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)
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/config.h>
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/jbd.h>
25 #include <linux/ext3_fs.h>
26 #include <linux/ext3_jbd.h>
27 #include <linux/slab.h>
28 #include <linux/init.h>
29 #include <linux/blkdev.h>
30 #include <linux/parser.h>
31 #include <linux/smp_lock.h>
32 #include <linux/buffer_head.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 <asm/uaccess.h>
42 static int ext3_load_journal(struct super_block *, struct ext3_super_block *);
43 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
45 static void ext3_commit_super (struct super_block * sb,
46 struct ext3_super_block * es,
48 static void ext3_mark_recovery_complete(struct super_block * sb,
49 struct ext3_super_block * es);
50 static void ext3_clear_journal_err(struct super_block * sb,
51 struct ext3_super_block * es);
52 static int ext3_sync_fs(struct super_block *sb, int wait);
55 * Wrappers for journal_start/end.
57 * The only special thing we need to do here is to make sure that all
58 * journal_end calls result in the superblock being marked dirty, so
59 * that sync() will call the filesystem's write_super callback if
62 handle_t *ext3_journal_start(struct inode *inode, int nblocks)
66 if (inode->i_sb->s_flags & MS_RDONLY)
67 return ERR_PTR(-EROFS);
69 /* Special case here: if the journal has aborted behind our
70 * backs (eg. EIO in the commit thread), then we still need to
71 * take the FS itself readonly cleanly. */
72 journal = EXT3_JOURNAL(inode);
73 if (is_journal_aborted(journal)) {
74 ext3_abort(inode->i_sb, __FUNCTION__,
75 "Detected aborted journal");
76 return ERR_PTR(-EROFS);
79 return journal_start(journal, nblocks);
83 * The only special thing we need to do here is to make sure that all
84 * journal_stop calls result in the superblock being marked dirty, so
85 * that sync() will call the filesystem's write_super callback if
88 int __ext3_journal_stop(const char *where, handle_t *handle)
90 struct super_block *sb;
94 sb = handle->h_transaction->t_journal->j_private;
96 rc = journal_stop(handle);
101 __ext3_std_error(sb, where, err);
105 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
106 struct buffer_head *bh, handle_t *handle, int err)
109 const char *errstr = ext3_decode_error(NULL, err, nbuf);
111 printk(KERN_ERR "%s: aborting transaction: %s in %s",
112 caller, errstr, err_fn);
115 BUFFER_TRACE(bh, "abort");
116 journal_abort_handle(handle);
121 static char error_buf[1024];
123 /* Deal with the reporting of failure conditions on a filesystem such as
124 * inconsistencies detected or read IO failures.
126 * On ext2, we can store the error state of the filesystem in the
127 * superblock. That is not possible on ext3, because we may have other
128 * write ordering constraints on the superblock which prevent us from
129 * writing it out straight away; and given that the journal is about to
130 * be aborted, we can't rely on the current, or future, transactions to
131 * write out the superblock safely.
133 * We'll just use the journal_abort() error code to record an error in
134 * the journal instead. On recovery, the journal will compain about
135 * that error until we've noted it down and cleared it.
138 static void ext3_handle_error(struct super_block *sb)
140 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
142 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
143 es->s_state |= cpu_to_le32(EXT3_ERROR_FS);
145 if (sb->s_flags & MS_RDONLY)
148 if (test_opt (sb, ERRORS_PANIC))
149 panic ("EXT3-fs (device %s): panic forced after error\n",
151 if (test_opt (sb, ERRORS_RO)) {
152 printk (KERN_CRIT "Remounting filesystem read-only\n");
153 sb->s_flags |= MS_RDONLY;
155 journal_t *journal = EXT3_SB(sb)->s_journal;
157 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
159 journal_abort(journal, -EIO);
161 ext3_commit_super(sb, es, 1);
164 void ext3_error (struct super_block * sb, const char * function,
165 const char * fmt, ...)
