?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: Use sb_any_quota_loaded() instead of sb_any_quota_enabled()
[linux-2.6/btrfs-unstable.git] / fs / ext4 / super.c
blob49fcf8864e7655a012a500bc6cedf3824f3532a2
1 /*
2 * linux/fs/ext4/super.c
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/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
42
43 #include "ext4.h"
44 #include "ext4_jbd2.h"
45 #include "xattr.h"
46 #include "acl.h"
47 #include "namei.h"
48 #include "group.h"
49
50 struct proc_dir_entry *ext4_proc_root;
51
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
55 unsigned int);
56 static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
64 char nbuf[16]);
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
70
71
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
74 {
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 }
79
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
82 {
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 }
87
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 }
95
96 void ext4_block_bitmap_set(struct super_block *sb,
97 struct ext4_group_desc *bg, ext4_fsblk_t blk)
98 {
99 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
100 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
102 }
103
104 void ext4_inode_bitmap_set(struct super_block *sb,
105 struct ext4_group_desc *bg, ext4_fsblk_t blk)
106 {
107 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
108 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
110 }
111
112 void ext4_inode_table_set(struct super_block *sb,
113 struct ext4_group_desc *bg, ext4_fsblk_t blk)
114 {
115 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
116 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 }
119
120 /*
121 * Wrappers for jbd2_journal_start/end.
122 *
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
126 * appropriate.
127 */
128 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 {
130 journal_t *journal;
131
132 if (sb->s_flags & MS_RDONLY)
133 return ERR_PTR(-EROFS);
134
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
138 journal = EXT4_SB(sb)->s_journal;
139 if (is_journal_aborted(journal)) {
140 ext4_abort(sb, __func__,
141 "Detected aborted journal");
142 return ERR_PTR(-EROFS);
143 }
144
145 return jbd2_journal_start(journal, nblocks);
146 }
147
148 /*
149 * The only special thing we need to do here is to make sure that all
150 * jbd2_journal_stop calls result in the superblock being marked dirty, so
151 * that sync() will call the filesystem's write_super callback if
152 * appropriate.
153 */
154 int __ext4_journal_stop(const char *where, handle_t *handle)
155 {
156 struct super_block *sb;
157 int err;
158 int rc;
159
160 sb = handle->h_transaction->t_journal->j_private;
161 err = handle->h_err;
162 rc = jbd2_journal_stop(handle);
163
164 if (!err)
165 err = rc;
166 if (err)
167 __ext4_std_error(sb, where, err);
168 return err;
169 }
170
171 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
172 struct buffer_head *bh, handle_t *handle, int err)
173 {
174 char nbuf[16];
175 const char *errstr = ext4_decode_error(NULL, err, nbuf);
176
177 if (bh)
178 BUFFER_TRACE(bh, "abort");
179
180 if (!handle->h_err)
181 handle->h_err = err;
182
183 if (is_handle_aborted(handle))
184 return;
185
186 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 caller, errstr, err_fn);
188
189 jbd2_journal_abort_handle(handle);
190 }
191
192 /* Deal with the reporting of failure conditions on a filesystem such as
193 * inconsistencies detected or read IO failures.
194 *
195 * On ext2, we can store the error state of the filesystem in the
196 * superblock. That is not possible on ext4, because we may have other
197 * write ordering constraints on the superblock which prevent us from
198 * writing it out straight away; and given that the journal is about to
199 * be aborted, we can't rely on the current, or future, transactions to
200 * write out the superblock safely.
201 *
202 * We'll just use the jbd2_journal_abort() error code to record an error in
203 * the journal instead. On recovery, the journal will compain about
204 * that error until we've noted it down and cleared it.
205 */
206
207 static void ext4_handle_error(struct super_block *sb)
208 {
209 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
210
211 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
213
214 if (sb->s_flags & MS_RDONLY)
215 return;
216
217 if (!test_opt(sb, ERRORS_CONT)) {
218 journal_t *journal = EXT4_SB(sb)->s_journal;
219
220 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
221 if (journal)
222 jbd2_journal_abort(journal, -EIO);
223 }
224 if (test_opt(sb, ERRORS_RO)) {
225 printk(KERN_CRIT "Remounting filesystem read-only\n");
226 sb->s_flags |= MS_RDONLY;
227 }
228 ext4_commit_super(sb, es, 1);
229 if (test_opt(sb, ERRORS_PANIC))
230 panic("EXT4-fs (device %s): panic forced after error\n",
231 sb->s_id);
232 }
233
234 void ext4_error(struct super_block *sb, const char *function,
235 const char *fmt, ...)
236 {
237 va_list args;
238
239 va_start(args, fmt);
240 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
241 vprintk(fmt, args);
242 printk("\n");
243 va_end(args);
244
245 ext4_handle_error(sb);
246 }
247
248 static const char *ext4_decode_error(struct super_block *sb, int errno,
249 char nbuf[16])
250 {
251 char *errstr = NULL;
252
253 switch (errno) {
254 case -EIO:
255 errstr = "IO failure";
256 break;
257 case -ENOMEM:
258 errstr = "Out of memory";
259 break;
260 case -EROFS:
261 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
262 errstr = "Journal has aborted";
263 else
264 errstr = "Readonly filesystem";
265 break;
266 default:
267 /* If the caller passed in an extra buffer for unknown
268 * errors, textualise them now. Else we just return
269 * NULL. */
270 if (nbuf) {
271 /* Check for truncated error codes... */
272 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 errstr = nbuf;
274 }
275 break;
276 }
277
278 return errstr;
279 }
280
281 /* __ext4_std_error decodes expected errors from journaling functions
282 * automatically and invokes the appropriate error response. */
283
284 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
285 {
286 char nbuf[16];
287 const char *errstr;
288
289 /* Special case: if the error is EROFS, and we're not already
290 * inside a transaction, then there's really no point in logging
291 * an error. */
292 if (errno == -EROFS && journal_current_handle() == NULL &&
293 (sb->s_flags & MS_RDONLY))
294 return;
295
296 errstr = ext4_decode_error(sb, errno, nbuf);
297 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 sb->s_id, function, errstr);
299
300 ext4_handle_error(sb);
301 }
302
303 /*
304 * ext4_abort is a much stronger failure handler than ext4_error. The
305 * abort function may be used to deal with unrecoverable failures such
306 * as journal IO errors or ENOMEM at a critical moment in log management.
307 *
308 * We unconditionally force the filesystem into an ABORT|READONLY state,
309 * unless the error response on the fs has been set to panic in which
310 * case we take the easy way out and panic immediately.
311 */
312
313 void ext4_abort(struct super_block *sb, const char *function,
314 const char *fmt, ...)
315 {
316 va_list args;
317
318 printk(KERN_CRIT "ext4_abort called.\n");
319
320 va_start(args, fmt);
321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
322 vprintk(fmt, args);
323 printk("\n");
324 va_end(args);
325
326 if (test_opt(sb, ERRORS_PANIC))
327 panic("EXT4-fs panic from previous error\n");
328
329 if (sb->s_flags & MS_RDONLY)
330 return;
331
332 printk(KERN_CRIT "Remounting filesystem read-only\n");
333 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
334 sb->s_flags |= MS_RDONLY;
335 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
336 if (EXT4_SB(sb)->s_journal)
337 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 }
339
340 void ext4_warning(struct super_block *sb, const char *function,
341 const char *fmt, ...)
342 {
343 va_list args;
344
345 va_start(args, fmt);
346 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
347 sb->s_id, function);
348 vprintk(fmt, args);
349 printk("\n");
350 va_end(args);
351 }
352
353 void ext4_update_dynamic_rev(struct super_block *sb)
354 {
355 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
356
357 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 return;
359
360 ext4_warning(sb, __func__,
361 "updating to rev %d because of new feature flag, "
362 "running e2fsck is recommended",
363 EXT4_DYNAMIC_REV);
364
365 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
366 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
367 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
368 /* leave es->s_feature_*compat flags alone */
369 /* es->s_uuid will be set by e2fsck if empty */
370
371 /*
372 * The rest of the superblock fields should be zero, and if not it
373 * means they are likely already in use, so leave them alone. We
374 * can leave it up to e2fsck to clean up any inconsistencies there.
375 */
376 }
377
378 /*
379 * Open the external journal device
380 */
381 static struct block_device *ext4_blkdev_get(dev_t dev)
382 {
383 struct block_device *bdev;
384 char b[BDEVNAME_SIZE];
385
386 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
387 if (IS_ERR(bdev))
388 goto fail;
389 return bdev;
390
391 fail:
392 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
393 __bdevname(dev, b), PTR_ERR(bdev));
394 return NULL;
395 }
396
397 /*
398 * Release the journal device
399 */
400 static int ext4_blkdev_put(struct block_device *bdev)
401 {
402 bd_release(bdev);
403 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
404 }
405
406 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
407 {
408 struct block_device *bdev;
409 int ret = -ENODEV;
410
411 bdev = sbi->journal_bdev;
412 if (bdev) {
413 ret = ext4_blkdev_put(bdev);
414 sbi->journal_bdev = NULL;
415 }
416 return ret;
417 }
418
419 static inline struct inode *orphan_list_entry(struct list_head *l)
420 {
421 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
422 }
423
424 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
425 {
426 struct list_head *l;
427
428 printk(KERN_ERR "sb orphan head is %d\n",
429 le32_to_cpu(sbi->s_es->s_last_orphan));
430
431 printk(KERN_ERR "sb_info orphan list:\n");
432 list_for_each(l, &sbi->s_orphan) {
433 struct inode *inode = orphan_list_entry(l);
434 printk(KERN_ERR " "
435 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
436 inode->i_sb->s_id, inode->i_ino, inode,
437 inode->i_mode, inode->i_nlink,
438 NEXT_ORPHAN(inode));
439 }
440 }
441
442 static void ext4_put_super(struct super_block *sb)
443 {
444 struct ext4_sb_info *sbi = EXT4_SB(sb);
445 struct ext4_super_block *es = sbi->s_es;
446 int i, err;
447
448 ext4_mb_release(sb);
449 ext4_ext_release(sb);
450 ext4_xattr_put_super(sb);
451 err = jbd2_journal_destroy(sbi->s_journal);
452 sbi->s_journal = NULL;
453 if (err < 0)
454 ext4_abort(sb, __func__, "Couldn't clean up the journal");
455
456 if (!(sb->s_flags & MS_RDONLY)) {
457 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
458 es->s_state = cpu_to_le16(sbi->s_mount_state);
459 ext4_commit_super(sb, es, 1);
460 }
461 if (sbi->s_proc) {
462 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
463 remove_proc_entry(sb->s_id, ext4_proc_root);
464 }
465
466 for (i = 0; i < sbi->s_gdb_count; i++)
467 brelse(sbi->s_group_desc[i]);
468 kfree(sbi->s_group_desc);
469 kfree(sbi->s_flex_groups);
470 percpu_counter_destroy(&sbi->s_freeblocks_counter);
471 percpu_counter_destroy(&sbi->s_freeinodes_counter);
472 percpu_counter_destroy(&sbi->s_dirs_counter);
473 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
474 brelse(sbi->s_sbh);
475 #ifdef CONFIG_QUOTA
476 for (i = 0; i < MAXQUOTAS; i++)
477 kfree(sbi->s_qf_names[i]);
478 #endif
479
480 /* Debugging code just in case the in-memory inode orphan list
481 * isn't empty. The on-disk one can be non-empty if we've
482 * detected an error and taken the fs readonly, but the
483 * in-memory list had better be clean by this point. */
484 if (!list_empty(&sbi->s_orphan))
485 dump_orphan_list(sb, sbi);
486 J_ASSERT(list_empty(&sbi->s_orphan));
487
488 invalidate_bdev(sb->s_bdev);
489 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
490 /*
491 * Invalidate the journal device's buffers. We don't want them
492 * floating about in memory - the physical journal device may
493 * hotswapped, and it breaks the `ro-after' testing code.
