts->numtail = 1;      /* negated option */
1233                         break;
1234                 case Opt_nonumtail_yes:         /* empty or 1 or yes or true */
1235                         opts->numtail = 0;      /* negated option */
1236                         break;
1237                 case Opt_rodir:
1238                         opts->rodir = 1;
1239                         break;
1240                 case Opt_discard:
1241                         opts->discard = 1;
1242                         break;
1243
1244                 /* obsolete mount options */
1245                 case Opt_obsolete:
1246                         fat_msg(sb, KERN_INFO, "\"%s\" option is obsolete, "
1247                                "not supported now", p);
1248                         break;
1249                 /* unknown option */
1250                 default:
1251                         if (!silent) {
1252                                 fat_msg(sb, KERN_ERR,
1253                                        "Unrecognized mount option \"%s\" "
1254                                        "or missing value", p);
1255                         }
1256                         return -EINVAL;
1257                 }
1258         }
1259
1260 out:
1261         /* UTF-8 doesn't provide FAT semantics */
1262         if (!strcmp(opts->iocharset, "utf8")) {
1263                 fat_msg(sb, KERN_WARNING, "utf8 is not a recommended IO charset"
1264                        " for FAT filesystems, filesystem will be "
1265                        "case sensitive!");
1266         }
1267
1268         /* If user doesn't specify allow_utime, it's initialized from dmask. */
1269         if (opts->allow_utime == (unsigned short)-1)
1270                 opts->allow_utime = ~opts->fs_dmask & (S_IWGRP | S_IWOTH);
1271         if (opts->unicode_xlate)
1272                 opts->utf8 = 0;
1273         if (opts->nfs == FAT_NFS_NOSTALE_RO) {
1274                 sb->s_flags |= MS_RDONLY;
1275                 sb->s_export_op = &fat_export_ops_nostale;
1276         }
1277
1278         return 0;
1279 }
1280
1281 static int fat_read_root(struct inode *inode)
1282 {
1283         struct super_block *sb = inode->i_sb;
1284         struct msdos_sb_info *sbi = MSDOS_SB(sb);
1285         int error;
1286
1287         MSDOS_I(inode)->i_pos = MSDOS_ROOT_INO;
1288         inode->i_uid = sbi->options.fs_uid;
1289         inode->i_gid = sbi->options.fs_gid;
1290         inode->i_version++;
1291         inode->i_generation = 0;
1292         inode->i_mode = fat_make_mode(sbi, ATTR_DIR, S_IRWXUGO);
1293         inode->i_op = sbi->dir_ops;
1294         inode->i_fop = &fat_dir_operations;
1295         if (sbi->fat_bits == 32) {
1296                 MSDOS_I(inode)->i_start = sbi->root_cluster;
1297                 error = fat_calc_dir_size(inode);
1298                 if (error < 0)
1299                         return error;
1300         } else {
1301                 MSDOS_I(inode)->i_start = 0;
1302                 inode->i_size = sbi->dir_entries * sizeof(struct msdos_dir_entry);
1303         }
1304         inode->i_blocks = ((inode->i_size + (sbi->cluster_size - 1))
1305                            & ~((loff_t)sbi->cluster_size - 1)) >> 9;
1306         MSDOS_I(inode)->i_logstart = 0;
1307         MSDOS_I(inode)->mmu_private = inode->i_size;
1308
1309         fat_save_attrs(inode, ATTR_DIR);
1310         inode->i_mtime.tv_sec = inode->i_atime.tv_sec = inode->i_ctime.tv_sec = 0;
1311         inode->i_mtime.tv_nsec = inode->i_atime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1312         set_nlink(inode, fat_subdirs(inode)+2);
1313
1314         return 0;
1315 }
1316
1317 static unsigned long calc_fat_clusters(struct super_block *sb)
1318 {
1319         struct msdos_sb_info *sbi = MSDOS_SB(sb);
1320
1321         /* Divide first to avoid overflow */
1322         if (sbi->fat_bits != 12) {
1323                 unsigned long ent_per_sec = sb->s_blocksize * 8 / sbi->fat_bits;
1324                 return ent_per_sec * sbi->fat_length;
1325         }
1326
1327         return sbi->fat_length * sb->s_blocksize * 8 / sbi->fat_bits;
1328 }
1329
1330 static bool fat_bpb_is_zero(struct fat_boot_sector *b)
1331 {
1332         if (get_unaligned_le16(&b->sector_size))
1333                 return false;
1334         if (b->sec_per_clus)
1335                 return false;
1336         if (b->reserved)
1337                 return false;