169 va_start (args, fmt);
170 vsprintf (error_buf, fmt, args);
173 printk (KERN_CRIT "EXT3-fs error (device %s): %s: %s\n",
174 sb->s_id, function, error_buf);
176 ext3_handle_error(sb);
179 const char *ext3_decode_error(struct super_block * sb, int errno, char nbuf[16])
185 errstr = "IO failure";
188 errstr = "Out of memory";
191 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
192 errstr = "Journal has aborted";
194 errstr = "Readonly filesystem";
197 /* If the caller passed in an extra buffer for unknown
198 * errors, textualise them now. Else we just return
201 /* Check for truncated error codes... */
202 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
211 /* __ext3_std_error decodes expected errors from journaling functions
212 * automatically and invokes the appropriate error response. */
214 void __ext3_std_error (struct super_block * sb, const char * function,
218 const char *errstr = ext3_decode_error(sb, errno, nbuf);
220 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
221 sb->s_id, function, errstr);
223 ext3_handle_error(sb);
227 * ext3_abort is a much stronger failure handler than ext3_error. The
228 * abort function may be used to deal with unrecoverable failures such
229 * as journal IO errors or ENOMEM at a critical moment in log management.
231 * We unconditionally force the filesystem into an ABORT|READONLY state,
232 * unless the error response on the fs has been set to panic in which
233 * case we take the easy way out and panic immediately.
236 void ext3_abort (struct super_block * sb, const char * function,
237 const char * fmt, ...)
241 printk (KERN_CRIT "ext3_abort called.\n");
243 va_start (args, fmt);
244 vsprintf (error_buf, fmt, args);
247 if (test_opt (sb, ERRORS_PANIC))
248 panic ("EXT3-fs panic (device %s): %s: %s\n",
249 sb->s_id, function, error_buf);
251 printk (KERN_CRIT "EXT3-fs abort (device %s): %s: %s\n",
252 sb->s_id, function, error_buf);
254 if (sb->s_flags & MS_RDONLY)
257 printk (KERN_CRIT "Remounting filesystem read-only\n");
258 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
259 sb->s_flags |= MS_RDONLY;
260 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
261 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
264 /* Deal with the reporting of failure conditions while running, such as
265 * inconsistencies in operation or invalid system states.
267 * Use ext3_error() for cases of invalid filesystem states, as that will
268 * record an error on disk and force a filesystem check on the next boot.
270 NORET_TYPE void ext3_panic (struct super_block * sb, const char * function,
271 const char * fmt, ...)
275 va_start (args, fmt);
276 vsprintf (error_buf, fmt, args);
279 /* this is to prevent panic from syncing this filesystem */
280 /* AKPM: is this sufficient? */
281 sb->s_flags |= MS_RDONLY;
282 panic ("EXT3-fs panic (device %s): %s: %s\n",
283 sb->s_id, function, error_buf);
286 void ext3_warning (struct super_block * sb, const char * function,
287 const char * fmt, ...)
291 va_start (args, fmt);
292 vsprintf (error_buf, fmt, args);
294 printk (KERN_WARNING "EXT3-fs warning (device %s): %s: %s\n",
295 sb->s_id, function, error_buf);
298 void ext3_update_dynamic_rev(struct super_block *sb)
300 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
302 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 ext3_warning(sb, __FUNCTION__,
306 "updating to rev %d because of new feature flag, "
307 "running e2fsck is recommended",
310 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
311 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
312 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
313 /* leave es->s_feature_*compat flags alone */
314 /* es->s_uuid will be set by e2fsck if empty */
317 * The rest of the superblock fields should be zero, and if not it
318 * means they are likely already in use, so leave them alone. We
319 * can leave it up to e2fsck to clean up any inconsistencies there.
