?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
GUI: Fix Tomato RAF theme for all builds. Compilation typo.
[tomato.git] / release / src-rt-6.x.4708 / linux / linux-2.6.36 / fs / ext3 / super.c
blobf82d00ad294f8f93b4d982da2dc5f6fe42d9aba8
1 /*
2 * linux/fs/ext3/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/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
52
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
71
72 /*
73 * Wrappers for journal_start/end.
74 *
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
78 * appropriate.
79 */
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 {
82 journal_t *journal;
83
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
86
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
95 }
96
97 return journal_start(journal, nblocks);
98 }
99
100 /*
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
104 * appropriate.
105 */
106 int __ext3_journal_stop(const char *where, handle_t *handle)
107 {
108 struct super_block *sb;
109 int err;
110 int rc;
111
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
115
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
121 }
122
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
125 {
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
131
132 if (!handle->h_err)
133 handle->h_err = err;
134
135 if (is_handle_aborted(handle))
136 return;
137
138 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
140
141 journal_abort_handle(handle);
142 }
143
144 void ext3_msg(struct super_block *sb, const char *prefix,
145 const char *fmt, ...)
146 {
147 va_list args;
148
149 va_start(args, fmt);
150 printk("%sEXT3-fs (%s): ", prefix, sb->s_id);
151 vprintk(fmt, args);
152 printk("\n");
153 va_end(args);
154 }
155
156 /* Deal with the reporting of failure conditions on a filesystem such as
157 * inconsistencies detected or read IO failures.
158 *
159 * On ext2, we can store the error state of the filesystem in the
160 * superblock. That is not possible on ext3, because we may have other
161 * write ordering constraints on the superblock which prevent us from
162 * writing it out straight away; and given that the journal is about to
163 * be aborted, we can't rely on the current, or future, transactions to
164 * write out the superblock safely.
165 *
166 * We'll just use the journal_abort() error code to record an error in
167 * the journal instead. On recovery, the journal will complain about
168 * that error until we've noted it down and cleared it.
169 */
170
171 static void ext3_handle_error(struct super_block *sb)
172 {
173 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
174
175 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
176 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
177
178 if (sb->s_flags & MS_RDONLY)
179 return;
180
181 if (!test_opt (sb, ERRORS_CONT)) {
182 journal_t *journal = EXT3_SB(sb)->s_journal;
183
184 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
185 if (journal)
186 journal_abort(journal, -EIO);
187 }
188 if (test_opt (sb, ERRORS_RO)) {
189 ext3_msg(sb, KERN_CRIT,
190 "error: remounting filesystem read-only");
191 sb->s_flags |= MS_RDONLY;
192 }
193 ext3_commit_super(sb, es, 1);
194 if (test_opt(sb, ERRORS_PANIC))
195 panic("EXT3-fs (%s): panic forced after error\n",
196 sb->s_id);
197 }
198
199 void ext3_error (struct super_block * sb, const char * function,
200 const char * fmt, ...)
201 {
202 va_list args;
203
204 va_start(args, fmt);
205 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
206 vprintk(fmt, args);
207 printk("\n");
208 va_end(args);
209
210 ext3_handle_error(sb);
211 }
212
213 static const char *ext3_decode_error(struct super_block * sb, int errno,
214 char nbuf[16])
215 {
216 char *errstr = NULL;
217
218 switch (errno) {
219 case -EIO:
220 errstr = "IO failure";
221 break;
222 case -ENOMEM:
223 errstr = "Out of memory";
224 break;
225 case -EROFS:
226 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
227 errstr = "Journal has aborted";
228 else
229 errstr = "Readonly filesystem";
230 break;
231 default:
232 /* If the caller passed in an extra buffer for unknown
233 * errors, textualise them now. Else we just return
234 * NULL. */
235 if (nbuf) {
236 /* Check for truncated error codes... */
237 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
238 errstr = nbuf;
239 }
240 break;
241 }
242
243 return errstr;
244 }
245
246 /* __ext3_std_error decodes expected errors from journaling functions
247 * automatically and invokes the appropriate error response. */
248
249 void __ext3_std_error (struct super_block * sb, const char * function,
250 int errno)
251 {
252 char nbuf[16];
253 const char *errstr;
254
255 /* Special case: if the error is EROFS, and we're not already
256 * inside a transaction, then there's really no point in logging
257 * an error. */
258 if (errno == -EROFS && journal_current_handle() == NULL &&
259 (sb->s_flags & MS_RDONLY))
260 return;
261
262 errstr = ext3_decode_error(sb, errno, nbuf);
263 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
264
265 ext3_handle_error(sb);
266 }
267
268 /*
269 * ext3_abort is a much stronger failure handler than ext3_error. The
270 * abort function may be used to deal with unrecoverable failures such
271 * as journal IO errors or ENOMEM at a critical moment in log management.
272 *
273 * We unconditionally force the filesystem into an ABORT|READONLY state,
274 * unless the error response on the fs has been set to panic in which
275 * case we take the easy way out and panic immediately.
276 */
277
278 void ext3_abort (struct super_block * sb, const char * function,
279 const char * fmt, ...)
280 {
281 va_list args;
282
283 va_start(args, fmt);
284 printk(KERN_CRIT "EXT3-fs (%s): error: %s: ", sb->s_id, function);
285 vprintk(fmt, args);
286 printk("\n");
287 va_end(args);
288
289 if (test_opt(sb, ERRORS_PANIC))
290 panic("EXT3-fs: panic from previous error\n");
291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
295 ext3_msg(sb, KERN_CRIT,
296 "error: remounting filesystem read-only");
297 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
298 sb->s_flags |= MS_RDONLY;
299 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
300 if (EXT3_SB(sb)->s_journal)
301 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
302 }
303
304 void ext3_warning (struct super_block * sb, const char * function,
305 const char * fmt, ...)
306 {
307 va_list args;
308
309 va_start(args, fmt);
310 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: ",
311 sb->s_id, function);
312 vprintk(fmt, args);
313 printk("\n");
314 va_end(args);
315 }
316
317 void ext3_update_dynamic_rev(struct super_block *sb)
318 {
319 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
320
321 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
322 return;
323
324 ext3_msg(sb, KERN_WARNING,
325 "warning: updating to rev %d because of "
326 "new feature flag, running e2fsck is recommended",
327 EXT3_DYNAMIC_REV);
328
329 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
330 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
331 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
332 /* leave es->s_feature_*compat flags alone */
333 /* es->s_uuid will be set by e2fsck if empty */
334
335 /*
336 * The rest of the superblock fields should be zero, and if not it
337 * means they are likely already in use, so leave them alone. We
338 * can leave it up to e2fsck to clean up any inconsistencies there.
339 */
340 }
341
342 /*
343 * Open the external journal device
344 */
345 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
346 {
347 struct block_device *bdev;
348 char b[BDEVNAME_SIZE];
349
350 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
351 if (IS_ERR(bdev))
352 goto fail;
353 return bdev;
354
355 fail:
356 ext3_msg(sb, "error: failed to open journal device %s: %ld",
357 __bdevname(dev, b), PTR_ERR(bdev));
358
359 return NULL;
360 }
361
362 /*
363 * Release the journal device
364 */
365 static int ext3_blkdev_put(struct block_device *bdev)
366 {
367 bd_release(bdev);
368 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
369 }
370
371 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
372 {
373 struct block_device *bdev;
374 int ret = -ENODEV;
375
376 bdev = sbi->journal_bdev;
377 if (bdev) {
378 ret = ext3_blkdev_put(bdev);
379 sbi->journal_bdev = NULL;
380 }
381 return ret;
382 }
383
384 static inline struct inode *orphan_list_entry(struct list_head *l)
385 {
386 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
387 }
388
389 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
390 {
391 struct list_head *l;
392
393 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
394 le32_to_cpu(sbi->s_es->s_last_orphan));
395
396 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
397 list_for_each(l, &sbi->s_orphan) {
398 struct inode *inode = orphan_list_entry(l);
399 ext3_msg(sb, KERN_ERR, " "
400 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
401 inode->i_sb->s_id, inode->i_ino, inode,
402 inode->i_mode, inode->i_nlink,
403 NEXT_ORPHAN(inode));
404 }
405 }
406
407 static void ext3_put_super (struct super_block * sb)
408 {
409 struct ext3_sb_info *sbi = EXT3_SB(sb);
410 struct ext3_super_block *es = sbi->s_es;
411 int i, err;
412
413 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
414
415 lock_kernel();
416
417 ext3_xattr_put_super(sb);
418 err = journal_destroy(sbi->s_journal);
419 sbi->s_journal = NULL;
420 if (err < 0)
421 ext3_abort(sb, __func__, "Couldn't clean up the journal");
422
423 if (!(sb->s_flags & MS_RDONLY)) {
424 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
425 es->s_state = cpu_to_le16(sbi->s_mount_state);
426 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
427 mark_buffer_dirty(sbi->s_sbh);
428 ext3_commit_super(sb, es, 1);
429 }
430
431 for (i = 0; i < sbi->s_gdb_count; i++)
432 brelse(sbi->s_group_desc[i]);
433 kfree(sbi->s_group_desc);
434 percpu_counter_destroy(&sbi->s_freeblocks_counter);
435 percpu_counter_destroy(&sbi->s_freeinodes_counter);
436 percpu_counter_destroy(&sbi->s_dirs_counter);
437 brelse(sbi->s_sbh);
438 #ifdef CONFIG_QUOTA
439 for (i = 0; i < MAXQUOTAS; i++)
440 kfree(sbi->s_qf_names[i]);
441 #endif
442
443 /* Debugging code just in case the in-memory inode orphan list
444 * isn't empty. The on-disk one can be non-empty if we've
445 * detected an error and taken the fs readonly, but the
446 * in-memory list had better be clean by this point. */
447 if (!list_empty(&sbi->s_orphan))
448 dump_orphan_list(sb, sbi);
449 J_ASSERT(list_empty(&sbi->s_orphan));
450
451 invalidate_bdev(sb->s_bdev);
452 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
453 /*
454 * Invalidate the journal device's buffers. We don't want them
455 * floating about in memory - the physical journal device may
456 * hotswapped, and it breaks the `ro-after' testing code.
