Merge branch 'v6v7' into devel
[linux/fpc-iii.git] / fs / ext3 / super.c
blob85c8cc8f24732c59c52943b6b8810780c2e94428
1 /*
2 * linux/fs/ext3/super.c
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)
9 * from
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
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/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/log2.h>
40 #include <asm/uaccess.h>
42 #include "xattr.h"
43 #include "acl.h"
44 #include "namei.h"
46 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
47 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
48 #else
49 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
50 #endif
52 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
53 unsigned long journal_devnum);
54 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
55 unsigned int);
56 static int ext3_commit_super(struct super_block *sb,
57 struct ext3_super_block *es,
58 int sync);
59 static void ext3_mark_recovery_complete(struct super_block * sb,
60 struct ext3_super_block * es);
61 static void ext3_clear_journal_err(struct super_block * sb,
62 struct ext3_super_block * es);
63 static int ext3_sync_fs(struct super_block *sb, int wait);
64 static const char *ext3_decode_error(struct super_block * sb, int errno,
65 char nbuf[16]);
66 static int ext3_remount (struct super_block * sb, int * flags, char * data);
67 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
68 static int ext3_unfreeze(struct super_block *sb);
69 static int ext3_freeze(struct super_block *sb);
72 * Wrappers for journal_start/end.
74 * The only special thing we need to do here is to make sure that all
75 * journal_end calls result in the superblock being marked dirty, so
76 * that sync() will call the filesystem's write_super callback if
77 * appropriate.
79 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 journal_t *journal;
83 if (sb->s_flags & MS_RDONLY)
84 return ERR_PTR(-EROFS);
86 /* Special case here: if the journal has aborted behind our
87 * backs (eg. EIO in the commit thread), then we still need to
88 * take the FS itself readonly cleanly. */
89 journal = EXT3_SB(sb)->s_journal;
90 if (is_journal_aborted(journal)) {
91 ext3_abort(sb, __func__,
92 "Detected aborted journal");
93 return ERR_PTR(-EROFS);
96 return journal_start(journal, nblocks);
100 * The only special thing we need to do here is to make sure that all
101 * journal_stop calls result in the superblock being marked dirty, so
102 * that sync() will call the filesystem's write_super callback if
103 * appropriate.
105 int __ext3_journal_stop(const char *where, handle_t *handle)
107 struct super_block *sb;
108 int err;
109 int rc;
111 sb = handle->h_transaction->t_journal->j_private;
112 err = handle->h_err;
113 rc = journal_stop(handle);
115 if (!err)
116 err = rc;
117 if (err)
118 __ext3_std_error(sb, where, err);
119 return err;
122 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
123 struct buffer_head *bh, handle_t *handle, int err)
125 char nbuf[16];
126 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128 if (bh)
129 BUFFER_TRACE(bh, "abort");
131 if (!handle->h_err)
132 handle->h_err = err;
134 if (is_handle_aborted(handle))
135 return;
137 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
138 caller, errstr, err_fn);
140 journal_abort_handle(handle);
143 void ext3_msg(struct super_block *sb, const char *prefix,
144 const char *fmt, ...)
146 struct va_format vaf;
147 va_list args;
149 va_start(args, fmt);
151 vaf.fmt = fmt;
152 vaf.va = &args;
154 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
156 va_end(args);
159 /* Deal with the reporting of failure conditions on a filesystem such as
160 * inconsistencies detected or read IO failures.
162 * On ext2, we can store the error state of the filesystem in the
163 * superblock. That is not possible on ext3, because we may have other
164 * write ordering constraints on the superblock which prevent us from
165 * writing it out straight away; and given that the journal is about to
166 * be aborted, we can't rely on the current, or future, transactions to
167 * write out the superblock safely.
169 * We'll just use the journal_abort() error code to record an error in
170 * the journal instead. On recovery, the journal will complain about
171 * that error until we've noted it down and cleared it.
174 static void ext3_handle_error(struct super_block *sb)
176 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
178 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
179 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
181 if (sb->s_flags & MS_RDONLY)
182 return;
184 if (!test_opt (sb, ERRORS_CONT)) {
185 journal_t *journal = EXT3_SB(sb)->s_journal;
187 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
188 if (journal)
189 journal_abort(journal, -EIO);
191 if (test_opt (sb, ERRORS_RO)) {
192 ext3_msg(sb, KERN_CRIT,
193 "error: remounting filesystem read-only");
194 sb->s_flags |= MS_RDONLY;
196 ext3_commit_super(sb, es, 1);
197 if (test_opt(sb, ERRORS_PANIC))
198 panic("EXT3-fs (%s): panic forced after error\n",
199 sb->s_id);
202 void ext3_error(struct super_block *sb, const char *function,
203 const char *fmt, ...)
205 struct va_format vaf;
206 va_list args;
208 va_start(args, fmt);
210 vaf.fmt = fmt;
211 vaf.va = &args;
213 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
214 sb->s_id, function, &vaf);
216 va_end(args);
218 ext3_handle_error(sb);
221 static const char *ext3_decode_error(struct super_block * sb, int errno,
222 char nbuf[16])
224 char *errstr = NULL;
226 switch (errno) {
227 case -EIO:
228 errstr = "IO failure";
229 break;
230 case -ENOMEM:
231 errstr = "Out of memory";
232 break;
233 case -EROFS:
234 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
235 errstr = "Journal has aborted";
236 else
237 errstr = "Readonly filesystem";
238 break;
239 default:
240 /* If the caller passed in an extra buffer for unknown
241 * errors, textualise them now. Else we just return
242 * NULL. */
243 if (nbuf) {
244 /* Check for truncated error codes... */
245 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
246 errstr = nbuf;
248 break;
251 return errstr;
254 /* __ext3_std_error decodes expected errors from journaling functions
255 * automatically and invokes the appropriate error response. */
257 void __ext3_std_error (struct super_block * sb, const char * function,
258 int errno)
260 char nbuf[16];
261 const char *errstr;
263 /* Special case: if the error is EROFS, and we're not already
264 * inside a transaction, then there's really no point in logging
265 * an error. */
266 if (errno == -EROFS && journal_current_handle() == NULL &&
267 (sb->s_flags & MS_RDONLY))
268 return;
270 errstr = ext3_decode_error(sb, errno, nbuf);
271 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
273 ext3_handle_error(sb);
277 * ext3_abort is a much stronger failure handler than ext3_error. The
278 * abort function may be used to deal with unrecoverable failures such
279 * as journal IO errors or ENOMEM at a critical moment in log management.
281 * We unconditionally force the filesystem into an ABORT|READONLY state,
282 * unless the error response on the fs has been set to panic in which
283 * case we take the easy way out and panic immediately.
286 void ext3_abort(struct super_block *sb, const char *function,
287 const char *fmt, ...)
289 struct va_format vaf;
290 va_list args;
292 va_start(args, fmt);
294 vaf.fmt = fmt;
295 vaf.va = &args;
297 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
298 sb->s_id, function, &vaf);
300 va_end(args);
302 if (test_opt(sb, ERRORS_PANIC))
303 panic("EXT3-fs: panic from previous error\n");
305 if (sb->s_flags & MS_RDONLY)
306 return;
308 ext3_msg(sb, KERN_CRIT,
309 "error: remounting filesystem read-only");
310 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
311 sb->s_flags |= MS_RDONLY;
312 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
313 if (EXT3_SB(sb)->s_journal)
314 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
317 void ext3_warning(struct super_block *sb, const char *function,
318 const char *fmt, ...)
320 struct va_format vaf;
321 va_list args;
323 va_start(args, fmt);
325 vaf.fmt = fmt;
326 vaf.va = &args;
328 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
329 sb->s_id, function, &vaf);
331 va_end(args);
334 void ext3_update_dynamic_rev(struct super_block *sb)
336 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
338 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
339 return;
341 ext3_msg(sb, KERN_WARNING,
342 "warning: updating to rev %d because of "
343 "new feature flag, running e2fsck is recommended",
344 EXT3_DYNAMIC_REV);
346 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
347 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
348 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
349 /* leave es->s_feature_*compat flags alone */
350 /* es->s_uuid will be set by e2fsck if empty */
353 * The rest of the superblock fields should be zero, and if not it
354 * means they are likely already in use, so leave them alone. We
355 * can leave it up to e2fsck to clean up any inconsistencies there.
360 * Open the external journal device
362 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
364 struct block_device *bdev;
365 char b[BDEVNAME_SIZE];
367 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
368 if (IS_ERR(bdev))
369 goto fail;
370 return bdev;
372 fail:
373 ext3_msg(sb, "error: failed to open journal device %s: %ld",
374 __bdevname(dev, b), PTR_ERR(bdev));
376 return NULL;
380 * Release the journal device
382 static int ext3_blkdev_put(struct block_device *bdev)
384 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
387 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
389 struct block_device *bdev;
390 int ret = -ENODEV;
392 bdev = sbi->journal_bdev;
393 if (bdev) {
394 ret = ext3_blkdev_put(bdev);
395 sbi->journal_bdev = NULL;
397 return ret;
400 static inline struct inode *orphan_list_entry(struct list_head *l)
402 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
405 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
407 struct list_head *l;
409 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
410 le32_to_cpu(sbi->s_es->s_last_orphan));
412 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
413 list_for_each(l, &sbi->s_orphan) {
414 struct inode *inode = orphan_list_entry(l);
415 ext3_msg(sb, KERN_ERR, " "
416 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
417 inode->i_sb->s_id, inode->i_ino, inode,
418 inode->i_mode, inode->i_nlink,
419 NEXT_ORPHAN(inode));
423 static void ext3_put_super (struct super_block * sb)
425 struct ext3_sb_info *sbi = EXT3_SB(sb);
426 struct ext3_super_block *es = sbi->s_es;
427 int i, err;
429 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
430 ext3_xattr_put_super(sb);
431 err = journal_destroy(sbi->s_journal);
432 sbi->s_journal = NULL;
433 if (err < 0)
434 ext3_abort(sb, __func__, "Couldn't clean up the journal");
436 if (!(sb->s_flags & MS_RDONLY)) {
437 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
438 es->s_state = cpu_to_le16(sbi->s_mount_state);
439 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
440 mark_buffer_dirty(sbi->s_sbh);
441 ext3_commit_super(sb, es, 1);
444 for (i = 0; i < sbi->s_gdb_count; i++)
445 brelse(sbi->s_group_desc[i]);
446 kfree(sbi->s_group_desc);
447 percpu_counter_destroy(&sbi->s_freeblocks_counter);
448 percpu_counter_destroy(&sbi->s_freeinodes_counter);
449 percpu_counter_destroy(&sbi->s_dirs_counter);
450 brelse(sbi->s_sbh);
451 #ifdef CONFIG_QUOTA
452 for (i = 0; i < MAXQUOTAS; i++)
453 kfree(sbi->s_qf_names[i]);
454 #endif
456 /* Debugging code just in case the in-memory inode orphan list
457 * isn't empty. The on-disk one can be non-empty if we've
458 * detected an error and taken the fs readonly, but the
459 * in-memory list had better be clean by this point. */
460 if (!list_empty(&sbi->s_orphan))
461 dump_orphan_list(sb, sbi);
462 J_ASSERT(list_empty(&sbi->s_orphan));
464 invalidate_bdev(sb->s_bdev);
465 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
467 * Invalidate the journal device's buffers. We don't want them
468 * floating about in memory - the physical journal device may
469 * hotswapped, and it breaks the `ro-after' testing code.