324 * Open the external journal device
326 static struct block_device *ext3_blkdev_get(dev_t dev)
328 struct block_device *bdev;
329 char b[BDEVNAME_SIZE];
331 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
337 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
338 __bdevname(dev, b), PTR_ERR(bdev));
343 * Release the journal device
345 static int ext3_blkdev_put(struct block_device *bdev)
348 return blkdev_put(bdev);
351 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
353 struct block_device *bdev;
356 bdev = sbi->journal_bdev;
358 ret = ext3_blkdev_put(bdev);
359 sbi->journal_bdev = NULL;
364 static inline struct inode *orphan_list_entry(struct list_head *l)
366 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
373 printk(KERN_ERR "sb orphan head is %d\n",
374 le32_to_cpu(sbi->s_es->s_last_orphan));
376 printk(KERN_ERR "sb_info orphan list:\n");
377 list_for_each(l, &sbi->s_orphan) {
378 struct inode *inode = orphan_list_entry(l);
380 "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
381 inode->i_sb->s_id, inode->i_ino, inode,
382 inode->i_mode, inode->i_nlink,
383 le32_to_cpu(NEXT_ORPHAN(inode)));
387 void ext3_put_super (struct super_block * sb)
389 struct ext3_sb_info *sbi = EXT3_SB(sb);
390 struct ext3_super_block *es = sbi->s_es;
393 ext3_xattr_put_super(sb);
394 journal_destroy(sbi->s_journal);
395 if (!(sb->s_flags & MS_RDONLY)) {
396 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
397 es->s_state = le16_to_cpu(sbi->s_mount_state);
398 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
399 mark_buffer_dirty(sbi->s_sbh);
400 ext3_commit_super(sb, es, 1);
403 for (i = 0; i < sbi->s_gdb_count; i++)
404 brelse(sbi->s_group_desc[i]);
405 kfree(sbi->s_group_desc);
409 for (i = 0; i < MAXQUOTAS; i++) {
410 if (sbi->s_qf_names[i])
411 kfree(sbi->s_qf_names[i]);
415 /* Debugging code just in case the in-memory inode orphan list
416 * isn't empty. The on-disk one can be non-empty if we've
417 * detected an error and taken the fs readonly, but the
418 * in-memory list had better be clean by this point. */
419 if (!list_empty(&sbi->s_orphan))
420 dump_orphan_list(sb, sbi);
421 J_ASSERT(list_empty(&sbi->s_orphan));
423 invalidate_bdev(sb->s_bdev, 0);
424 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
426 * Invalidate the journal device's buffers. We don't want them
427 * floating about in memory - the physical journal device may
428 * hotswapped, and it breaks the `ro-after' testing code.
430 sync_blockdev(sbi->journal_bdev);
431 invalidate_bdev(sbi->journal_bdev, 0);
432 ext3_blkdev_remove(sbi);
434 sb->s_fs_info = NULL;
439 static kmem_cache_t *ext3_inode_cachep;
442 * Called inside transaction, so use GFP_NOFS
444 static struct inode *ext3_alloc_inode(struct super_block *sb)
446 struct ext3_inode_info *ei;
448 ei = kmem_cache_alloc(ext3_inode_cachep, SLAB_NOFS);
451 #ifdef CONFIG_EXT3_FS_POSIX_ACL
452 ei->i_acl = EXT3_ACL_NOT_CACHED;
453 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
455 ei->vfs_inode.i_version = 1;
456 return &ei->vfs_inode;
459 static void ext3_destroy_inode(struct inode *inode)
461 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
464 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
466 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
468 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
469 SLAB_CTOR_CONSTRUCTOR) {
470 INIT_LIST_HEAD(&ei->i_orphan);
471 #ifdef CONFIG_EXT3_FS_XATTR
472 init_rwsem(&ei->xattr_sem);
474 init_MUTEX(&ei->truncate_sem);
475 inode_init_once(&ei->vfs_inode);
479 static int init_inodecache(void)
481 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
482 sizeof(struct ext3_inode_info),
483 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
485 if (ext3_inode_cachep == NULL)
490 static void destroy_inodecache(void)
492 if (kmem_cache_destroy(ext3_inode_cachep))
493 printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
496 #ifdef CONFIG_EXT3_FS_POSIX_ACL
498 static void ext3_clear_inode(struct inode *inode)
500 if (EXT3_I(inode)->i_acl &&
501 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
502 posix_acl_release(EXT3_I(inode)->i_acl);
503 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
505 if (EXT3_I(inode)->i_default_acl &&
506 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
507 posix_acl_release(EXT3_I(inode)->i_default_acl);
508 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