494 */
495 sync_blockdev(sbi->journal_bdev);
496 invalidate_bdev(sbi->journal_bdev);
497 ext4_blkdev_remove(sbi);
498 }
499 sb->s_fs_info = NULL;
500 kfree(sbi);
501 return;
502 }
503
504 static struct kmem_cache *ext4_inode_cachep;
505
506 /*
507 * Called inside transaction, so use GFP_NOFS
508 */
509 static struct inode *ext4_alloc_inode(struct super_block *sb)
510 {
511 struct ext4_inode_info *ei;
512
513 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
514 if (!ei)
515 return NULL;
516 #ifdef CONFIG_EXT4_FS_POSIX_ACL
517 ei->i_acl = EXT4_ACL_NOT_CACHED;
518 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
519 #endif
520 ei->vfs_inode.i_version = 1;
521 ei->vfs_inode.i_data.writeback_index = 0;
522 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
523 INIT_LIST_HEAD(&ei->i_prealloc_list);
524 spin_lock_init(&ei->i_prealloc_lock);
525 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
526 ei->i_reserved_data_blocks = 0;
527 ei->i_reserved_meta_blocks = 0;
528 ei->i_allocated_meta_blocks = 0;
529 ei->i_delalloc_reserved_flag = 0;
530 spin_lock_init(&(ei->i_block_reservation_lock));
531 return &ei->vfs_inode;
532 }
533
534 static void ext4_destroy_inode(struct inode *inode)
535 {
536 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
537 printk("EXT4 Inode %p: orphan list check failed!\n",
538 EXT4_I(inode));
539 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
540 EXT4_I(inode), sizeof(struct ext4_inode_info),
541 true);
542 dump_stack();
543 }
544 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
545 }
546
547 static void init_once(void *foo)
548 {
549 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
550
551 INIT_LIST_HEAD(&ei->i_orphan);
552 #ifdef CONFIG_EXT4_FS_XATTR
553 init_rwsem(&ei->xattr_sem);
554 #endif
555 init_rwsem(&ei->i_data_sem);
556 inode_init_once(&ei->vfs_inode);
557 }
558
559 static int init_inodecache(void)
560 {
561 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
562 sizeof(struct ext4_inode_info),
563 0, (SLAB_RECLAIM_ACCOUNT|
564 SLAB_MEM_SPREAD),
565 init_once);
566 if (ext4_inode_cachep == NULL)
567 return -ENOMEM;
568 return 0;
569 }
570
571 static void destroy_inodecache(void)
572 {
573 kmem_cache_destroy(ext4_inode_cachep);
574 }
575
576 static void ext4_clear_inode(struct inode *inode)
577 {
578 #ifdef CONFIG_EXT4_FS_POSIX_ACL
579 if (EXT4_I(inode)->i_acl &&
580 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
581 posix_acl_release(EXT4_I(inode)->i_acl);
582 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
583 }
584 if (EXT4_I(inode)->i_default_acl &&
585 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
586 posix_acl_release(EXT4_I(inode)->i_default_acl);
587 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
588 }
589 #endif
590 ext4_discard_preallocations(inode);
591 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
592 &EXT4_I(inode)->jinode);
593 }
594
595 static inline void ext4_show_quota_options(struct seq_file *seq,
596 struct super_block *sb)
597 {
598 #if defined(CONFIG_QUOTA)
599 struct ext4_sb_info *sbi = EXT4_SB(sb);
600
601 if (sbi->s_jquota_fmt)
602 seq_printf(seq, ",jqfmt=%s",
603 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
604
605 if (sbi->s_qf_names[USRQUOTA])
606 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
607
608 if (sbi->s_qf_names[GRPQUOTA])
609 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
610
611 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
612 seq_puts(seq, ",usrquota");
613
614 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
615 seq_puts(seq, ",grpquota");
616 #endif
617 }
618
619 /*
620 * Show an option if
621 * - it's set to a non-default value OR
622 * - if the per-sb default is different from the global default
623 */
624 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
625 {
626 int def_errors;
627 unsigned long def_mount_opts;
628 struct super_block *sb = vfs->mnt_sb;
629 struct ext4_sb_info *sbi = EXT4_SB(sb);
630 struct ext4_super_block *es = sbi->s_es;
631
632 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
633 def_errors = le16_to_cpu(es->s_errors);
634
635 if (sbi->s_sb_block != 1)
636 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
637 if (test_opt(sb, MINIX_DF))
638 seq_puts(seq, ",minixdf");
639 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
640 seq_puts(seq, ",grpid");
641 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
642 seq_puts(seq, ",nogrpid");
643 if (sbi->s_resuid != EXT4_DEF_RESUID ||
644 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
645 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
646 }
647 if (sbi->s_resgid != EXT4_DEF_RESGID ||
648 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
649 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
650 }
651 if (test_opt(sb, ERRORS_RO)) {
652 if (def_errors == EXT4_ERRORS_PANIC ||
653 def_errors == EXT4_ERRORS_CONTINUE) {
654 seq_puts(seq, ",errors=remount-ro");
655 }
656 }
657 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
658 seq_puts(seq, ",errors=continue");
659 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
660 seq_puts(seq, ",errors=panic");
661 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
662 seq_puts(seq, ",nouid32");
663 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
664 seq_puts(seq, ",debug");
665 if (test_opt(sb, OLDALLOC))
666 seq_puts(seq, ",oldalloc");
667 #ifdef CONFIG_EXT4_FS_XATTR
668 if (test_opt(sb, XATTR_USER) &&
669 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
670 seq_puts(seq, ",user_xattr");
671 if (!test_opt(sb, XATTR_USER) &&
672 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
673 seq_puts(seq, ",nouser_xattr");
674 }
675 #endif
676 #ifdef CONFIG_EXT4_FS_POSIX_ACL
677 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
678 seq_puts(seq, ",acl");
679 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
680 seq_puts(seq, ",noacl");
681 #endif
682 if (!test_opt(sb, RESERVATION))
683 seq_puts(seq, ",noreservation");
684 if (sbi->s_commit_interval) {
685 seq_printf(seq, ",commit=%u",
686 (unsigned) (sbi->s_commit_interval / HZ));
687 }
688 /*
689 * We're changing the default of barrier mount option, so
690 * let's always display its mount state so it's clear what its
691 * status is.
692 */
693 seq_puts(seq, ",barrier=");
694 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
695 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
696 seq_puts(seq, ",journal_async_commit");
697 if (test_opt(sb, NOBH))
698 seq_puts(seq, ",nobh");
699 if (!test_opt(sb, EXTENTS))
700 seq_puts(seq, ",noextents");
701 if (test_opt(sb, I_VERSION))
702 seq_puts(seq, ",i_version");
703 if (!test_opt(sb, DELALLOC))
704 seq_puts(seq, ",nodelalloc");
705
706
707 if (sbi->s_stripe)
708 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
709 /*
710 * journal mode get enabled in different ways
711 * So just print the value even if we didn't specify it
712 */
713 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
714 seq_puts(seq, ",data=journal");
715 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
716 seq_puts(seq, ",data=ordered");
717 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
718 seq_puts(seq, ",data=writeback");
719
720 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
721 seq_printf(seq, ",inode_readahead_blks=%u",
722 sbi->s_inode_readahead_blks);
723
724 if (test_opt(sb, DATA_ERR_ABORT))
725 seq_puts(seq, ",data_err=abort");
726
727 ext4_show_quota_options(seq, sb);
728 return 0;
729 }
730
731
732 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
733 u64 ino, u32 generation)
734 {
735 struct inode *inode;
736
737 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
738 return ERR_PTR(-ESTALE);
739 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
740 return ERR_PTR(-ESTALE);
741
742 /* iget isn't really right if the inode is currently unallocated!!
743 *
744 * ext4_read_inode will return a bad_inode if the inode had been
745 * deleted, so we should be safe.
746 *
747 * Currently we don't know the generation for parent directory, so
748 * a generation of 0 means "accept any"
749 */
750 inode = ext4_iget(sb, ino);
751 if (IS_ERR(inode))
752 return ERR_CAST(inode);
753 if (generation && inode->i_generation != generation) {
754 iput(inode);
755 return ERR_PTR(-ESTALE);
756 }
757
758 return inode;
759 }
760
761 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
762 int fh_len, int fh_type)
763 {
764 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
765 ext4_nfs_get_inode);
766 }
767
768 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
769 int fh_len, int fh_type)
770 {
771 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
772 ext4_nfs_get_inode);
773 }
774
775 #ifdef CONFIG_QUOTA
776 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
777 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
778
779 static int ext4_dquot_initialize(struct inode *inode, int type);
780 static int ext4_dquot_drop(struct inode *inode);
781 static int ext4_write_dquot(struct dquot *dquot);
782 static int ext4_acquire_dquot(struct dquot *dquot);
783 static int ext4_release_dquot(struct dquot *dquot);
784 static int ext4_mark_dquot_dirty(struct dquot *dquot);
785 static int ext4_write_info(struct super_block *sb, int type);
786 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
787 char *path, int remount);
788 static int ext4_quota_on_mount(struct super_block *sb, int type);
789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
790 size_t len, loff_t off);
791 static ssize_t ext4_quota_write(struct super_block *sb, int type,
792 const char *data, size_t len, loff_t off);
793
794 static struct dquot_operations ext4_quota_operations = {
795 .initialize = ext4_dquot_initialize,
796 .drop = ext4_dquot_drop,
797 .alloc_space = dquot_alloc_space,
798 .alloc_inode = dquot_alloc_inode,
799 .free_space = dquot_free_space,
800 .free_inode = dquot_free_inode,
801 .transfer = dquot_transfer,
802 .write_dquot = ext4_write_dquot,
803 .acquire_dquot = ext4_acquire_dquot,
804 .release_dquot = ext4_release_dquot,
805 .mark_dirty = ext4_mark_dquot_dirty,
806 .write_info = ext4_write_info
807 };
808
809 static struct quotactl_ops ext4_qctl_operations = {
810 .quota_on = ext4_quota_on,
811 .quota_off = vfs_quota_off,
812 .quota_sync = vfs_quota_sync,
813 .get_info = vfs_get_dqinfo,
814 .set_info = vfs_set_dqinfo,
815 .get_dqblk = vfs_get_dqblk,
816 .set_dqblk = vfs_set_dqblk
817 };
818 #endif
819
820 static const struct super_operations ext4_sops = {
821 .alloc_inode = ext4_alloc_inode,
822 .destroy_inode = ext4_destroy_inode,
823 .write_inode = ext4_write_inode,
824 .dirty_inode = ext4_dirty_inode,
825 .delete_inode = ext4_delete_inode,
826 .put_super = ext4_put_super,
827 .write_super = ext4_write_super,
828 .sync_fs = ext4_sync_fs,
829 .write_super_lockfs = ext4_write_super_lockfs,
830 .unlockfs = ext4_unlockfs,
831 .statfs = ext4_statfs,
832 .remount_fs = ext4_remount,
833 .clear_inode = ext4_clear_inode,
834 .show_options = ext4_show_options,
835 #ifdef CONFIG_QUOTA
836 .quota_read = ext4_quota_read,
837 .quota_write = ext4_quota_write,
838 #endif
839 };
840
841 static const struct export_operations ext4_export_ops = {
842 .fh_to_dentry = ext4_fh_to_dentry,
843 .fh_to_parent = ext4_fh_to_parent,
844 .get_parent = ext4_get_parent,
845 };
846
847 enum {
848 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
849 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
850 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
851 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
852 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
853 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
854 Opt_journal_checksum, Opt_journal_async_commit,
855 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
856 Opt_data_err_abort, Opt_data_err_ignore,
857 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
858 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
859 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
860 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
861 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
862 Opt_inode_readahead_blks
863 };
864
865 static const match_table_t tokens = {
866 {Opt_bsd_df, "bsddf"},
867 {Opt_minix_df, "minixdf"},
868 {Opt_grpid, "grpid"},
869 {Opt_grpid, "bsdgroups"},
870 {Opt_nogrpid, "nogrpid"},
871 {Opt_nogrpid, "sysvgroups"},
872 {Opt_resgid, "resgid=%u"},
873 {Opt_resuid, "resuid=%u"},
874 {Opt_sb, "sb=%u"},
875 {Opt_err_cont, "errors=continue"},
876 {Opt_err_panic, "errors=panic"},
877 {Opt_err_ro, "errors=remount-ro"},
878 {Opt_nouid32, "nouid32"},
879 {Opt_debug, "debug"},
880 {Opt_oldalloc, "oldalloc"},
881 {Opt_orlov, "orlov"},
882 {Opt_user_xattr, "user_xattr"},
883 {Opt_nouser_xattr, "nouser_xattr"},
884 {Opt_acl, "acl"},
885 {Opt_noacl, "noacl"},
886 {Opt_reservation, "reservation"},
887 {Opt_noreservation, "noreservation"},
888 {Opt_noload, "noload"},
889 {Opt_nobh, "nobh"},
890 {Opt_bh, "bh"},
891 {Opt_commit, "commit=%u"},
892 {Opt_journal_update, "journal=update"},
893 {Opt_journal_inum, "journal=%u"},
894 {Opt_journal_dev, "journal_dev=%u"},
895 {Opt_journal_checksum, "journal_checksum"},
896 {Opt_journal_async_commit, "journal_async_commit"},
897 {Opt_abort, "abort"},
898 {Opt_data_journal, "data=journal"},
899 {Opt_data_ordered, "data=ordered"},
900 {Opt_data_writeback, "data=writeback"},
901 {Opt_data_err_abort, "data_err=abort"},
902 {Opt_data_err_ignore, "data_err=ignore"},
903 {Opt_offusrjquota, "usrjquota="},
904 {Opt_usrjquota, "usrjquota=%s"},
905 {Opt_offgrpjquota, "grpjquota="},
906 {Opt_grpjquota, "grpjquota=%s"},
907 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
908 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
909 {Opt_grpquota, "grpquota"},
910 {Opt_noquota, "noquota"},
911 {Opt_quota, "quota"},
912 {Opt_usrquota, "usrquota"},
913 {Opt_barrier, "barrier=%u"},
914 {Opt_extents, "extents"},