1338         if (b->fats)
1339                 return false;
1340         if (get_unaligned_le16(&b->dir_entries))
1341                 return false;
1342         if (get_unaligned_le16(&b->sectors))
1343                 return false;
1344         if (b->media)
1345                 return false;
1346         if (b->fat_length)
1347                 return false;
1348         if (b->secs_track)
1349                 return false;
1350         if (b->heads)
1351                 return false;
1352         return true;
1353 }
1354
1355 static int fat_read_bpb(struct super_block *sb, struct fat_boot_sector *b,
1356         int silent, struct fat_bios_param_block *bpb)
1357 {
1358         int error = -EINVAL;
1359
1360         /* Read in BPB ... */
1361         memset(bpb, 0, sizeof(*bpb));
1362         bpb->fat_sector_size = get_unaligned_le16(&b->sector_size);
1363         bpb->fat_sec_per_clus = b->sec_per_clus;
1364         bpb->fat_reserved = le16_to_cpu(b->reserved);
1365         bpb->fat_fats = b->fats;
1366         bpb->fat_dir_entries = get_unaligned_le16(&b->dir_entries);
1367         bpb->fat_sectors = get_unaligned_le16(&b->sectors);
1368         bpb->fat_fat_length = le16_to_cpu(b->fat_length);
1369         bpb->fat_total_sect = le32_to_cpu(b->total_sect);
1370
1371         bpb->fat16_state = b->fat16.state;
1372         bpb->fat16_vol_id = get_unaligned_le32(b->fat16.vol_id);
1373
1374         bpb->fat32_length = le32_to_cpu(b->fat32.length);
1375         bpb->fat32_root_cluster = le32_to_cpu(b->fat32.root_cluster);
1376         bpb->fat32_info_sector = le16_to_cpu(b->fat32.info_sector);
1377         bpb->fat32_state = b->fat32.state;
1378         bpb->fat32_vol_id = get_unaligned_le32(b->fat32.vol_id);
1379
1380         /* Validate this looks like a FAT filesystem BPB */
1381         if (!bpb->fat_reserved) {
1382                 if (!silent)
1383                         fat_msg(sb, KERN_ERR,
1384                                 "bogus number of reserved sectors");
1385                 goto out;
1386         }
1387         if (!bpb->fat_fats) {
1388                 if (!silent)
1389                         fat_msg(sb, KERN_ERR, "bogus number of FAT structure");
1390                 goto out;
1391         }
1392
1393         /*
1394          * Earlier we checked here that b->secs_track and b->head are nonzero,
1395          * but it turns out valid FAT filesystems can have zero there.
1396          */
1397
1398         if (!fat_valid_media(b->media)) {
1399                 if (!silent)
1400                         fat_msg(sb, KERN_ERR, "invalid media value (0x%02x)",
1401                                 (unsigned)b->media);
1402                 goto out;
1403         }
1404
1405         if (!is_power_of_2(bpb->fat_sector_size)
1406             || (bpb->fat_sector_size < 512)
1407             || (bpb->fat_sector_size > 4096)) {
1408                 if (!silent)
1409                         fat_msg(sb, KERN_ERR, "bogus logical sector size %u",
1410                                (unsigned)bpb->fat_sector_size);
1411                 goto out;
1412         }
1413
1414         if (!is_power_of_2(bpb->fat_sec_per_clus)) {
1415                 if (!silent)
1416                         fat_msg(sb, KERN_ERR, "bogus sectors per cluster %u",
1417                                 (unsigned)bpb->fat_sec_per_clus);
1418                 goto out;
1419         }
1420
1421         error = 0;
1422
1423 out:
1424         return error;
1425 }
1426
1427 static int fat_read_static_bpb(struct super_block *sb,
1428         struct fat_boot_sector *b, int silent,
1429         struct fat_bios_param_block *bpb)
1430 {
1431         static const char *notdos1x = "This doesn't look like a DOS 1.x volume";
1432
1433         struct fat_floppy_defaults *fdefaults = NULL;
1434         int error = -EINVAL;
1435         sector_t bd_sects;
1436         unsigned i;
1437
1438         bd_sects = i_size_read(sb->s_bdev->bd_inode) / SECTOR_SIZE;
1439
1440         /* 16-bit DOS 1.x reliably wrote bootstrap short-jmp code */
1441         if (b->ignored[0] != 0xeb || b->ignored[2] != 0x90) {
1442                 if (!silent)
1443                         fat_msg(sb, KERN_ERR,
1444                                 "%s; no bootstrapping code", notdos1x);
1445                 goto out;
1446         }
1447
1448         /*
1449          * If any value in this region is non-zero, it isn't archaic
1450          * DOS.