513 # define ext3_clear_inode NULL
518 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
519 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
521 static int ext3_dquot_initialize(struct inode *inode, int type);
522 static int ext3_dquot_drop(struct inode *inode);
523 static int ext3_write_dquot(struct dquot *dquot);
524 static int ext3_acquire_dquot(struct dquot *dquot);
525 static int ext3_release_dquot(struct dquot *dquot);
526 static int ext3_mark_dquot_dirty(struct dquot *dquot);
527 static int ext3_write_info(struct super_block *sb, int type);
528 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
529 static int ext3_quota_on_mount(struct super_block *sb, int type);
530 static int ext3_quota_off_mount(struct super_block *sb, int type);
532 static struct dquot_operations ext3_quota_operations = {
533 .initialize = ext3_dquot_initialize,
534 .drop = ext3_dquot_drop,
535 .alloc_space = dquot_alloc_space,
536 .alloc_inode = dquot_alloc_inode,
537 .free_space = dquot_free_space,
538 .free_inode = dquot_free_inode,
539 .transfer = dquot_transfer,
540 .write_dquot = ext3_write_dquot,
541 .acquire_dquot = ext3_acquire_dquot,
542 .release_dquot = ext3_release_dquot,
543 .mark_dirty = ext3_mark_dquot_dirty,
544 .write_info = ext3_write_info
547 static struct quotactl_ops ext3_qctl_operations = {
548 .quota_on = ext3_quota_on,
549 .quota_off = vfs_quota_off,
550 .quota_sync = vfs_quota_sync,
551 .get_info = vfs_get_dqinfo,
552 .set_info = vfs_set_dqinfo,
553 .get_dqblk = vfs_get_dqblk,
554 .set_dqblk = vfs_set_dqblk
558 static struct super_operations ext3_sops = {
559 .alloc_inode = ext3_alloc_inode,
560 .destroy_inode = ext3_destroy_inode,
561 .read_inode = ext3_read_inode,
562 .write_inode = ext3_write_inode,
563 .dirty_inode = ext3_dirty_inode,
564 .put_inode = ext3_put_inode,
565 .delete_inode = ext3_delete_inode,
566 .put_super = ext3_put_super,
567 .write_super = ext3_write_super,
568 .sync_fs = ext3_sync_fs,
569 .write_super_lockfs = ext3_write_super_lockfs,
570 .unlockfs = ext3_unlockfs,
571 .statfs = ext3_statfs,
572 .remount_fs = ext3_remount,
573 .clear_inode = ext3_clear_inode,
576 struct dentry *ext3_get_parent(struct dentry *child);
577 static struct export_operations ext3_export_ops = {
578 .get_parent = ext3_get_parent,
582 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
583 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
584 Opt_nouid32, Opt_check, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
585 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, Opt_noload,
586 Opt_commit, Opt_journal_update, Opt_journal_inum,
587 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
588 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
589 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0,
590 Opt_ignore, Opt_barrier, Opt_err,
593 static match_table_t tokens = {
594 {Opt_bsd_df, "bsddf"},
595 {Opt_minix_df, "minixdf"},
596 {Opt_grpid, "grpid"},
597 {Opt_grpid, "bsdgroups"},
598 {Opt_nogrpid, "nogrpid"},
599 {Opt_nogrpid, "sysvgroups"},
600 {Opt_resgid, "resgid=%u"},
601 {Opt_resuid, "resuid=%u"},
603 {Opt_err_cont, "errors=continue"},
604 {Opt_err_panic, "errors=panic"},
605 {Opt_err_ro, "errors=remount-ro"},
606 {Opt_nouid32, "nouid32"},
607 {Opt_nocheck, "nocheck"},
608 {Opt_nocheck, "check=none"},
609 {Opt_check, "check"},
610 {Opt_debug, "debug"},
611 {Opt_oldalloc, "oldalloc"},
612 {Opt_orlov, "orlov"},
613 {Opt_user_xattr, "user_xattr"},
614 {Opt_nouser_xattr, "nouser_xattr"},
616 {Opt_noacl, "noacl"},
617 {Opt_noload, "noload"},
618 {Opt_commit, "commit=%u"},
619 {Opt_journal_update, "journal=update"},
620 {Opt_journal_inum, "journal=%u"},
621 {Opt_abort, "abort"},
622 {Opt_data_journal, "data=journal"},
623 {Opt_data_ordered, "data=ordered"},
624 {Opt_data_writeback, "data=writeback"},
625 {Opt_offusrjquota, "usrjquota="},
626 {Opt_usrjquota, "usrjquota=%s"},
627 {Opt_offgrpjquota, "grpjquota="},
628 {Opt_grpjquota, "grpjquota=%s"},
629 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
630 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
631 {Opt_ignore, "grpquota"},
632 {Opt_ignore, "noquota"},
633 {Opt_ignore, "quota"},
634 {Opt_ignore, "usrquota"},
635 {Opt_barrier, "barrier=%u"},
639 static unsigned long get_sb_block(void **data)
641 unsigned long sb_block;
642 char *options = (char *) *data;