915 {Opt_noextents, "noextents"},
916 {Opt_i_version, "i_version"},
917 {Opt_stripe, "stripe=%u"},
918 {Opt_resize, "resize"},
919 {Opt_delalloc, "delalloc"},
920 {Opt_nodelalloc, "nodelalloc"},
921 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
922 {Opt_err, NULL},
923 };
924
925 static ext4_fsblk_t get_sb_block(void **data)
926 {
927 ext4_fsblk_t sb_block;
928 char *options = (char *) *data;
929
930 if (!options || strncmp(options, "sb=", 3) != 0)
931 return 1; /* Default location */
932 options += 3;
933 /*todo: use simple_strtoll with >32bit ext4 */
934 sb_block = simple_strtoul(options, &options, 0);
935 if (*options && *options != ',') {
936 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
937 (char *) *data);
938 return 1;
939 }
940 if (*options == ',')
941 options++;
942 *data = (void *) options;
943 return sb_block;
944 }
945
946 static int parse_options(char *options, struct super_block *sb,
947 unsigned int *inum, unsigned long *journal_devnum,
948 ext4_fsblk_t *n_blocks_count, int is_remount)
949 {
950 struct ext4_sb_info *sbi = EXT4_SB(sb);
951 char *p;
952 substring_t args[MAX_OPT_ARGS];
953 int data_opt = 0;
954 int option;
955 #ifdef CONFIG_QUOTA
956 int qtype, qfmt;
957 char *qname;
958 #endif
959 ext4_fsblk_t last_block;
960
961 if (!options)
962 return 1;
963
964 while ((p = strsep(&options, ",")) != NULL) {
965 int token;
966 if (!*p)
967 continue;
968
969 token = match_token(p, tokens, args);
970 switch (token) {
971 case Opt_bsd_df:
972 clear_opt(sbi->s_mount_opt, MINIX_DF);
973 break;
974 case Opt_minix_df:
975 set_opt(sbi->s_mount_opt, MINIX_DF);
976 break;
977 case Opt_grpid:
978 set_opt(sbi->s_mount_opt, GRPID);
979 break;
980 case Opt_nogrpid:
981 clear_opt(sbi->s_mount_opt, GRPID);
982 break;
983 case Opt_resuid:
984 if (match_int(&args[0], &option))
985 return 0;
986 sbi->s_resuid = option;
987 break;
988 case Opt_resgid:
989 if (match_int(&args[0], &option))
990 return 0;
991 sbi->s_resgid = option;
992 break;
993 case Opt_sb:
994 /* handled by get_sb_block() instead of here */
995 /* *sb_block = match_int(&args[0]); */
996 break;
997 case Opt_err_panic:
998 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
999 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1000 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1001 break;
1002 case Opt_err_ro:
1003 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1004 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1005 set_opt(sbi->s_mount_opt, ERRORS_RO);
1006 break;
1007 case Opt_err_cont:
1008 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1009 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1010 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1011 break;
1012 case Opt_nouid32:
1013 set_opt(sbi->s_mount_opt, NO_UID32);
1014 break;
1015 case Opt_debug:
1016 set_opt(sbi->s_mount_opt, DEBUG);
1017 break;
1018 case Opt_oldalloc:
1019 set_opt(sbi->s_mount_opt, OLDALLOC);
1020 break;
1021 case Opt_orlov:
1022 clear_opt(sbi->s_mount_opt, OLDALLOC);
1023 break;
1024 #ifdef CONFIG_EXT4_FS_XATTR
1025 case Opt_user_xattr:
1026 set_opt(sbi->s_mount_opt, XATTR_USER);
1027 break;
1028 case Opt_nouser_xattr:
1029 clear_opt(sbi->s_mount_opt, XATTR_USER);
1030 break;
1031 #else
1032 case Opt_user_xattr:
1033 case Opt_nouser_xattr:
1034 printk(KERN_ERR "EXT4 (no)user_xattr options "
1035 "not supported\n");
1036 break;
1037 #endif
1038 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1039 case Opt_acl:
1040 set_opt(sbi->s_mount_opt, POSIX_ACL);
1041 break;
1042 case Opt_noacl:
1043 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1044 break;
1045 #else
1046 case Opt_acl:
1047 case Opt_noacl:
1048 printk(KERN_ERR "EXT4 (no)acl options "
1049 "not supported\n");
1050 break;
1051 #endif
1052 case Opt_reservation:
1053 set_opt(sbi->s_mount_opt, RESERVATION);
1054 break;
1055 case Opt_noreservation:
1056 clear_opt(sbi->s_mount_opt, RESERVATION);
1057 break;
1058 case Opt_journal_update:
1059 /* @@@ FIXME */
1060 /* Eventually we will want to be able to create
1061 a journal file here. For now, only allow the
1062 user to specify an existing inode to be the
1063 journal file. */
1064 if (is_remount) {
1065 printk(KERN_ERR "EXT4-fs: cannot specify "
1066 "journal on remount\n");
1067 return 0;
1068 }
1069 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1070 break;
1071 case Opt_journal_inum:
1072 if (is_remount) {
1073 printk(KERN_ERR "EXT4-fs: cannot specify "
1074 "journal on remount\n");
1075 return 0;
1076 }
1077 if (match_int(&args[0], &option))
1078 return 0;
1079 *inum = option;
1080 break;
1081 case Opt_journal_dev:
1082 if (is_remount) {
1083 printk(KERN_ERR "EXT4-fs: cannot specify "
1084 "journal on remount\n");
1085 return 0;
1086 }
1087 if (match_int(&args[0], &option))
1088 return 0;
1089 *journal_devnum = option;
1090 break;
1091 case Opt_journal_checksum:
1092 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1093 break;
1094 case Opt_journal_async_commit:
1095 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1096 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1097 break;
1098 case Opt_noload:
1099 set_opt(sbi->s_mount_opt, NOLOAD);
1100 break;
1101 case Opt_commit:
1102 if (match_int(&args[0], &option))
1103 return 0;
1104 if (option < 0)
1105 return 0;
1106 if (option == 0)
1107 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1108 sbi->s_commit_interval = HZ * option;
1109 break;
1110 case Opt_data_journal:
1111 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1112 goto datacheck;
1113 case Opt_data_ordered:
1114 data_opt = EXT4_MOUNT_ORDERED_DATA;
1115 goto datacheck;
1116 case Opt_data_writeback:
1117 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1118 datacheck:
1119 if (is_remount) {
1120 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1121 != data_opt) {
1122 printk(KERN_ERR
1123 "EXT4-fs: cannot change data "
1124 "mode on remount\n");
1125 return 0;
1126 }
1127 } else {
1128 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1129 sbi->s_mount_opt |= data_opt;
1130 }
1131 break;
1132 case Opt_data_err_abort:
1133 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1134 break;
1135 case Opt_data_err_ignore:
1136 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1137 break;
1138 #ifdef CONFIG_QUOTA
1139 case Opt_usrjquota:
1140 qtype = USRQUOTA;
1141 goto set_qf_name;
1142 case Opt_grpjquota:
1143 qtype = GRPQUOTA;
1144 set_qf_name:
1145 if (sb_any_quota_loaded(sb) &&
1146 !sbi->s_qf_names[qtype]) {
1147 printk(KERN_ERR
1148 "EXT4-fs: Cannot change journaled "
1149 "quota options when quota turned on.\n");
1150 return 0;
1151 }
1152 qname = match_strdup(&args[0]);
1153 if (!qname) {
1154 printk(KERN_ERR
1155 "EXT4-fs: not enough memory for "
1156 "storing quotafile name.\n");
1157 return 0;
1158 }
1159 if (sbi->s_qf_names[qtype] &&
1160 strcmp(sbi->s_qf_names[qtype], qname)) {
1161 printk(KERN_ERR
1162 "EXT4-fs: %s quota file already "
1163 "specified.\n", QTYPE2NAME(qtype));
1164 kfree(qname);
1165 return 0;
1166 }
1167 sbi->s_qf_names[qtype] = qname;
1168 if (strchr(sbi->s_qf_names[qtype], '/')) {
1169 printk(KERN_ERR
1170 "EXT4-fs: quotafile must be on "
1171 "filesystem root.\n");
1172 kfree(sbi->s_qf_names[qtype]);
1173 sbi->s_qf_names[qtype] = NULL;
1174 return 0;
1175 }
1176 set_opt(sbi->s_mount_opt, QUOTA);
1177 break;
1178 case Opt_offusrjquota:
1179 qtype = USRQUOTA;
1180 goto clear_qf_name;
1181 case Opt_offgrpjquota:
1182 qtype = GRPQUOTA;
1183 clear_qf_name:
1184 if (sb_any_quota_loaded(sb) &&
1185 sbi->s_qf_names[qtype]) {
1186 printk(KERN_ERR "EXT4-fs: Cannot change "
1187 "journaled quota options when "
1188 "quota turned on.\n");
1189 return 0;
1190 }
1191 /*
1192 * The space will be released later when all options
1193 * are confirmed to be correct
1194 */
1195 sbi->s_qf_names[qtype] = NULL;
1196 break;
1197 case Opt_jqfmt_vfsold:
1198 qfmt = QFMT_VFS_OLD;
1199 goto set_qf_format;
1200 case Opt_jqfmt_vfsv0:
1201 qfmt = QFMT_VFS_V0;
1202 set_qf_format:
1203 if (sb_any_quota_loaded(sb) &&
1204 sbi->s_jquota_fmt != qfmt) {
1205 printk(KERN_ERR "EXT4-fs: Cannot change "
1206 "journaled quota options when "
1207 "quota turned on.\n");
1208 return 0;
1209 }
1210 sbi->s_jquota_fmt = qfmt;
1211 break;
1212 case Opt_quota:
1213 case Opt_usrquota:
1214 set_opt(sbi->s_mount_opt, QUOTA);
1215 set_opt(sbi->s_mount_opt, USRQUOTA);
1216 break;
1217 case Opt_grpquota:
1218 set_opt(sbi->s_mount_opt, QUOTA);
1219 set_opt(sbi->s_mount_opt, GRPQUOTA);
1220 break;
1221 case Opt_noquota:
1222 if (sb_any_quota_loaded(sb)) {
1223 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1224 "options when quota turned on.\n");
1225 return 0;
1226 }
1227 clear_opt(sbi->s_mount_opt, QUOTA);
1228 clear_opt(sbi->s_mount_opt, USRQUOTA);
1229 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1230 break;
1231 #else
1232 case Opt_quota:
1233 case Opt_usrquota:
1234 case Opt_grpquota:
1235 printk(KERN_ERR
1236 "EXT4-fs: quota options not supported.\n");
1237 break;
1238 case Opt_usrjquota:
1239 case Opt_grpjquota:
1240 case Opt_offusrjquota:
1241 case Opt_offgrpjquota:
1242 case Opt_jqfmt_vfsold:
1243 case Opt_jqfmt_vfsv0:
1244 printk(KERN_ERR
1245 "EXT4-fs: journaled quota options not "
1246 "supported.\n");
1247 break;
1248 case Opt_noquota:
1249 break;
1250 #endif
1251 case Opt_abort:
1252 set_opt(sbi->s_mount_opt, ABORT);
1253 break;
1254 case Opt_barrier:
1255 if (match_int(&args[0], &option))
1256 return 0;
1257 if (option)
1258 set_opt(sbi->s_mount_opt, BARRIER);
1259 else
1260 clear_opt(sbi->s_mount_opt, BARRIER);
1261 break;
1262 case Opt_ignore:
1263 break;
1264 case Opt_resize:
1265 if (!is_remount) {
1266 printk("EXT4-fs: resize option only available "
1267 "for remount\n");
1268 return 0;
1269 }
1270 if (match_int(&args[0], &option) != 0)
1271 return 0;
1272 *n_blocks_count = option;
1273 break;
1274 case Opt_nobh:
1275 set_opt(sbi->s_mount_opt, NOBH);
1276 break;
1277 case Opt_bh:
1278 clear_opt(sbi->s_mount_opt, NOBH);
1279 break;
1280 case Opt_extents:
1281 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1282 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1283 ext4_warning(sb, __func__,
1284 "extents feature not enabled "
1285 "on this filesystem, use tune2fs\n");
1286 return 0;
1287 }
1288 set_opt(sbi->s_mount_opt, EXTENTS);
1289 break;
1290 case Opt_noextents:
1291 /*
1292 * When e2fsprogs support resizing an already existing
1293 * ext3 file system to greater than 2**32 we need to
1294 * add support to block allocator to handle growing
1295 * already existing block mapped inode so that blocks
1296 * allocated for them fall within 2**32
1297 */
1298 last_block = ext4_blocks_count(sbi->s_es) - 1;
1299 if (last_block > 0xffffffffULL) {
1300 printk(KERN_ERR "EXT4-fs: Filesystem too "
1301 "large to mount with "
1302 "-o noextents options\n");
1303 return 0;
1304 }
1305 clear_opt(sbi->s_mount_opt, EXTENTS);
1306 break;
1307 case Opt_i_version:
1308 set_opt(sbi->s_mount_opt, I_VERSION);
1309 sb->s_flags |= MS_I_VERSION;
1310 break;
1311 case Opt_nodelalloc:
1312 clear_opt(sbi->s_mount_opt, DELALLOC);
1313 break;
1314 case Opt_stripe:
1315 if (match_int(&args[0], &option))
1316 return 0;
1317 if (option < 0)
1318 return 0;
1319 sbi->s_stripe = option;
1320 break;
1321 case Opt_delalloc:
1322 set_opt(sbi->s_mount_opt, DELALLOC);
1323 break;
1324 case Opt_inode_readahead_blks:
1325 if (match_int(&args[0], &option))
1326 return 0;
1327 if (option < 0 || option > (1 << 30))
1328 return 0;
1329 sbi->s_inode_readahead_blks = option;
1330 break;
1331 default:
1332 printk(KERN_ERR
1333 "EXT4-fs: Unrecognized mount option \"%s\" "
1334 "or missing value\n", p);
1335 return 0;
1336 }
1337 }
1338 #ifdef CONFIG_QUOTA
1339 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1340 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1341 sbi->s_qf_names[USRQUOTA])
1342 clear_opt(sbi->s_mount_opt, USRQUOTA);
1343
1344 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1345 sbi->s_qf_names[GRPQUOTA])
1346 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1347
1348 if ((sbi->s_qf_names[USRQUOTA] &&
1349 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1350 (sbi->s_qf_names[GRPQUOTA] &&
1351 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1352 printk(KERN_ERR "EXT4-fs: old and new quota "
1353 "format mixing.\n");
1354 return 0;
1355 }
1356
1357 if (!sbi->s_jquota_fmt) {
1358 printk(KERN_ERR "EXT4-fs: journaled quota format "
1359 "not specified.\n");
1360 return 0;
1361 }
1362 } else {
1363 if (sbi->s_jquota_fmt) {
1364 printk(KERN_ERR "EXT4-fs: journaled quota format "
1365 "specified with no journaling "
1366 "enabled.\n");
1367 return 0;
1368 }
1369 }
1370 #endif
1371 return 1;
1372 }
1373
1374 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1375 int read_only)
1376 {
1377 struct ext4_sb_info *sbi = EXT4_SB(sb);
1378 int res = 0;
1379
1380 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1381 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1382 "forcing read-only mode\n");
1383 res = MS_RDONLY;
1384 }
1385 if (read_only)
1386 return res;
1387 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1388 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1389 "running e2fsck is recommended\n");
1390 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1391 printk(KERN_WARNING
1392 "EXT4-fs warning: mounting fs with errors, "
1393 "running e2fsck is recommended\n");
1394 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1395 le16_to_cpu(es->s_mnt_count) >=
1396 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1397 printk(KERN_WARNING
1398 "EXT4-fs warning: maximal mount count reached, "
1399 "running e2fsck is recommended\n");
1400 else if (le32_to_cpu(es->s_checkinterval) &&
1401 (le32_to_cpu(es->s_lastcheck) +
1402 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1403 printk(KERN_WARNING
1404 "EXT4-fs warning: checktime reached, "