457 */
458 sync_blockdev(sbi->journal_bdev);
459 invalidate_bdev(sbi->journal_bdev);
460 ext3_blkdev_remove(sbi);
461 }
462 sb->s_fs_info = NULL;
463 kfree(sbi->s_blockgroup_lock);
464 kfree(sbi);
465
466 unlock_kernel();
467 }
468
469 static struct kmem_cache *ext3_inode_cachep;
470
471 /*
472 * Called inside transaction, so use GFP_NOFS
473 */
474 static struct inode *ext3_alloc_inode(struct super_block *sb)
475 {
476 struct ext3_inode_info *ei;
477
478 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
479 if (!ei)
480 return NULL;
481 ei->i_block_alloc_info = NULL;
482 ei->vfs_inode.i_version = 1;
483 atomic_set(&ei->i_datasync_tid, 0);
484 atomic_set(&ei->i_sync_tid, 0);
485 return &ei->vfs_inode;
486 }
487
488 static void ext3_destroy_inode(struct inode *inode)
489 {
490 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
491 printk("EXT3 Inode %p: orphan list check failed!\n",
492 EXT3_I(inode));
493 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
494 EXT3_I(inode), sizeof(struct ext3_inode_info),
495 false);
496 dump_stack();
497 }
498 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
499 }
500
501 static void init_once(void *foo)
502 {
503 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
504
505 INIT_LIST_HEAD(&ei->i_orphan);
506 #ifdef CONFIG_EXT3_FS_XATTR
507 init_rwsem(&ei->xattr_sem);
508 #endif
509 mutex_init(&ei->truncate_mutex);
510 inode_init_once(&ei->vfs_inode);
511 }
512
513 static int init_inodecache(void)
514 {
515 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
516 sizeof(struct ext3_inode_info),
517 0, (SLAB_RECLAIM_ACCOUNT|
518 SLAB_MEM_SPREAD),
519 init_once);
520 if (ext3_inode_cachep == NULL)
521 return -ENOMEM;
522 return 0;
523 }
524
525 static void destroy_inodecache(void)
526 {
527 kmem_cache_destroy(ext3_inode_cachep);
528 }
529
530 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
531 {
532 #if defined(CONFIG_QUOTA)
533 struct ext3_sb_info *sbi = EXT3_SB(sb);
534
535 if (sbi->s_jquota_fmt) {
536 char *fmtname = "";
537
538 switch (sbi->s_jquota_fmt) {
539 case QFMT_VFS_OLD:
540 fmtname = "vfsold";
541 break;
542 case QFMT_VFS_V0:
543 fmtname = "vfsv0";
544 break;
545 case QFMT_VFS_V1:
546 fmtname = "vfsv1";
547 break;
548 }
549 seq_printf(seq, ",jqfmt=%s", fmtname);
550 }
551
552 if (sbi->s_qf_names[USRQUOTA])
553 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
554
555 if (sbi->s_qf_names[GRPQUOTA])
556 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
557
558 if (test_opt(sb, USRQUOTA))
559 seq_puts(seq, ",usrquota");
560
561 if (test_opt(sb, GRPQUOTA))
562 seq_puts(seq, ",grpquota");
563 #endif
564 }
565
566 static char *data_mode_string(unsigned long mode)
567 {
568 switch (mode) {
569 case EXT3_MOUNT_JOURNAL_DATA:
570 return "journal";
571 case EXT3_MOUNT_ORDERED_DATA:
572 return "ordered";
573 case EXT3_MOUNT_WRITEBACK_DATA:
574 return "writeback";
575 }
576 return "unknown";
577 }
578
579 /*
580 * Show an option if
581 * - it's set to a non-default value OR
582 * - if the per-sb default is different from the global default
583 */
584 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
585 {
586 struct super_block *sb = vfs->mnt_sb;
587 struct ext3_sb_info *sbi = EXT3_SB(sb);
588 struct ext3_super_block *es = sbi->s_es;
589 unsigned long def_mount_opts;
590
591 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
592
593 if (sbi->s_sb_block != 1)
594 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
595 if (test_opt(sb, MINIX_DF))
596 seq_puts(seq, ",minixdf");
597 if (test_opt(sb, GRPID))
598 seq_puts(seq, ",grpid");
599 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
600 seq_puts(seq, ",nogrpid");
601 if (sbi->s_resuid != EXT3_DEF_RESUID ||
602 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
603 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
604 }
605 if (sbi->s_resgid != EXT3_DEF_RESGID ||
606 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
607 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
608 }
609 if (test_opt(sb, ERRORS_RO)) {
610 int def_errors = le16_to_cpu(es->s_errors);
611
612 if (def_errors == EXT3_ERRORS_PANIC ||
613 def_errors == EXT3_ERRORS_CONTINUE) {
614 seq_puts(seq, ",errors=remount-ro");
615 }
616 }
617 if (test_opt(sb, ERRORS_CONT))
618 seq_puts(seq, ",errors=continue");
619 if (test_opt(sb, ERRORS_PANIC))
620 seq_puts(seq, ",errors=panic");
621 if (test_opt(sb, NO_UID32))
622 seq_puts(seq, ",nouid32");
623 if (test_opt(sb, DEBUG))
624 seq_puts(seq, ",debug");
625 if (test_opt(sb, OLDALLOC))
626 seq_puts(seq, ",oldalloc");
627 #ifdef CONFIG_EXT3_FS_XATTR
628 if (test_opt(sb, XATTR_USER))
629 seq_puts(seq, ",user_xattr");
630 if (!test_opt(sb, XATTR_USER) &&
631 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
632 seq_puts(seq, ",nouser_xattr");
633 }
634 #endif
635 #ifdef CONFIG_EXT3_FS_POSIX_ACL
636 if (test_opt(sb, POSIX_ACL))
637 seq_puts(seq, ",acl");
638 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
639 seq_puts(seq, ",noacl");
640 #endif
641 if (!test_opt(sb, RESERVATION))
642 seq_puts(seq, ",noreservation");
643 if (sbi->s_commit_interval) {
644 seq_printf(seq, ",commit=%u",
645 (unsigned) (sbi->s_commit_interval / HZ));
646 }
647
648 /*
649 * Always display barrier state so it's clear what the status is.
650 */
651 seq_puts(seq, ",barrier=");
652 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
653 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
654 if (test_opt(sb, DATA_ERR_ABORT))
655 seq_puts(seq, ",data_err=abort");
656
657 if (test_opt(sb, NOLOAD))
658 seq_puts(seq, ",norecovery");
659
660 ext3_show_quota_options(seq, sb);
661
662 return 0;
663 }
664
665
666 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
667 u64 ino, u32 generation)
668 {
669 struct inode *inode;
670
671 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
672 return ERR_PTR(-ESTALE);
673 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
674 return ERR_PTR(-ESTALE);
675
676 /* iget isn't really right if the inode is currently unallocated!!
677 *
678 * ext3_read_inode will return a bad_inode if the inode had been
679 * deleted, so we should be safe.
680 *
681 * Currently we don't know the generation for parent directory, so
682 * a generation of 0 means "accept any"
683 */
684 inode = ext3_iget(sb, ino);
685 if (IS_ERR(inode))
686 return ERR_CAST(inode);
687 if (generation && inode->i_generation != generation) {
688 iput(inode);
689 return ERR_PTR(-ESTALE);
690 }
691
692 return inode;
693 }
694
695 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
696 int fh_len, int fh_type)
697 {
698 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
699 ext3_nfs_get_inode);
700 }
701
702 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
703 int fh_len, int fh_type)
704 {
705 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
706 ext3_nfs_get_inode);
707 }
708
709 /*
710 * Try to release metadata pages (indirect blocks, directories) which are
711 * mapped via the block device. Since these pages could have journal heads
712 * which would prevent try_to_free_buffers() from freeing them, we must use
713 * jbd layer's try_to_free_buffers() function to release them.
714 */
715 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
716 gfp_t wait)
717 {
718 journal_t *journal = EXT3_SB(sb)->s_journal;
719
720 WARN_ON(PageChecked(page));
721 if (!page_has_buffers(page))
722 return 0;
723 if (journal)
724 return journal_try_to_free_buffers(journal, page,
725 wait & ~__GFP_WAIT);
726 return try_to_free_buffers(page);
727 }
728
729 #ifdef CONFIG_QUOTA
730 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
731 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
732
733 static int ext3_write_dquot(struct dquot *dquot);
734 static int ext3_acquire_dquot(struct dquot *dquot);
735 static int ext3_release_dquot(struct dquot *dquot);
736 static int ext3_mark_dquot_dirty(struct dquot *dquot);
737 static int ext3_write_info(struct super_block *sb, int type);
738 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
739 char *path);
740 static int ext3_quota_on_mount(struct super_block *sb, int type);
741 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
742 size_t len, loff_t off);
743 static ssize_t ext3_quota_write(struct super_block *sb, int type,
744 const char *data, size_t len, loff_t off);
745
746 static const struct dquot_operations ext3_quota_operations = {
747 .write_dquot = ext3_write_dquot,
748 .acquire_dquot = ext3_acquire_dquot,
749 .release_dquot = ext3_release_dquot,
750 .mark_dirty = ext3_mark_dquot_dirty,
751 .write_info = ext3_write_info,
752 .alloc_dquot = dquot_alloc,
753 .destroy_dquot = dquot_destroy,
754 };
755
756 static const struct quotactl_ops ext3_qctl_operations = {
757 .quota_on = ext3_quota_on,
758 .quota_off = dquot_quota_off,
759 .quota_sync = dquot_quota_sync,
760 .get_info = dquot_get_dqinfo,
761 .set_info = dquot_set_dqinfo,
762 .get_dqblk = dquot_get_dqblk,
763 .set_dqblk = dquot_set_dqblk
764 };
765 #endif
766
767 static const struct super_operations ext3_sops = {
768 .alloc_inode = ext3_alloc_inode,
769 .destroy_inode = ext3_destroy_inode,
770 .write_inode = ext3_write_inode,
771 .dirty_inode = ext3_dirty_inode,
772 .evict_inode = ext3_evict_inode,
773 .put_super = ext3_put_super,
774 .sync_fs = ext3_sync_fs,
775 .freeze_fs = ext3_freeze,
776 .unfreeze_fs = ext3_unfreeze,
777 .statfs = ext3_statfs,
778 .remount_fs = ext3_remount,
779 .show_options = ext3_show_options,
780 #ifdef CONFIG_QUOTA
781 .quota_read = ext3_quota_read,
782 .quota_write = ext3_quota_write,
783 #endif
784 .bdev_try_to_free_page = bdev_try_to_free_page,
785 };
786
787 static const struct export_operations ext3_export_ops = {
788 .fh_to_dentry = ext3_fh_to_dentry,
789 .fh_to_parent = ext3_fh_to_parent,
790 .get_parent = ext3_get_parent,
791 };
792
793 enum {
794 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
795 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
796 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
797 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
798 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
799 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
800 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
801 Opt_data_err_abort, Opt_data_err_ignore,
802 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
803 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
804 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
805 Opt_resize, Opt_usrquota, Opt_grpquota
806 };
807
808 static const match_table_t tokens = {
809 {Opt_bsd_df, "bsddf"},
810 {Opt_minix_df, "minixdf"},
811 {Opt_grpid, "grpid"},