471 sync_blockdev(sbi->journal_bdev);
472 invalidate_bdev(sbi->journal_bdev);
473 ext3_blkdev_remove(sbi);
475 sb->s_fs_info = NULL;
476 kfree(sbi->s_blockgroup_lock);
477 kfree(sbi);
480 static struct kmem_cache *ext3_inode_cachep;
483 * Called inside transaction, so use GFP_NOFS
485 static struct inode *ext3_alloc_inode(struct super_block *sb)
487 struct ext3_inode_info *ei;
489 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
490 if (!ei)
491 return NULL;
492 ei->i_block_alloc_info = NULL;
493 ei->vfs_inode.i_version = 1;
494 atomic_set(&ei->i_datasync_tid, 0);
495 atomic_set(&ei->i_sync_tid, 0);
496 return &ei->vfs_inode;
499 static void ext3_i_callback(struct rcu_head *head)
501 struct inode *inode = container_of(head, struct inode, i_rcu);
502 INIT_LIST_HEAD(&inode->i_dentry);
503 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
506 static void ext3_destroy_inode(struct inode *inode)
508 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
509 printk("EXT3 Inode %p: orphan list check failed!\n",
510 EXT3_I(inode));
511 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
512 EXT3_I(inode), sizeof(struct ext3_inode_info),
513 false);
514 dump_stack();
516 call_rcu(&inode->i_rcu, ext3_i_callback);
519 static void init_once(void *foo)
521 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
523 INIT_LIST_HEAD(&ei->i_orphan);
524 #ifdef CONFIG_EXT3_FS_XATTR
525 init_rwsem(&ei->xattr_sem);
526 #endif
527 mutex_init(&ei->truncate_mutex);
528 inode_init_once(&ei->vfs_inode);
531 static int init_inodecache(void)
533 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
534 sizeof(struct ext3_inode_info),
535 0, (SLAB_RECLAIM_ACCOUNT|
536 SLAB_MEM_SPREAD),
537 init_once);
538 if (ext3_inode_cachep == NULL)
539 return -ENOMEM;
540 return 0;
543 static void destroy_inodecache(void)
545 kmem_cache_destroy(ext3_inode_cachep);
548 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
550 #if defined(CONFIG_QUOTA)
551 struct ext3_sb_info *sbi = EXT3_SB(sb);
553 if (sbi->s_jquota_fmt) {
554 char *fmtname = "";
556 switch (sbi->s_jquota_fmt) {
557 case QFMT_VFS_OLD:
558 fmtname = "vfsold";
559 break;
560 case QFMT_VFS_V0:
561 fmtname = "vfsv0";
562 break;
563 case QFMT_VFS_V1:
564 fmtname = "vfsv1";
565 break;
567 seq_printf(seq, ",jqfmt=%s", fmtname);
570 if (sbi->s_qf_names[USRQUOTA])
571 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
573 if (sbi->s_qf_names[GRPQUOTA])
574 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
576 if (test_opt(sb, USRQUOTA))
577 seq_puts(seq, ",usrquota");
579 if (test_opt(sb, GRPQUOTA))
580 seq_puts(seq, ",grpquota");
581 #endif
584 static char *data_mode_string(unsigned long mode)
586 switch (mode) {
587 case EXT3_MOUNT_JOURNAL_DATA:
588 return "journal";
589 case EXT3_MOUNT_ORDERED_DATA:
590 return "ordered";
591 case EXT3_MOUNT_WRITEBACK_DATA:
592 return "writeback";
594 return "unknown";
598 * Show an option if
599 * - it's set to a non-default value OR
600 * - if the per-sb default is different from the global default
602 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
604 struct super_block *sb = vfs->mnt_sb;
605 struct ext3_sb_info *sbi = EXT3_SB(sb);
606 struct ext3_super_block *es = sbi->s_es;
607 unsigned long def_mount_opts;
609 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
611 if (sbi->s_sb_block != 1)
612 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
613 if (test_opt(sb, MINIX_DF))
614 seq_puts(seq, ",minixdf");
615 if (test_opt(sb, GRPID))
616 seq_puts(seq, ",grpid");
617 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
618 seq_puts(seq, ",nogrpid");
619 if (sbi->s_resuid != EXT3_DEF_RESUID ||
620 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
621 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
623 if (sbi->s_resgid != EXT3_DEF_RESGID ||
624 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
625 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
627 if (test_opt(sb, ERRORS_RO)) {
628 int def_errors = le16_to_cpu(es->s_errors);
630 if (def_errors == EXT3_ERRORS_PANIC ||
631 def_errors == EXT3_ERRORS_CONTINUE) {
632 seq_puts(seq, ",errors=remount-ro");
635 if (test_opt(sb, ERRORS_CONT))
636 seq_puts(seq, ",errors=continue");
637 if (test_opt(sb, ERRORS_PANIC))
638 seq_puts(seq, ",errors=panic");
639 if (test_opt(sb, NO_UID32))
640 seq_puts(seq, ",nouid32");
641 if (test_opt(sb, DEBUG))
642 seq_puts(seq, ",debug");
643 if (test_opt(sb, OLDALLOC))
644 seq_puts(seq, ",oldalloc");
645 #ifdef CONFIG_EXT3_FS_XATTR
646 if (test_opt(sb, XATTR_USER))
647 seq_puts(seq, ",user_xattr");
648 if (!test_opt(sb, XATTR_USER) &&
649 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
650 seq_puts(seq, ",nouser_xattr");
652 #endif
653 #ifdef CONFIG_EXT3_FS_POSIX_ACL
654 if (test_opt(sb, POSIX_ACL))
655 seq_puts(seq, ",acl");
656 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
657 seq_puts(seq, ",noacl");
658 #endif
659 if (!test_opt(sb, RESERVATION))
660 seq_puts(seq, ",noreservation");
661 if (sbi->s_commit_interval) {
662 seq_printf(seq, ",commit=%u",
663 (unsigned) (sbi->s_commit_interval / HZ));
667 * Always display barrier state so it's clear what the status is.
669 seq_puts(seq, ",barrier=");
670 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
671 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
672 if (test_opt(sb, DATA_ERR_ABORT))
673 seq_puts(seq, ",data_err=abort");
675 if (test_opt(sb, NOLOAD))
676 seq_puts(seq, ",norecovery");
678 ext3_show_quota_options(seq, sb);
680 return 0;
684 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
685 u64 ino, u32 generation)
687 struct inode *inode;
689 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
690 return ERR_PTR(-ESTALE);
691 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
692 return ERR_PTR(-ESTALE);
694 /* iget isn't really right if the inode is currently unallocated!!
696 * ext3_read_inode will return a bad_inode if the inode had been
697 * deleted, so we should be safe.
699 * Currently we don't know the generation for parent directory, so
700 * a generation of 0 means "accept any"
702 inode = ext3_iget(sb, ino);
703 if (IS_ERR(inode))
704 return ERR_CAST(inode);
705 if (generation && inode->i_generation != generation) {
706 iput(inode);
707 return ERR_PTR(-ESTALE);
710 return inode;
713 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
714 int fh_len, int fh_type)
716 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
717 ext3_nfs_get_inode);
720 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
721 int fh_len, int fh_type)
723 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
724 ext3_nfs_get_inode);
728 * Try to release metadata pages (indirect blocks, directories) which are
729 * mapped via the block device. Since these pages could have journal heads
730 * which would prevent try_to_free_buffers() from freeing them, we must use
731 * jbd layer's try_to_free_buffers() function to release them.