1405 "running e2fsck is recommended\n");
1406 #if 0
1407 /* @@@ We _will_ want to clear the valid bit if we find
1408 * inconsistencies, to force a fsck at reboot. But for
1409 * a plain journaled filesystem we can keep it set as
1410 * valid forever! :)
1411 */
1412 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1413 #endif
1414 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1415 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1416 le16_add_cpu(&es->s_mnt_count, 1);
1417 es->s_mtime = cpu_to_le32(get_seconds());
1418 ext4_update_dynamic_rev(sb);
1419 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1420
1421 ext4_commit_super(sb, es, 1);
1422 if (test_opt(sb, DEBUG))
1423 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1424 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1425 sb->s_blocksize,
1426 sbi->s_groups_count,
1427 EXT4_BLOCKS_PER_GROUP(sb),
1428 EXT4_INODES_PER_GROUP(sb),
1429 sbi->s_mount_opt);
1430
1431 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1432 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1433 "external", EXT4_SB(sb)->s_journal->j_devname);
1434 return res;
1435 }
1436
1437 static int ext4_fill_flex_info(struct super_block *sb)
1438 {
1439 struct ext4_sb_info *sbi = EXT4_SB(sb);
1440 struct ext4_group_desc *gdp = NULL;
1441 struct buffer_head *bh;
1442 ext4_group_t flex_group_count;
1443 ext4_group_t flex_group;
1444 int groups_per_flex = 0;
1445 __u64 block_bitmap = 0;
1446 int i;
1447
1448 if (!sbi->s_es->s_log_groups_per_flex) {
1449 sbi->s_log_groups_per_flex = 0;
1450 return 1;
1451 }
1452
1453 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1454 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1455
1456 /* We allocate both existing and potentially added groups */
1457 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1458 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1459 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1460 sbi->s_flex_groups = kzalloc(flex_group_count *
1461 sizeof(struct flex_groups), GFP_KERNEL);
1462 if (sbi->s_flex_groups == NULL) {
1463 printk(KERN_ERR "EXT4-fs: not enough memory for "
1464 "%lu flex groups\n", flex_group_count);
1465 goto failed;
1466 }
1467
1468 gdp = ext4_get_group_desc(sb, 1, &bh);
1469 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1470
1471 for (i = 0; i < sbi->s_groups_count; i++) {
1472 gdp = ext4_get_group_desc(sb, i, &bh);
1473
1474 flex_group = ext4_flex_group(sbi, i);
1475 sbi->s_flex_groups[flex_group].free_inodes +=
1476 le16_to_cpu(gdp->bg_free_inodes_count);
1477 sbi->s_flex_groups[flex_group].free_blocks +=
1478 le16_to_cpu(gdp->bg_free_blocks_count);
1479 }
1480
1481 return 1;
1482 failed:
1483 return 0;
1484 }
1485
1486 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1487 struct ext4_group_desc *gdp)
1488 {
1489 __u16 crc = 0;
1490
1491 if (sbi->s_es->s_feature_ro_compat &
1492 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1493 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1494 __le32 le_group = cpu_to_le32(block_group);
1495
1496 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1497 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1498 crc = crc16(crc, (__u8 *)gdp, offset);
1499 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1500 /* for checksum of struct ext4_group_desc do the rest...*/
1501 if ((sbi->s_es->s_feature_incompat &
1502 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1503 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1504 crc = crc16(crc, (__u8 *)gdp + offset,
1505 le16_to_cpu(sbi->s_es->s_desc_size) -
1506 offset);
1507 }
1508
1509 return cpu_to_le16(crc);
1510 }
1511
1512 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1513 struct ext4_group_desc *gdp)
1514 {
1515 if ((sbi->s_es->s_feature_ro_compat &
1516 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1517 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1518 return 0;
1519
1520 return 1;
1521 }
1522
1523 /* Called at mount-time, super-block is locked */
1524 static int ext4_check_descriptors(struct super_block *sb)
1525 {
1526 struct ext4_sb_info *sbi = EXT4_SB(sb);
1527 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1528 ext4_fsblk_t last_block;
1529 ext4_fsblk_t block_bitmap;
1530 ext4_fsblk_t inode_bitmap;
1531 ext4_fsblk_t inode_table;
1532 int flexbg_flag = 0;
1533 ext4_group_t i;
1534
1535 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1536 flexbg_flag = 1;
1537
1538 ext4_debug("Checking group descriptors");
1539
1540 for (i = 0; i < sbi->s_groups_count; i++) {
1541 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1542
1543 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1544 last_block = ext4_blocks_count(sbi->s_es) - 1;
1545 else
1546 last_block = first_block +
1547 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1548
1549 block_bitmap = ext4_block_bitmap(sb, gdp);
1550 if (block_bitmap < first_block || block_bitmap > last_block) {
1551 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1552 "Block bitmap for group %lu not in group "
1553 "(block %llu)!\n", i, block_bitmap);
1554 return 0;
1555 }
1556 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1557 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1558 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1559 "Inode bitmap for group %lu not in group "
1560 "(block %llu)!\n", i, inode_bitmap);
1561 return 0;
1562 }
1563 inode_table = ext4_inode_table(sb, gdp);
1564 if (inode_table < first_block ||
1565 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1566 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1567 "Inode table for group %lu not in group "
1568 "(block %llu)!\n", i, inode_table);
1569 return 0;
1570 }
1571 spin_lock(sb_bgl_lock(sbi, i));
1572 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1573 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1574 "Checksum for group %lu failed (%u!=%u)\n",
1575 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1576 gdp)), le16_to_cpu(gdp->bg_checksum));
1577 if (!(sb->s_flags & MS_RDONLY)) {
1578 spin_unlock(sb_bgl_lock(sbi, i));
1579 return 0;
1580 }
1581 }
1582 spin_unlock(sb_bgl_lock(sbi, i));
1583 if (!flexbg_flag)
1584 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1585 }
1586
1587 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1588 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1589 return 1;
1590 }
1591
1592 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1593 * the superblock) which were deleted from all directories, but held open by
1594 * a process at the time of a crash. We walk the list and try to delete these
1595 * inodes at recovery time (only with a read-write filesystem).
1596 *
1597 * In order to keep the orphan inode chain consistent during traversal (in
1598 * case of crash during recovery), we link each inode into the superblock
1599 * orphan list_head and handle it the same way as an inode deletion during
1600 * normal operation (which journals the operations for us).
1601 *
1602 * We only do an iget() and an iput() on each inode, which is very safe if we
1603 * accidentally point at an in-use or already deleted inode. The worst that
1604 * can happen in this case is that we get a "bit already cleared" message from
1605 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1606 * e2fsck was run on this filesystem, and it must have already done the orphan
1607 * inode cleanup for us, so we can safely abort without any further action.
1608 */
1609 static void ext4_orphan_cleanup(struct super_block *sb,
1610 struct ext4_super_block *es)
1611 {
1612 unsigned int s_flags = sb->s_flags;
1613 int nr_orphans = 0, nr_truncates = 0;
1614 #ifdef CONFIG_QUOTA
1615 int i;
1616 #endif
1617 if (!es->s_last_orphan) {
1618 jbd_debug(4, "no orphan inodes to clean up\n");
1619 return;
1620 }
1621
1622 if (bdev_read_only(sb->s_bdev)) {
1623 printk(KERN_ERR "EXT4-fs: write access "
1624 "unavailable, skipping orphan cleanup.\n");
1625 return;
1626 }
1627
1628 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1629 if (es->s_last_orphan)
1630 jbd_debug(1, "Errors on filesystem, "
1631 "clearing orphan list.\n");
1632 es->s_last_orphan = 0;
1633 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1634 return;
1635 }
1636
1637 if (s_flags & MS_RDONLY) {
1638 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1639 sb->s_id);
1640 sb->s_flags &= ~MS_RDONLY;
1641 }
1642 #ifdef CONFIG_QUOTA
1643 /* Needed for iput() to work correctly and not trash data */
1644 sb->s_flags |= MS_ACTIVE;
1645 /* Turn on quotas so that they are updated correctly */
1646 for (i = 0; i < MAXQUOTAS; i++) {
1647 if (EXT4_SB(sb)->s_qf_names[i]) {
1648 int ret = ext4_quota_on_mount(sb, i);
1649 if (ret < 0)
1650 printk(KERN_ERR
1651 "EXT4-fs: Cannot turn on journaled "
1652 "quota: error %d\n", ret);
1653 }
1654 }
1655 #endif
1656
1657 while (es->s_last_orphan) {
1658 struct inode *inode;
1659
1660 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1661 if (IS_ERR(inode)) {
1662 es->s_last_orphan = 0;
1663 break;
1664 }
1665
1666 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1667 DQUOT_INIT(inode);
1668 if (inode->i_nlink) {
1669 printk(KERN_DEBUG
1670 "%s: truncating inode %lu to %lld bytes\n",
1671 __func__, inode->i_ino, inode->i_size);
1672 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1673 inode->i_ino, inode->i_size);
1674 ext4_truncate(inode);
1675 nr_truncates++;
1676 } else {
1677 printk(KERN_DEBUG
1678 "%s: deleting unreferenced inode %lu\n",
1679 __func__, inode->i_ino);
1680 jbd_debug(2, "deleting unreferenced inode %lu\n",
1681 inode->i_ino);
1682 nr_orphans++;
1683 }
1684 iput(inode); /* The delete magic happens here! */
1685 }
1686
1687 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1688
1689 if (nr_orphans)
1690 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1691 sb->s_id, PLURAL(nr_orphans));
1692 if (nr_truncates)
1693 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1694 sb->s_id, PLURAL(nr_truncates));
1695 #ifdef CONFIG_QUOTA
1696 /* Turn quotas off */
1697 for (i = 0; i < MAXQUOTAS; i++) {
1698 if (sb_dqopt(sb)->files[i])
1699 vfs_quota_off(sb, i, 0);
1700 }
1701 #endif
1702 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1703 }
1704 /*
1705 * Maximal extent format file size.
1706 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1707 * extent format containers, within a sector_t, and within i_blocks
1708 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1709 * so that won't be a limiting factor.
1710 *
1711 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1712 */
1713 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1714 {
1715 loff_t res;
1716 loff_t upper_limit = MAX_LFS_FILESIZE;
1717
1718 /* small i_blocks in vfs inode? */
1719 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1720 /*
1721 * CONFIG_LBD is not enabled implies the inode
1722 * i_block represent total blocks in 512 bytes
1723 * 32 == size of vfs inode i_blocks * 8
1724 */
1725 upper_limit = (1LL << 32) - 1;
1726
1727 /* total blocks in file system block size */
1728 upper_limit >>= (blkbits - 9);
1729 upper_limit <<= blkbits;
1730 }
1731
1732 /* 32-bit extent-start container, ee_block */
1733 res = 1LL << 32;
1734 res <<= blkbits;
1735 res -= 1;
1736
1737 /* Sanity check against vm- & vfs- imposed limits */
1738 if (res > upper_limit)
1739 res = upper_limit;
1740
1741 return res;
1742 }
1743
1744 /*
1745 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1746 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1747 * We need to be 1 filesystem block less than the 2^48 sector limit.
1748 */
1749 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1750 {
1751 loff_t res = EXT4_NDIR_BLOCKS;
1752 int meta_blocks;
1753 loff_t upper_limit;
1754 /* This is calculated to be the largest file size for a
1755 * dense, bitmapped file such that the total number of
1756 * sectors in the file, including data and all indirect blocks,
1757 * does not exceed 2^48 -1
1758 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1759 * total number of 512 bytes blocks of the file
1760 */
1761
1762 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1763 /*
1764 * !has_huge_files or CONFIG_LBD is not enabled
1765 * implies the inode i_block represent total blocks in
1766 * 512 bytes 32 == size of vfs inode i_blocks * 8
1767 */
1768 upper_limit = (1LL << 32) - 1;
1769
1770 /* total blocks in file system block size */
1771 upper_limit >>= (bits - 9);
1772
1773 } else {
1774 /*
1775 * We use 48 bit ext4_inode i_blocks
1776 * With EXT4_HUGE_FILE_FL set the i_blocks
1777 * represent total number of blocks in
1778 * file system block size
1779 */
1780 upper_limit = (1LL << 48) - 1;
1781
1782 }
1783
1784 /* indirect blocks */
1785 meta_blocks = 1;
1786 /* double indirect blocks */
1787 meta_blocks += 1 + (1LL << (bits-2));
1788 /* tripple indirect blocks */
1789 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1790
1791 upper_limit -= meta_blocks;
1792 upper_limit <<= bits;
1793
1794 res += 1LL << (bits-2);
1795 res += 1LL << (2*(bits-2));
1796 res += 1LL << (3*(bits-2));
1797 res <<= bits;
1798 if (res > upper_limit)
1799 res = upper_limit;
1800
1801 if (res > MAX_LFS_FILESIZE)
1802 res = MAX_LFS_FILESIZE;
1803
1804 return res;
1805 }
1806
1807 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1808 ext4_fsblk_t logical_sb_block, int nr)
1809 {
1810 struct ext4_sb_info *sbi = EXT4_SB(sb);
1811 ext4_group_t bg, first_meta_bg;
1812 int has_super = 0;
1813
1814 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1815
1816 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1817 nr < first_meta_bg)
1818 return logical_sb_block + nr + 1;
1819 bg = sbi->s_desc_per_block * nr;
1820 if (ext4_bg_has_super(sb, bg))
1821 has_super = 1;
1822 return (has_super + ext4_group_first_block_no(sb, bg));
1823 }
1824
1825 /**
1826 * ext4_get_stripe_size: Get the stripe size.