1451          */
1452         if (!fat_bpb_is_zero(b)) {
1453                 if (!silent)
1454                         fat_msg(sb, KERN_ERR,
1455                                 "%s; DOS 2.x BPB is non-zero", notdos1x);
1456                 goto out;
1457         }
1458
1459         for (i = 0; i < ARRAY_SIZE(floppy_defaults); i++) {
1460                 if (floppy_defaults[i].nr_sectors == bd_sects) {
1461                         fdefaults = &floppy_defaults[i];
1462                         break;
1463                 }
1464         }
1465
1466         if (fdefaults == NULL) {
1467                 if (!silent)
1468                         fat_msg(sb, KERN_WARNING,
1469                                 "This looks like a DOS 1.x volume, but isn't a recognized floppy size (%llu sectors)",
1470                                 (u64)bd_sects);
1471                 goto out;
1472         }
1473
1474         if (!silent)
1475                 fat_msg(sb, KERN_INFO,
1476                         "This looks like a DOS 1.x volume; assuming default BPB values");
1477
1478         memset(bpb, 0, sizeof(*bpb));
1479         bpb->fat_sector_size = SECTOR_SIZE;
1480         bpb->fat_sec_per_clus = fdefaults->sec_per_clus;
1481         bpb->fat_reserved = 1;
1482         bpb->fat_fats = 2;
1483         bpb->fat_dir_entries = fdefaults->dir_entries;
1484         bpb->fat_sectors = fdefaults->nr_sectors;
1485         bpb->fat_fat_length = fdefaults->fat_length;
1486
1487         error = 0;
1488
1489 out:
1490         return error;
1491 }
1492
1493 /*
1494  * Read the super block of an MS-DOS FS.
1495  */
1496 int fat_fill_super(struct super_block *sb, void *data, int silent, int isvfat,
1497                    void (*setup)(struct super_block *))
1498 {
1499         struct inode *root_inode = NULL, *fat_inode = NULL;
1500         struct inode *fsinfo_inode = NULL;
1501         struct buffer_head *bh;
1502         struct fat_bios_param_block bpb;
1503         struct msdos_sb_info *sbi;
1504         u16 logical_sector_size;
1505         u32 total_sectors, total_clusters, fat_clusters, rootdir_sectors;
1506         int debug;
1507         long error;
1508         char buf[50];
1509
1510         /*
1511          * GFP_KERNEL is ok here, because while we do hold the
1512          * supeblock lock, memory pressure can't call back into
1513          * the filesystem, since we're only just about to mount
1514          * it and have no inodes etc active!
1515          */
1516         sbi = kzalloc(sizeof(struct msdos_sb_info), GFP_KERNEL);
1517         if (!sbi)
1518                 return -ENOMEM;
1519         sb->s_fs_info = sbi;
1520
1521         sb->s_flags |= MS_NODIRATIME;
1522         sb->s_magic = MSDOS_SUPER_MAGIC;
1523         sb->s_op = &fat_sops;
1524         sb->s_export_op = &fat_export_ops;
1525         mutex_init(&sbi->nfs_build_inode_lock);
1526         ratelimit_state_init(&sbi->ratelimit, DEFAULT_RATELIMIT_INTERVAL,
1527                              DEFAULT_RATELIMIT_BURST);
1528
1529         error = parse_options(sb, data, isvfat, silent, &debug, &sbi->options);
1530         if (error)
1531                 goto out_fail;
1532
1533         setup(sb); /* flavour-specific stuff that needs options */
1534
1535         error = -EIO;
1536         sb_min_blocksize(sb, 512);
1537         bh = sb_bread(sb, 0);
1538         if (bh == NULL) {
1539                 fat_msg(sb, KERN_ERR, "unable to read boot sector");
1540                 goto out_fail;
1541         }
1542
1543         error = fat_read_bpb(sb, (struct fat_boot_sector *)bh->b_data, silent,