644 if (!options || strncmp(options, "sb=", 3) != 0)
645 return 1; /* Default location */
647 sb_block = simple_strtoul(options, &options, 0);
648 if (*options && *options != ',') {
649 printk("EXT3-fs: Invalid sb specification: %s\n",
655 *data = (void *) options;
659 static int parse_options (char * options, struct super_block *sb,
660 unsigned long * inum, int is_remount)
662 struct ext3_sb_info *sbi = EXT3_SB(sb);
664 substring_t args[MAX_OPT_ARGS];
674 while ((p = strsep (&options, ",")) != NULL) {
679 token = match_token(p, tokens, args);
682 clear_opt (sbi->s_mount_opt, MINIX_DF);
685 set_opt (sbi->s_mount_opt, MINIX_DF);
688 set_opt (sbi->s_mount_opt, GRPID);
691 clear_opt (sbi->s_mount_opt, GRPID);
694 if (match_int(&args[0], &option))
696 sbi->s_resuid = option;
699 if (match_int(&args[0], &option))
701 sbi->s_resgid = option;
704 /* handled by get_sb_block() instead of here */
705 /* *sb_block = match_int(&args[0]); */
708 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
709 clear_opt (sbi->s_mount_opt, ERRORS_RO);
710 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
713 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
714 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
715 set_opt (sbi->s_mount_opt, ERRORS_RO);
718 clear_opt (sbi->s_mount_opt, ERRORS_RO);
719 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
720 set_opt (sbi->s_mount_opt, ERRORS_CONT);
723 set_opt (sbi->s_mount_opt, NO_UID32);
726 #ifdef CONFIG_EXT3_CHECK
727 set_opt (sbi->s_mount_opt, CHECK);
730 "EXT3 Check option not supported\n");
734 clear_opt (sbi->s_mount_opt, CHECK);
737 set_opt (sbi->s_mount_opt, DEBUG);
740 set_opt (sbi->s_mount_opt, OLDALLOC);
743 clear_opt (sbi->s_mount_opt, OLDALLOC);
745 #ifdef CONFIG_EXT3_FS_XATTR
747 set_opt (sbi->s_mount_opt, XATTR_USER);
749 case Opt_nouser_xattr:
750 clear_opt (sbi->s_mount_opt, XATTR_USER);
754 case Opt_nouser_xattr:
755 printk("EXT3 (no)user_xattr options not supported\n");
758 #ifdef CONFIG_EXT3_FS_POSIX_ACL
760 set_opt(sbi->s_mount_opt, POSIX_ACL);
763 clear_opt(sbi->s_mount_opt, POSIX_ACL);
768 printk("EXT3 (no)acl options not supported\n");
771 case Opt_journal_update:
773 /* Eventually we will want to be able to create
774 a journal file here. For now, only allow the
775 user to specify an existing inode to be the
778 printk(KERN_ERR "EXT3-fs: cannot specify "
779 "journal on remount\n");
782 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
784 case Opt_journal_inum:
786 printk(KERN_ERR "EXT3-fs: cannot specify "
787 "journal on remount\n");
790 if (match_int(&args[0], &option))
795 set_opt (sbi->s_mount_opt, NOLOAD);
798 if (match_int(&args[0], &option))
803 option = JBD_DEFAULT_MAX_COMMIT_AGE;
804 sbi->s_commit_interval = HZ * option;
806 case Opt_data_journal:
807 data_opt = EXT3_MOUNT_JOURNAL_DATA;
809 case Opt_data_ordered:
810 data_opt = EXT3_MOUNT_ORDERED_DATA;
812 case Opt_data_writeback:
813 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
816 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
819 "EXT3-fs: cannot change data "
820 "mode on remount\n");
824 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
825 sbi->s_mount_opt |= data_opt;
835 if (sb_any_quota_enabled(sb)) {
837 "EXT3-fs: Cannot change journalled "
838 "quota options when quota turned on.\n");
841 if (sbi->s_qf_names[qtype]) {
843 "EXT3-fs: %s quota file already "
844 "specified.\n", QTYPE2NAME(qtype));
847 sbi->s_qf_names[qtype] = match_strdup(&args[0]);
848 if (!sbi->s_qf_names[qtype]) {
850 "EXT3-fs: not enough memory for "
851 "storing quotafile name.\n");
854 if (strchr(sbi->s_qf_names[qtype], '/')) {
856 "EXT3-fs: quotafile must be on "
857 "filesystem root.\n");
858 kfree(sbi->s_qf_names[qtype]);
859 sbi->s_qf_names[qtype] = NULL;
863 case Opt_offusrjquota:
866 case Opt_offgrpjquota:
869 if (sb_any_quota_enabled(sb)) {
870 printk(KERN_ERR "EXT3-fs: Cannot change "
871 "journalled quota options when "
872 "quota turned on.\n");
875 if (sbi->s_qf_names[qtype]) {
876 kfree(sbi->s_qf_names[qtype]);
877 sbi->s_qf_names[qtype] = NULL;
880 case Opt_jqfmt_vfsold:
881 sbi->s_jquota_fmt = QFMT_VFS_OLD;
883 case Opt_jqfmt_vfsv0:
884 sbi->s_jquota_fmt = QFMT_VFS_V0;
889 case Opt_offusrjquota:
890 case Opt_offgrpjquota:
891 case Opt_jqfmt_vfsold:
892 case Opt_jqfmt_vfsv0:
894 "EXT3-fs: journalled quota options not "
899 set_opt(sbi->s_mount_opt, ABORT);