1827 * @sbi: In memory super block info
1828 *
1829 * If we have specified it via mount option, then
1830 * use the mount option value. If the value specified at mount time is
1831 * greater than the blocks per group use the super block value.
1832 * If the super block value is greater than blocks per group return 0.
1833 * Allocator needs it be less than blocks per group.
1834 *
1835 */
1836 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1837 {
1838 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1839 unsigned long stripe_width =
1840 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1841
1842 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1843 return sbi->s_stripe;
1844
1845 if (stripe_width <= sbi->s_blocks_per_group)
1846 return stripe_width;
1847
1848 if (stride <= sbi->s_blocks_per_group)
1849 return stride;
1850
1851 return 0;
1852 }
1853
1854 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1855 __releases(kernel_lock)
1856 __acquires(kernel_lock)
1857
1858 {
1859 struct buffer_head *bh;
1860 struct ext4_super_block *es = NULL;
1861 struct ext4_sb_info *sbi;
1862 ext4_fsblk_t block;
1863 ext4_fsblk_t sb_block = get_sb_block(&data);
1864 ext4_fsblk_t logical_sb_block;
1865 unsigned long offset = 0;
1866 unsigned int journal_inum = 0;
1867 unsigned long journal_devnum = 0;
1868 unsigned long def_mount_opts;
1869 struct inode *root;
1870 char *cp;
1871 int ret = -EINVAL;
1872 int blocksize;
1873 int db_count;
1874 int i;
1875 int needs_recovery, has_huge_files;
1876 __le32 features;
1877 __u64 blocks_count;
1878 int err;
1879
1880 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1881 if (!sbi)
1882 return -ENOMEM;
1883 sb->s_fs_info = sbi;
1884 sbi->s_mount_opt = 0;
1885 sbi->s_resuid = EXT4_DEF_RESUID;
1886 sbi->s_resgid = EXT4_DEF_RESGID;
1887 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1888 sbi->s_sb_block = sb_block;
1889
1890 unlock_kernel();
1891
1892 /* Cleanup superblock name */
1893 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1894 *cp = '!';
1895
1896 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1897 if (!blocksize) {
1898 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1899 goto out_fail;
1900 }
1901
1902 /*
1903 * The ext4 superblock will not be buffer aligned for other than 1kB
1904 * block sizes. We need to calculate the offset from buffer start.
1905 */
1906 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1907 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1908 offset = do_div(logical_sb_block, blocksize);
1909 } else {
1910 logical_sb_block = sb_block;
1911 }
1912
1913 if (!(bh = sb_bread(sb, logical_sb_block))) {
1914 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1915 goto out_fail;
1916 }
1917 /*
1918 * Note: s_es must be initialized as soon as possible because
1919 * some ext4 macro-instructions depend on its value
1920 */
1921 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1922 sbi->s_es = es;
1923 sb->s_magic = le16_to_cpu(es->s_magic);
1924 if (sb->s_magic != EXT4_SUPER_MAGIC)
1925 goto cantfind_ext4;
1926
1927 /* Set defaults before we parse the mount options */
1928 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1929 if (def_mount_opts & EXT4_DEFM_DEBUG)
1930 set_opt(sbi->s_mount_opt, DEBUG);
1931 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1932 set_opt(sbi->s_mount_opt, GRPID);
1933 if (def_mount_opts & EXT4_DEFM_UID16)
1934 set_opt(sbi->s_mount_opt, NO_UID32);
1935 #ifdef CONFIG_EXT4_FS_XATTR
1936 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1937 set_opt(sbi->s_mount_opt, XATTR_USER);
1938 #endif
1939 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1940 if (def_mount_opts & EXT4_DEFM_ACL)
1941 set_opt(sbi->s_mount_opt, POSIX_ACL);
1942 #endif
1943 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1944 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1945 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1946 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1947 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1948 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1949
1950 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1951 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1952 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1953 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1954 else
1955 set_opt(sbi->s_mount_opt, ERRORS_RO);
1956
1957 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1958 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1959
1960 set_opt(sbi->s_mount_opt, RESERVATION);
1961 set_opt(sbi->s_mount_opt, BARRIER);
1962
1963 /*
1964 * turn on extents feature by default in ext4 filesystem
1965 * only if feature flag already set by mkfs or tune2fs.
1966 * Use -o noextents to turn it off
1967 */
1968 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1969 set_opt(sbi->s_mount_opt, EXTENTS);
1970 else
1971 ext4_warning(sb, __func__,
1972 "extents feature not enabled on this filesystem, "
1973 "use tune2fs.\n");
1974
1975 /*
1976 * enable delayed allocation by default
1977 * Use -o nodelalloc to turn it off
1978 */
1979 set_opt(sbi->s_mount_opt, DELALLOC);
1980
1981
1982 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
1983 NULL, 0))
1984 goto failed_mount;
1985
1986 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1987 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1988
1989 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1990 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1991 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1992 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1993 printk(KERN_WARNING
1994 "EXT4-fs warning: feature flags set on rev 0 fs, "
1995 "running e2fsck is recommended\n");
1996
1997 /*
1998 * Check feature flags regardless of the revision level, since we
1999 * previously didn't change the revision level when setting the flags,
2000 * so there is a chance incompat flags are set on a rev 0 filesystem.
2001 */
2002 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2003 if (features) {
2004 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2005 "unsupported optional features (%x).\n",
2006 sb->s_id, le32_to_cpu(features));
2007 goto failed_mount;
2008 }
2009 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2010 if (!(sb->s_flags & MS_RDONLY) && features) {
2011 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2012 "unsupported optional features (%x).\n",
2013 sb->s_id, le32_to_cpu(features));
2014 goto failed_mount;
2015 }
2016 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2017 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2018 if (has_huge_files) {
2019 /*
2020 * Large file size enabled file system can only be
2021 * mount if kernel is build with CONFIG_LBD
2022 */
2023 if (sizeof(root->i_blocks) < sizeof(u64) &&
2024 !(sb->s_flags & MS_RDONLY)) {
2025 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2026 "files cannot be mounted read-write "
2027 "without CONFIG_LBD.\n", sb->s_id);
2028 goto failed_mount;
2029 }
2030 }
2031 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2032
2033 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2034 blocksize > EXT4_MAX_BLOCK_SIZE) {
2035 printk(KERN_ERR
2036 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2037 blocksize, sb->s_id);
2038 goto failed_mount;
2039 }
2040
2041 if (sb->s_blocksize != blocksize) {
2042
2043 /* Validate the filesystem blocksize */
2044 if (!sb_set_blocksize(sb, blocksize)) {
2045 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2046 blocksize);
2047 goto failed_mount;
2048 }
2049
2050 brelse(bh);
2051 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2052 offset = do_div(logical_sb_block, blocksize);
2053 bh = sb_bread(sb, logical_sb_block);
2054 if (!bh) {
2055 printk(KERN_ERR
2056 "EXT4-fs: Can't read superblock on 2nd try.\n");
2057 goto failed_mount;
2058 }
2059 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2060 sbi->s_es = es;
2061 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2062 printk(KERN_ERR
2063 "EXT4-fs: Magic mismatch, very weird !\n");
2064 goto failed_mount;
2065 }
2066 }
2067
2068 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2069 has_huge_files);
2070 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2071
2072 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2073 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2074 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2075 } else {
2076 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2077 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2078 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2079 (!is_power_of_2(sbi->s_inode_size)) ||
2080 (sbi->s_inode_size > blocksize)) {
2081 printk(KERN_ERR
2082 "EXT4-fs: unsupported inode size: %d\n",
2083 sbi->s_inode_size);
2084 goto failed_mount;
2085 }
2086 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2087 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2088 }
2089 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2090 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2091 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2092 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2093 !is_power_of_2(sbi->s_desc_size)) {
2094 printk(KERN_ERR
2095 "EXT4-fs: unsupported descriptor size %lu\n",
2096 sbi->s_desc_size);
2097 goto failed_mount;
2098 }
2099 } else
2100 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2101 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2102 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2103 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2104 goto cantfind_ext4;
2105 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2106 if (sbi->s_inodes_per_block == 0)
2107 goto cantfind_ext4;
2108 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2109 sbi->s_inodes_per_block;
2110 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2111 sbi->s_sbh = bh;
2112 sbi->s_mount_state = le16_to_cpu(es->s_state);
2113 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2114 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2115 for (i = 0; i < 4; i++)
2116 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2117 sbi->s_def_hash_version = es->s_def_hash_version;
2118
2119 if (sbi->s_blocks_per_group > blocksize * 8) {
2120 printk(KERN_ERR
2121 "EXT4-fs: #blocks per group too big: %lu\n",
2122 sbi->s_blocks_per_group);
2123 goto failed_mount;
2124 }
2125 if (sbi->s_inodes_per_group > blocksize * 8) {
2126 printk(KERN_ERR
2127 "EXT4-fs: #inodes per group too big: %lu\n",
2128 sbi->s_inodes_per_group);
2129 goto failed_mount;
2130 }
2131
2132 if (ext4_blocks_count(es) >
2133 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2134 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2135 " too large to mount safely\n", sb->s_id);
2136 if (sizeof(sector_t) < 8)
2137 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2138 "enabled\n");
2139 goto failed_mount;
2140 }
2141
2142 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2143 goto cantfind_ext4;
2144
2145 /* ensure blocks_count calculation below doesn't sign-extend */
2146 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2147 le32_to_cpu(es->s_first_data_block) + 1) {
2148 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2149 "first data block %u, blocks per group %lu\n",
2150 ext4_blocks_count(es),
2151 le32_to_cpu(es->s_first_data_block),
2152 EXT4_BLOCKS_PER_GROUP(sb));
2153 goto failed_mount;
2154 }
2155 blocks_count = (ext4_blocks_count(es) -
2156 le32_to_cpu(es->s_first_data_block) +
2157 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2158 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2159 sbi->s_groups_count = blocks_count;
2160 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2161 EXT4_DESC_PER_BLOCK(sb);
2162 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2163 GFP_KERNEL);
2164 if (sbi->s_group_desc == NULL) {
2165 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2166 goto failed_mount;
2167 }
2168
2169 #ifdef CONFIG_PROC_FS
2170 if (ext4_proc_root)
2171 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2172
2173 if (sbi->s_proc)
2174 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2175 &ext4_ui_proc_fops,
2176 &sbi->s_inode_readahead_blks);
2177 #endif
2178
2179 bgl_lock_init(&sbi->s_blockgroup_lock);
2180
2181 for (i = 0; i < db_count; i++) {
2182 block = descriptor_loc(sb, logical_sb_block, i);
2183 sbi->s_group_desc[i] = sb_bread(sb, block);
2184 if (!sbi->s_group_desc[i]) {
2185 printk(KERN_ERR "EXT4-fs: "
2186 "can't read group descriptor %d\n", i);
2187 db_count = i;
2188 goto failed_mount2;
2189 }
2190 }
2191 if (!ext4_check_descriptors(sb)) {
2192 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2193 goto failed_mount2;
2194 }
2195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2196 if (!ext4_fill_flex_info(sb)) {
2197 printk(KERN_ERR
2198 "EXT4-fs: unable to initialize "
2199 "flex_bg meta info!\n");
2200 goto failed_mount2;
2201 }
2202
2203 sbi->s_gdb_count = db_count;
2204 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2205 spin_lock_init(&sbi->s_next_gen_lock);
2206
2207 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2208 ext4_count_free_blocks(sb));
2209 if (!err) {
2210 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2211 ext4_count_free_inodes(sb));
2212 }
2213 if (!err) {
2214 err = percpu_counter_init(&sbi->s_dirs_counter,
2215 ext4_count_dirs(sb));
2216 }
2217 if (!err) {
2218 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2219 }
2220 if (err) {
2221 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2222 goto failed_mount3;
2223 }
2224
2225 sbi->s_stripe = ext4_get_stripe_size(sbi);
2226
2227 /*
2228 * set up enough so that it can read an inode
2229 */
2230 sb->s_op = &ext4_sops;
2231 sb->s_export_op = &ext4_export_ops;
2232 sb->s_xattr = ext4_xattr_handlers;
2233 #ifdef CONFIG_QUOTA
2234 sb->s_qcop = &ext4_qctl_operations;
2235 sb->dq_op = &ext4_quota_operations;
2236 #endif
2237 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2238
2239 sb->s_root = NULL;
2240
2241 needs_recovery = (es->s_last_orphan != 0 ||
2242 EXT4_HAS_INCOMPAT_FEATURE(sb,
2243 EXT4_FEATURE_INCOMPAT_RECOVER));
2244
2245 /*
2246 * The first inode we look at is the journal inode. Don't try
2247 * root first: it may be modified in the journal!