812 {Opt_grpid, "bsdgroups"},
813 {Opt_nogrpid, "nogrpid"},
814 {Opt_nogrpid, "sysvgroups"},
815 {Opt_resgid, "resgid=%u"},
816 {Opt_resuid, "resuid=%u"},
817 {Opt_sb, "sb=%u"},
818 {Opt_err_cont, "errors=continue"},
819 {Opt_err_panic, "errors=panic"},
820 {Opt_err_ro, "errors=remount-ro"},
821 {Opt_nouid32, "nouid32"},
822 {Opt_nocheck, "nocheck"},
823 {Opt_nocheck, "check=none"},
824 {Opt_debug, "debug"},
825 {Opt_oldalloc, "oldalloc"},
826 {Opt_orlov, "orlov"},
827 {Opt_user_xattr, "user_xattr"},
828 {Opt_nouser_xattr, "nouser_xattr"},
829 {Opt_acl, "acl"},
830 {Opt_noacl, "noacl"},
831 {Opt_reservation, "reservation"},
832 {Opt_noreservation, "noreservation"},
833 {Opt_noload, "noload"},
834 {Opt_noload, "norecovery"},
835 {Opt_nobh, "nobh"},
836 {Opt_bh, "bh"},
837 {Opt_commit, "commit=%u"},
838 {Opt_journal_update, "journal=update"},
839 {Opt_journal_inum, "journal=%u"},
840 {Opt_journal_dev, "journal_dev=%u"},
841 {Opt_abort, "abort"},
842 {Opt_data_journal, "data=journal"},
843 {Opt_data_ordered, "data=ordered"},
844 {Opt_data_writeback, "data=writeback"},
845 {Opt_data_err_abort, "data_err=abort"},
846 {Opt_data_err_ignore, "data_err=ignore"},
847 {Opt_offusrjquota, "usrjquota="},
848 {Opt_usrjquota, "usrjquota=%s"},
849 {Opt_offgrpjquota, "grpjquota="},
850 {Opt_grpjquota, "grpjquota=%s"},
851 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
852 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
853 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
854 {Opt_grpquota, "grpquota"},
855 {Opt_noquota, "noquota"},
856 {Opt_quota, "quota"},
857 {Opt_usrquota, "usrquota"},
858 {Opt_barrier, "barrier=%u"},
859 {Opt_barrier, "barrier"},
860 {Opt_nobarrier, "nobarrier"},
861 {Opt_resize, "resize"},
862 {Opt_err, NULL},
863 };
864
865 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
866 {
867 ext3_fsblk_t sb_block;
868 char *options = (char *) *data;
869
870 if (!options || strncmp(options, "sb=", 3) != 0)
871 return 1; /* Default location */
872 options += 3;
873 /*todo: use simple_strtoll with >32bit ext3 */
874 sb_block = simple_strtoul(options, &options, 0);
875 if (*options && *options != ',') {
876 ext3_msg(sb, "error: invalid sb specification: %s",
877 (char *) *data);
878 return 1;
879 }
880 if (*options == ',')
881 options++;
882 *data = (void *) options;
883 return sb_block;
884 }
885
886 #ifdef CONFIG_QUOTA
887 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
888 {
889 struct ext3_sb_info *sbi = EXT3_SB(sb);
890 char *qname;
891
892 if (sb_any_quota_loaded(sb) &&
893 !sbi->s_qf_names[qtype]) {
894 ext3_msg(sb, KERN_ERR,
895 "Cannot change journaled "
896 "quota options when quota turned on");
897 return 0;
898 }
899 qname = match_strdup(args);
900 if (!qname) {
901 ext3_msg(sb, KERN_ERR,
902 "Not enough memory for storing quotafile name");
903 return 0;
904 }
905 if (sbi->s_qf_names[qtype] &&
906 strcmp(sbi->s_qf_names[qtype], qname)) {
907 ext3_msg(sb, KERN_ERR,
908 "%s quota file already specified", QTYPE2NAME(qtype));
909 kfree(qname);
910 return 0;
911 }
912 sbi->s_qf_names[qtype] = qname;
913 if (strchr(sbi->s_qf_names[qtype], '/')) {
914 ext3_msg(sb, KERN_ERR,
915 "quotafile must be on filesystem root");
916 kfree(sbi->s_qf_names[qtype]);
917 sbi->s_qf_names[qtype] = NULL;
918 return 0;
919 }
920 set_opt(sbi->s_mount_opt, QUOTA);
921 return 1;
922 }
923
924 static int clear_qf_name(struct super_block *sb, int qtype) {
925
926 struct ext3_sb_info *sbi = EXT3_SB(sb);
927
928 if (sb_any_quota_loaded(sb) &&
929 sbi->s_qf_names[qtype]) {
930 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
931 " when quota turned on");
932 return 0;
933 }
934 /*
935 * The space will be released later when all options are confirmed
936 * to be correct
937 */
938 sbi->s_qf_names[qtype] = NULL;
939 return 1;
940 }
941 #endif
942
943 static int parse_options (char *options, struct super_block *sb,
944 unsigned int *inum, unsigned long *journal_devnum,
945 ext3_fsblk_t *n_blocks_count, int is_remount)
946 {
947 struct ext3_sb_info *sbi = EXT3_SB(sb);
948 char * p;
949 substring_t args[MAX_OPT_ARGS];
950 int data_opt = 0;
951 int option;
952 #ifdef CONFIG_QUOTA
953 int qfmt;
954 #endif
955
956 if (!options)
957 return 1;
958
959 while ((p = strsep (&options, ",")) != NULL) {
960 int token;
961 if (!*p)
962 continue;
963 /*
964 * Initialize args struct so we know whether arg was
965 * found; some options take optional arguments.
966 */
967 args[0].to = args[0].from = 0;
968 token = match_token(p, tokens, args);
969 switch (token) {
970 case Opt_bsd_df:
971 clear_opt (sbi->s_mount_opt, MINIX_DF);
972 break;
973 case Opt_minix_df:
974 set_opt (sbi->s_mount_opt, MINIX_DF);
975 break;
976 case Opt_grpid:
977 set_opt (sbi->s_mount_opt, GRPID);
978 break;
979 case Opt_nogrpid:
980 clear_opt (sbi->s_mount_opt, GRPID);
981 break;
982 case Opt_resuid:
983 if (match_int(&args[0], &option))
984 return 0;
985 sbi->s_resuid = option;
986 break;
987 case Opt_resgid:
988 if (match_int(&args[0], &option))
989 return 0;
990 sbi->s_resgid = option;
991 break;
992 case Opt_sb:
993 /* handled by get_sb_block() instead of here */
994 /* *sb_block = match_int(&args[0]); */
995 break;
996 case Opt_err_panic:
997 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
998 clear_opt (sbi->s_mount_opt, ERRORS_RO);
999 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1000 break;
1001 case Opt_err_ro:
1002 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1003 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1004 set_opt (sbi->s_mount_opt, ERRORS_RO);
1005 break;
1006 case Opt_err_cont:
1007 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1008 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1009 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1010 break;
1011 case Opt_nouid32:
1012 set_opt (sbi->s_mount_opt, NO_UID32);
1013 break;
1014 case Opt_nocheck:
1015 clear_opt (sbi->s_mount_opt, CHECK);
1016 break;
1017 case Opt_debug:
1018 set_opt (sbi->s_mount_opt, DEBUG);
1019 break;
1020 case Opt_oldalloc:
1021 set_opt (sbi->s_mount_opt, OLDALLOC);
1022 break;
1023 case Opt_orlov:
1024 clear_opt (sbi->s_mount_opt, OLDALLOC);
1025 break;
1026 #ifdef CONFIG_EXT3_FS_XATTR
1027 case Opt_user_xattr:
1028 set_opt (sbi->s_mount_opt, XATTR_USER);
1029 break;
1030 case Opt_nouser_xattr:
1031 clear_opt (sbi->s_mount_opt, XATTR_USER);
1032 break;
1033 #else
1034 case Opt_user_xattr:
1035 case Opt_nouser_xattr:
1036 ext3_msg(sb, KERN_INFO,
1037 "(no)user_xattr options not supported");
1038 break;
1039 #endif
1040 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1041 case Opt_acl:
1042 set_opt(sbi->s_mount_opt, POSIX_ACL);
1043 break;
1044 case Opt_noacl:
1045 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1046 break;
1047 #else
1048 case Opt_acl:
1049 case Opt_noacl:
1050 ext3_msg(sb, KERN_INFO,
1051 "(no)acl options not supported");
1052 break;
1053 #endif
1054 case Opt_reservation:
1055 set_opt(sbi->s_mount_opt, RESERVATION);
1056 break;
1057 case Opt_noreservation:
1058 clear_opt(sbi->s_mount_opt, RESERVATION);
1059 break;
1060 case Opt_journal_update:
1061 /* Eventually we will want to be able to create
1062 a journal file here. For now, only allow the
1063 user to specify an existing inode to be the
1064 journal file. */
1065 if (is_remount) {
1066 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1067 "journal on remount");
1068 return 0;
1069 }
1070 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1071 break;
1072 case Opt_journal_inum:
1073 if (is_remount) {
1074 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1075 "journal on remount");
1076 return 0;
1077 }
1078 if (match_int(&args[0], &option))
1079 return 0;
1080 *inum = option;
1081 break;
1082 case Opt_journal_dev:
1083 if (is_remount) {
1084 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1085 "journal on remount");
1086 return 0;
1087 }
1088 if (match_int(&args[0], &option))
1089 return 0;
1090 *journal_devnum = option;
1091 break;
1092 case Opt_noload:
1093 set_opt (sbi->s_mount_opt, NOLOAD);
1094 break;
1095 case Opt_commit:
1096 if (match_int(&args[0], &option))
1097 return 0;
1098 if (option < 0)
1099 return 0;
1100 if (option == 0)
1101 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1102 sbi->s_commit_interval = HZ * option;
1103 break;
1104 case Opt_data_journal:
1105 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1106 goto datacheck;
1107 case Opt_data_ordered:
1108 data_opt = EXT3_MOUNT_ORDERED_DATA;
1109 goto datacheck;
1110 case Opt_data_writeback:
1111 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1112 datacheck:
1113 if (is_remount) {
1114 if (test_opt(sb, DATA_FLAGS) == data_opt)
1115 break;
1116 ext3_msg(sb, KERN_ERR,
1117 "error: cannot change "
1118 "data mode on remount. The filesystem "
1119 "is mounted in data=%s mode and you "
1120 "try to remount it in data=%s mode.",
1121 data_mode_string(test_opt(sb,
1122 DATA_FLAGS)),
1123 data_mode_string(data_opt));
1124 return 0;
1125 } else {
1126 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1127 sbi->s_mount_opt |= data_opt;
1128 }
1129 break;
1130 case Opt_data_err_abort:
1131 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1132 break;
1133 case Opt_data_err_ignore:
1134 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1135 break;
1136 #ifdef CONFIG_QUOTA
1137 case Opt_usrjquota:
1138 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1139 return 0;
1140 break;
1141 case Opt_grpjquota:
1142 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1143 return 0;
1144 break;
1145 case Opt_offusrjquota:
1146 if (!clear_qf_name(sb, USRQUOTA))
1147 return 0;
1148 break;
1149 case Opt_offgrpjquota:
1150 if (!clear_qf_name(sb, GRPQUOTA))
1151 return 0;
1152 break;
1153 case Opt_jqfmt_vfsold:
1154 qfmt = QFMT_VFS_OLD;
1155 goto set_qf_format;
1156 case Opt_jqfmt_vfsv0:
1157 qfmt = QFMT_VFS_V0;
1158 goto set_qf_format;
1159 case Opt_jqfmt_vfsv1:
1160 qfmt = QFMT_VFS_V1;
1161 set_qf_format:
1162 if (sb_any_quota_loaded(sb) &&
1163 sbi->s_jquota_fmt != qfmt) {
1164 ext3_msg(sb, KERN_ERR, "error: cannot change "
1165 "journaled quota options when "
1166 "quota turned on.");
1167 return 0;
1168 }
1169 sbi->s_jquota_fmt = qfmt;
1170 break;
1171 case Opt_quota:
1172 case Opt_usrquota:
1173 set_opt(sbi->s_mount_opt, QUOTA);
1174 set_opt(sbi->s_mount_opt, USRQUOTA);
1175 break;
1176 case Opt_grpquota:
1177 set_opt(sbi->s_mount_opt, QUOTA);
1178 set_opt(sbi->s_mount_opt, GRPQUOTA);
1179 break;
1180 case Opt_noquota:
1181 if (sb_any_quota_loaded(sb)) {
1182 ext3_msg(sb, KERN_ERR, "error: cannot change "