733 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
734 gfp_t wait)
736 journal_t *journal = EXT3_SB(sb)->s_journal;
738 WARN_ON(PageChecked(page));
739 if (!page_has_buffers(page))
740 return 0;
741 if (journal)
742 return journal_try_to_free_buffers(journal, page,
743 wait & ~__GFP_WAIT);
744 return try_to_free_buffers(page);
747 #ifdef CONFIG_QUOTA
748 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
749 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
751 static int ext3_write_dquot(struct dquot *dquot);
752 static int ext3_acquire_dquot(struct dquot *dquot);
753 static int ext3_release_dquot(struct dquot *dquot);
754 static int ext3_mark_dquot_dirty(struct dquot *dquot);
755 static int ext3_write_info(struct super_block *sb, int type);
756 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
757 struct path *path);
758 static int ext3_quota_on_mount(struct super_block *sb, int type);
759 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
760 size_t len, loff_t off);
761 static ssize_t ext3_quota_write(struct super_block *sb, int type,
762 const char *data, size_t len, loff_t off);
764 static const struct dquot_operations ext3_quota_operations = {
765 .write_dquot = ext3_write_dquot,
766 .acquire_dquot = ext3_acquire_dquot,
767 .release_dquot = ext3_release_dquot,
768 .mark_dirty = ext3_mark_dquot_dirty,
769 .write_info = ext3_write_info,
770 .alloc_dquot = dquot_alloc,
771 .destroy_dquot = dquot_destroy,
774 static const struct quotactl_ops ext3_qctl_operations = {
775 .quota_on = ext3_quota_on,
776 .quota_off = dquot_quota_off,
777 .quota_sync = dquot_quota_sync,
778 .get_info = dquot_get_dqinfo,
779 .set_info = dquot_set_dqinfo,
780 .get_dqblk = dquot_get_dqblk,
781 .set_dqblk = dquot_set_dqblk
783 #endif
785 static const struct super_operations ext3_sops = {
786 .alloc_inode = ext3_alloc_inode,
787 .destroy_inode = ext3_destroy_inode,
788 .write_inode = ext3_write_inode,
789 .dirty_inode = ext3_dirty_inode,
790 .evict_inode = ext3_evict_inode,
791 .put_super = ext3_put_super,
792 .sync_fs = ext3_sync_fs,
793 .freeze_fs = ext3_freeze,
794 .unfreeze_fs = ext3_unfreeze,
795 .statfs = ext3_statfs,
796 .remount_fs = ext3_remount,
797 .show_options = ext3_show_options,
798 #ifdef CONFIG_QUOTA
799 .quota_read = ext3_quota_read,
800 .quota_write = ext3_quota_write,
801 #endif
802 .bdev_try_to_free_page = bdev_try_to_free_page,
805 static const struct export_operations ext3_export_ops = {
806 .fh_to_dentry = ext3_fh_to_dentry,
807 .fh_to_parent = ext3_fh_to_parent,
808 .get_parent = ext3_get_parent,
811 enum {
812 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
813 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
814 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
815 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
816 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
817 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
819 Opt_data_err_abort, Opt_data_err_ignore,
820 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
821 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
822 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
823 Opt_resize, Opt_usrquota, Opt_grpquota
826 static const match_table_t tokens = {
827 {Opt_bsd_df, "bsddf"},
828 {Opt_minix_df, "minixdf"},
829 {Opt_grpid, "grpid"},
830 {Opt_grpid, "bsdgroups"},
831 {Opt_nogrpid, "nogrpid"},
832 {Opt_nogrpid, "sysvgroups"},
833 {Opt_resgid, "resgid=%u"},
834 {Opt_resuid, "resuid=%u"},
835 {Opt_sb, "sb=%u"},
836 {Opt_err_cont, "errors=continue"},
837 {Opt_err_panic, "errors=panic"},
838 {Opt_err_ro, "errors=remount-ro"},
839 {Opt_nouid32, "nouid32"},
840 {Opt_nocheck, "nocheck"},
841 {Opt_nocheck, "check=none"},
842 {Opt_debug, "debug"},
843 {Opt_oldalloc, "oldalloc"},
844 {Opt_orlov, "orlov"},
845 {Opt_user_xattr, "user_xattr"},
846 {Opt_nouser_xattr, "nouser_xattr"},
847 {Opt_acl, "acl"},
848 {Opt_noacl, "noacl"},
849 {Opt_reservation, "reservation"},
850 {Opt_noreservation, "noreservation"},
851 {Opt_noload, "noload"},
852 {Opt_noload, "norecovery"},
853 {Opt_nobh, "nobh"},
854 {Opt_bh, "bh"},
855 {Opt_commit, "commit=%u"},
856 {Opt_journal_update, "journal=update"},
857 {Opt_journal_inum, "journal=%u"},
858 {Opt_journal_dev, "journal_dev=%u"},
859 {Opt_abort, "abort"},
860 {Opt_data_journal, "data=journal"},
861 {Opt_data_ordered, "data=ordered"},
862 {Opt_data_writeback, "data=writeback"},
863 {Opt_data_err_abort, "data_err=abort"},
864 {Opt_data_err_ignore, "data_err=ignore"},
865 {Opt_offusrjquota, "usrjquota="},
866 {Opt_usrjquota, "usrjquota=%s"},
867 {Opt_offgrpjquota, "grpjquota="},
868 {Opt_grpjquota, "grpjquota=%s"},
869 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
870 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
871 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
872 {Opt_grpquota, "grpquota"},
873 {Opt_noquota, "noquota"},
874 {Opt_quota, "quota"},
875 {Opt_usrquota, "usrquota"},
876 {Opt_barrier, "barrier=%u"},
877 {Opt_barrier, "barrier"},
878 {Opt_nobarrier, "nobarrier"},
879 {Opt_resize, "resize"},
880 {Opt_err, NULL},
883 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
885 ext3_fsblk_t sb_block;
886 char *options = (char *) *data;
888 if (!options || strncmp(options, "sb=", 3) != 0)
889 return 1; /* Default location */
890 options += 3;
891 /*todo: use simple_strtoll with >32bit ext3 */
892 sb_block = simple_strtoul(options, &options, 0);
893 if (*options && *options != ',') {
894 ext3_msg(sb, "error: invalid sb specification: %s",
895 (char *) *data);
896 return 1;
898 if (*options == ',')
899 options++;
900 *data = (void *) options;
901 return sb_block;
904 #ifdef CONFIG_QUOTA
905 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
907 struct ext3_sb_info *sbi = EXT3_SB(sb);
908 char *qname;
910 if (sb_any_quota_loaded(sb) &&
911 !sbi->s_qf_names[qtype]) {
912 ext3_msg(sb, KERN_ERR,
913 "Cannot change journaled "
914 "quota options when quota turned on");
915 return 0;
917 qname = match_strdup(args);
918 if (!qname) {
919 ext3_msg(sb, KERN_ERR,
920 "Not enough memory for storing quotafile name");
921 return 0;
923 if (sbi->s_qf_names[qtype] &&
924 strcmp(sbi->s_qf_names[qtype], qname)) {
925 ext3_msg(sb, KERN_ERR,
926 "%s quota file already specified", QTYPE2NAME(qtype));
927 kfree(qname);
928 return 0;
930 sbi->s_qf_names[qtype] = qname;
931 if (strchr(sbi->s_qf_names[qtype], '/')) {
932 ext3_msg(sb, KERN_ERR,
933 "quotafile must be on filesystem root");
934 kfree(sbi->s_qf_names[qtype]);
935 sbi->s_qf_names[qtype] = NULL;
936 return 0;
938 set_opt(sbi->s_mount_opt, QUOTA);
939 return 1;
942 static int clear_qf_name(struct super_block *sb, int qtype) {
944 struct ext3_sb_info *sbi = EXT3_SB(sb);
946 if (sb_any_quota_loaded(sb) &&
947 sbi->s_qf_names[qtype]) {
948 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
949 " when quota turned on");
950 return 0;
953 * The space will be released later when all options are confirmed
954 * to be correct
956 sbi->s_qf_names[qtype] = NULL;
957 return 1;
959 #endif
961 static int parse_options (char *options, struct super_block *sb,
962 unsigned int *inum, unsigned long *journal_devnum,
963 ext3_fsblk_t *n_blocks_count, int is_remount)
965 struct ext3_sb_info *sbi = EXT3_SB(sb);
966 char * p;
967 substring_t args[MAX_OPT_ARGS];
968 int data_opt = 0;
969 int option;
970 #ifdef CONFIG_QUOTA
971 int qfmt;
972 #endif
974 if (!options)
975 return 1;
977 while ((p = strsep (&options, ",")) != NULL) {
978 int token;
979 if (!*p)
980 continue;
982 * Initialize args struct so we know whether arg was
983 * found; some options take optional arguments.
985 args[0].to = args[0].from = 0;
986 token = match_token(p, tokens, args);
987 switch (token) {
988 case Opt_bsd_df:
989 clear_opt (sbi->s_mount_opt, MINIX_DF);
990 break;
991 case Opt_minix_df:
992 set_opt (sbi->s_mount_opt, MINIX_DF);
993 break;
994 case Opt_grpid:
995 set_opt (sbi->s_mount_opt, GRPID);
996 break;
997 case Opt_nogrpid:
998 clear_opt (sbi->s_mount_opt, GRPID);
999 break;
1000 case Opt_resuid:
1001 if (match_int(&args[0], &option))
1002 return 0;
1003 sbi->s_resuid = option;
1004 break;
1005 case Opt_resgid:
1006 if (match_int(&args[0], &option))
1007 return 0;
1008 sbi->s_resgid = option;
1009 break;
1010 case Opt_sb:
1011 /* handled by get_sb_block() instead of here */
1012 /* *sb_block = match_int(&args[0]); */
1013 break;
1014 case Opt_err_panic:
1015 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1016 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1017 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1018 break;
1019 case Opt_err_ro:
1020 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1021 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1022 set_opt (sbi->s_mount_opt, ERRORS_RO);
1023 break;
1024 case Opt_err_cont:
1025 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1026 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1027 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1028 break;
1029 case Opt_nouid32:
1030 set_opt (sbi->s_mount_opt, NO_UID32);
1031 break;
1032 case Opt_nocheck:
1033 clear_opt (sbi->s_mount_opt, CHECK);
1034 break;
1035 case Opt_debug:
1036 set_opt (sbi->s_mount_opt, DEBUG);
1037 break;
1038 case Opt_oldalloc:
1039 set_opt (sbi->s_mount_opt, OLDALLOC);
1040 break;
1041 case Opt_orlov:
1042 clear_opt (sbi->s_mount_opt, OLDALLOC);
1043 break;
1044 #ifdef CONFIG_EXT3_FS_XATTR
1045 case Opt_user_xattr:
1046 set_opt (sbi->s_mount_opt, XATTR_USER);
1047 break;
1048 case Opt_nouser_xattr:
1049 clear_opt (sbi->s_mount_opt, XATTR_USER);
1050 break;
1051 #else
1052 case Opt_user_xattr:
1053 case Opt_nouser_xattr:
1054 ext3_msg(sb, KERN_INFO,
1055 "(no)user_xattr options not supported");
1056 break;
1057 #endif
1058 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1059 case Opt_acl:
1060 set_opt(sbi->s_mount_opt, POSIX_ACL);
1061 break;
1062 case Opt_noacl:
1063 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1064 break;
1065 #else
1066 case Opt_acl:
1067 case Opt_noacl:
1068 ext3_msg(sb, KERN_INFO,
1069 "(no)acl options not supported");
1070 break;
1071 #endif
1072 case Opt_reservation:
1073 set_opt(sbi->s_mount_opt, RESERVATION);
1074 break;
1075 case Opt_noreservation:
1076 clear_opt(sbi->s_mount_opt, RESERVATION);
1077 break;
1078 case Opt_journal_update:
1079 /* @@@ FIXME */
1080 /* Eventually we will want to be able to create
1081 a journal file here. For now, only allow the
1082 user to specify an existing inode to be the
1083 journal file. */
1084 if (is_remount) {
1085 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1086 "journal on remount");
1087 return 0;
1089 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1090 break;
1091 case Opt_journal_inum:
1092 if (is_remount) {
1093 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1094 "journal on remount");
1095 return 0;
1097 if (match_int(&args[0], &option))
1098 return 0;
1099 *inum = option;
1100 break;
1101 case Opt_journal_dev:
1102 if (is_remount) {
1103 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1104 "journal on remount");
1105 return 0;
1107 if (match_int(&args[0], &option))
1108 return 0;
1109 *journal_devnum = option;
1110 break;
1111 case Opt_noload:
1112 set_opt (sbi->s_mount_opt, NOLOAD);
1113 break;
1114 case Opt_commit:
1115 if (match_int(&args[0], &option))
1116 return 0;
1117 if (option < 0)
1118 return 0;
1119 if (option == 0)
1120 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1121 sbi->s_commit_interval = HZ * option;
1122 break;
1123 case Opt_data_journal:
1124 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1125 goto datacheck;
1126 case Opt_data_ordered:
1127 data_opt = EXT3_MOUNT_ORDERED_DATA;
1128 goto datacheck;
1129 case Opt_data_writeback:
1130 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1131 datacheck:
1132 if (is_remount) {
1133 if (test_opt(sb, DATA_FLAGS) == data_opt)
1134 break;
1135 ext3_msg(sb, KERN_ERR,
1136 "error: cannot change "
1137 "data mode on remount. The filesystem "
1138 "is mounted in data=%s mode and you "
1139 "try to remount it in data=%s mode.",
1140 data_mode_string(test_opt(sb,
1141 DATA_FLAGS)),