2248 */
2249 if (!test_opt(sb, NOLOAD) &&
2250 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2251 if (ext4_load_journal(sb, es, journal_devnum))
2252 goto failed_mount3;
2253 if (!(sb->s_flags & MS_RDONLY) &&
2254 EXT4_SB(sb)->s_journal->j_failed_commit) {
2255 printk(KERN_CRIT "EXT4-fs error (device %s): "
2256 "ext4_fill_super: Journal transaction "
2257 "%u is corrupt\n", sb->s_id,
2258 EXT4_SB(sb)->s_journal->j_failed_commit);
2259 if (test_opt(sb, ERRORS_RO)) {
2260 printk(KERN_CRIT
2261 "Mounting filesystem read-only\n");
2262 sb->s_flags |= MS_RDONLY;
2263 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2264 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2265 }
2266 if (test_opt(sb, ERRORS_PANIC)) {
2267 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2268 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2269 ext4_commit_super(sb, es, 1);
2270 printk(KERN_CRIT
2271 "EXT4-fs (device %s): mount failed\n",
2272 sb->s_id);
2273 goto failed_mount4;
2274 }
2275 }
2276 } else if (journal_inum) {
2277 if (ext4_create_journal(sb, es, journal_inum))
2278 goto failed_mount3;
2279 } else {
2280 if (!silent)
2281 printk(KERN_ERR
2282 "ext4: No journal on filesystem on %s\n",
2283 sb->s_id);
2284 goto failed_mount3;
2285 }
2286
2287 if (ext4_blocks_count(es) > 0xffffffffULL &&
2288 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2289 JBD2_FEATURE_INCOMPAT_64BIT)) {
2290 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2291 goto failed_mount4;
2292 }
2293
2294 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2295 jbd2_journal_set_features(sbi->s_journal,
2296 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2297 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2298 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2299 jbd2_journal_set_features(sbi->s_journal,
2300 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2301 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2303 } else {
2304 jbd2_journal_clear_features(sbi->s_journal,
2305 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2306 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2307 }
2308
2309 /* We have now updated the journal if required, so we can
2310 * validate the data journaling mode. */
2311 switch (test_opt(sb, DATA_FLAGS)) {
2312 case 0:
2313 /* No mode set, assume a default based on the journal
2314 * capabilities: ORDERED_DATA if the journal can
2315 * cope, else JOURNAL_DATA
2316 */
2317 if (jbd2_journal_check_available_features
2318 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2319 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2320 else
2321 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2322 break;
2323
2324 case EXT4_MOUNT_ORDERED_DATA:
2325 case EXT4_MOUNT_WRITEBACK_DATA:
2326 if (!jbd2_journal_check_available_features
2327 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2328 printk(KERN_ERR "EXT4-fs: Journal does not support "
2329 "requested data journaling mode\n");
2330 goto failed_mount4;
2331 }
2332 default:
2333 break;
2334 }
2335
2336 if (test_opt(sb, NOBH)) {
2337 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2338 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2339 "its supported only with writeback mode\n");
2340 clear_opt(sbi->s_mount_opt, NOBH);
2341 }
2342 }
2343 /*
2344 * The jbd2_journal_load will have done any necessary log recovery,
2345 * so we can safely mount the rest of the filesystem now.
2346 */
2347
2348 root = ext4_iget(sb, EXT4_ROOT_INO);
2349 if (IS_ERR(root)) {
2350 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2351 ret = PTR_ERR(root);
2352 goto failed_mount4;
2353 }
2354 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2355 iput(root);
2356 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2357 goto failed_mount4;
2358 }
2359 sb->s_root = d_alloc_root(root);
2360 if (!sb->s_root) {
2361 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2362 iput(root);
2363 ret = -ENOMEM;
2364 goto failed_mount4;
2365 }
2366
2367 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2368
2369 /* determine the minimum size of new large inodes, if present */
2370 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2371 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2372 EXT4_GOOD_OLD_INODE_SIZE;
2373 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2374 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2375 if (sbi->s_want_extra_isize <
2376 le16_to_cpu(es->s_want_extra_isize))
2377 sbi->s_want_extra_isize =
2378 le16_to_cpu(es->s_want_extra_isize);
2379 if (sbi->s_want_extra_isize <
2380 le16_to_cpu(es->s_min_extra_isize))
2381 sbi->s_want_extra_isize =
2382 le16_to_cpu(es->s_min_extra_isize);
2383 }
2384 }
2385 /* Check if enough inode space is available */
2386 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2387 sbi->s_inode_size) {
2388 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2389 EXT4_GOOD_OLD_INODE_SIZE;
2390 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2391 "available.\n");
2392 }
2393
2394 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2395 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2396 "requested data journaling mode\n");
2397 clear_opt(sbi->s_mount_opt, DELALLOC);
2398 } else if (test_opt(sb, DELALLOC))
2399 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2400
2401 ext4_ext_init(sb);
2402 err = ext4_mb_init(sb, needs_recovery);
2403 if (err) {
2404 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2405 err);
2406 goto failed_mount4;
2407 }
2408
2409 /*
2410 * akpm: core read_super() calls in here with the superblock locked.
2411 * That deadlocks, because orphan cleanup needs to lock the superblock
2412 * in numerous places. Here we just pop the lock - it's relatively
2413 * harmless, because we are now ready to accept write_super() requests,
2414 * and aviro says that's the only reason for hanging onto the
2415 * superblock lock.
2416 */
2417 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2418 ext4_orphan_cleanup(sb, es);
2419 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2420 if (needs_recovery)
2421 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2422 ext4_mark_recovery_complete(sb, es);
2423 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2424 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2425 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2426 "writeback");
2427
2428 lock_kernel();
2429 return 0;
2430
2431 cantfind_ext4:
2432 if (!silent)
2433 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2434 sb->s_id);
2435 goto failed_mount;
2436
2437 failed_mount4:
2438 jbd2_journal_destroy(sbi->s_journal);
2439 sbi->s_journal = NULL;
2440 failed_mount3:
2441 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2442 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2443 percpu_counter_destroy(&sbi->s_dirs_counter);
2444 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2445 failed_mount2:
2446 for (i = 0; i < db_count; i++)
2447 brelse(sbi->s_group_desc[i]);
2448 kfree(sbi->s_group_desc);
2449 failed_mount:
2450 if (sbi->s_proc) {
2451 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2452 remove_proc_entry(sb->s_id, ext4_proc_root);
2453 }
2454 #ifdef CONFIG_QUOTA
2455 for (i = 0; i < MAXQUOTAS; i++)
2456 kfree(sbi->s_qf_names[i]);
2457 #endif
2458 ext4_blkdev_remove(sbi);
2459 brelse(bh);
2460 out_fail:
2461 sb->s_fs_info = NULL;
2462 kfree(sbi);
2463 lock_kernel();
2464 return ret;
2465 }
2466
2467 /*
2468 * Setup any per-fs journal parameters now. We'll do this both on
2469 * initial mount, once the journal has been initialised but before we've
2470 * done any recovery; and again on any subsequent remount.
2471 */
2472 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2473 {
2474 struct ext4_sb_info *sbi = EXT4_SB(sb);
2475
2476 if (sbi->s_commit_interval)
2477 journal->j_commit_interval = sbi->s_commit_interval;
2478 /* We could also set up an ext4-specific default for the commit
2479 * interval here, but for now we'll just fall back to the jbd
2480 * default. */
2481
2482 spin_lock(&journal->j_state_lock);
2483 if (test_opt(sb, BARRIER))
2484 journal->j_flags |= JBD2_BARRIER;
2485 else
2486 journal->j_flags &= ~JBD2_BARRIER;
2487 if (test_opt(sb, DATA_ERR_ABORT))
2488 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2489 else
2490 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2491 spin_unlock(&journal->j_state_lock);
2492 }
2493
2494 static journal_t *ext4_get_journal(struct super_block *sb,
2495 unsigned int journal_inum)
2496 {
2497 struct inode *journal_inode;
2498 journal_t *journal;
2499
2500 /* First, test for the existence of a valid inode on disk. Bad
2501 * things happen if we iget() an unused inode, as the subsequent
2502 * iput() will try to delete it. */
2503
2504 journal_inode = ext4_iget(sb, journal_inum);
2505 if (IS_ERR(journal_inode)) {
2506 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2507 return NULL;
2508 }
2509 if (!journal_inode->i_nlink) {
2510 make_bad_inode(journal_inode);
2511 iput(journal_inode);
2512 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2513 return NULL;
2514 }
2515
2516 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2517 journal_inode, journal_inode->i_size);
2518 if (!S_ISREG(journal_inode->i_mode)) {
2519 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2520 iput(journal_inode);
2521 return NULL;
2522 }
2523
2524 journal = jbd2_journal_init_inode(journal_inode);
2525 if (!journal) {
2526 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2527 iput(journal_inode);
2528 return NULL;
2529 }
2530 journal->j_private = sb;
2531 ext4_init_journal_params(sb, journal);
2532 return journal;
2533 }
2534
2535 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2536 dev_t j_dev)
2537 {
2538 struct buffer_head *bh;
2539 journal_t *journal;
2540 ext4_fsblk_t start;
2541 ext4_fsblk_t len;
2542 int hblock, blocksize;
2543 ext4_fsblk_t sb_block;
2544 unsigned long offset;
2545 struct ext4_super_block *es;
2546 struct block_device *bdev;
2547
2548 bdev = ext4_blkdev_get(j_dev);
2549 if (bdev == NULL)
2550 return NULL;
2551
2552 if (bd_claim(bdev, sb)) {
2553 printk(KERN_ERR
2554 "EXT4: failed to claim external journal device.\n");
2555 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2556 return NULL;
2557 }
2558
2559 blocksize = sb->s_blocksize;
2560 hblock = bdev_hardsect_size(bdev);
2561 if (blocksize < hblock) {
2562 printk(KERN_ERR
2563 "EXT4-fs: blocksize too small for journal device.\n");
2564 goto out_bdev;
2565 }
2566
2567 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2568 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2569 set_blocksize(bdev, blocksize);
2570 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2571 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2572 "external journal\n");
2573 goto out_bdev;
2574 }
2575
2576 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2577 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2578 !(le32_to_cpu(es->s_feature_incompat) &
2579 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2580 printk(KERN_ERR "EXT4-fs: external journal has "
2581 "bad superblock\n");
2582 brelse(bh);
2583 goto out_bdev;
2584 }
2585
2586 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2587 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2588 brelse(bh);
2589 goto out_bdev;
2590 }
2591
2592 len = ext4_blocks_count(es);
2593 start = sb_block + 1;
2594 brelse(bh); /* we're done with the superblock */
2595
2596 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2597 start, len, blocksize);
2598 if (!journal) {
2599 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2600 goto out_bdev;
2601 }
2602 journal->j_private = sb;
2603 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2604 wait_on_buffer(journal->j_sb_buffer);
2605 if (!buffer_uptodate(journal->j_sb_buffer)) {
2606 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2607 goto out_journal;
2608 }
2609 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2610 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2611 "user (unsupported) - %d\n",
2612 be32_to_cpu(journal->j_superblock->s_nr_users));
2613 goto out_journal;
2614 }
2615 EXT4_SB(sb)->journal_bdev = bdev;
2616 ext4_init_journal_params(sb, journal);
2617 return journal;
2618 out_journal:
2619 jbd2_journal_destroy(journal);
2620 out_bdev:
2621 ext4_blkdev_put(bdev);
2622 return NULL;
2623 }
2624
2625 static int ext4_load_journal(struct super_block *sb,
2626 struct ext4_super_block *es,
2627 unsigned long journal_devnum)
2628 {
2629 journal_t *journal;
2630 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2631 dev_t journal_dev;
2632 int err = 0;
2633 int really_read_only;
2634
2635 if (journal_devnum &&
2636 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2637 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2638 "numbers have changed\n");
2639 journal_dev = new_decode_dev(journal_devnum);
2640 } else
2641 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2642
2643 really_read_only = bdev_read_only(sb->s_bdev);
2644
2645 /*
2646 * Are we loading a blank journal or performing recovery after a
2647 * crash? For recovery, we need to check in advance whether we
2648 * can get read-write access to the device.
2649 */
2650
2651 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2652 if (sb->s_flags & MS_RDONLY) {
2653 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2654 "required on readonly filesystem.\n");
2655 if (really_read_only) {
2656 printk(KERN_ERR "EXT4-fs: write access "
2657 "unavailable, cannot proceed.\n");
2658 return -EROFS;
2659 }
2660 printk(KERN_INFO "EXT4-fs: write access will "
2661 "be enabled during recovery.\n");
2662 }
2663 }
2664
2665 if (journal_inum && journal_dev) {
2666 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2667 "and inode journals!\n");
2668 return -EINVAL;
2669 }
2670
2671 if (journal_inum) {
2672 if (!(journal = ext4_get_journal(sb, journal_inum)))
2673 return -EINVAL;
2674 } else {
2675 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2676 return -EINVAL;
2677 }
2678
2679 if (journal->j_flags & JBD2_BARRIER)
2680 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2681 else
2682 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2683
2684 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2685 err = jbd2_journal_update_format(journal);
2686 if (err) {
2687 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2688 jbd2_journal_destroy(journal);
2689 return err;
2690 }
2691 }
2692
2693 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2694 err = jbd2_journal_wipe(journal, !really_read_only);
2695 if (!err)
2696 err = jbd2_journal_load(journal);
2697
2698 if (err) {
2699 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2700 jbd2_journal_destroy(journal);
2701 return err;
2702 }
2703
2704 EXT4_SB(sb)->s_journal = journal;
2705 ext4_clear_journal_err(sb, es);
2706
2707 if (journal_devnum &&
2708 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2709 es->s_journal_dev = cpu_to_le32(journal_devnum);
2710 sb->s_dirt = 1;
2711
2712 /* Make sure we flush the recovery flag to disk. */
2713 ext4_commit_super(sb, es, 1);
2714 }
2715
2716 return 0;
2717 }
2718
2719 static int ext4_create_journal(struct super_block *sb,
2720 struct ext4_super_block *es,
2721 unsigned int journal_inum)
2722 {
2723 journal_t *journal;
2724 int err;
2725
2726 if (sb->s_flags & MS_RDONLY) {
2727 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2728 "create journal.\n");
2729 return -EROFS;
2730 }
2731
2732 journal = ext4_get_journal(sb, journal_inum);
2733 if (!journal)
2734 return -EINVAL;
2735
2736 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2737 journal_inum);
2738
2739 err = jbd2_journal_create(journal);
2740 if (err) {
2741 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2742 jbd2_journal_destroy(journal);
2743 return -EIO;
2744 }
2745
2746 EXT4_SB(sb)->s_journal = journal;
2747
2748 ext4_update_dynamic_rev(sb);
2749 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2750 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2751
2752 es->s_journal_inum = cpu_to_le32(journal_inum);
2753 sb->s_dirt = 1;
2754
2755 /* Make sure we flush the recovery flag to disk. */
2756 ext4_commit_super(sb, es, 1);
2757
2758 return 0;
2759 }
2760
2761 static void ext4_commit_super(struct super_block *sb,
2762 struct ext4_super_block *es, int sync)
2763 {
2764 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2765
2766 if (!sbh)
2767 return;
2768 if (buffer_write_io_error(sbh)) {
2769 /*
2770 * Oh, dear. A previous attempt to write the
2771 * superblock failed. This could happen because the
2772 * USB device was yanked out. Or it could happen to
2773 * be a transient write error and maybe the block will
2774 * be remapped. Nothing we can do but to retry the
2775 * write and hope for the best.