1183 "quota options when quota turned on.");
1184 return 0;
1185 }
1186 clear_opt(sbi->s_mount_opt, QUOTA);
1187 clear_opt(sbi->s_mount_opt, USRQUOTA);
1188 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1189 break;
1190 #else
1191 case Opt_quota:
1192 case Opt_usrquota:
1193 case Opt_grpquota:
1194 ext3_msg(sb, KERN_ERR,
1195 "error: quota options not supported.");
1196 break;
1197 case Opt_usrjquota:
1198 case Opt_grpjquota:
1199 case Opt_offusrjquota:
1200 case Opt_offgrpjquota:
1201 case Opt_jqfmt_vfsold:
1202 case Opt_jqfmt_vfsv0:
1203 case Opt_jqfmt_vfsv1:
1204 ext3_msg(sb, KERN_ERR,
1205 "error: journaled quota options not "
1206 "supported.");
1207 break;
1208 case Opt_noquota:
1209 break;
1210 #endif
1211 case Opt_abort:
1212 set_opt(sbi->s_mount_opt, ABORT);
1213 break;
1214 case Opt_nobarrier:
1215 clear_opt(sbi->s_mount_opt, BARRIER);
1216 break;
1217 case Opt_barrier:
1218 if (args[0].from) {
1219 if (match_int(&args[0], &option))
1220 return 0;
1221 } else
1222 option = 1; /* No argument, default to 1 */
1223 if (option)
1224 set_opt(sbi->s_mount_opt, BARRIER);
1225 else
1226 clear_opt(sbi->s_mount_opt, BARRIER);
1227 break;
1228 case Opt_ignore:
1229 break;
1230 case Opt_resize:
1231 if (!is_remount) {
1232 ext3_msg(sb, KERN_ERR,
1233 "error: resize option only available "
1234 "for remount");
1235 return 0;
1236 }
1237 if (match_int(&args[0], &option) != 0)
1238 return 0;
1239 *n_blocks_count = option;
1240 break;
1241 case Opt_nobh:
1242 ext3_msg(sb, KERN_WARNING,
1243 "warning: ignoring deprecated nobh option");
1244 break;
1245 case Opt_bh:
1246 ext3_msg(sb, KERN_WARNING,
1247 "warning: ignoring deprecated bh option");
1248 break;
1249 default:
1250 ext3_msg(sb, KERN_ERR,
1251 "error: unrecognized mount option \"%s\" "
1252 "or missing value", p);
1253 return 0;
1254 }
1255 }
1256 #ifdef CONFIG_QUOTA
1257 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1258 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1259 clear_opt(sbi->s_mount_opt, USRQUOTA);
1260 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1261 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1262
1263 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1264 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1265 "format mixing.");
1266 return 0;
1267 }
1268
1269 if (!sbi->s_jquota_fmt) {
1270 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1271 "not specified.");
1272 return 0;
1273 }
1274 } else {
1275 if (sbi->s_jquota_fmt) {
1276 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1277 "specified with no journaling "
1278 "enabled.");
1279 return 0;
1280 }
1281 }
1282 #endif
1283 return 1;
1284 }
1285
1286 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1287 int read_only)
1288 {
1289 struct ext3_sb_info *sbi = EXT3_SB(sb);
1290 int res = 0;
1291
1292 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1293 ext3_msg(sb, KERN_ERR,
1294 "error: revision level too high, "
1295 "forcing read-only mode");
1296 res = MS_RDONLY;
1297 }
1298 if (read_only)
1299 return res;
1300 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1301 ext3_msg(sb, KERN_WARNING,
1302 "warning: mounting unchecked fs, "
1303 "running e2fsck is recommended");
1304 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1305 ext3_msg(sb, KERN_WARNING,
1306 "warning: mounting fs with errors, "
1307 "running e2fsck is recommended");
1308 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1309 le16_to_cpu(es->s_mnt_count) >=
1310 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1311 ext3_msg(sb, KERN_WARNING,
1312 "warning: maximal mount count reached, "
1313 "running e2fsck is recommended");
1314 else if (le32_to_cpu(es->s_checkinterval) &&
1315 (le32_to_cpu(es->s_lastcheck) +
1316 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1317 ext3_msg(sb, KERN_WARNING,
1318 "warning: checktime reached, "
1319 "running e2fsck is recommended");
1320 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1321 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1322 le16_add_cpu(&es->s_mnt_count, 1);
1323 es->s_mtime = cpu_to_le32(get_seconds());
1324 ext3_update_dynamic_rev(sb);
1325 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1326
1327 ext3_commit_super(sb, es, 1);
1328 if (test_opt(sb, DEBUG))
1329 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1330 "bpg=%lu, ipg=%lu, mo=%04lx]",
1331 sb->s_blocksize,
1332 sbi->s_groups_count,
1333 EXT3_BLOCKS_PER_GROUP(sb),
1334 EXT3_INODES_PER_GROUP(sb),
1335 sbi->s_mount_opt);
1336
1337 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1338 char b[BDEVNAME_SIZE];
1339 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1340 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1341 } else {
1342 ext3_msg(sb, KERN_INFO, "using internal journal");
1343 }
1344 return res;
1345 }
1346
1347 /* Called at mount-time, super-block is locked */
1348 static int ext3_check_descriptors(struct super_block *sb)
1349 {
1350 struct ext3_sb_info *sbi = EXT3_SB(sb);
1351 int i;
1352
1353 ext3_debug ("Checking group descriptors");
1354
1355 for (i = 0; i < sbi->s_groups_count; i++) {
1356 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1357 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1358 ext3_fsblk_t last_block;
1359
1360 if (i == sbi->s_groups_count - 1)
1361 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1362 else
1363 last_block = first_block +
1364 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1365
1366 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1367 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1368 {
1369 ext3_error (sb, "ext3_check_descriptors",
1370 "Block bitmap for group %d"
1371 " not in group (block %lu)!",
1372 i, (unsigned long)
1373 le32_to_cpu(gdp->bg_block_bitmap));
1374 return 0;
1375 }
1376 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1377 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1378 {
1379 ext3_error (sb, "ext3_check_descriptors",
1380 "Inode bitmap for group %d"
1381 " not in group (block %lu)!",
1382 i, (unsigned long)
1383 le32_to_cpu(gdp->bg_inode_bitmap));
1384 return 0;
1385 }
1386 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1387 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1388 last_block)
1389 {
1390 ext3_error (sb, "ext3_check_descriptors",
1391 "Inode table for group %d"
1392 " not in group (block %lu)!",
1393 i, (unsigned long)
1394 le32_to_cpu(gdp->bg_inode_table));
1395 return 0;
1396 }
1397 }
1398
1399 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1400 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1401 return 1;
1402 }
1403
1404
1405 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1406 * the superblock) which were deleted from all directories, but held open by
1407 * a process at the time of a crash. We walk the list and try to delete these
1408 * inodes at recovery time (only with a read-write filesystem).
1409 *
1410 * In order to keep the orphan inode chain consistent during traversal (in
1411 * case of crash during recovery), we link each inode into the superblock
1412 * orphan list_head and handle it the same way as an inode deletion during
1413 * normal operation (which journals the operations for us).
1414 *
1415 * We only do an iget() and an iput() on each inode, which is very safe if we
1416 * accidentally point at an in-use or already deleted inode. The worst that
1417 * can happen in this case is that we get a "bit already cleared" message from
1418 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1419 * e2fsck was run on this filesystem, and it must have already done the orphan
1420 * inode cleanup for us, so we can safely abort without any further action.
1421 */
1422 static void ext3_orphan_cleanup (struct super_block * sb,
1423 struct ext3_super_block * es)
1424 {
1425 unsigned int s_flags = sb->s_flags;
1426 int nr_orphans = 0, nr_truncates = 0;
1427 #ifdef CONFIG_QUOTA
1428 int i;
1429 #endif
1430 if (!es->s_last_orphan) {
1431 jbd_debug(4, "no orphan inodes to clean up\n");
1432 return;
1433 }
1434
1435 if (bdev_read_only(sb->s_bdev)) {
1436 ext3_msg(sb, KERN_ERR, "error: write access "
1437 "unavailable, skipping orphan cleanup.");
1438 return;
1439 }
1440
1441 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1442 if (es->s_last_orphan)
1443 jbd_debug(1, "Errors on filesystem, "
1444 "clearing orphan list.\n");
1445 es->s_last_orphan = 0;
1446 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1447 return;
1448 }
1449
1450 if (s_flags & MS_RDONLY) {
1451 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1452 sb->s_flags &= ~MS_RDONLY;
1453 }
1454 #ifdef CONFIG_QUOTA
1455 /* Needed for iput() to work correctly and not trash data */
1456 sb->s_flags |= MS_ACTIVE;
1457 /* Turn on quotas so that they are updated correctly */
1458 for (i = 0; i < MAXQUOTAS; i++) {
1459 if (EXT3_SB(sb)->s_qf_names[i]) {
1460 int ret = ext3_quota_on_mount(sb, i);
1461 if (ret < 0)
1462 ext3_msg(sb, KERN_ERR,
1463 "error: cannot turn on journaled "
1464 "quota: %d", ret);
1465 }
1466 }
1467 #endif
1468
1469 while (es->s_last_orphan) {
1470 struct inode *inode;
1471
1472 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1473 if (IS_ERR(inode)) {
1474 es->s_last_orphan = 0;
1475 break;
1476 }
1477
1478 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1479 dquot_initialize(inode);
1480 if (inode->i_nlink) {
1481 printk(KERN_DEBUG
1482 "%s: truncating inode %lu to %Ld bytes\n",
1483 __func__, inode->i_ino, inode->i_size);
1484 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1485 inode->i_ino, inode->i_size);
1486 ext3_truncate(inode);
1487 nr_truncates++;
1488 } else {
1489 printk(KERN_DEBUG
1490 "%s: deleting unreferenced inode %lu\n",
1491 __func__, inode->i_ino);
1492 jbd_debug(2, "deleting unreferenced inode %lu\n",
1493 inode->i_ino);
1494 nr_orphans++;
1495 }
1496 iput(inode); /* The delete magic happens here! */
1497 }
1498
1499 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1500
1501 if (nr_orphans)
1502 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1503 PLURAL(nr_orphans));
1504 if (nr_truncates)
1505 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1506 PLURAL(nr_truncates));
1507 #ifdef CONFIG_QUOTA
1508 /* Turn quotas off */
1509 for (i = 0; i < MAXQUOTAS; i++) {
1510 if (sb_dqopt(sb)->files[i])
1511 dquot_quota_off(sb, i);
1512 }
1513 #endif
1514 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1515 }
1516
1517 /*
1518 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1519 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1520 * We need to be 1 filesystem block less than the 2^32 sector limit.