1142 data_mode_string(data_opt));
1143 return 0;
1144 } else {
1145 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1146 sbi->s_mount_opt |= data_opt;
1148 break;
1149 case Opt_data_err_abort:
1150 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1151 break;
1152 case Opt_data_err_ignore:
1153 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1154 break;
1155 #ifdef CONFIG_QUOTA
1156 case Opt_usrjquota:
1157 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1158 return 0;
1159 break;
1160 case Opt_grpjquota:
1161 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1162 return 0;
1163 break;
1164 case Opt_offusrjquota:
1165 if (!clear_qf_name(sb, USRQUOTA))
1166 return 0;
1167 break;
1168 case Opt_offgrpjquota:
1169 if (!clear_qf_name(sb, GRPQUOTA))
1170 return 0;
1171 break;
1172 case Opt_jqfmt_vfsold:
1173 qfmt = QFMT_VFS_OLD;
1174 goto set_qf_format;
1175 case Opt_jqfmt_vfsv0:
1176 qfmt = QFMT_VFS_V0;
1177 goto set_qf_format;
1178 case Opt_jqfmt_vfsv1:
1179 qfmt = QFMT_VFS_V1;
1180 set_qf_format:
1181 if (sb_any_quota_loaded(sb) &&
1182 sbi->s_jquota_fmt != qfmt) {
1183 ext3_msg(sb, KERN_ERR, "error: cannot change "
1184 "journaled quota options when "
1185 "quota turned on.");
1186 return 0;
1188 sbi->s_jquota_fmt = qfmt;
1189 break;
1190 case Opt_quota:
1191 case Opt_usrquota:
1192 set_opt(sbi->s_mount_opt, QUOTA);
1193 set_opt(sbi->s_mount_opt, USRQUOTA);
1194 break;
1195 case Opt_grpquota:
1196 set_opt(sbi->s_mount_opt, QUOTA);
1197 set_opt(sbi->s_mount_opt, GRPQUOTA);
1198 break;
1199 case Opt_noquota:
1200 if (sb_any_quota_loaded(sb)) {
1201 ext3_msg(sb, KERN_ERR, "error: cannot change "
1202 "quota options when quota turned on.");
1203 return 0;
1205 clear_opt(sbi->s_mount_opt, QUOTA);
1206 clear_opt(sbi->s_mount_opt, USRQUOTA);
1207 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1208 break;
1209 #else
1210 case Opt_quota:
1211 case Opt_usrquota:
1212 case Opt_grpquota:
1213 ext3_msg(sb, KERN_ERR,
1214 "error: quota options not supported.");
1215 break;
1216 case Opt_usrjquota:
1217 case Opt_grpjquota:
1218 case Opt_offusrjquota:
1219 case Opt_offgrpjquota:
1220 case Opt_jqfmt_vfsold:
1221 case Opt_jqfmt_vfsv0:
1222 case Opt_jqfmt_vfsv1:
1223 ext3_msg(sb, KERN_ERR,
1224 "error: journaled quota options not "
1225 "supported.");
1226 break;
1227 case Opt_noquota:
1228 break;
1229 #endif
1230 case Opt_abort:
1231 set_opt(sbi->s_mount_opt, ABORT);
1232 break;
1233 case Opt_nobarrier:
1234 clear_opt(sbi->s_mount_opt, BARRIER);
1235 break;
1236 case Opt_barrier:
1237 if (args[0].from) {
1238 if (match_int(&args[0], &option))
1239 return 0;
1240 } else
1241 option = 1; /* No argument, default to 1 */
1242 if (option)
1243 set_opt(sbi->s_mount_opt, BARRIER);
1244 else
1245 clear_opt(sbi->s_mount_opt, BARRIER);
1246 break;
1247 case Opt_ignore:
1248 break;
1249 case Opt_resize:
1250 if (!is_remount) {
1251 ext3_msg(sb, KERN_ERR,
1252 "error: resize option only available "
1253 "for remount");
1254 return 0;
1256 if (match_int(&args[0], &option) != 0)
1257 return 0;
1258 *n_blocks_count = option;
1259 break;
1260 case Opt_nobh:
1261 ext3_msg(sb, KERN_WARNING,
1262 "warning: ignoring deprecated nobh option");
1263 break;
1264 case Opt_bh:
1265 ext3_msg(sb, KERN_WARNING,
1266 "warning: ignoring deprecated bh option");
1267 break;
1268 default:
1269 ext3_msg(sb, KERN_ERR,
1270 "error: unrecognized mount option \"%s\" "
1271 "or missing value", p);
1272 return 0;
1275 #ifdef CONFIG_QUOTA
1276 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1277 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1278 clear_opt(sbi->s_mount_opt, USRQUOTA);
1279 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1280 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1282 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1283 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1284 "format mixing.");
1285 return 0;
1288 if (!sbi->s_jquota_fmt) {
1289 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1290 "not specified.");
1291 return 0;
1293 } else {
1294 if (sbi->s_jquota_fmt) {
1295 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1296 "specified with no journaling "
1297 "enabled.");
1298 return 0;
1301 #endif
1302 return 1;
1305 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1306 int read_only)
1308 struct ext3_sb_info *sbi = EXT3_SB(sb);
1309 int res = 0;
1311 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1312 ext3_msg(sb, KERN_ERR,
1313 "error: revision level too high, "
1314 "forcing read-only mode");
1315 res = MS_RDONLY;
1317 if (read_only)
1318 return res;
1319 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1320 ext3_msg(sb, KERN_WARNING,
1321 "warning: mounting unchecked fs, "
1322 "running e2fsck is recommended");
1323 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1324 ext3_msg(sb, KERN_WARNING,
1325 "warning: mounting fs with errors, "
1326 "running e2fsck is recommended");
1327 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1328 le16_to_cpu(es->s_mnt_count) >=
1329 le16_to_cpu(es->s_max_mnt_count))
1330 ext3_msg(sb, KERN_WARNING,
1331 "warning: maximal mount count reached, "
1332 "running e2fsck is recommended");
1333 else if (le32_to_cpu(es->s_checkinterval) &&
1334 (le32_to_cpu(es->s_lastcheck) +
1335 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1336 ext3_msg(sb, KERN_WARNING,
1337 "warning: checktime reached, "
1338 "running e2fsck is recommended");
1339 #if 0
1340 /* @@@ We _will_ want to clear the valid bit if we find
1341 inconsistencies, to force a fsck at reboot. But for
1342 a plain journaled filesystem we can keep it set as
1343 valid forever! :) */
1344 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1345 #endif
1346 if (!le16_to_cpu(es->s_max_mnt_count))
1347 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1348 le16_add_cpu(&es->s_mnt_count, 1);
1349 es->s_mtime = cpu_to_le32(get_seconds());
1350 ext3_update_dynamic_rev(sb);
1351 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1353 ext3_commit_super(sb, es, 1);
1354 if (test_opt(sb, DEBUG))
1355 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1356 "bpg=%lu, ipg=%lu, mo=%04lx]",
1357 sb->s_blocksize,
1358 sbi->s_groups_count,
1359 EXT3_BLOCKS_PER_GROUP(sb),
1360 EXT3_INODES_PER_GROUP(sb),
1361 sbi->s_mount_opt);
1363 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1364 char b[BDEVNAME_SIZE];
1365 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1366 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1367 } else {
1368 ext3_msg(sb, KERN_INFO, "using internal journal");
1370 return res;
1373 /* Called at mount-time, super-block is locked */
1374 static int ext3_check_descriptors(struct super_block *sb)
1376 struct ext3_sb_info *sbi = EXT3_SB(sb);
1377 int i;
1379 ext3_debug ("Checking group descriptors");
1381 for (i = 0; i < sbi->s_groups_count; i++) {
1382 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1383 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1384 ext3_fsblk_t last_block;
1386 if (i == sbi->s_groups_count - 1)
1387 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1388 else
1389 last_block = first_block +
1390 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1392 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1393 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1395 ext3_error (sb, "ext3_check_descriptors",
1396 "Block bitmap for group %d"
1397 " not in group (block %lu)!",
1398 i, (unsigned long)
1399 le32_to_cpu(gdp->bg_block_bitmap));
1400 return 0;
1402 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1403 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1405 ext3_error (sb, "ext3_check_descriptors",
1406 "Inode bitmap for group %d"
1407 " not in group (block %lu)!",
1408 i, (unsigned long)
1409 le32_to_cpu(gdp->bg_inode_bitmap));
1410 return 0;
1412 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1413 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1414 last_block)
1416 ext3_error (sb, "ext3_check_descriptors",
1417 "Inode table for group %d"
1418 " not in group (block %lu)!",
1419 i, (unsigned long)
1420 le32_to_cpu(gdp->bg_inode_table));
1421 return 0;
1425 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1426 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1427 return 1;
1431 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1432 * the superblock) which were deleted from all directories, but held open by
1433 * a process at the time of a crash. We walk the list and try to delete these
1434 * inodes at recovery time (only with a read-write filesystem).
1436 * In order to keep the orphan inode chain consistent during traversal (in
1437 * case of crash during recovery), we link each inode into the superblock
1438 * orphan list_head and handle it the same way as an inode deletion during
1439 * normal operation (which journals the operations for us).
1441 * We only do an iget() and an iput() on each inode, which is very safe if we
1442 * accidentally point at an in-use or already deleted inode. The worst that
1443 * can happen in this case is that we get a "bit already cleared" message from
1444 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1445 * e2fsck was run on this filesystem, and it must have already done the orphan
1446 * inode cleanup for us, so we can safely abort without any further action.
1448 static void ext3_orphan_cleanup (struct super_block * sb,
1449 struct ext3_super_block * es)
1451 unsigned int s_flags = sb->s_flags;
1452 int nr_orphans = 0, nr_truncates = 0;
1453 #ifdef CONFIG_QUOTA
1454 int i;
1455 #endif
1456 if (!es->s_last_orphan) {
1457 jbd_debug(4, "no orphan inodes to clean up\n");
1458 return;
1461 if (bdev_read_only(sb->s_bdev)) {
1462 ext3_msg(sb, KERN_ERR, "error: write access "
1463 "unavailable, skipping orphan cleanup.");
1464 return;
1467 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1468 if (es->s_last_orphan)
1469 jbd_debug(1, "Errors on filesystem, "
1470 "clearing orphan list.\n");
1471 es->s_last_orphan = 0;
1472 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1473 return;
1476 if (s_flags & MS_RDONLY) {
1477 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1478 sb->s_flags &= ~MS_RDONLY;
1480 #ifdef CONFIG_QUOTA
1481 /* Needed for iput() to work correctly and not trash data */
1482 sb->s_flags |= MS_ACTIVE;
1483 /* Turn on quotas so that they are updated correctly */
1484 for (i = 0; i < MAXQUOTAS; i++) {
1485 if (EXT3_SB(sb)->s_qf_names[i]) {
1486 int ret = ext3_quota_on_mount(sb, i);
1487 if (ret < 0)
1488 ext3_msg(sb, KERN_ERR,
1489 "error: cannot turn on journaled "
1490 "quota: %d", ret);
1493 #endif
1495 while (es->s_last_orphan) {
1496 struct inode *inode;
1498 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1499 if (IS_ERR(inode)) {
1500 es->s_last_orphan = 0;
1501 break;
1504 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1505 dquot_initialize(inode);
1506 if (inode->i_nlink) {
1507 printk(KERN_DEBUG
1508 "%s: truncating inode %lu to %Ld bytes\n",
1509 __func__, inode->i_ino, inode->i_size);
1510 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1511 inode->i_ino, inode->i_size);
1512 ext3_truncate(inode);
1513 nr_truncates++;
1514 } else {
1515 printk(KERN_DEBUG
1516 "%s: deleting unreferenced inode %lu\n",
1517 __func__, inode->i_ino);
1518 jbd_debug(2, "deleting unreferenced inode %lu\n",
1519 inode->i_ino);
1520 nr_orphans++;
1522 iput(inode); /* The delete magic happens here! */
1525 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1527 if (nr_orphans)
1528 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1529 PLURAL(nr_orphans));
1530 if (nr_truncates)
1531 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1532 PLURAL(nr_truncates));
1533 #ifdef CONFIG_QUOTA
1534 /* Turn quotas off */
1535 for (i = 0; i < MAXQUOTAS; i++) {
1536 if (sb_dqopt(sb)->files[i])
1537 dquot_quota_off(sb, i);
1539 #endif
1540 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1544 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1545 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1546 * We need to be 1 filesystem block less than the 2^32 sector limit.