2776 */
2777 printk(KERN_ERR "ext4: previous I/O error to "
2778 "superblock detected for %s.\n", sb->s_id);
2779 clear_buffer_write_io_error(sbh);
2780 set_buffer_uptodate(sbh);
2781 }
2782 es->s_wtime = cpu_to_le32(get_seconds());
2783 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2784 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2785 BUFFER_TRACE(sbh, "marking dirty");
2786 mark_buffer_dirty(sbh);
2787 if (sync) {
2788 sync_dirty_buffer(sbh);
2789 if (buffer_write_io_error(sbh)) {
2790 printk(KERN_ERR "ext4: I/O error while writing "
2791 "superblock for %s.\n", sb->s_id);
2792 clear_buffer_write_io_error(sbh);
2793 set_buffer_uptodate(sbh);
2794 }
2795 }
2796 }
2797
2798
2799 /*
2800 * Have we just finished recovery? If so, and if we are mounting (or
2801 * remounting) the filesystem readonly, then we will end up with a
2802 * consistent fs on disk. Record that fact.
2803 */
2804 static void ext4_mark_recovery_complete(struct super_block *sb,
2805 struct ext4_super_block *es)
2806 {
2807 journal_t *journal = EXT4_SB(sb)->s_journal;
2808
2809 jbd2_journal_lock_updates(journal);
2810 if (jbd2_journal_flush(journal) < 0)
2811 goto out;
2812
2813 lock_super(sb);
2814 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2815 sb->s_flags & MS_RDONLY) {
2816 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2817 sb->s_dirt = 0;
2818 ext4_commit_super(sb, es, 1);
2819 }
2820 unlock_super(sb);
2821
2822 out:
2823 jbd2_journal_unlock_updates(journal);
2824 }
2825
2826 /*
2827 * If we are mounting (or read-write remounting) a filesystem whose journal
2828 * has recorded an error from a previous lifetime, move that error to the
2829 * main filesystem now.
2830 */
2831 static void ext4_clear_journal_err(struct super_block *sb,
2832 struct ext4_super_block *es)
2833 {
2834 journal_t *journal;
2835 int j_errno;
2836 const char *errstr;
2837
2838 journal = EXT4_SB(sb)->s_journal;
2839
2840 /*
2841 * Now check for any error status which may have been recorded in the
2842 * journal by a prior ext4_error() or ext4_abort()
2843 */
2844
2845 j_errno = jbd2_journal_errno(journal);
2846 if (j_errno) {
2847 char nbuf[16];
2848
2849 errstr = ext4_decode_error(sb, j_errno, nbuf);
2850 ext4_warning(sb, __func__, "Filesystem error recorded "
2851 "from previous mount: %s", errstr);
2852 ext4_warning(sb, __func__, "Marking fs in need of "
2853 "filesystem check.");
2854
2855 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2856 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2857 ext4_commit_super(sb, es, 1);
2858
2859 jbd2_journal_clear_err(journal);
2860 }
2861 }
2862
2863 /*
2864 * Force the running and committing transactions to commit,
2865 * and wait on the commit.
2866 */
2867 int ext4_force_commit(struct super_block *sb)
2868 {
2869 journal_t *journal;
2870 int ret;
2871
2872 if (sb->s_flags & MS_RDONLY)
2873 return 0;
2874
2875 journal = EXT4_SB(sb)->s_journal;
2876 sb->s_dirt = 0;
2877 ret = ext4_journal_force_commit(journal);
2878 return ret;
2879 }
2880
2881 /*
2882 * Ext4 always journals updates to the superblock itself, so we don't
2883 * have to propagate any other updates to the superblock on disk at this
2884 * point. (We can probably nuke this function altogether, and remove
2885 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2886 */
2887 static void ext4_write_super(struct super_block *sb)
2888 {
2889 if (mutex_trylock(&sb->s_lock) != 0)
2890 BUG();
2891 sb->s_dirt = 0;
2892 }
2893
2894 static int ext4_sync_fs(struct super_block *sb, int wait)
2895 {
2896 int ret = 0;
2897
2898 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2899 sb->s_dirt = 0;
2900 if (wait)
2901 ret = ext4_force_commit(sb);
2902 else
2903 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
2904 return ret;
2905 }
2906
2907 /*
2908 * LVM calls this function before a (read-only) snapshot is created. This
2909 * gives us a chance to flush the journal completely and mark the fs clean.
2910 */
2911 static void ext4_write_super_lockfs(struct super_block *sb)
2912 {
2913 sb->s_dirt = 0;
2914
2915 if (!(sb->s_flags & MS_RDONLY)) {
2916 journal_t *journal = EXT4_SB(sb)->s_journal;
2917
2918 /* Now we set up the journal barrier. */
2919 jbd2_journal_lock_updates(journal);
2920
2921 /*
2922 * We don't want to clear needs_recovery flag when we failed
2923 * to flush the journal.
2924 */
2925 if (jbd2_journal_flush(journal) < 0)
2926 return;
2927
2928 /* Journal blocked and flushed, clear needs_recovery flag. */
2929 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2930 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2931 }
2932 }
2933
2934 /*
2935 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2936 * flag here, even though the filesystem is not technically dirty yet.
2937 */
2938 static void ext4_unlockfs(struct super_block *sb)
2939 {
2940 if (!(sb->s_flags & MS_RDONLY)) {
2941 lock_super(sb);
2942 /* Reser the needs_recovery flag before the fs is unlocked. */
2943 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2944 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2945 unlock_super(sb);
2946 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2947 }
2948 }
2949
2950 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2951 {
2952 struct ext4_super_block *es;
2953 struct ext4_sb_info *sbi = EXT4_SB(sb);
2954 ext4_fsblk_t n_blocks_count = 0;
2955 unsigned long old_sb_flags;
2956 struct ext4_mount_options old_opts;
2957 ext4_group_t g;
2958 int err;
2959 #ifdef CONFIG_QUOTA
2960 int i;
2961 #endif
2962
2963 /* Store the original options */
2964 old_sb_flags = sb->s_flags;
2965 old_opts.s_mount_opt = sbi->s_mount_opt;
2966 old_opts.s_resuid = sbi->s_resuid;
2967 old_opts.s_resgid = sbi->s_resgid;
2968 old_opts.s_commit_interval = sbi->s_commit_interval;
2969 #ifdef CONFIG_QUOTA
2970 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2971 for (i = 0; i < MAXQUOTAS; i++)
2972 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2973 #endif
2974
2975 /*
2976 * Allow the "check" option to be passed as a remount option.
2977 */
2978 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2979 err = -EINVAL;
2980 goto restore_opts;
2981 }
2982
2983 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2984 ext4_abort(sb, __func__, "Abort forced by user");
2985
2986 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2987 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2988
2989 es = sbi->s_es;
2990
2991 ext4_init_journal_params(sb, sbi->s_journal);
2992
2993 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2994 n_blocks_count > ext4_blocks_count(es)) {
2995 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2996 err = -EROFS;
2997 goto restore_opts;
2998 }
2999
3000 if (*flags & MS_RDONLY) {
3001 /*
3002 * First of all, the unconditional stuff we have to do
3003 * to disable replay of the journal when we next remount
3004 */
3005 sb->s_flags |= MS_RDONLY;
3006
3007 /*
3008 * OK, test if we are remounting a valid rw partition
3009 * readonly, and if so set the rdonly flag and then
3010 * mark the partition as valid again.
3011 */
3012 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3013 (sbi->s_mount_state & EXT4_VALID_FS))
3014 es->s_state = cpu_to_le16(sbi->s_mount_state);
3015
3016 /*
3017 * We have to unlock super so that we can wait for
3018 * transactions.
3019 */
3020 unlock_super(sb);
3021 ext4_mark_recovery_complete(sb, es);
3022 lock_super(sb);
3023 } else {
3024 __le32 ret;
3025 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3026 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3027 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3028 "remount RDWR because of unsupported "
3029 "optional features (%x).\n",
3030 sb->s_id, le32_to_cpu(ret));
3031 err = -EROFS;
3032 goto restore_opts;
3033 }
3034
3035 /*
3036 * Make sure the group descriptor checksums
3037 * are sane. If they aren't, refuse to
3038 * remount r/w.
3039 */
3040 for (g = 0; g < sbi->s_groups_count; g++) {
3041 struct ext4_group_desc *gdp =
3042 ext4_get_group_desc(sb, g, NULL);
3043
3044 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3045 printk(KERN_ERR
3046 "EXT4-fs: ext4_remount: "
3047 "Checksum for group %lu failed (%u!=%u)\n",
3048 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3049 le16_to_cpu(gdp->bg_checksum));
3050 err = -EINVAL;
3051 goto restore_opts;
3052 }
3053 }
3054
3055 /*
3056 * If we have an unprocessed orphan list hanging
3057 * around from a previously readonly bdev mount,
3058 * require a full umount/remount for now.
3059 */
3060 if (es->s_last_orphan) {
3061 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3062 "remount RDWR because of unprocessed "
3063 "orphan inode list. Please "
3064 "umount/remount instead.\n",
3065 sb->s_id);
3066 err = -EINVAL;
3067 goto restore_opts;
3068 }
3069
3070 /*
3071 * Mounting a RDONLY partition read-write, so reread
3072 * and store the current valid flag. (It may have
3073 * been changed by e2fsck since we originally mounted
3074 * the partition.)
3075 */
3076 ext4_clear_journal_err(sb, es);
3077 sbi->s_mount_state = le16_to_cpu(es->s_state);
3078 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3079 goto restore_opts;
3080 if (!ext4_setup_super(sb, es, 0))
3081 sb->s_flags &= ~MS_RDONLY;
3082 }
3083 }
3084 #ifdef CONFIG_QUOTA
3085 /* Release old quota file names */
3086 for (i = 0; i < MAXQUOTAS; i++)
3087 if (old_opts.s_qf_names[i] &&
3088 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3089 kfree(old_opts.s_qf_names[i]);
3090 #endif
3091 return 0;
3092 restore_opts:
3093 sb->s_flags = old_sb_flags;
3094 sbi->s_mount_opt = old_opts.s_mount_opt;
3095 sbi->s_resuid = old_opts.s_resuid;
3096 sbi->s_resgid = old_opts.s_resgid;
3097 sbi->s_commit_interval = old_opts.s_commit_interval;
3098 #ifdef CONFIG_QUOTA
3099 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3100 for (i = 0; i < MAXQUOTAS; i++) {
3101 if (sbi->s_qf_names[i] &&
3102 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3103 kfree(sbi->s_qf_names[i]);
3104 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3105 }
3106 #endif
3107 return err;
3108 }
3109
3110 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3111 {
3112 struct super_block *sb = dentry->d_sb;
3113 struct ext4_sb_info *sbi = EXT4_SB(sb);
3114 struct ext4_super_block *es = sbi->s_es;
3115 u64 fsid;
3116
3117 if (test_opt(sb, MINIX_DF)) {
3118 sbi->s_overhead_last = 0;
3119 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3120 ext4_group_t ngroups = sbi->s_groups_count, i;
3121 ext4_fsblk_t overhead = 0;
3122 smp_rmb();
3123
3124 /*
3125 * Compute the overhead (FS structures). This is constant
3126 * for a given filesystem unless the number of block groups
3127 * changes so we cache the previous value until it does.
3128 */
3129
3130 /*
3131 * All of the blocks before first_data_block are
3132 * overhead
3133 */
3134 overhead = le32_to_cpu(es->s_first_data_block);
3135
3136 /*
3137 * Add the overhead attributed to the superblock and
3138 * block group descriptors. If the sparse superblocks
3139 * feature is turned on, then not all groups have this.
3140 */
3141 for (i = 0; i < ngroups; i++) {
3142 overhead += ext4_bg_has_super(sb, i) +
3143 ext4_bg_num_gdb(sb, i);
3144 cond_resched();
3145 }
3146
3147 /*
3148 * Every block group has an inode bitmap, a block
3149 * bitmap, and an inode table.