1544                 &bpb);
1545         if (error == -EINVAL && sbi->options.dos1xfloppy)
1546                 error = fat_read_static_bpb(sb,
1547                         (struct fat_boot_sector *)bh->b_data, silent, &bpb);
1548         brelse(bh);
1549
1550         if (error == -EINVAL)
1551                 goto out_invalid;
1552         else if (error)
1553                 goto out_fail;
1554
1555         logical_sector_size = bpb.fat_sector_size;
1556         sbi->sec_per_clus = bpb.fat_sec_per_clus;
1557
1558         error = -EIO;
1559         if (logical_sector_size < sb->s_blocksize) {
1560                 fat_msg(sb, KERN_ERR, "logical sector size too small for device"
1561                        " (logical sector size = %u)", logical_sector_size);
1562                 goto out_fail;
1563         }
1564
1565         if (logical_sector_size > sb->s_blocksize) {
1566                 struct buffer_head *bh_resize;
1567
1568                 if (!sb_set_blocksize(sb, logical_sector_size)) {
1569                         fat_msg(sb, KERN_ERR, "unable to set blocksize %u",
1570                                logical_sector_size);
1571                         goto out_fail;
1572                 }
1573
1574                 /* Verify that the larger boot sector is fully readable */
1575                 bh_resize = sb_bread(sb, 0);
1576                 if (bh_resize == NULL) {
1577                         fat_msg(sb, KERN_ERR, "unable to read boot sector"
1578                                " (logical sector size = %lu)",
1579                                sb->s_blocksize);
1580                         goto out_fail;
1581                 }
1582                 brelse(bh_resize);
1583         }
1584
1585         mutex_init(&sbi->s_lock);
1586         sbi->cluster_size = sb->s_blocksize * sbi->sec_per_clus;
1587         sbi->cluster_bits = ffs(sbi->cluster_size) - 1;
1588         sbi->fats = bpb.fat_fats;
1589         sbi->fat_bits = 0;              /* Don't know yet */
1590         sbi->fat_start = bpb.fat_reserved;
1591         sbi->fat_length = bpb.fat_fat_length;
1592         sbi->root_cluster = 0;
1593         sbi->free_clusters = -1;        /* Don't know yet */
1594         sbi->free_clus_valid = 0;
1595         sbi->prev_free = FAT_START_ENT;
1596         sb->s_maxbytes = 0xffffffff;
1597
1598         if (!sbi->fat_length && bpb.fat32_length) {
1599                 struct fat_boot_fsinfo *fsinfo;
1600                 struct buffer_head *fsinfo_bh;
1601
1602                 /* Must be FAT32 */
1603                 sbi->fat_bits = 32;
1604                 sbi->fat_length = bpb.fat32_length;
1605                 sbi->root_cluster = bpb.fat32_root_cluster;
1606
1607                 /* MC - if info_sector is 0, don't multiply by 0 */
1608                 sbi->fsinfo_sector = bpb.fat32_info_sector;
1609                 if (sbi->fsinfo_sector == 0)
1610                         sbi->fsinfo_sector = 1;
1611
1612                 fsinfo_bh = sb_bread(sb, sbi->fsinfo_sector);
1613                 if (fsinfo_bh == NULL) {
1614                         fat_msg(sb, KERN_ERR, "bread failed, FSINFO block"
1615                                " (sector = %lu)", sbi->fsinfo_sector);
1616                         goto out_fail;
1617                 }
1618
1619                 fsinfo = (struct fat_boot_fsinfo *)fsinfo_bh->b_data;
1620                 if (!IS_FSINFO(fsinfo)) {
1621                         fat_msg(sb, KERN_WARNING, "Invalid FSINFO signature: "
1622                                "0x%08x, 0x%08x (sector = %lu)",
1623                                le32_to_cpu(fsinfo->signature1),