1548 static loff_t ext3_max_size(int bits)
1550 loff_t res = EXT3_NDIR_BLOCKS;
1551 int meta_blocks;
1552 loff_t upper_limit;
1554 /* This is calculated to be the largest file size for a
1555 * dense, file such that the total number of
1556 * sectors in the file, including data and all indirect blocks,
1557 * does not exceed 2^32 -1
1558 * __u32 i_blocks representing the total number of
1559 * 512 bytes blocks of the file
1561 upper_limit = (1LL << 32) - 1;
1563 /* total blocks in file system block size */
1564 upper_limit >>= (bits - 9);
1567 /* indirect blocks */
1568 meta_blocks = 1;
1569 /* double indirect blocks */
1570 meta_blocks += 1 + (1LL << (bits-2));
1571 /* tripple indirect blocks */
1572 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1574 upper_limit -= meta_blocks;
1575 upper_limit <<= bits;
1577 res += 1LL << (bits-2);
1578 res += 1LL << (2*(bits-2));
1579 res += 1LL << (3*(bits-2));
1580 res <<= bits;
1581 if (res > upper_limit)
1582 res = upper_limit;
1584 if (res > MAX_LFS_FILESIZE)
1585 res = MAX_LFS_FILESIZE;
1587 return res;
1590 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1591 ext3_fsblk_t logic_sb_block,
1592 int nr)
1594 struct ext3_sb_info *sbi = EXT3_SB(sb);
1595 unsigned long bg, first_meta_bg;
1596 int has_super = 0;
1598 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1600 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1601 nr < first_meta_bg)
1602 return (logic_sb_block + nr + 1);
1603 bg = sbi->s_desc_per_block * nr;
1604 if (ext3_bg_has_super(sb, bg))
1605 has_super = 1;
1606 return (has_super + ext3_group_first_block_no(sb, bg));
1610 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1612 struct buffer_head * bh;
1613 struct ext3_super_block *es = NULL;
1614 struct ext3_sb_info *sbi;
1615 ext3_fsblk_t block;
1616 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1617 ext3_fsblk_t logic_sb_block;
1618 unsigned long offset = 0;
1619 unsigned int journal_inum = 0;
1620 unsigned long journal_devnum = 0;
1621 unsigned long def_mount_opts;
1622 struct inode *root;
1623 int blocksize;
1624 int hblock;
1625 int db_count;
1626 int i;
1627 int needs_recovery;
1628 int ret = -EINVAL;
1629 __le32 features;
1630 int err;
1632 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1633 if (!sbi)
1634 return -ENOMEM;
1636 sbi->s_blockgroup_lock =
1637 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1638 if (!sbi->s_blockgroup_lock) {
1639 kfree(sbi);
1640 return -ENOMEM;
1642 sb->s_fs_info = sbi;
1643 sbi->s_mount_opt = 0;
1644 sbi->s_resuid = EXT3_DEF_RESUID;
1645 sbi->s_resgid = EXT3_DEF_RESGID;
1646 sbi->s_sb_block = sb_block;
1648 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1649 if (!blocksize) {
1650 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1651 goto out_fail;
1655 * The ext3 superblock will not be buffer aligned for other than 1kB
1656 * block sizes. We need to calculate the offset from buffer start.
1658 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1659 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1660 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1661 } else {
1662 logic_sb_block = sb_block;
1665 if (!(bh = sb_bread(sb, logic_sb_block))) {
1666 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1667 goto out_fail;
1670 * Note: s_es must be initialized as soon as possible because
1671 * some ext3 macro-instructions depend on its value
1673 es = (struct ext3_super_block *) (bh->b_data + offset);
1674 sbi->s_es = es;
1675 sb->s_magic = le16_to_cpu(es->s_magic);
1676 if (sb->s_magic != EXT3_SUPER_MAGIC)
1677 goto cantfind_ext3;
1679 /* Set defaults before we parse the mount options */
1680 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1681 if (def_mount_opts & EXT3_DEFM_DEBUG)
1682 set_opt(sbi->s_mount_opt, DEBUG);
1683 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1684 set_opt(sbi->s_mount_opt, GRPID);
1685 if (def_mount_opts & EXT3_DEFM_UID16)
1686 set_opt(sbi->s_mount_opt, NO_UID32);
1687 #ifdef CONFIG_EXT3_FS_XATTR
1688 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1689 set_opt(sbi->s_mount_opt, XATTR_USER);
1690 #endif
1691 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1692 if (def_mount_opts & EXT3_DEFM_ACL)
1693 set_opt(sbi->s_mount_opt, POSIX_ACL);
1694 #endif
1695 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1696 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1697 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1698 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1699 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1700 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1702 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1703 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1704 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1705 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1706 else
1707 set_opt(sbi->s_mount_opt, ERRORS_RO);
1709 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1710 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1712 set_opt(sbi->s_mount_opt, RESERVATION);
1714 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1715 NULL, 0))
1716 goto failed_mount;
1718 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1719 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1721 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1722 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1723 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1724 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1725 ext3_msg(sb, KERN_WARNING,
1726 "warning: feature flags set on rev 0 fs, "
1727 "running e2fsck is recommended");
1729 * Check feature flags regardless of the revision level, since we
1730 * previously didn't change the revision level when setting the flags,
1731 * so there is a chance incompat flags are set on a rev 0 filesystem.
1733 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1734 if (features) {
1735 ext3_msg(sb, KERN_ERR,
1736 "error: couldn't mount because of unsupported "
1737 "optional features (%x)", le32_to_cpu(features));
1738 goto failed_mount;
1740 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1741 if (!(sb->s_flags & MS_RDONLY) && features) {
1742 ext3_msg(sb, KERN_ERR,
1743 "error: couldn't mount RDWR because of unsupported "
1744 "optional features (%x)", le32_to_cpu(features));
1745 goto failed_mount;
1747 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1749 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1750 blocksize > EXT3_MAX_BLOCK_SIZE) {
1751 ext3_msg(sb, KERN_ERR,
1752 "error: couldn't mount because of unsupported "
1753 "filesystem blocksize %d", blocksize);
1754 goto failed_mount;
1757 hblock = bdev_logical_block_size(sb->s_bdev);
1758 if (sb->s_blocksize != blocksize) {
1760 * Make sure the blocksize for the filesystem is larger
1761 * than the hardware sectorsize for the machine.
1763 if (blocksize < hblock) {
1764 ext3_msg(sb, KERN_ERR,
1765 "error: fsblocksize %d too small for "
1766 "hardware sectorsize %d", blocksize, hblock);
1767 goto failed_mount;
1770 brelse (bh);
1771 if (!sb_set_blocksize(sb, blocksize)) {
1772 ext3_msg(sb, KERN_ERR,
1773 "error: bad blocksize %d", blocksize);
1774 goto out_fail;
1776 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1777 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1778 bh = sb_bread(sb, logic_sb_block);
1779 if (!bh) {
1780 ext3_msg(sb, KERN_ERR,
1781 "error: can't read superblock on 2nd try");
1782 goto failed_mount;
1784 es = (struct ext3_super_block *)(bh->b_data + offset);
1785 sbi->s_es = es;
1786 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1787 ext3_msg(sb, KERN_ERR,
1788 "error: magic mismatch");
1789 goto failed_mount;
1793 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1795 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1796 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1797 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1798 } else {
1799 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1800 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1801 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1802 (!is_power_of_2(sbi->s_inode_size)) ||
1803 (sbi->s_inode_size > blocksize)) {
1804 ext3_msg(sb, KERN_ERR,
1805 "error: unsupported inode size: %d",
1806 sbi->s_inode_size);
1807 goto failed_mount;
1810 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1811 le32_to_cpu(es->s_log_frag_size);
1812 if (blocksize != sbi->s_frag_size) {
1813 ext3_msg(sb, KERN_ERR,
1814 "error: fragsize %lu != blocksize %u (unsupported)",
1815 sbi->s_frag_size, blocksize);
1816 goto failed_mount;
1818 sbi->s_frags_per_block = 1;
1819 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1820 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1821 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1822 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1823 goto cantfind_ext3;
1824 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1825 if (sbi->s_inodes_per_block == 0)
1826 goto cantfind_ext3;
1827 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1828 sbi->s_inodes_per_block;
1829 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1830 sbi->s_sbh = bh;
1831 sbi->s_mount_state = le16_to_cpu(es->s_state);
1832 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1833 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1834 for (i=0; i < 4; i++)
1835 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1836 sbi->s_def_hash_version = es->s_def_hash_version;
1837 i = le32_to_cpu(es->s_flags);
1838 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1839 sbi->s_hash_unsigned = 3;
1840 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1841 #ifdef __CHAR_UNSIGNED__
1842 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1843 sbi->s_hash_unsigned = 3;
1844 #else
1845 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1846 #endif
1849 if (sbi->s_blocks_per_group > blocksize * 8) {
1850 ext3_msg(sb, KERN_ERR,
1851 "#blocks per group too big: %lu",
1852 sbi->s_blocks_per_group);
1853 goto failed_mount;
1855 if (sbi->s_frags_per_group > blocksize * 8) {
1856 ext3_msg(sb, KERN_ERR,
1857 "error: #fragments per group too big: %lu",
1858 sbi->s_frags_per_group);
1859 goto failed_mount;
1861 if (sbi->s_inodes_per_group > blocksize * 8) {
1862 ext3_msg(sb, KERN_ERR,
1863 "error: #inodes per group too big: %lu",
1864 sbi->s_inodes_per_group);
1865 goto failed_mount;
1868 err = generic_check_addressable(sb->s_blocksize_bits,
1869 le32_to_cpu(es->s_blocks_count));
1870 if (err) {
1871 ext3_msg(sb, KERN_ERR,
1872 "error: filesystem is too large to mount safely");
1873 if (sizeof(sector_t) < 8)
1874 ext3_msg(sb, KERN_ERR,
1875 "error: CONFIG_LBDAF not enabled");
1876 ret = err;
1877 goto failed_mount;
1880 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1881 goto cantfind_ext3;
1882 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1883 le32_to_cpu(es->s_first_data_block) - 1)
1884 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1885 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1886 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1887 GFP_KERNEL);
1888 if (sbi->s_group_desc == NULL) {
1889 ext3_msg(sb, KERN_ERR,
1890 "error: not enough memory");
1891 ret = -ENOMEM;
1892 goto failed_mount;
1895 bgl_lock_init(sbi->s_blockgroup_lock);
1897 for (i = 0; i < db_count; i++) {
1898 block = descriptor_loc(sb, logic_sb_block, i);
1899 sbi->s_group_desc[i] = sb_bread(sb, block);
1900 if (!sbi->s_group_desc[i]) {
1901 ext3_msg(sb, KERN_ERR,
1902 "error: can't read group descriptor %d", i);
1903 db_count = i;
1904 goto failed_mount2;
1907 if (!ext3_check_descriptors (sb)) {
1908 ext3_msg(sb, KERN_ERR,
1909 "error: group descriptors corrupted");
1910 goto failed_mount2;
1912 sbi->s_gdb_count = db_count;
1913 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1914 spin_lock_init(&sbi->s_next_gen_lock);
1916 /* per fileystem reservation list head & lock */
1917 spin_lock_init(&sbi->s_rsv_window_lock);
1918 sbi->s_rsv_window_root = RB_ROOT;
1919 /* Add a single, static dummy reservation to the start of the
1920 * reservation window list --- it gives us a placeholder for
1921 * append-at-start-of-list which makes the allocation logic
1922 * _much_ simpler. */
1923 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1924 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1925 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1926 sbi->s_rsv_window_head.rsv_goal_size = 0;
1927 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1930 * set up enough so that it can read an inode
1932 sb->s_op = &ext3_sops;
1933 sb->s_export_op = &ext3_export_ops;
1934 sb->s_xattr = ext3_xattr_handlers;
1935 #ifdef CONFIG_QUOTA
1936 sb->s_qcop = &ext3_qctl_operations;
1937 sb->dq_op = &ext3_quota_operations;
1938 #endif
1939 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1940 mutex_init(&sbi->s_orphan_lock);
1941 mutex_init(&sbi->s_resize_lock);
1943 sb->s_root = NULL;
1945 needs_recovery = (es->s_last_orphan != 0 ||
1946 EXT3_HAS_INCOMPAT_FEATURE(sb,
1947 EXT3_FEATURE_INCOMPAT_RECOVER));
1950 * The first inode we look at is the journal inode. Don't try
1951 * root first: it may be modified in the journal!