1521 */
1522 static loff_t ext3_max_size(int bits)
1523 {
1524 loff_t res = EXT3_NDIR_BLOCKS;
1525 int meta_blocks;
1526 loff_t upper_limit;
1527
1528 /* This is calculated to be the largest file size for a
1529 * dense, file such that the total number of
1530 * sectors in the file, including data and all indirect blocks,
1531 * does not exceed 2^32 -1
1532 * __u32 i_blocks representing the total number of
1533 * 512 bytes blocks of the file
1534 */
1535 upper_limit = (1LL << 32) - 1;
1536
1537 /* total blocks in file system block size */
1538 upper_limit >>= (bits - 9);
1539
1540
1541 /* indirect blocks */
1542 meta_blocks = 1;
1543 /* double indirect blocks */
1544 meta_blocks += 1 + (1LL << (bits-2));
1545 /* tripple indirect blocks */
1546 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1547
1548 upper_limit -= meta_blocks;
1549 upper_limit <<= bits;
1550
1551 res += 1LL << (bits-2);
1552 res += 1LL << (2*(bits-2));
1553 res += 1LL << (3*(bits-2));
1554 res <<= bits;
1555 if (res > upper_limit)
1556 res = upper_limit;
1557
1558 if (res > MAX_LFS_FILESIZE)
1559 res = MAX_LFS_FILESIZE;
1560
1561 return res;
1562 }
1563
1564 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1565 ext3_fsblk_t logic_sb_block,
1566 int nr)
1567 {
1568 struct ext3_sb_info *sbi = EXT3_SB(sb);
1569 unsigned long bg, first_meta_bg;
1570 int has_super = 0;
1571
1572 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1573
1574 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1575 nr < first_meta_bg)
1576 return (logic_sb_block + nr + 1);
1577 bg = sbi->s_desc_per_block * nr;
1578 if (ext3_bg_has_super(sb, bg))
1579 has_super = 1;
1580 return (has_super + ext3_group_first_block_no(sb, bg));
1581 }
1582
1583
1584 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1585 {
1586 struct buffer_head * bh;
1587 struct ext3_super_block *es = NULL;
1588 struct ext3_sb_info *sbi;
1589 ext3_fsblk_t block;
1590 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1591 ext3_fsblk_t logic_sb_block;
1592 unsigned long offset = 0;
1593 unsigned int journal_inum = 0;
1594 unsigned long journal_devnum = 0;
1595 unsigned long def_mount_opts;
1596 struct inode *root;
1597 int blocksize;
1598 int hblock;
1599 int db_count;
1600 int i;
1601 int needs_recovery;
1602 int ret = -EINVAL;
1603 __le32 features;
1604 int err;
1605
1606 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1607 if (!sbi)
1608 return -ENOMEM;
1609
1610 sbi->s_blockgroup_lock =
1611 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1612 if (!sbi->s_blockgroup_lock) {
1613 kfree(sbi);
1614 return -ENOMEM;
1615 }
1616 sb->s_fs_info = sbi;
1617 sbi->s_mount_opt = 0;
1618 sbi->s_resuid = EXT3_DEF_RESUID;
1619 sbi->s_resgid = EXT3_DEF_RESGID;
1620 sbi->s_sb_block = sb_block;
1621
1622 unlock_kernel();
1623
1624 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1625 if (!blocksize) {
1626 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1627 goto out_fail;
1628 }
1629
1630 /*
1631 * The ext3 superblock will not be buffer aligned for other than 1kB
1632 * block sizes. We need to calculate the offset from buffer start.
1633 */
1634 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1635 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1636 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1637 } else {
1638 logic_sb_block = sb_block;
1639 }
1640
1641 if (!(bh = sb_bread(sb, logic_sb_block))) {
1642 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1643 goto out_fail;
1644 }
1645 /*
1646 * Note: s_es must be initialized as soon as possible because
1647 * some ext3 macro-instructions depend on its value
1648 */
1649 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1650 sbi->s_es = es;
1651 sb->s_magic = le16_to_cpu(es->s_magic);
1652 if (sb->s_magic != EXT3_SUPER_MAGIC)
1653 goto cantfind_ext3;
1654
1655 /* Set defaults before we parse the mount options */
1656 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1657 if (def_mount_opts & EXT3_DEFM_DEBUG)
1658 set_opt(sbi->s_mount_opt, DEBUG);
1659 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1660 set_opt(sbi->s_mount_opt, GRPID);
1661 if (def_mount_opts & EXT3_DEFM_UID16)
1662 set_opt(sbi->s_mount_opt, NO_UID32);
1663 #ifdef CONFIG_EXT3_FS_XATTR
1664 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1665 set_opt(sbi->s_mount_opt, XATTR_USER);
1666 #endif
1667 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1668 if (def_mount_opts & EXT3_DEFM_ACL)
1669 set_opt(sbi->s_mount_opt, POSIX_ACL);
1670 #endif
1671 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1672 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1673 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1674 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1675 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1676 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1677
1678 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1679 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1680 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1681 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1682 else
1683 set_opt(sbi->s_mount_opt, ERRORS_RO);
1684
1685 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1686 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1687
1688 set_opt(sbi->s_mount_opt, RESERVATION);
1689
1690 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1691 NULL, 0))
1692 goto failed_mount;
1693
1694 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1695 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1696
1697 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1698 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1699 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1700 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1701 ext3_msg(sb, KERN_WARNING,
1702 "warning: feature flags set on rev 0 fs, "
1703 "running e2fsck is recommended");
1704 /*
1705 * Check feature flags regardless of the revision level, since we
1706 * previously didn't change the revision level when setting the flags,
1707 * so there is a chance incompat flags are set on a rev 0 filesystem.
1708 */
1709 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1710 if (features) {
1711 ext3_msg(sb, KERN_ERR,
1712 "error: couldn't mount because of unsupported "
1713 "optional features (%x)", le32_to_cpu(features));
1714 goto failed_mount;
1715 }
1716 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1717 if (!(sb->s_flags & MS_RDONLY) && features) {
1718 ext3_msg(sb, KERN_ERR,
1719 "error: couldn't mount RDWR because of unsupported "
1720 "optional features (%x)", le32_to_cpu(features));
1721 goto failed_mount;
1722 }
1723 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1724
1725 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1726 blocksize > EXT3_MAX_BLOCK_SIZE) {
1727 ext3_msg(sb, KERN_ERR,
1728 "error: couldn't mount because of unsupported "
1729 "filesystem blocksize %d", blocksize);
1730 goto failed_mount;
1731 }
1732
1733 hblock = bdev_logical_block_size(sb->s_bdev);
1734 if (sb->s_blocksize != blocksize) {
1735 /*
1736 * Make sure the blocksize for the filesystem is larger
1737 * than the hardware sectorsize for the machine.
1738 */
1739 if (blocksize < hblock) {
1740 ext3_msg(sb, KERN_ERR,
1741 "error: fsblocksize %d too small for "
1742 "hardware sectorsize %d", blocksize, hblock);
1743 goto failed_mount;
1744 }
1745
1746 brelse (bh);
1747 if (!sb_set_blocksize(sb, blocksize)) {
1748 ext3_msg(sb, KERN_ERR,
1749 "error: bad blocksize %d", blocksize);
1750 goto out_fail;
1751 }
1752 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1753 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1754 bh = sb_bread(sb, logic_sb_block);
1755 if (!bh) {
1756 ext3_msg(sb, KERN_ERR,
1757 "error: can't read superblock on 2nd try");
1758 goto failed_mount;
1759 }
1760 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1761 sbi->s_es = es;
1762 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1763 ext3_msg(sb, KERN_ERR,
1764 "error: magic mismatch");
1765 goto failed_mount;
1766 }
1767 }
1768
1769 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1770
1771 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1772 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1773 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1774 } else {
1775 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1776 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1777 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1778 (!is_power_of_2(sbi->s_inode_size)) ||
1779 (sbi->s_inode_size > blocksize)) {
1780 ext3_msg(sb, KERN_ERR,
1781 "error: unsupported inode size: %d",
1782 sbi->s_inode_size);
1783 goto failed_mount;
1784 }
1785 }
1786 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1787 le32_to_cpu(es->s_log_frag_size);
1788 if (blocksize != sbi->s_frag_size) {
1789 ext3_msg(sb, KERN_ERR,
1790 "error: fragsize %lu != blocksize %u (unsupported)",
1791 sbi->s_frag_size, blocksize);
1792 goto failed_mount;
1793 }
1794 sbi->s_frags_per_block = 1;
1795 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1796 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1797 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1798 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1799 goto cantfind_ext3;
1800 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1801 if (sbi->s_inodes_per_block == 0)
1802 goto cantfind_ext3;
1803 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1804 sbi->s_inodes_per_block;
1805 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1806 sbi->s_sbh = bh;
1807 sbi->s_mount_state = le16_to_cpu(es->s_state);
1808 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1809 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1810 for (i=0; i < 4; i++)
1811 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1812 sbi->s_def_hash_version = es->s_def_hash_version;
1813 i = le32_to_cpu(es->s_flags);
1814 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1815 sbi->s_hash_unsigned = 3;
1816 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1817 #ifdef __CHAR_UNSIGNED__
1818 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1819 sbi->s_hash_unsigned = 3;
1820 #else
1821 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1822 #endif
1823 }
1824
1825 if (sbi->s_blocks_per_group > blocksize * 8) {
1826 ext3_msg(sb, KERN_ERR,
1827 "#blocks per group too big: %lu",
1828 sbi->s_blocks_per_group);
1829 goto failed_mount;
1830 }
1831 if (sbi->s_frags_per_group > blocksize * 8) {
1832 ext3_msg(sb, KERN_ERR,
1833 "error: #fragments per group too big: %lu",
1834 sbi->s_frags_per_group);
1835 goto failed_mount;
1836 }
1837 if (sbi->s_inodes_per_group > blocksize * 8) {
1838 ext3_msg(sb, KERN_ERR,
1839 "error: #inodes per group too big: %lu",
1840 sbi->s_inodes_per_group);
1841 goto failed_mount;
1842 }
1843
1844 if (le32_to_cpu(es->s_blocks_count) >
1845 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1846 ext3_msg(sb, KERN_ERR,
1847 "error: filesystem is too large to mount safely");
1848 if (sizeof(sector_t) < 8)
1849 ext3_msg(sb, KERN_ERR,
1850 "error: CONFIG_LBDAF not enabled");
1851 goto failed_mount;
1852 }
1853
1854 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1855 goto cantfind_ext3;
1856 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1857 le32_to_cpu(es->s_first_data_block) - 1)
1858 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1859 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1860 EXT3_DESC_PER_BLOCK(sb);
1861 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1862 GFP_KERNEL);
1863 if (sbi->s_group_desc == NULL) {
1864 ext3_msg(sb, KERN_ERR,
1865 "error: not enough memory");
1866 goto failed_mount;
1867 }
1868
1869 bgl_lock_init(sbi->s_blockgroup_lock);
1870
1871 for (i = 0; i < db_count; i++) {
1872 block = descriptor_loc(sb, logic_sb_block, i);
1873 sbi->s_group_desc[i] = sb_bread(sb, block);
1874 if (!sbi->s_group_desc[i]) {
1875 ext3_msg(sb, KERN_ERR,
1876 "error: can't read group descriptor %d", i);
1877 db_count = i;
1878 goto failed_mount2;
1879 }
1880 }
1881 if (!ext3_check_descriptors (sb)) {
1882 ext3_msg(sb, KERN_ERR,
1883 "error: group descriptors corrupted");
1884 goto failed_mount2;
1885 }
1886 sbi->s_gdb_count = db_count;
1887 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1888 spin_lock_init(&sbi->s_next_gen_lock);
1889
1890 /* per fileystem reservation list head & lock */
1891 spin_lock_init(&sbi->s_rsv_window_lock);
1892 sbi->s_rsv_window_root = RB_ROOT;
1893 /* Add a single, static dummy reservation to the start of the
1894 * reservation window list --- it gives us a placeholder for
1895 * append-at-start-of-list which makes the allocation logic
1896 * _much_ simpler. */
1897 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1898 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1899 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1900 sbi->s_rsv_window_head.rsv_goal_size = 0;
1901 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1902
1903 /*
1904 * set up enough so that it can read an inode
1905 */
1906 sb->s_op = &ext3_sops;
1907 sb->s_export_op = &ext3_export_ops;
1908 sb->s_xattr = ext3_xattr_handlers;
1909 #ifdef CONFIG_QUOTA
1910 sb->s_qcop = &ext3_qctl_operations;
1911 sb->dq_op = &ext3_quota_operations;
1912 #endif
1913 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1914 mutex_init(&sbi->s_orphan_lock);
1915 mutex_init(&sbi->s_resize_lock);
1916
1917 sb->s_root = NULL;
1918
1919 needs_recovery = (es->s_last_orphan != 0 ||
1920 EXT3_HAS_INCOMPAT_FEATURE(sb,
1921 EXT3_FEATURE_INCOMPAT_RECOVER));
1922
1923 /*
1924 * The first inode we look at is the journal inode. Don't try
1925 * root first: it may be modified in the journal!