1624                                le32_to_cpu(fsinfo->signature2),
1625                                sbi->fsinfo_sector);
1626                 } else {
1627                         if (sbi->options.usefree)
1628                                 sbi->free_clus_valid = 1;
1629                         sbi->free_clusters = le32_to_cpu(fsinfo->free_clusters);
1630                         sbi->prev_free = le32_to_cpu(fsinfo->next_cluster);
1631                 }
1632
1633                 brelse(fsinfo_bh);
1634         }
1635
1636         /* interpret volume ID as a little endian 32 bit integer */
1637         if (sbi->fat_bits == 32)
1638                 sbi->vol_id = bpb.fat32_vol_id;
1639         else /* fat 16 or 12 */
1640                 sbi->vol_id = bpb.fat16_vol_id;
1641
1642         sbi->dir_per_block = sb->s_blocksize / sizeof(struct msdos_dir_entry);
1643         sbi->dir_per_block_bits = ffs(sbi->dir_per_block) - 1;
1644
1645         sbi->dir_start = sbi->fat_start + sbi->fats * sbi->fat_length;
1646         sbi->dir_entries = bpb.fat_dir_entries;
1647         if (sbi->dir_entries & (sbi->dir_per_block - 1)) {
1648                 if (!silent)
1649                         fat_msg(sb, KERN_ERR, "bogus directory-entries per block"
1650                                " (%u)", sbi->dir_entries);
1651                 goto out_invalid;
1652         }
1653
1654         rootdir_sectors = sbi->dir_entries
1655                 * sizeof(struct msdos_dir_entry) / sb->s_blocksize;
1656         sbi->data_start = sbi->dir_start + rootdir_sectors;
1657         total_sectors = bpb.fat_sectors;
1658         if (total_sectors == 0)
1659                 total_sectors = bpb.fat_total_sect;
1660
1661         total_clusters = (total_sectors - sbi->data_start) / sbi->sec_per_clus;
1662
1663         if (sbi->fat_bits != 32)
1664                 sbi->fat_bits = (total_clusters > MAX_FAT12) ? 16 : 12;
1665
1666         /* some OSes set FAT_STATE_DIRTY and clean it on unmount. */
1667         if (sbi->fat_bits == 32)
1668                 sbi->dirty = bpb.fat32_state & FAT_STATE_DIRTY;
1669         else /* fat 16 or 12 */
1670                 sbi->dirty = bpb.fat16_state & FAT_STATE_DIRTY;
1671
1672         /* check that FAT table does not overflow */
1673         fat_clusters = calc_fat_clusters(sb);
1674         total_clusters = min(total_clusters, fat_clusters - FAT_START_ENT);
1675         if (total_clusters > MAX_FAT(sb)) {
1676                 if (!silent)
1677                         fat_msg(sb, KERN_ERR, "count of clusters too big (%u)",
1678                                total_clusters);
1679                 goto out_invalid;
1680         }
1681
1682         sbi->max_cluster = total_clusters + FAT_START_ENT;
1683         /* check the free_clusters, it's not necessarily correct */
1684         if (sbi->free_clusters != -1 && sbi->free_clusters > total_clusters)
1685                 sbi->free_clusters = -1;
1686         /* check the prev_free, it's not necessarily correct */
1687         sbi->prev_free %= sbi->max_cluster;
1688         if (sbi->prev_free < FAT_START_ENT)
1689                 sbi->prev_free = FAT_START_ENT;
1690
1691         /* set up enough so that it can read an inode */
1692         fat_hash_init(sb);
1693         dir_hash_init(sb);
1694         fat_ent_access_init(sb);
1695
1696         /*
1697          * The low byte of FAT's first entry must have same value with
1698          * media-field.  But in real world, too many devices is
1699          * writing wrong value.  So, removed that validity check.