1953 if (!test_opt(sb, NOLOAD) &&
1954 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1955 if (ext3_load_journal(sb, es, journal_devnum))
1956 goto failed_mount2;
1957 } else if (journal_inum) {
1958 if (ext3_create_journal(sb, es, journal_inum))
1959 goto failed_mount2;
1960 } else {
1961 if (!silent)
1962 ext3_msg(sb, KERN_ERR,
1963 "error: no journal found. "
1964 "mounting ext3 over ext2?");
1965 goto failed_mount2;
1967 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1968 ext3_count_free_blocks(sb));
1969 if (!err) {
1970 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1971 ext3_count_free_inodes(sb));
1973 if (!err) {
1974 err = percpu_counter_init(&sbi->s_dirs_counter,
1975 ext3_count_dirs(sb));
1977 if (err) {
1978 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1979 ret = err;
1980 goto failed_mount3;
1983 /* We have now updated the journal if required, so we can
1984 * validate the data journaling mode. */
1985 switch (test_opt(sb, DATA_FLAGS)) {
1986 case 0:
1987 /* No mode set, assume a default based on the journal
1988 capabilities: ORDERED_DATA if the journal can
1989 cope, else JOURNAL_DATA */
1990 if (journal_check_available_features
1991 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1992 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1993 else
1994 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1995 break;
1997 case EXT3_MOUNT_ORDERED_DATA:
1998 case EXT3_MOUNT_WRITEBACK_DATA:
1999 if (!journal_check_available_features
2000 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2001 ext3_msg(sb, KERN_ERR,
2002 "error: journal does not support "
2003 "requested data journaling mode");
2004 goto failed_mount3;
2006 default:
2007 break;
2011 * The journal_load will have done any necessary log recovery,
2012 * so we can safely mount the rest of the filesystem now.
2015 root = ext3_iget(sb, EXT3_ROOT_INO);
2016 if (IS_ERR(root)) {
2017 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2018 ret = PTR_ERR(root);
2019 goto failed_mount3;
2021 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2022 iput(root);
2023 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2024 goto failed_mount3;
2026 sb->s_root = d_alloc_root(root);
2027 if (!sb->s_root) {
2028 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2029 iput(root);
2030 ret = -ENOMEM;
2031 goto failed_mount3;
2034 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2036 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2037 ext3_orphan_cleanup(sb, es);
2038 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2039 if (needs_recovery)
2040 ext3_msg(sb, KERN_INFO, "recovery complete");
2041 ext3_mark_recovery_complete(sb, es);
2042 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2043 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2044 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2045 "writeback");
2047 return 0;
2049 cantfind_ext3:
2050 if (!silent)
2051 ext3_msg(sb, KERN_INFO,
2052 "error: can't find ext3 filesystem on dev %s.",
2053 sb->s_id);
2054 goto failed_mount;
2056 failed_mount3:
2057 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2058 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2059 percpu_counter_destroy(&sbi->s_dirs_counter);
2060 journal_destroy(sbi->s_journal);
2061 failed_mount2:
2062 for (i = 0; i < db_count; i++)
2063 brelse(sbi->s_group_desc[i]);
2064 kfree(sbi->s_group_desc);
2065 failed_mount:
2066 #ifdef CONFIG_QUOTA
2067 for (i = 0; i < MAXQUOTAS; i++)
2068 kfree(sbi->s_qf_names[i]);
2069 #endif
2070 ext3_blkdev_remove(sbi);
2071 brelse(bh);
2072 out_fail:
2073 sb->s_fs_info = NULL;
2074 kfree(sbi->s_blockgroup_lock);
2075 kfree(sbi);
2076 return ret;
2080 * Setup any per-fs journal parameters now. We'll do this both on
2081 * initial mount, once the journal has been initialised but before we've
2082 * done any recovery; and again on any subsequent remount.
2084 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2086 struct ext3_sb_info *sbi = EXT3_SB(sb);
2088 if (sbi->s_commit_interval)
2089 journal->j_commit_interval = sbi->s_commit_interval;
2090 /* We could also set up an ext3-specific default for the commit
2091 * interval here, but for now we'll just fall back to the jbd
2092 * default. */
2094 spin_lock(&journal->j_state_lock);
2095 if (test_opt(sb, BARRIER))
2096 journal->j_flags |= JFS_BARRIER;
2097 else
2098 journal->j_flags &= ~JFS_BARRIER;
2099 if (test_opt(sb, DATA_ERR_ABORT))
2100 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2101 else
2102 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2103 spin_unlock(&journal->j_state_lock);
2106 static journal_t *ext3_get_journal(struct super_block *sb,
2107 unsigned int journal_inum)
2109 struct inode *journal_inode;
2110 journal_t *journal;
2112 /* First, test for the existence of a valid inode on disk. Bad
2113 * things happen if we iget() an unused inode, as the subsequent
2114 * iput() will try to delete it. */
2116 journal_inode = ext3_iget(sb, journal_inum);
2117 if (IS_ERR(journal_inode)) {
2118 ext3_msg(sb, KERN_ERR, "error: no journal found");
2119 return NULL;
2121 if (!journal_inode->i_nlink) {
2122 make_bad_inode(journal_inode);
2123 iput(journal_inode);
2124 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2125 return NULL;
2128 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2129 journal_inode, journal_inode->i_size);
2130 if (!S_ISREG(journal_inode->i_mode)) {
2131 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2132 iput(journal_inode);
2133 return NULL;
2136 journal = journal_init_inode(journal_inode);
2137 if (!journal) {
2138 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2139 iput(journal_inode);
2140 return NULL;
2142 journal->j_private = sb;
2143 ext3_init_journal_params(sb, journal);
2144 return journal;
2147 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2148 dev_t j_dev)
2150 struct buffer_head * bh;
2151 journal_t *journal;
2152 ext3_fsblk_t start;
2153 ext3_fsblk_t len;
2154 int hblock, blocksize;
2155 ext3_fsblk_t sb_block;
2156 unsigned long offset;
2157 struct ext3_super_block * es;
2158 struct block_device *bdev;
2160 bdev = ext3_blkdev_get(j_dev, sb);
2161 if (bdev == NULL)
2162 return NULL;
2164 blocksize = sb->s_blocksize;
2165 hblock = bdev_logical_block_size(bdev);
2166 if (blocksize < hblock) {
2167 ext3_msg(sb, KERN_ERR,
2168 "error: blocksize too small for journal device");
2169 goto out_bdev;
2172 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2173 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2174 set_blocksize(bdev, blocksize);
2175 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2176 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2177 "external journal");
2178 goto out_bdev;
2181 es = (struct ext3_super_block *) (bh->b_data + offset);
2182 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2183 !(le32_to_cpu(es->s_feature_incompat) &
2184 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2185 ext3_msg(sb, KERN_ERR, "error: external journal has "
2186 "bad superblock");
2187 brelse(bh);
2188 goto out_bdev;
2191 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2192 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2193 brelse(bh);
2194 goto out_bdev;
2197 len = le32_to_cpu(es->s_blocks_count);
2198 start = sb_block + 1;
2199 brelse(bh); /* we're done with the superblock */
2201 journal = journal_init_dev(bdev, sb->s_bdev,
2202 start, len, blocksize);
2203 if (!journal) {
2204 ext3_msg(sb, KERN_ERR,
2205 "error: failed to create device journal");
2206 goto out_bdev;
2208 journal->j_private = sb;
2209 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2210 wait_on_buffer(journal->j_sb_buffer);
2211 if (!buffer_uptodate(journal->j_sb_buffer)) {
2212 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2213 goto out_journal;
2215 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2216 ext3_msg(sb, KERN_ERR,
2217 "error: external journal has more than one "
2218 "user (unsupported) - %d",
2219 be32_to_cpu(journal->j_superblock->s_nr_users));
2220 goto out_journal;
2222 EXT3_SB(sb)->journal_bdev = bdev;
2223 ext3_init_journal_params(sb, journal);
2224 return journal;
2225 out_journal:
2226 journal_destroy(journal);
2227 out_bdev:
2228 ext3_blkdev_put(bdev);
2229 return NULL;
2232 static int ext3_load_journal(struct super_block *sb,
2233 struct ext3_super_block *es,
2234 unsigned long journal_devnum)
2236 journal_t *journal;
2237 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2238 dev_t journal_dev;
2239 int err = 0;
2240 int really_read_only;
2242 if (journal_devnum &&
2243 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2244 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2245 "numbers have changed");
2246 journal_dev = new_decode_dev(journal_devnum);
2247 } else
2248 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2250 really_read_only = bdev_read_only(sb->s_bdev);
2253 * Are we loading a blank journal or performing recovery after a
2254 * crash? For recovery, we need to check in advance whether we
2255 * can get read-write access to the device.