902 if (match_int(&args[0], &option))
905 set_opt(sbi->s_mount_opt, BARRIER);
907 clear_opt(sbi->s_mount_opt, BARRIER);
913 "EXT3-fs: Unrecognized mount option \"%s\" "
914 "or missing value\n", p);
919 if (!sbi->s_jquota_fmt && (sbi->s_qf_names[USRQUOTA] ||
920 sbi->s_qf_names[GRPQUOTA])) {
922 "EXT3-fs: journalled quota format not specified.\n");
930 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
933 struct ext3_sb_info *sbi = EXT3_SB(sb);
936 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
937 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
938 "forcing read-only mode\n");
943 if (!(sbi->s_mount_state & EXT3_VALID_FS))
944 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
945 "running e2fsck is recommended\n");
946 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
948 "EXT3-fs warning: mounting fs with errors, "
949 "running e2fsck is recommended\n");
950 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
951 le16_to_cpu(es->s_mnt_count) >=
952 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
954 "EXT3-fs warning: maximal mount count reached, "
955 "running e2fsck is recommended\n");
956 else if (le32_to_cpu(es->s_checkinterval) &&
957 (le32_to_cpu(es->s_lastcheck) +
958 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
960 "EXT3-fs warning: checktime reached, "
961 "running e2fsck is recommended\n");
963 /* @@@ We _will_ want to clear the valid bit if we find
964 inconsistencies, to force a fsck at reboot. But for
965 a plain journaled filesystem we can keep it set as
967 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
969 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
970 es->s_max_mnt_count =
971 (__s16) cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
972 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
973 es->s_mtime = cpu_to_le32(get_seconds());
974 ext3_update_dynamic_rev(sb);
975 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
977 ext3_commit_super(sb, es, 1);
978 if (test_opt(sb, DEBUG))
979 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
980 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
983 EXT3_BLOCKS_PER_GROUP(sb),
984 EXT3_INODES_PER_GROUP(sb),
987 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
988 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
989 char b[BDEVNAME_SIZE];
991 printk("external journal on %s\n",
992 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
994 printk("internal journal\n");
996 #ifdef CONFIG_EXT3_CHECK
997 if (test_opt (sb, CHECK)) {
998 ext3_check_blocks_bitmap (sb);
999 ext3_check_inodes_bitmap (sb);
1005 static int ext3_check_descriptors (struct super_block * sb)
1007 struct ext3_sb_info *sbi = EXT3_SB(sb);
1008 unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
1009 struct ext3_group_desc * gdp = NULL;
1013 ext3_debug ("Checking group descriptors");
1015 for (i = 0; i < sbi->s_groups_count; i++)
1017 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1018 gdp = (struct ext3_group_desc *)
1019 sbi->s_group_desc[desc_block++]->b_data;
1020 if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
1021 le32_to_cpu(gdp->bg_block_bitmap) >=
1022 block + EXT3_BLOCKS_PER_GROUP(sb))
1024 ext3_error (sb, "ext3_check_descriptors",
1025 "Block bitmap for group %d"
1026 " not in group (block %lu)!",
1028 le32_to_cpu(gdp->bg_block_bitmap));
1031 if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
1032 le32_to_cpu(gdp->bg_inode_bitmap) >=
1033 block + EXT3_BLOCKS_PER_GROUP(sb))
1035 ext3_error (sb, "ext3_check_descriptors",
1036 "Inode bitmap for group %d"
1037 " not in group (block %lu)!",
1039 le32_to_cpu(gdp->bg_inode_bitmap));
1042 if (le32_to_cpu(gdp->bg_inode_table) < block ||
1043 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
1044 block + EXT3_BLOCKS_PER_GROUP(sb))
1046 ext3_error (sb, "ext3_check_descriptors",
1047 "Inode table for group %d"
1048 " not in group (block %lu)!",
1050 le32_to_cpu(gdp->bg_inode_table));
1053 block += EXT3_BLOCKS_PER_GROUP(sb);
1057 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1058 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1063 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1064 * the superblock) which were deleted from all directories, but held open by
1065 * a process at the time of a crash. We walk the list and try to delete these
1066 * inodes at recovery time (only with a read-write filesystem).