1926 */
1927 if (!test_opt(sb, NOLOAD) &&
1928 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1929 if (ext3_load_journal(sb, es, journal_devnum))
1930 goto failed_mount2;
1931 } else if (journal_inum) {
1932 if (ext3_create_journal(sb, es, journal_inum))
1933 goto failed_mount2;
1934 } else {
1935 if (!silent)
1936 ext3_msg(sb, KERN_ERR,
1937 "error: no journal found. "
1938 "mounting ext3 over ext2?");
1939 goto failed_mount2;
1940 }
1941 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1942 ext3_count_free_blocks(sb));
1943 if (!err) {
1944 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1945 ext3_count_free_inodes(sb));
1946 }
1947 if (!err) {
1948 err = percpu_counter_init(&sbi->s_dirs_counter,
1949 ext3_count_dirs(sb));
1950 }
1951 if (err) {
1952 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1953 goto failed_mount3;
1954 }
1955
1956 /* We have now updated the journal if required, so we can
1957 * validate the data journaling mode. */
1958 switch (test_opt(sb, DATA_FLAGS)) {
1959 case 0:
1960 /* No mode set, assume a default based on the journal
1961 capabilities: ORDERED_DATA if the journal can
1962 cope, else JOURNAL_DATA */
1963 if (journal_check_available_features
1964 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1965 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1966 else
1967 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1968 break;
1969
1970 case EXT3_MOUNT_ORDERED_DATA:
1971 case EXT3_MOUNT_WRITEBACK_DATA:
1972 if (!journal_check_available_features
1973 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1974 ext3_msg(sb, KERN_ERR,
1975 "error: journal does not support "
1976 "requested data journaling mode");
1977 goto failed_mount3;
1978 }
1979 default:
1980 break;
1981 }
1982
1983 /*
1984 * The journal_load will have done any necessary log recovery,
1985 * so we can safely mount the rest of the filesystem now.
1986 */
1987
1988 root = ext3_iget(sb, EXT3_ROOT_INO);
1989 if (IS_ERR(root)) {
1990 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
1991 ret = PTR_ERR(root);
1992 goto failed_mount3;
1993 }
1994 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1995 iput(root);
1996 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1997 goto failed_mount3;
1998 }
1999 sb->s_root = d_alloc_root(root);
2000 if (!sb->s_root) {
2001 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2002 iput(root);
2003 ret = -ENOMEM;
2004 goto failed_mount3;
2005 }
2006
2007 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2008
2009 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2010 ext3_orphan_cleanup(sb, es);
2011 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2012 if (needs_recovery)
2013 ext3_msg(sb, KERN_INFO, "recovery complete");
2014 ext3_mark_recovery_complete(sb, es);
2015 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2016 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2017 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2018 "writeback");
2019
2020 lock_kernel();
2021 return 0;
2022
2023 cantfind_ext3:
2024 if (!silent)
2025 ext3_msg(sb, KERN_INFO,
2026 "error: can't find ext3 filesystem on dev %s.",
2027 sb->s_id);
2028 goto failed_mount;
2029
2030 failed_mount3:
2031 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2032 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2033 percpu_counter_destroy(&sbi->s_dirs_counter);
2034 journal_destroy(sbi->s_journal);
2035 failed_mount2:
2036 for (i = 0; i < db_count; i++)
2037 brelse(sbi->s_group_desc[i]);
2038 kfree(sbi->s_group_desc);
2039 failed_mount:
2040 #ifdef CONFIG_QUOTA
2041 for (i = 0; i < MAXQUOTAS; i++)
2042 kfree(sbi->s_qf_names[i]);
2043 #endif
2044 ext3_blkdev_remove(sbi);
2045 brelse(bh);
2046 out_fail:
2047 sb->s_fs_info = NULL;
2048 kfree(sbi->s_blockgroup_lock);
2049 kfree(sbi);
2050 lock_kernel();
2051 return ret;
2052 }
2053
2054 /*
2055 * Setup any per-fs journal parameters now. We'll do this both on
2056 * initial mount, once the journal has been initialised but before we've
2057 * done any recovery; and again on any subsequent remount.
2058 */
2059 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2060 {
2061 struct ext3_sb_info *sbi = EXT3_SB(sb);
2062
2063 if (sbi->s_commit_interval)
2064 journal->j_commit_interval = sbi->s_commit_interval;
2065 /* We could also set up an ext3-specific default for the commit
2066 * interval here, but for now we'll just fall back to the jbd
2067 * default. */
2068
2069 spin_lock(&journal->j_state_lock);
2070 if (test_opt(sb, BARRIER))
2071 journal->j_flags |= JFS_BARRIER;
2072 else
2073 journal->j_flags &= ~JFS_BARRIER;
2074 if (test_opt(sb, DATA_ERR_ABORT))
2075 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2076 else
2077 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2078 spin_unlock(&journal->j_state_lock);
2079 }
2080
2081 static journal_t *ext3_get_journal(struct super_block *sb,
2082 unsigned int journal_inum)
2083 {
2084 struct inode *journal_inode;
2085 journal_t *journal;
2086
2087 /* First, test for the existence of a valid inode on disk. Bad
2088 * things happen if we iget() an unused inode, as the subsequent
2089 * iput() will try to delete it. */
2090
2091 journal_inode = ext3_iget(sb, journal_inum);
2092 if (IS_ERR(journal_inode)) {
2093 ext3_msg(sb, KERN_ERR, "error: no journal found");
2094 return NULL;
2095 }
2096 if (!journal_inode->i_nlink) {
2097 make_bad_inode(journal_inode);
2098 iput(journal_inode);
2099 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2100 return NULL;
2101 }
2102
2103 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2104 journal_inode, journal_inode->i_size);
2105 if (!S_ISREG(journal_inode->i_mode)) {
2106 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2107 iput(journal_inode);
2108 return NULL;
2109 }
2110
2111 journal = journal_init_inode(journal_inode);
2112 if (!journal) {
2113 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2114 iput(journal_inode);
2115 return NULL;
2116 }
2117 journal->j_private = sb;
2118 ext3_init_journal_params(sb, journal);
2119 return journal;
2120 }
2121
2122 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2123 dev_t j_dev)
2124 {
2125 struct buffer_head * bh;
2126 journal_t *journal;
2127 ext3_fsblk_t start;
2128 ext3_fsblk_t len;
2129 int hblock, blocksize;
2130 ext3_fsblk_t sb_block;
2131 unsigned long offset;
2132 struct ext3_super_block * es;
2133 struct block_device *bdev;
2134
2135 bdev = ext3_blkdev_get(j_dev, sb);
2136 if (bdev == NULL)
2137 return NULL;
2138
2139 if (bd_claim(bdev, sb)) {
2140 ext3_msg(sb, KERN_ERR,
2141 "error: failed to claim external journal device");
2142 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2143 return NULL;
2144 }
2145
2146 blocksize = sb->s_blocksize;
2147 hblock = bdev_logical_block_size(bdev);
2148 if (blocksize < hblock) {
2149 ext3_msg(sb, KERN_ERR,
2150 "error: blocksize too small for journal device");
2151 goto out_bdev;
2152 }
2153
2154 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2155 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2156 set_blocksize(bdev, blocksize);
2157 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2158 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2159 "external journal");
2160 goto out_bdev;
2161 }
2162
2163 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2164 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2165 !(le32_to_cpu(es->s_feature_incompat) &
2166 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2167 ext3_msg(sb, KERN_ERR, "error: external journal has "
2168 "bad superblock");
2169 brelse(bh);
2170 goto out_bdev;
2171 }
2172
2173 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2174 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2175 brelse(bh);
2176 goto out_bdev;
2177 }
2178
2179 len = le32_to_cpu(es->s_blocks_count);
2180 start = sb_block + 1;
2181 brelse(bh); /* we're done with the superblock */
2182
2183 journal = journal_init_dev(bdev, sb->s_bdev,
2184 start, len, blocksize);
2185 if (!journal) {
2186 ext3_msg(sb, KERN_ERR,
2187 "error: failed to create device journal");
2188 goto out_bdev;
2189 }
2190 journal->j_private = sb;
2191 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2192 wait_on_buffer(journal->j_sb_buffer);
2193 if (!buffer_uptodate(journal->j_sb_buffer)) {
2194 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2195 goto out_journal;
2196 }
2197 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2198 ext3_msg(sb, KERN_ERR,
2199 "error: external journal has more than one "
2200 "user (unsupported) - %d",
2201 be32_to_cpu(journal->j_superblock->s_nr_users));
2202 goto out_journal;
2203 }
2204 EXT3_SB(sb)->journal_bdev = bdev;
2205 ext3_init_journal_params(sb, journal);
2206 return journal;
2207 out_journal:
2208 journal_destroy(journal);
2209 out_bdev:
2210 ext3_blkdev_put(bdev);
2211 return NULL;
2212 }
2213
2214 static int ext3_load_journal(struct super_block *sb,
2215 struct ext3_super_block *es,
2216 unsigned long journal_devnum)
2217 {
2218 journal_t *journal;
2219 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2220 dev_t journal_dev;
2221 int err = 0;
2222 int really_read_only;
2223
2224 if (journal_devnum &&
2225 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2226 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2227 "numbers have changed");
2228 journal_dev = new_decode_dev(journal_devnum);
2229 } else
2230 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2231
2232 really_read_only = bdev_read_only(sb->s_bdev);
2233
2234 /*
2235 * Are we loading a blank journal or performing recovery after a
2236 * crash? For recovery, we need to check in advance whether we
2237 * can get read-write access to the device.