2258 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2259 if (sb->s_flags & MS_RDONLY) {
2260 ext3_msg(sb, KERN_INFO,
2261 "recovery required on readonly filesystem");
2262 if (really_read_only) {
2263 ext3_msg(sb, KERN_ERR, "error: write access "
2264 "unavailable, cannot proceed");
2265 return -EROFS;
2267 ext3_msg(sb, KERN_INFO,
2268 "write access will be enabled during recovery");
2272 if (journal_inum && journal_dev) {
2273 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2274 "and inode journals");
2275 return -EINVAL;
2278 if (journal_inum) {
2279 if (!(journal = ext3_get_journal(sb, journal_inum)))
2280 return -EINVAL;
2281 } else {
2282 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2283 return -EINVAL;
2286 if (!(journal->j_flags & JFS_BARRIER))
2287 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2289 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2290 err = journal_update_format(journal);
2291 if (err) {
2292 ext3_msg(sb, KERN_ERR, "error updating journal");
2293 journal_destroy(journal);
2294 return err;
2298 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2299 err = journal_wipe(journal, !really_read_only);
2300 if (!err)
2301 err = journal_load(journal);
2303 if (err) {
2304 ext3_msg(sb, KERN_ERR, "error loading journal");
2305 journal_destroy(journal);
2306 return err;
2309 EXT3_SB(sb)->s_journal = journal;
2310 ext3_clear_journal_err(sb, es);
2312 if (!really_read_only && journal_devnum &&
2313 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2314 es->s_journal_dev = cpu_to_le32(journal_devnum);
2316 /* Make sure we flush the recovery flag to disk. */
2317 ext3_commit_super(sb, es, 1);
2320 return 0;
2323 static int ext3_create_journal(struct super_block *sb,
2324 struct ext3_super_block *es,
2325 unsigned int journal_inum)
2327 journal_t *journal;
2328 int err;
2330 if (sb->s_flags & MS_RDONLY) {
2331 ext3_msg(sb, KERN_ERR,
2332 "error: readonly filesystem when trying to "
2333 "create journal");
2334 return -EROFS;
2337 journal = ext3_get_journal(sb, journal_inum);
2338 if (!journal)
2339 return -EINVAL;
2341 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2342 journal_inum);
2344 err = journal_create(journal);
2345 if (err) {
2346 ext3_msg(sb, KERN_ERR, "error creating journal");
2347 journal_destroy(journal);
2348 return -EIO;
2351 EXT3_SB(sb)->s_journal = journal;
2353 ext3_update_dynamic_rev(sb);
2354 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2355 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2357 es->s_journal_inum = cpu_to_le32(journal_inum);
2359 /* Make sure we flush the recovery flag to disk. */
2360 ext3_commit_super(sb, es, 1);
2362 return 0;
2365 static int ext3_commit_super(struct super_block *sb,
2366 struct ext3_super_block *es,
2367 int sync)
2369 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2370 int error = 0;
2372 if (!sbh)
2373 return error;
2375 if (buffer_write_io_error(sbh)) {
2377 * Oh, dear. A previous attempt to write the
2378 * superblock failed. This could happen because the
2379 * USB device was yanked out. Or it could happen to
2380 * be a transient write error and maybe the block will
2381 * be remapped. Nothing we can do but to retry the
2382 * write and hope for the best.
2384 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2385 "superblock detected");
2386 clear_buffer_write_io_error(sbh);
2387 set_buffer_uptodate(sbh);
2390 * If the file system is mounted read-only, don't update the
2391 * superblock write time. This avoids updating the superblock
2392 * write time when we are mounting the root file system
2393 * read/only but we need to replay the journal; at that point,
2394 * for people who are east of GMT and who make their clock
2395 * tick in localtime for Windows bug-for-bug compatibility,
2396 * the clock is set in the future, and this will cause e2fsck
2397 * to complain and force a full file system check.
2399 if (!(sb->s_flags & MS_RDONLY))
2400 es->s_wtime = cpu_to_le32(get_seconds());
2401 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2402 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2403 BUFFER_TRACE(sbh, "marking dirty");
2404 mark_buffer_dirty(sbh);
2405 if (sync) {
2406 error = sync_dirty_buffer(sbh);
2407 if (buffer_write_io_error(sbh)) {
2408 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2409 "superblock");
2410 clear_buffer_write_io_error(sbh);
2411 set_buffer_uptodate(sbh);
2414 return error;
2419 * Have we just finished recovery? If so, and if we are mounting (or
2420 * remounting) the filesystem readonly, then we will end up with a
2421 * consistent fs on disk. Record that fact.
2423 static void ext3_mark_recovery_complete(struct super_block * sb,
2424 struct ext3_super_block * es)
2426 journal_t *journal = EXT3_SB(sb)->s_journal;
2428 journal_lock_updates(journal);
2429 if (journal_flush(journal) < 0)
2430 goto out;
2432 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2433 sb->s_flags & MS_RDONLY) {
2434 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2435 ext3_commit_super(sb, es, 1);
2438 out:
2439 journal_unlock_updates(journal);
2443 * If we are mounting (or read-write remounting) a filesystem whose journal
2444 * has recorded an error from a previous lifetime, move that error to the
2445 * main filesystem now.
2447 static void ext3_clear_journal_err(struct super_block *sb,
2448 struct ext3_super_block *es)
2450 journal_t *journal;
2451 int j_errno;
2452 const char *errstr;
2454 journal = EXT3_SB(sb)->s_journal;
2457 * Now check for any error status which may have been recorded in the
2458 * journal by a prior ext3_error() or ext3_abort()
2461 j_errno = journal_errno(journal);
2462 if (j_errno) {
2463 char nbuf[16];
2465 errstr = ext3_decode_error(sb, j_errno, nbuf);
2466 ext3_warning(sb, __func__, "Filesystem error recorded "
2467 "from previous mount: %s", errstr);
2468 ext3_warning(sb, __func__, "Marking fs in need of "
2469 "filesystem check.");
2471 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2472 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2473 ext3_commit_super (sb, es, 1);
2475 journal_clear_err(journal);
2480 * Force the running and committing transactions to commit,
2481 * and wait on the commit.
2483 int ext3_force_commit(struct super_block *sb)
2485 journal_t *journal;
2486 int ret;
2488 if (sb->s_flags & MS_RDONLY)
2489 return 0;
2491 journal = EXT3_SB(sb)->s_journal;
2492 ret = ext3_journal_force_commit(journal);
2493 return ret;
2496 static int ext3_sync_fs(struct super_block *sb, int wait)
2498 tid_t target;
2500 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2501 if (wait)
2502 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2504 return 0;
2508 * LVM calls this function before a (read-only) snapshot is created. This
2509 * gives us a chance to flush the journal completely and mark the fs clean.
2511 static int ext3_freeze(struct super_block *sb)
2513 int error = 0;
2514 journal_t *journal;
2516 if (!(sb->s_flags & MS_RDONLY)) {
2517 journal = EXT3_SB(sb)->s_journal;
2519 /* Now we set up the journal barrier. */
2520 journal_lock_updates(journal);
2523 * We don't want to clear needs_recovery flag when we failed
2524 * to flush the journal.
2526 error = journal_flush(journal);
2527 if (error < 0)
2528 goto out;
2530 /* Journal blocked and flushed, clear needs_recovery flag. */
2531 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2532 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2533 if (error)
2534 goto out;
2536 return 0;
2538 out:
2539 journal_unlock_updates(journal);
2540 return error;
2544 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2545 * flag here, even though the filesystem is not technically dirty yet.
2547 static int ext3_unfreeze(struct super_block *sb)
2549 if (!(sb->s_flags & MS_RDONLY)) {
2550 lock_super(sb);
2551 /* Reser the needs_recovery flag before the fs is unlocked. */
2552 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2553 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2554 unlock_super(sb);
2555 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2557 return 0;
2560 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2562 struct ext3_super_block * es;
2563 struct ext3_sb_info *sbi = EXT3_SB(sb);
2564 ext3_fsblk_t n_blocks_count = 0;
2565 unsigned long old_sb_flags;
2566 struct ext3_mount_options old_opts;
2567 int enable_quota = 0;
2568 int err;
2569 #ifdef CONFIG_QUOTA
2570 int i;
2571 #endif
2573 /* Store the original options */
2574 lock_super(sb);
2575 old_sb_flags = sb->s_flags;
2576 old_opts.s_mount_opt = sbi->s_mount_opt;
2577 old_opts.s_resuid = sbi->s_resuid;
2578 old_opts.s_resgid = sbi->s_resgid;
2579 old_opts.s_commit_interval = sbi->s_commit_interval;
2580 #ifdef CONFIG_QUOTA
2581 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2582 for (i = 0; i < MAXQUOTAS; i++)
2583 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2584 #endif
2587 * Allow the "check" option to be passed as a remount option.
2589 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2590 err = -EINVAL;
2591 goto restore_opts;
2594 if (test_opt(sb, ABORT))
2595 ext3_abort(sb, __func__, "Abort forced by user");
2597 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2598 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2600 es = sbi->s_es;
2602 ext3_init_journal_params(sb, sbi->s_journal);
2604 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2605 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2606 if (test_opt(sb, ABORT)) {
2607 err = -EROFS;
2608 goto restore_opts;
2611 if (*flags & MS_RDONLY) {
2612 err = dquot_suspend(sb, -1);
2613 if (err < 0)
2614 goto restore_opts;
2617 * First of all, the unconditional stuff we have to do
2618 * to disable replay of the journal when we next remount
2620 sb->s_flags |= MS_RDONLY;
2623 * OK, test if we are remounting a valid rw partition
2624 * readonly, and if so set the rdonly flag and then
2625 * mark the partition as valid again.
2627 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2628 (sbi->s_mount_state & EXT3_VALID_FS))
2629 es->s_state = cpu_to_le16(sbi->s_mount_state);
2631 ext3_mark_recovery_complete(sb, es);
2632 } else {
2633 __le32 ret;
2634 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2635 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2636 ext3_msg(sb, KERN_WARNING,
2637 "warning: couldn't remount RDWR "
2638 "because of unsupported optional "
2639 "features (%x)", le32_to_cpu(ret));
2640 err = -EROFS;
2641 goto restore_opts;
2645 * If we have an unprocessed orphan list hanging
2646 * around from a previously readonly bdev mount,
2647 * require a full umount/remount for now.
2649 if (es->s_last_orphan) {
2650 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2651 "remount RDWR because of unprocessed "
2652 "orphan inode list. Please "
2653 "umount/remount instead.");
2654 err = -EINVAL;
2655 goto restore_opts;
2659 * Mounting a RDONLY partition read-write, so reread
2660 * and store the current valid flag. (It may have
2661 * been changed by e2fsck since we originally mounted
2662 * the partition.)