1700          *
1701          * if (FAT_FIRST_ENT(sb, media) != first)
1702          */
1703
1704         error = -EINVAL;
1705         sprintf(buf, "cp%d", sbi->options.codepage);
1706         sbi->nls_disk = load_nls(buf);
1707         if (!sbi->nls_disk) {
1708                 fat_msg(sb, KERN_ERR, "codepage %s not found", buf);
1709                 goto out_fail;
1710         }
1711
1712         /* FIXME: utf8 is using iocharset for upper/lower conversion */
1713         if (sbi->options.isvfat) {
1714                 sbi->nls_io = load_nls(sbi->options.iocharset);
1715                 if (!sbi->nls_io) {
1716                         fat_msg(sb, KERN_ERR, "IO charset %s not found",
1717                                sbi->options.iocharset);
1718                         goto out_fail;
1719                 }
1720         }
1721
1722         error = -ENOMEM;
1723         fat_inode = new_inode(sb);
1724         if (!fat_inode)
1725                 goto out_fail;
1726         MSDOS_I(fat_inode)->i_pos = 0;
1727         sbi->fat_inode = fat_inode;
1728
1729         fsinfo_inode = new_inode(sb);
1730         if (!fsinfo_inode)
1731                 goto out_fail;
1732         fsinfo_inode->i_ino = MSDOS_FSINFO_INO;
1733         sbi->fsinfo_inode = fsinfo_inode;
1734         insert_inode_hash(fsinfo_inode);
1735
1736         root_inode = new_inode(sb);
1737         if (!root_inode)
1738                 goto out_fail;
1739         root_inode->i_ino = MSDOS_ROOT_INO;
1740         root_inode->i_version = 1;
1741         error = fat_read_root(root_inode);
1742         if (error < 0) {
1743                 iput(root_inode);
1744                 goto out_fail;
1745         }
1746         error = -ENOMEM;
1747         insert_inode_hash(root_inode);
1748         fat_attach(root_inode, 0);
1749         sb->s_root = d_make_root(root_inode);
1750         if (!sb->s_root) {
1751                 fat_msg(sb, KERN_ERR, "get root inode failed");
1752                 goto out_fail;
1753         }
1754
1755         if (sbi->options.discard) {
1756                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1757                 if (!blk_queue_discard(q))
1758                         fat_msg(sb, KERN_WARNING,
1759                                         "mounting with \"discard\" option, but "
1760                                         "the device does not support discard");
1761         }
1762
1763         fat_set_state(sb, 1, 0);
1764         return 0;
1765
1766 out_invalid:
1767         error = -EINVAL;
1768         if (!silent)
1769                 fat_msg(sb, KERN_INFO, "Can't find a valid FAT filesystem");
1770
1771 out_fail:
1772         if (fsinfo_inode)
1773                 iput(fsinfo_inode);
1774         if (fat_inode)
1775                 iput(fat_inode);
1776         unload_nls(sbi->nls_io);
1777         unload_nls(sbi->nls_disk);
1778         if (sbi->options.iocharset != fat_default_iocharset)
1779                 kfree(sbi->options.iocharset);
1780         sb->s_fs_info = NULL;
1781         kfree(sbi);
1782         return error;
1783 }
1784
1785 EXPORT_SYMBOL_GPL(fat_fill_super);
1786
1787 /*
1788  * helper function for fat_flush_inodes.  This writes both the inode
1789  * and the file data blocks, waiting for in flight data blocks before
1790  * the start of the call.  It does not wait for any io started
1791  * during the call
1792  */
1793 static int writeback_inode(struct inode *inode)
1794 {
1795
1796         int ret;
1797
1798         /* if we used wait=1, sync_inode_metadata waits for the io for the
1799         * inode to finish.  So wait=0 is sent down to sync_inode_metadata
1800         * and filemap_fdatawrite is used for the data blocks
1801         */
1802         ret = sync_inode_metadata(inode, 0);
1803         if (!ret)
1804                 ret = filemap_fdatawrite(inode->i_mapping);
1805         return ret;
1806 }
1807
1808 /*
1809  * write data and metadata corresponding to i1 and i2.  The io is
1810  * started but we do not wait for any of it to finish.
1811  *
1812  * filemap_flush is used for the block device, so if there is a dirty
1813  * page for a block already in flight, we will not wait and start the
1814  * io over again
1815  */
1816 int fat_flush_inodes(struct super_block *sb, struct inode *i1, struct inode *i2)
1817 {
1818         int ret = 0;
1819         if (!MSDOS_SB(sb)->options.flush)
1820                 return 0;
1821         if (i1)
1822                 ret = writeback_inode(i1);
1823         if (!ret && i2)
1824                 ret = writeback_inode(i2);
1825         if (!ret) {
1826                 struct address_space *mapping = sb->s_bdev->bd_inode->i_mapping;
1827                 ret = filemap_flush(mapping);
1828         }
1829         return ret;
1830 }
1831 EXPORT_SYMBOL_GPL(fat_flush_inodes);
1832
1833 static int __init init_fat_fs(void)
1834 {
1835         int err;
1836
1837         err = fat_cache_init();
1838         if (err)
1839                 return err;
1840
1841         err = fat_init_inodecache();
1842         if (err)
1843                 goto failed;
1844
1845         return 0;
1846
1847 failed:
1848         fat_cache_destroy();
1849         return err;
1850 }
1851
1852 static void __exit exit_fat_fs(void)
1853 {
1854         fat_cache_destroy();
1855         fat_destroy_inodecache();
1856 }
1857
1858 module_init(init_fat_fs)
1859 module_exit(exit_fat_fs)
1860
1861 MODULE_LICENSE("GPL");