2238 */
2239
2240 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2241 if (sb->s_flags & MS_RDONLY) {
2242 ext3_msg(sb, KERN_INFO,
2243 "recovery required on readonly filesystem");
2244 if (really_read_only) {
2245 ext3_msg(sb, KERN_ERR, "error: write access "
2246 "unavailable, cannot proceed");
2247 return -EROFS;
2248 }
2249 ext3_msg(sb, KERN_INFO,
2250 "write access will be enabled during recovery");
2251 }
2252 }
2253
2254 if (journal_inum && journal_dev) {
2255 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2256 "and inode journals");
2257 return -EINVAL;
2258 }
2259
2260 if (journal_inum) {
2261 if (!(journal = ext3_get_journal(sb, journal_inum)))
2262 return -EINVAL;
2263 } else {
2264 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2265 return -EINVAL;
2266 }
2267
2268 if (!(journal->j_flags & JFS_BARRIER))
2269 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2270
2271 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2272 err = journal_update_format(journal);
2273 if (err) {
2274 ext3_msg(sb, KERN_ERR, "error updating journal");
2275 journal_destroy(journal);
2276 return err;
2277 }
2278 }
2279
2280 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2281 err = journal_wipe(journal, !really_read_only);
2282 if (!err)
2283 err = journal_load(journal);
2284
2285 if (err) {
2286 ext3_msg(sb, KERN_ERR, "error loading journal");
2287 journal_destroy(journal);
2288 return err;
2289 }
2290
2291 EXT3_SB(sb)->s_journal = journal;
2292 ext3_clear_journal_err(sb, es);
2293
2294 if (journal_devnum &&
2295 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2296 es->s_journal_dev = cpu_to_le32(journal_devnum);
2297
2298 /* Make sure we flush the recovery flag to disk. */
2299 ext3_commit_super(sb, es, 1);
2300 }
2301
2302 return 0;
2303 }
2304
2305 static int ext3_create_journal(struct super_block *sb,
2306 struct ext3_super_block *es,
2307 unsigned int journal_inum)
2308 {
2309 journal_t *journal;
2310 int err;
2311
2312 if (sb->s_flags & MS_RDONLY) {
2313 ext3_msg(sb, KERN_ERR,
2314 "error: readonly filesystem when trying to "
2315 "create journal");
2316 return -EROFS;
2317 }
2318
2319 journal = ext3_get_journal(sb, journal_inum);
2320 if (!journal)
2321 return -EINVAL;
2322
2323 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2324 journal_inum);
2325
2326 err = journal_create(journal);
2327 if (err) {
2328 ext3_msg(sb, KERN_ERR, "error creating journal");
2329 journal_destroy(journal);
2330 return -EIO;
2331 }
2332
2333 EXT3_SB(sb)->s_journal = journal;
2334
2335 ext3_update_dynamic_rev(sb);
2336 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2337 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2338
2339 es->s_journal_inum = cpu_to_le32(journal_inum);
2340
2341 /* Make sure we flush the recovery flag to disk. */
2342 ext3_commit_super(sb, es, 1);
2343
2344 return 0;
2345 }
2346
2347 static int ext3_commit_super(struct super_block *sb,
2348 struct ext3_super_block *es,
2349 int sync)
2350 {
2351 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2352 int error = 0;
2353
2354 if (!sbh)
2355 return error;
2356 /*
2357 * If the file system is mounted read-only, don't update the
2358 * superblock write time. This avoids updating the superblock
2359 * write time when we are mounting the root file system
2360 * read/only but we need to replay the journal; at that point,
2361 * for people who are east of GMT and who make their clock
2362 * tick in localtime for Windows bug-for-bug compatibility,
2363 * the clock is set in the future, and this will cause e2fsck
2364 * to complain and force a full file system check.
2365 */
2366 if (!(sb->s_flags & MS_RDONLY))
2367 es->s_wtime = cpu_to_le32(get_seconds());
2368 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2369 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2370 BUFFER_TRACE(sbh, "marking dirty");
2371 mark_buffer_dirty(sbh);
2372 if (sync)
2373 error = sync_dirty_buffer(sbh);
2374 return error;
2375 }
2376
2377
2378 /*
2379 * Have we just finished recovery? If so, and if we are mounting (or
2380 * remounting) the filesystem readonly, then we will end up with a
2381 * consistent fs on disk. Record that fact.
2382 */
2383 static void ext3_mark_recovery_complete(struct super_block * sb,
2384 struct ext3_super_block * es)
2385 {
2386 journal_t *journal = EXT3_SB(sb)->s_journal;
2387
2388 journal_lock_updates(journal);
2389 if (journal_flush(journal) < 0)
2390 goto out;
2391
2392 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2393 sb->s_flags & MS_RDONLY) {
2394 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2395 ext3_commit_super(sb, es, 1);
2396 }
2397
2398 out:
2399 journal_unlock_updates(journal);
2400 }
2401
2402 /*
2403 * If we are mounting (or read-write remounting) a filesystem whose journal
2404 * has recorded an error from a previous lifetime, move that error to the
2405 * main filesystem now.
2406 */
2407 static void ext3_clear_journal_err(struct super_block *sb,
2408 struct ext3_super_block *es)
2409 {
2410 journal_t *journal;
2411 int j_errno;
2412 const char *errstr;
2413
2414 journal = EXT3_SB(sb)->s_journal;
2415
2416 /*
2417 * Now check for any error status which may have been recorded in the
2418 * journal by a prior ext3_error() or ext3_abort()
2419 */
2420
2421 j_errno = journal_errno(journal);
2422 if (j_errno) {
2423 char nbuf[16];
2424
2425 errstr = ext3_decode_error(sb, j_errno, nbuf);
2426 ext3_warning(sb, __func__, "Filesystem error recorded "
2427 "from previous mount: %s", errstr);
2428 ext3_warning(sb, __func__, "Marking fs in need of "
2429 "filesystem check.");
2430
2431 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2432 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2433 ext3_commit_super (sb, es, 1);
2434
2435 journal_clear_err(journal);
2436 }
2437 }
2438
2439 /*
2440 * Force the running and committing transactions to commit,
2441 * and wait on the commit.
2442 */
2443 int ext3_force_commit(struct super_block *sb)
2444 {
2445 journal_t *journal;
2446 int ret;
2447
2448 if (sb->s_flags & MS_RDONLY)
2449 return 0;
2450
2451 journal = EXT3_SB(sb)->s_journal;
2452 ret = ext3_journal_force_commit(journal);
2453 return ret;
2454 }
2455
2456 static int ext3_sync_fs(struct super_block *sb, int wait)
2457 {
2458 tid_t target;
2459
2460 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2461 if (wait)
2462 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2463 }
2464 return 0;
2465 }
2466
2467 /*
2468 * LVM calls this function before a (read-only) snapshot is created. This
2469 * gives us a chance to flush the journal completely and mark the fs clean.
2470 */
2471 static int ext3_freeze(struct super_block *sb)
2472 {
2473 int error = 0;
2474 journal_t *journal;
2475
2476 if (!(sb->s_flags & MS_RDONLY)) {
2477 journal = EXT3_SB(sb)->s_journal;
2478
2479 /* Now we set up the journal barrier. */
2480 journal_lock_updates(journal);
2481
2482 /*
2483 * We don't want to clear needs_recovery flag when we failed
2484 * to flush the journal.
2485 */
2486 error = journal_flush(journal);
2487 if (error < 0)
2488 goto out;
2489
2490 /* Journal blocked and flushed, clear needs_recovery flag. */
2491 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2492 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2493 if (error)
2494 goto out;
2495 }
2496 return 0;
2497
2498 out:
2499 journal_unlock_updates(journal);
2500 return error;
2501 }
2502
2503 /*
2504 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2505 * flag here, even though the filesystem is not technically dirty yet.
2506 */
2507 static int ext3_unfreeze(struct super_block *sb)
2508 {
2509 if (!(sb->s_flags & MS_RDONLY)) {
2510 lock_super(sb);
2511 /* Reser the needs_recovery flag before the fs is unlocked. */
2512 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2513 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2514 unlock_super(sb);
2515 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2516 }
2517 return 0;
2518 }
2519
2520 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2521 {
2522 struct ext3_super_block * es;
2523 struct ext3_sb_info *sbi = EXT3_SB(sb);
2524 ext3_fsblk_t n_blocks_count = 0;
2525 unsigned long old_sb_flags;
2526 struct ext3_mount_options old_opts;
2527 int enable_quota = 0;
2528 int err;
2529 #ifdef CONFIG_QUOTA
2530 int i;
2531 #endif
2532
2533 lock_kernel();
2534
2535 /* Store the original options */
2536 lock_super(sb);
2537 old_sb_flags = sb->s_flags;
2538 old_opts.s_mount_opt = sbi->s_mount_opt;
2539 old_opts.s_resuid = sbi->s_resuid;
2540 old_opts.s_resgid = sbi->s_resgid;
2541 old_opts.s_commit_interval = sbi->s_commit_interval;
2542 #ifdef CONFIG_QUOTA
2543 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2544 for (i = 0; i < MAXQUOTAS; i++)
2545 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2546 #endif
2547
2548 /*
2549 * Allow the "check" option to be passed as a remount option.
2550 */
2551 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2552 err = -EINVAL;
2553 goto restore_opts;
2554 }
2555
2556 if (test_opt(sb, ABORT))
2557 ext3_abort(sb, __func__, "Abort forced by user");
2558
2559 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2560 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2561
2562 es = sbi->s_es;
2563
2564 ext3_init_journal_params(sb, sbi->s_journal);
2565
2566 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2567 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2568 if (test_opt(sb, ABORT)) {
2569 err = -EROFS;
2570 goto restore_opts;
2571 }
2572
2573 if (*flags & MS_RDONLY) {
2574 err = dquot_suspend(sb, -1);
2575 if (err < 0)
2576 goto restore_opts;
2577
2578 /*
2579 * First of all, the unconditional stuff we have to do
2580 * to disable replay of the journal when we next remount
2581 */
2582 sb->s_flags |= MS_RDONLY;
2583
2584 /*
2585 * OK, test if we are remounting a valid rw partition
2586 * readonly, and if so set the rdonly flag and then
2587 * mark the partition as valid again.
2588 */
2589 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2590 (sbi->s_mount_state & EXT3_VALID_FS))
2591 es->s_state = cpu_to_le16(sbi->s_mount_state);
2592
2593 ext3_mark_recovery_complete(sb, es);
2594 } else {
2595 __le32 ret;
2596 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2597 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2598 ext3_msg(sb, KERN_WARNING,
2599 "warning: couldn't remount RDWR "
2600 "because of unsupported optional "
2601 "features (%x)", le32_to_cpu(ret));
2602 err = -EROFS;
2603 goto restore_opts;
2604 }
2605
2606 /*
2607 * If we have an unprocessed orphan list hanging
2608 * around from a previously readonly bdev mount,
2609 * require a full umount/remount for now.
2610 */
2611 if (es->s_last_orphan) {
2612 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2613 "remount RDWR because of unprocessed "
2614 "orphan inode list. Please "
2615 "umount/remount instead.");
2616 err = -EINVAL;
2617 goto restore_opts;
2618 }
2619
2620 /*
2621 * Mounting a RDONLY partition read-write, so reread
2622 * and store the current valid flag. (It may have
2623 * been changed by e2fsck since we originally mounted
2624 * the partition.)
2625 */
2626 ext3_clear_journal_err(sb, es);
2627 sbi->s_mount_state = le16_to_cpu(es->s_state);
2628 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2629 goto restore_opts;
2630 if (!ext3_setup_super (sb, es, 0))
2631 sb->s_flags &= ~MS_RDONLY;
2632 enable_quota = 1;
2633 }
2634 }
2635 #ifdef CONFIG_QUOTA
2636 /* Release old quota file names */
2637 for (i = 0; i < MAXQUOTAS; i++)
2638 if (old_opts.s_qf_names[i] &&
2639 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2640 kfree(old_opts.s_qf_names[i]);
2641 #endif
2642 unlock_super(sb);
2643 unlock_kernel();
2644
2645 if (enable_quota)
2646 dquot_resume(sb, -1);
2647 return 0;
2648 restore_opts:
2649 sb->s_flags = old_sb_flags;
2650 sbi->s_mount_opt = old_opts.s_mount_opt;
2651 sbi->s_resuid = old_opts.s_resuid;
2652 sbi->s_resgid = old_opts.s_resgid;
2653 sbi->s_commit_interval = old_opts.s_commit_interval;
2654 #ifdef CONFIG_QUOTA
2655 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2656 for (i = 0; i < MAXQUOTAS; i++) {
2657 if (sbi->s_qf_names[i] &&
2658 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2659 kfree(sbi->s_qf_names[i]);
2660 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2661 }
2662 #endif
2663 unlock_super(sb);
2664 unlock_kernel();
2665 return err;
2666 }
2667
2668 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2669 {
2670 struct super_block *sb = dentry->d_sb;
2671 struct ext3_sb_info *sbi = EXT3_SB(sb);
2672 struct ext3_super_block *es = sbi->s_es;
2673 u64 fsid;
2674
2675 if (test_opt(sb, MINIX_DF)) {
2676 sbi->s_overhead_last = 0;
2677 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2678 unsigned long ngroups = sbi->s_groups_count, i;
2679 ext3_fsblk_t overhead = 0;
2680 smp_rmb();
2681
2682 /*
2683 * Compute the overhead (FS structures). This is constant
2684 * for a given filesystem unless the number of block groups
2685 * changes so we cache the previous value until it does.