3150 */
3151 overhead += ngroups * (2 + sbi->s_itb_per_group);
3152 sbi->s_overhead_last = overhead;
3153 smp_wmb();
3154 sbi->s_blocks_last = ext4_blocks_count(es);
3155 }
3156
3157 buf->f_type = EXT4_SUPER_MAGIC;
3158 buf->f_bsize = sb->s_blocksize;
3159 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3160 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3161 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3162 ext4_free_blocks_count_set(es, buf->f_bfree);
3163 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3164 if (buf->f_bfree < ext4_r_blocks_count(es))
3165 buf->f_bavail = 0;
3166 buf->f_files = le32_to_cpu(es->s_inodes_count);
3167 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3168 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3169 buf->f_namelen = EXT4_NAME_LEN;
3170 fsid = le64_to_cpup((void *)es->s_uuid) ^
3171 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3172 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3173 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3174 return 0;
3175 }
3176
3177 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3178 * is locked for write. Otherwise the are possible deadlocks:
3179 * Process 1 Process 2
3180 * ext4_create() quota_sync()
3181 * jbd2_journal_start() write_dquot()
3182 * DQUOT_INIT() down(dqio_mutex)
3183 * down(dqio_mutex) jbd2_journal_start()
3184 *
3185 */
3186
3187 #ifdef CONFIG_QUOTA
3188
3189 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3190 {
3191 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3192 }
3193
3194 static int ext4_dquot_initialize(struct inode *inode, int type)
3195 {
3196 handle_t *handle;
3197 int ret, err;
3198
3199 /* We may create quota structure so we need to reserve enough blocks */
3200 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3201 if (IS_ERR(handle))
3202 return PTR_ERR(handle);
3203 ret = dquot_initialize(inode, type);
3204 err = ext4_journal_stop(handle);
3205 if (!ret)
3206 ret = err;
3207 return ret;
3208 }
3209
3210 static int ext4_dquot_drop(struct inode *inode)
3211 {
3212 handle_t *handle;
3213 int ret, err;
3214
3215 /* We may delete quota structure so we need to reserve enough blocks */
3216 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3217 if (IS_ERR(handle)) {
3218 /*
3219 * We call dquot_drop() anyway to at least release references
3220 * to quota structures so that umount does not hang.
3221 */
3222 dquot_drop(inode);
3223 return PTR_ERR(handle);
3224 }
3225 ret = dquot_drop(inode);
3226 err = ext4_journal_stop(handle);
3227 if (!ret)
3228 ret = err;
3229 return ret;
3230 }
3231
3232 static int ext4_write_dquot(struct dquot *dquot)
3233 {
3234 int ret, err;
3235 handle_t *handle;
3236 struct inode *inode;
3237
3238 inode = dquot_to_inode(dquot);
3239 handle = ext4_journal_start(inode,
3240 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3241 if (IS_ERR(handle))
3242 return PTR_ERR(handle);
3243 ret = dquot_commit(dquot);
3244 err = ext4_journal_stop(handle);
3245 if (!ret)
3246 ret = err;
3247 return ret;
3248 }
3249
3250 static int ext4_acquire_dquot(struct dquot *dquot)
3251 {
3252 int ret, err;
3253 handle_t *handle;
3254
3255 handle = ext4_journal_start(dquot_to_inode(dquot),
3256 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3257 if (IS_ERR(handle))
3258 return PTR_ERR(handle);
3259 ret = dquot_acquire(dquot);
3260 err = ext4_journal_stop(handle);
3261 if (!ret)
3262 ret = err;
3263 return ret;
3264 }
3265
3266 static int ext4_release_dquot(struct dquot *dquot)
3267 {
3268 int ret, err;
3269 handle_t *handle;
3270
3271 handle = ext4_journal_start(dquot_to_inode(dquot),
3272 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3273 if (IS_ERR(handle)) {
3274 /* Release dquot anyway to avoid endless cycle in dqput() */
3275 dquot_release(dquot);
3276 return PTR_ERR(handle);
3277 }
3278 ret = dquot_release(dquot);
3279 err = ext4_journal_stop(handle);
3280 if (!ret)
3281 ret = err;
3282 return ret;
3283 }
3284
3285 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3286 {
3287 /* Are we journaling quotas? */
3288 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3289 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3290 dquot_mark_dquot_dirty(dquot);
3291 return ext4_write_dquot(dquot);
3292 } else {
3293 return dquot_mark_dquot_dirty(dquot);
3294 }
3295 }
3296
3297 static int ext4_write_info(struct super_block *sb, int type)
3298 {
3299 int ret, err;
3300 handle_t *handle;
3301
3302 /* Data block + inode block */
3303 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3304 if (IS_ERR(handle))
3305 return PTR_ERR(handle);
3306 ret = dquot_commit_info(sb, type);
3307 err = ext4_journal_stop(handle);
3308 if (!ret)
3309 ret = err;
3310 return ret;
3311 }
3312
3313 /*
3314 * Turn on quotas during mount time - we need to find
3315 * the quota file and such...
3316 */
3317 static int ext4_quota_on_mount(struct super_block *sb, int type)
3318 {
3319 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3320 EXT4_SB(sb)->s_jquota_fmt, type);
3321 }
3322
3323 /*
3324 * Standard function to be called on quota_on
3325 */
3326 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3327 char *name, int remount)
3328 {
3329 int err;
3330 struct path path;
3331
3332 if (!test_opt(sb, QUOTA))
3333 return -EINVAL;
3334 /* When remounting, no checks are needed and in fact, name is NULL */
3335 if (remount)
3336 return vfs_quota_on(sb, type, format_id, name, remount);
3337
3338 err = kern_path(name, LOOKUP_FOLLOW, &path);
3339 if (err)
3340 return err;
3341
3342 /* Quotafile not on the same filesystem? */
3343 if (path.mnt->mnt_sb != sb) {
3344 path_put(&path);
3345 return -EXDEV;
3346 }
3347 /* Journaling quota? */
3348 if (EXT4_SB(sb)->s_qf_names[type]) {
3349 /* Quotafile not in fs root? */
3350 if (path.dentry->d_parent != sb->s_root)
3351 printk(KERN_WARNING
3352 "EXT4-fs: Quota file not on filesystem root. "
3353 "Journaled quota will not work.\n");
3354 }
3355
3356 /*
3357 * When we journal data on quota file, we have to flush journal to see
3358 * all updates to the file when we bypass pagecache...
3359 */
3360 if (ext4_should_journal_data(path.dentry->d_inode)) {
3361 /*
3362 * We don't need to lock updates but journal_flush() could
3363 * otherwise be livelocked...
3364 */
3365 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3366 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3367 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3368 if (err) {
3369 path_put(&path);
3370 return err;
3371 }
3372 }
3373
3374 err = vfs_quota_on_path(sb, type, format_id, &path);
3375 path_put(&path);
3376 return err;
3377 }
3378
3379 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3380 * acquiring the locks... As quota files are never truncated and quota code
3381 * itself serializes the operations (and noone else should touch the files)
3382 * we don't have to be afraid of races */
3383 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3384 size_t len, loff_t off)
3385 {
3386 struct inode *inode = sb_dqopt(sb)->files[type];
3387 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3388 int err = 0;
3389 int offset = off & (sb->s_blocksize - 1);
3390 int tocopy;
3391 size_t toread;
3392 struct buffer_head *bh;
3393 loff_t i_size = i_size_read(inode);
3394
3395 if (off > i_size)
3396 return 0;
3397 if (off+len > i_size)
3398 len = i_size-off;
3399 toread = len;
3400 while (toread > 0) {
3401 tocopy = sb->s_blocksize - offset < toread ?
3402 sb->s_blocksize - offset : toread;
3403 bh = ext4_bread(NULL, inode, blk, 0, &err);
3404 if (err)
3405 return err;
3406 if (!bh) /* A hole? */
3407 memset(data, 0, tocopy);
3408 else
3409 memcpy(data, bh->b_data+offset, tocopy);
3410 brelse(bh);
3411 offset = 0;
3412 toread -= tocopy;
3413 data += tocopy;
3414 blk++;
3415 }
3416 return len;
3417 }
3418
3419 /* Write to quotafile (we know the transaction is already started and has
3420 * enough credits) */
3421 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3422 const char *data, size_t len, loff_t off)
3423 {
3424 struct inode *inode = sb_dqopt(sb)->files[type];
3425 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3426 int err = 0;
3427 int offset = off & (sb->s_blocksize - 1);
3428 int tocopy;
3429 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3430 size_t towrite = len;
3431 struct buffer_head *bh;
3432 handle_t *handle = journal_current_handle();
3433
3434 if (!handle) {
3435 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3436 " cancelled because transaction is not started.\n",
3437 (unsigned long long)off, (unsigned long long)len);
3438 return -EIO;
3439 }
3440 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3441 while (towrite > 0) {
3442 tocopy = sb->s_blocksize - offset < towrite ?
3443 sb->s_blocksize - offset : towrite;
3444 bh = ext4_bread(handle, inode, blk, 1, &err);
3445 if (!bh)
3446 goto out;
3447 if (journal_quota) {
3448 err = ext4_journal_get_write_access(handle, bh);
3449 if (err) {
3450 brelse(bh);
3451 goto out;
3452 }
3453 }
3454 lock_buffer(bh);
3455 memcpy(bh->b_data+offset, data, tocopy);
3456 flush_dcache_page(bh->b_page);
3457 unlock_buffer(bh);
3458 if (journal_quota)
3459 err = ext4_journal_dirty_metadata(handle, bh);
3460 else {
3461 /* Always do at least ordered writes for quotas */
3462 err = ext4_jbd2_file_inode(handle, inode);
3463 mark_buffer_dirty(bh);
3464 }
3465 brelse(bh);
3466 if (err)
3467 goto out;
3468 offset = 0;
3469 towrite -= tocopy;
3470 data += tocopy;
3471 blk++;
3472 }
3473 out:
3474 if (len == towrite) {
3475 mutex_unlock(&inode->i_mutex);
3476 return err;
3477 }
3478 if (inode->i_size < off+len-towrite) {
3479 i_size_write(inode, off+len-towrite);
3480 EXT4_I(inode)->i_disksize = inode->i_size;
3481 }
3482 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3483 ext4_mark_inode_dirty(handle, inode);
3484 mutex_unlock(&inode->i_mutex);
3485 return len - towrite;
3486 }
3487
3488 #endif
3489
3490 static int ext4_get_sb(struct file_system_type *fs_type,
3491 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3492 {
3493 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3494 }
3495
3496 #ifdef CONFIG_PROC_FS
3497 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3498 {
3499 unsigned int *p = m->private;
3500
3501 seq_printf(m, "%u\n", *p);
3502 return 0;
3503 }
3504
3505 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3506 {
3507 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3508 }
3509
3510 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3511 size_t cnt, loff_t *ppos)
3512 {
3513 unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3514 char str[32];
3515 unsigned long value;
3516
3517 if (cnt >= sizeof(str))
3518 return -EINVAL;
3519 if (copy_from_user(str, buf, cnt))
3520 return -EFAULT;
3521 value = simple_strtol(str, NULL, 0);
3522 if (value < 0)
3523 return -ERANGE;
3524 *p = value;
3525 return cnt;
3526 }
3527
3528 const struct file_operations ext4_ui_proc_fops = {
3529 .owner = THIS_MODULE,
3530 .open = ext4_ui_proc_open,
3531 .read = seq_read,
3532 .llseek = seq_lseek,
3533 .release = single_release,
3534 .write = ext4_ui_proc_write,
3535 };
3536 #endif
3537
3538 static struct file_system_type ext4_fs_type = {
3539 .owner = THIS_MODULE,
3540 .name = "ext4",
3541 .get_sb = ext4_get_sb,
3542 .kill_sb = kill_block_super,
3543 .fs_flags = FS_REQUIRES_DEV,
3544 };
3545
3546 #ifdef CONFIG_EXT4DEV_COMPAT
3547 static int ext4dev_get_sb(struct file_system_type *fs_type,
3548 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3549 {
3550 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3551 "to mount using ext4\n");
3552 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3553 "will go away by 2.6.31\n");
3554 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3555 }
3556
3557 static struct file_system_type ext4dev_fs_type = {
3558 .owner = THIS_MODULE,
3559 .name = "ext4dev",
3560 .get_sb = ext4dev_get_sb,
3561 .kill_sb = kill_block_super,
3562 .fs_flags = FS_REQUIRES_DEV,
3563 };
3564 MODULE_ALIAS("ext4dev");
3565 #endif
3566
3567 static int __init init_ext4_fs(void)
3568 {
3569 int err;
3570
3571 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3572 err = init_ext4_mballoc();
3573 if (err)
3574 return err;
3575
3576 err = init_ext4_xattr();
3577 if (err)
3578 goto out2;
3579 err = init_inodecache();
3580 if (err)
3581 goto out1;
3582 err = register_filesystem(&ext4_fs_type);
3583 if (err)
3584 goto out;
3585 #ifdef CONFIG_EXT4DEV_COMPAT
3586 err = register_filesystem(&ext4dev_fs_type);
3587 if (err) {
3588 unregister_filesystem(&ext4_fs_type);
3589 goto out;
3590 }
3591 #endif
3592 return 0;
3593 out:
3594 destroy_inodecache();
3595 out1:
3596 exit_ext4_xattr();
3597 out2:
3598 exit_ext4_mballoc();
3599 return err;
3600 }
3601
3602 static void __exit exit_ext4_fs(void)
3603 {
3604 unregister_filesystem(&ext4_fs_type);
3605 #ifdef CONFIG_EXT4DEV_COMPAT
3606 unregister_filesystem(&ext4dev_fs_type);
3607 #endif
3608 destroy_inodecache();
3609 exit_ext4_xattr();
3610 exit_ext4_mballoc();
3611 remove_proc_entry("fs/ext4", NULL);
3612 }
3613
3614 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3615 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3616 MODULE_LICENSE("GPL");
3617 module_init(init_ext4_fs)
3618 module_exit(exit_ext4_fs)
(deprecated) Btrfs devel tree