2664 ext3_clear_journal_err(sb, es);
2665 sbi->s_mount_state = le16_to_cpu(es->s_state);
2666 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2667 goto restore_opts;
2668 if (!ext3_setup_super (sb, es, 0))
2669 sb->s_flags &= ~MS_RDONLY;
2670 enable_quota = 1;
2673 #ifdef CONFIG_QUOTA
2674 /* Release old quota file names */
2675 for (i = 0; i < MAXQUOTAS; i++)
2676 if (old_opts.s_qf_names[i] &&
2677 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2678 kfree(old_opts.s_qf_names[i]);
2679 #endif
2680 unlock_super(sb);
2682 if (enable_quota)
2683 dquot_resume(sb, -1);
2684 return 0;
2685 restore_opts:
2686 sb->s_flags = old_sb_flags;
2687 sbi->s_mount_opt = old_opts.s_mount_opt;
2688 sbi->s_resuid = old_opts.s_resuid;
2689 sbi->s_resgid = old_opts.s_resgid;
2690 sbi->s_commit_interval = old_opts.s_commit_interval;
2691 #ifdef CONFIG_QUOTA
2692 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2693 for (i = 0; i < MAXQUOTAS; i++) {
2694 if (sbi->s_qf_names[i] &&
2695 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2696 kfree(sbi->s_qf_names[i]);
2697 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2699 #endif
2700 unlock_super(sb);
2701 return err;
2704 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2706 struct super_block *sb = dentry->d_sb;
2707 struct ext3_sb_info *sbi = EXT3_SB(sb);
2708 struct ext3_super_block *es = sbi->s_es;
2709 u64 fsid;
2711 if (test_opt(sb, MINIX_DF)) {
2712 sbi->s_overhead_last = 0;
2713 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2714 unsigned long ngroups = sbi->s_groups_count, i;
2715 ext3_fsblk_t overhead = 0;
2716 smp_rmb();
2719 * Compute the overhead (FS structures). This is constant
2720 * for a given filesystem unless the number of block groups
2721 * changes so we cache the previous value until it does.
2725 * All of the blocks before first_data_block are
2726 * overhead
2728 overhead = le32_to_cpu(es->s_first_data_block);
2731 * Add the overhead attributed to the superblock and
2732 * block group descriptors. If the sparse superblocks
2733 * feature is turned on, then not all groups have this.
2735 for (i = 0; i < ngroups; i++) {
2736 overhead += ext3_bg_has_super(sb, i) +
2737 ext3_bg_num_gdb(sb, i);
2738 cond_resched();
2742 * Every block group has an inode bitmap, a block
2743 * bitmap, and an inode table.
2745 overhead += ngroups * (2 + sbi->s_itb_per_group);
2746 sbi->s_overhead_last = overhead;
2747 smp_wmb();
2748 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2751 buf->f_type = EXT3_SUPER_MAGIC;
2752 buf->f_bsize = sb->s_blocksize;
2753 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2754 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2755 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2756 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2757 buf->f_bavail = 0;
2758 buf->f_files = le32_to_cpu(es->s_inodes_count);
2759 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2760 buf->f_namelen = EXT3_NAME_LEN;
2761 fsid = le64_to_cpup((void *)es->s_uuid) ^
2762 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2763 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2764 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2765 return 0;
2768 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2769 * is locked for write. Otherwise the are possible deadlocks:
2770 * Process 1 Process 2
2771 * ext3_create() quota_sync()
2772 * journal_start() write_dquot()
2773 * dquot_initialize() down(dqio_mutex)
2774 * down(dqio_mutex) journal_start()
2778 #ifdef CONFIG_QUOTA
2780 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2782 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2785 static int ext3_write_dquot(struct dquot *dquot)
2787 int ret, err;
2788 handle_t *handle;
2789 struct inode *inode;
2791 inode = dquot_to_inode(dquot);
2792 handle = ext3_journal_start(inode,
2793 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2794 if (IS_ERR(handle))
2795 return PTR_ERR(handle);
2796 ret = dquot_commit(dquot);
2797 err = ext3_journal_stop(handle);
2798 if (!ret)
2799 ret = err;
2800 return ret;
2803 static int ext3_acquire_dquot(struct dquot *dquot)
2805 int ret, err;
2806 handle_t *handle;
2808 handle = ext3_journal_start(dquot_to_inode(dquot),
2809 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2810 if (IS_ERR(handle))
2811 return PTR_ERR(handle);
2812 ret = dquot_acquire(dquot);
2813 err = ext3_journal_stop(handle);
2814 if (!ret)
2815 ret = err;
2816 return ret;
2819 static int ext3_release_dquot(struct dquot *dquot)
2821 int ret, err;
2822 handle_t *handle;
2824 handle = ext3_journal_start(dquot_to_inode(dquot),
2825 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2826 if (IS_ERR(handle)) {
2827 /* Release dquot anyway to avoid endless cycle in dqput() */
2828 dquot_release(dquot);
2829 return PTR_ERR(handle);
2831 ret = dquot_release(dquot);
2832 err = ext3_journal_stop(handle);
2833 if (!ret)
2834 ret = err;
2835 return ret;
2838 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2840 /* Are we journaling quotas? */
2841 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2842 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2843 dquot_mark_dquot_dirty(dquot);
2844 return ext3_write_dquot(dquot);
2845 } else {
2846 return dquot_mark_dquot_dirty(dquot);
2850 static int ext3_write_info(struct super_block *sb, int type)
2852 int ret, err;
2853 handle_t *handle;
2855 /* Data block + inode block */
2856 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2857 if (IS_ERR(handle))
2858 return PTR_ERR(handle);
2859 ret = dquot_commit_info(sb, type);
2860 err = ext3_journal_stop(handle);
2861 if (!ret)
2862 ret = err;
2863 return ret;
2867 * Turn on quotas during mount time - we need to find
2868 * the quota file and such...
2870 static int ext3_quota_on_mount(struct super_block *sb, int type)
2872 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2873 EXT3_SB(sb)->s_jquota_fmt, type);
2877 * Standard function to be called on quota_on
2879 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2880 struct path *path)
2882 int err;
2884 if (!test_opt(sb, QUOTA))
2885 return -EINVAL;
2887 /* Quotafile not on the same filesystem? */
2888 if (path->mnt->mnt_sb != sb)
2889 return -EXDEV;
2890 /* Journaling quota? */
2891 if (EXT3_SB(sb)->s_qf_names[type]) {
2892 /* Quotafile not of fs root? */
2893 if (path->dentry->d_parent != sb->s_root)
2894 ext3_msg(sb, KERN_WARNING,
2895 "warning: Quota file not on filesystem root. "
2896 "Journaled quota will not work.");
2900 * When we journal data on quota file, we have to flush journal to see
2901 * all updates to the file when we bypass pagecache...
2903 if (ext3_should_journal_data(path->dentry->d_inode)) {
2905 * We don't need to lock updates but journal_flush() could
2906 * otherwise be livelocked...
2908 journal_lock_updates(EXT3_SB(sb)->s_journal);
2909 err = journal_flush(EXT3_SB(sb)->s_journal);
2910 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2911 if (err)
2912 return err;
2915 return dquot_quota_on(sb, type, format_id, path);
2918 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2919 * acquiring the locks... As quota files are never truncated and quota code
2920 * itself serializes the operations (and noone else should touch the files)
2921 * we don't have to be afraid of races */
2922 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2923 size_t len, loff_t off)
2925 struct inode *inode = sb_dqopt(sb)->files[type];
2926 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2927 int err = 0;
2928 int offset = off & (sb->s_blocksize - 1);
2929 int tocopy;
2930 size_t toread;
2931 struct buffer_head *bh;
2932 loff_t i_size = i_size_read(inode);
2934 if (off > i_size)
2935 return 0;
2936 if (off+len > i_size)
2937 len = i_size-off;
2938 toread = len;
2939 while (toread > 0) {
2940 tocopy = sb->s_blocksize - offset < toread ?
2941 sb->s_blocksize - offset : toread;
2942 bh = ext3_bread(NULL, inode, blk, 0, &err);
2943 if (err)
2944 return err;
2945 if (!bh) /* A hole? */
2946 memset(data, 0, tocopy);
2947 else
2948 memcpy(data, bh->b_data+offset, tocopy);
2949 brelse(bh);
2950 offset = 0;
2951 toread -= tocopy;
2952 data += tocopy;
2953 blk++;
2955 return len;
2958 /* Write to quotafile (we know the transaction is already started and has
2959 * enough credits) */
2960 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2961 const char *data, size_t len, loff_t off)
2963 struct inode *inode = sb_dqopt(sb)->files[type];
2964 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2965 int err = 0;
2966 int offset = off & (sb->s_blocksize - 1);
2967 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2968 struct buffer_head *bh;
2969 handle_t *handle = journal_current_handle();
2971 if (!handle) {
2972 ext3_msg(sb, KERN_WARNING,
2973 "warning: quota write (off=%llu, len=%llu)"
2974 " cancelled because transaction is not started.",
2975 (unsigned long long)off, (unsigned long long)len);
2976 return -EIO;
2980 * Since we account only one data block in transaction credits,
2981 * then it is impossible to cross a block boundary.
2983 if (sb->s_blocksize - offset < len) {
2984 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2985 " cancelled because not block aligned",
2986 (unsigned long long)off, (unsigned long long)len);
2987 return -EIO;
2989 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2990 bh = ext3_bread(handle, inode, blk, 1, &err);
2991 if (!bh)
2992 goto out;
2993 if (journal_quota) {
2994 err = ext3_journal_get_write_access(handle, bh);
2995 if (err) {
2996 brelse(bh);
2997 goto out;
3000 lock_buffer(bh);
3001 memcpy(bh->b_data+offset, data, len);
3002 flush_dcache_page(bh->b_page);
3003 unlock_buffer(bh);
3004 if (journal_quota)
3005 err = ext3_journal_dirty_metadata(handle, bh);
3006 else {
3007 /* Always do at least ordered writes for quotas */
3008 err = ext3_journal_dirty_data(handle, bh);
3009 mark_buffer_dirty(bh);
3011 brelse(bh);
3012 out:
3013 if (err) {
3014 mutex_unlock(&inode->i_mutex);
3015 return err;
3017 if (inode->i_size < off + len) {
3018 i_size_write(inode, off + len);
3019 EXT3_I(inode)->i_disksize = inode->i_size;
3021 inode->i_version++;
3022 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3023 ext3_mark_inode_dirty(handle, inode);
3024 mutex_unlock(&inode->i_mutex);
3025 return len;
3028 #endif
3030 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3031 int flags, const char *dev_name, void *data)
3033 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3036 static struct file_system_type ext3_fs_type = {
3037 .owner = THIS_MODULE,
3038 .name = "ext3",
3039 .mount = ext3_mount,
3040 .kill_sb = kill_block_super,
3041 .fs_flags = FS_REQUIRES_DEV,
3044 static int __init init_ext3_fs(void)
3046 int err = init_ext3_xattr();
3047 if (err)
3048 return err;
3049 err = init_inodecache();
3050 if (err)
3051 goto out1;
3052 err = register_filesystem(&ext3_fs_type);
3053 if (err)
3054 goto out;
3055 return 0;
3056 out:
3057 destroy_inodecache();
3058 out1:
3059 exit_ext3_xattr();
3060 return err;
3063 static void __exit exit_ext3_fs(void)
3065 unregister_filesystem(&ext3_fs_type);
3066 destroy_inodecache();
3067 exit_ext3_xattr();
3070 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3071 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3072 MODULE_LICENSE("GPL");
3073 module_init(init_ext3_fs)
3074 module_exit(exit_ext3_fs)