2686 */
2687
2688 /*
2689 * All of the blocks before first_data_block are
2690 * overhead
2691 */
2692 overhead = le32_to_cpu(es->s_first_data_block);
2693
2694 /*
2695 * Add the overhead attributed to the superblock and
2696 * block group descriptors. If the sparse superblocks
2697 * feature is turned on, then not all groups have this.
2698 */
2699 for (i = 0; i < ngroups; i++) {
2700 overhead += ext3_bg_has_super(sb, i) +
2701 ext3_bg_num_gdb(sb, i);
2702 cond_resched();
2703 }
2704
2705 /*
2706 * Every block group has an inode bitmap, a block
2707 * bitmap, and an inode table.
2708 */
2709 overhead += ngroups * (2 + sbi->s_itb_per_group);
2710 sbi->s_overhead_last = overhead;
2711 smp_wmb();
2712 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2713 }
2714
2715 buf->f_type = EXT3_SUPER_MAGIC;
2716 buf->f_bsize = sb->s_blocksize;
2717 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2718 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2719 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2720 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2721 buf->f_bavail = 0;
2722 buf->f_files = le32_to_cpu(es->s_inodes_count);
2723 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2724 buf->f_namelen = EXT3_NAME_LEN;
2725 fsid = le64_to_cpup((void *)es->s_uuid) ^
2726 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2727 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2728 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2729 return 0;
2730 }
2731
2732 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2733 * is locked for write. Otherwise the are possible deadlocks:
2734 * Process 1 Process 2
2735 * ext3_create() quota_sync()
2736 * journal_start() write_dquot()
2737 * dquot_initialize() down(dqio_mutex)
2738 * down(dqio_mutex) journal_start()
2739 *
2740 */
2741
2742 #ifdef CONFIG_QUOTA
2743
2744 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2745 {
2746 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2747 }
2748
2749 static int ext3_write_dquot(struct dquot *dquot)
2750 {
2751 int ret, err;
2752 handle_t *handle;
2753 struct inode *inode;
2754
2755 inode = dquot_to_inode(dquot);
2756 handle = ext3_journal_start(inode,
2757 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2758 if (IS_ERR(handle))
2759 return PTR_ERR(handle);
2760 ret = dquot_commit(dquot);
2761 err = ext3_journal_stop(handle);
2762 if (!ret)
2763 ret = err;
2764 return ret;
2765 }
2766
2767 static int ext3_acquire_dquot(struct dquot *dquot)
2768 {
2769 int ret, err;
2770 handle_t *handle;
2771
2772 handle = ext3_journal_start(dquot_to_inode(dquot),
2773 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2774 if (IS_ERR(handle))
2775 return PTR_ERR(handle);
2776 ret = dquot_acquire(dquot);
2777 err = ext3_journal_stop(handle);
2778 if (!ret)
2779 ret = err;
2780 return ret;
2781 }
2782
2783 static int ext3_release_dquot(struct dquot *dquot)
2784 {
2785 int ret, err;
2786 handle_t *handle;
2787
2788 handle = ext3_journal_start(dquot_to_inode(dquot),
2789 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2790 if (IS_ERR(handle)) {
2791 /* Release dquot anyway to avoid endless cycle in dqput() */
2792 dquot_release(dquot);
2793 return PTR_ERR(handle);
2794 }
2795 ret = dquot_release(dquot);
2796 err = ext3_journal_stop(handle);
2797 if (!ret)
2798 ret = err;
2799 return ret;
2800 }
2801
2802 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2803 {
2804 /* Are we journaling quotas? */
2805 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2806 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2807 dquot_mark_dquot_dirty(dquot);
2808 return ext3_write_dquot(dquot);
2809 } else {
2810 return dquot_mark_dquot_dirty(dquot);
2811 }
2812 }
2813
2814 static int ext3_write_info(struct super_block *sb, int type)
2815 {
2816 int ret, err;
2817 handle_t *handle;
2818
2819 /* Data block + inode block */
2820 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2821 if (IS_ERR(handle))
2822 return PTR_ERR(handle);
2823 ret = dquot_commit_info(sb, type);
2824 err = ext3_journal_stop(handle);
2825 if (!ret)
2826 ret = err;
2827 return ret;
2828 }
2829
2830 /*
2831 * Turn on quotas during mount time - we need to find
2832 * the quota file and such...
2833 */
2834 static int ext3_quota_on_mount(struct super_block *sb, int type)
2835 {
2836 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2837 EXT3_SB(sb)->s_jquota_fmt, type);
2838 }
2839
2840 /*
2841 * Standard function to be called on quota_on
2842 */
2843 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2844 char *name)
2845 {
2846 int err;
2847 struct path path;
2848
2849 if (!test_opt(sb, QUOTA))
2850 return -EINVAL;
2851
2852 err = kern_path(name, LOOKUP_FOLLOW, &path);
2853 if (err)
2854 return err;
2855
2856 /* Quotafile not on the same filesystem? */
2857 if (path.mnt->mnt_sb != sb) {
2858 path_put(&path);
2859 return -EXDEV;
2860 }
2861 /* Journaling quota? */
2862 if (EXT3_SB(sb)->s_qf_names[type]) {
2863 /* Quotafile not of fs root? */
2864 if (path.dentry->d_parent != sb->s_root)
2865 ext3_msg(sb, KERN_WARNING,
2866 "warning: Quota file not on filesystem root. "
2867 "Journaled quota will not work.");
2868 }
2869
2870 /*
2871 * When we journal data on quota file, we have to flush journal to see
2872 * all updates to the file when we bypass pagecache...
2873 */
2874 if (ext3_should_journal_data(path.dentry->d_inode)) {
2875 /*
2876 * We don't need to lock updates but journal_flush() could
2877 * otherwise be livelocked...
2878 */
2879 journal_lock_updates(EXT3_SB(sb)->s_journal);
2880 err = journal_flush(EXT3_SB(sb)->s_journal);
2881 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2882 if (err) {
2883 path_put(&path);
2884 return err;
2885 }
2886 }
2887
2888 err = dquot_quota_on_path(sb, type, format_id, &path);
2889 path_put(&path);
2890 return err;
2891 }
2892
2893 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2894 * acquiring the locks... As quota files are never truncated and quota code
2895 * itself serializes the operations (and noone else should touch the files)
2896 * we don't have to be afraid of races */
2897 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2898 size_t len, loff_t off)
2899 {
2900 struct inode *inode = sb_dqopt(sb)->files[type];
2901 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2902 int err = 0;
2903 int offset = off & (sb->s_blocksize - 1);
2904 int tocopy;
2905 size_t toread;
2906 struct buffer_head *bh;
2907 loff_t i_size = i_size_read(inode);
2908
2909 if (off > i_size)
2910 return 0;
2911 if (off+len > i_size)
2912 len = i_size-off;
2913 toread = len;
2914 while (toread > 0) {
2915 tocopy = sb->s_blocksize - offset < toread ?
2916 sb->s_blocksize - offset : toread;
2917 bh = ext3_bread(NULL, inode, blk, 0, &err);
2918 if (err)
2919 return err;
2920 if (!bh) /* A hole? */
2921 memset(data, 0, tocopy);
2922 else
2923 memcpy(data, bh->b_data+offset, tocopy);
2924 brelse(bh);
2925 offset = 0;
2926 toread -= tocopy;
2927 data += tocopy;
2928 blk++;
2929 }
2930 return len;
2931 }
2932
2933 /* Write to quotafile (we know the transaction is already started and has
2934 * enough credits) */
2935 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2936 const char *data, size_t len, loff_t off)
2937 {
2938 struct inode *inode = sb_dqopt(sb)->files[type];
2939 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2940 int err = 0;
2941 int offset = off & (sb->s_blocksize - 1);
2942 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2943 struct buffer_head *bh;
2944 handle_t *handle = journal_current_handle();
2945
2946 if (!handle) {
2947 ext3_msg(sb, KERN_WARNING,
2948 "warning: quota write (off=%llu, len=%llu)"
2949 " cancelled because transaction is not started.",
2950 (unsigned long long)off, (unsigned long long)len);
2951 return -EIO;
2952 }
2953
2954 /*
2955 * Since we account only one data block in transaction credits,
2956 * then it is impossible to cross a block boundary.
2957 */
2958 if (sb->s_blocksize - offset < len) {
2959 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2960 " cancelled because not block aligned",
2961 (unsigned long long)off, (unsigned long long)len);
2962 return -EIO;
2963 }
2964 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2965 bh = ext3_bread(handle, inode, blk, 1, &err);
2966 if (!bh)
2967 goto out;
2968 if (journal_quota) {
2969 err = ext3_journal_get_write_access(handle, bh);
2970 if (err) {
2971 brelse(bh);
2972 goto out;
2973 }
2974 }
2975 lock_buffer(bh);
2976 memcpy(bh->b_data+offset, data, len);
2977 flush_dcache_page(bh->b_page);
2978 unlock_buffer(bh);
2979 if (journal_quota)
2980 err = ext3_journal_dirty_metadata(handle, bh);
2981 else {
2982 /* Always do at least ordered writes for quotas */
2983 err = ext3_journal_dirty_data(handle, bh);
2984 mark_buffer_dirty(bh);
2985 }
2986 brelse(bh);
2987 out:
2988 if (err) {
2989 mutex_unlock(&inode->i_mutex);
2990 return err;
2991 }
2992 if (inode->i_size < off + len) {
2993 i_size_write(inode, off + len);
2994 EXT3_I(inode)->i_disksize = inode->i_size;
2995 }
2996 inode->i_version++;
2997 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2998 ext3_mark_inode_dirty(handle, inode);
2999 mutex_unlock(&inode->i_mutex);
3000 return len;
3001 }
3002
3003 #endif
3004
3005 static int ext3_get_sb(struct file_system_type *fs_type,
3006 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3007 {
3008 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
3009 }
3010
3011 static struct file_system_type ext3_fs_type = {
3012 .owner = THIS_MODULE,
3013 .name = "ext3",
3014 .get_sb = ext3_get_sb,
3015 .kill_sb = kill_block_super,
3016 .fs_flags = FS_REQUIRES_DEV,
3017 };
3018
3019 static int __init init_ext3_fs(void)
3020 {
3021 int err = init_ext3_xattr();
3022 if (err)
3023 return err;
3024 err = init_inodecache();
3025 if (err)
3026 goto out1;
3027 err = register_filesystem(&ext3_fs_type);
3028 if (err)
3029 goto out;
3030 return 0;
3031 out:
3032 destroy_inodecache();
3033 out1:
3034 exit_ext3_xattr();
3035 return err;
3036 }
3037
3038 static void __exit exit_ext3_fs(void)
3039 {
3040 unregister_filesystem(&ext3_fs_type);
3041 destroy_inodecache();
3042 exit_ext3_xattr();
3043 }
3044
3045 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3046 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3047 MODULE_LICENSE("GPL");
3048 module_init(init_ext3_fs)
3049 module_exit(exit_ext3_fs)
Tomato firmware and modifications.