1068 * In order to keep the orphan inode chain consistent during traversal (in
1069 * case of crash during recovery), we link each inode into the superblock
1070 * orphan list_head and handle it the same way as an inode deletion during
1071 * normal operation (which journals the operations for us).
1073 * We only do an iget() and an iput() on each inode, which is very safe if we
1074 * accidentally point at an in-use or already deleted inode. The worst that
1075 * can happen in this case is that we get a "bit already cleared" message from
1076 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1077 * e2fsck was run on this filesystem, and it must have already done the orphan
1078 * inode cleanup for us, so we can safely abort without any further action.
1080 static void ext3_orphan_cleanup (struct super_block * sb,
1081 struct ext3_super_block * es)
1083 unsigned int s_flags = sb->s_flags;
1084 int nr_orphans = 0, nr_truncates = 0;
1088 if (!es->s_last_orphan) {
1089 jbd_debug(4, "no orphan inodes to clean up\n");
1093 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1094 if (es->s_last_orphan)
1095 jbd_debug(1, "Errors on filesystem, "
1096 "clearing orphan list.\n");
1097 es->s_last_orphan = 0;
1098 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1102 if (s_flags & MS_RDONLY) {
1103 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1105 sb->s_flags &= ~MS_RDONLY;
1108 /* Needed for iput() to work correctly and not trash data */
1109 sb->s_flags |= MS_ACTIVE;
1110 /* Turn on quotas so that they are updated correctly */
1111 for (i = 0; i < MAXQUOTAS; i++) {
1112 if (EXT3_SB(sb)->s_qf_names[i]) {
1113 int ret = ext3_quota_on_mount(sb, i);
1116 "EXT3-fs: Cannot turn on journalled "
1117 "quota: error %d\n", ret);
1122 while (es->s_last_orphan) {
1123 struct inode *inode;
1126 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1127 es->s_last_orphan = 0;
1131 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1133 if (inode->i_nlink) {
1135 "%s: truncating inode %ld to %Ld bytes\n",
1136 __FUNCTION__, inode->i_ino, inode->i_size);
1137 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1138 inode->i_ino, inode->i_size);
1139 ext3_truncate(inode);
1143 "%s: deleting unreferenced inode %ld\n",
1144 __FUNCTION__, inode->i_ino);
1145 jbd_debug(2, "deleting unreferenced inode %ld\n",
1149 iput(inode); /* The delete magic happens here! */
1152 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1155 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1156 sb->s_id, PLURAL(nr_orphans));
1158 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1159 sb->s_id, PLURAL(nr_truncates));
1161 /* Turn quotas off */
1162 for (i = 0; i < MAXQUOTAS; i++) {
1163 if (sb_dqopt(sb)->files[i])
1164 ext3_quota_off_mount(sb, i);
1167 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1170 #define log2(n) ffz(~(n))
1173 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1174 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1175 * We need to be 1 filesystem block less than the 2^32 sector limit.
1177 static loff_t ext3_max_size(int bits)
1179 loff_t res = EXT3_NDIR_BLOCKS;
1180 res += 1LL << (bits-2);
1181 res += 1LL << (2*(bits-2));
1182 res += 1LL << (3*(bits-2));
1184 if (res > (512LL << 32) - (1 << bits))
1185 res = (512LL << 32) - (1 << bits);
1189 static unsigned long descriptor_loc(struct super_block *sb,
1190 unsigned long logic_sb_block,
1193 struct ext3_sb_info *sbi = EXT3_SB(sb);
1194 unsigned long bg, first_data_block, first_meta_bg;
1197 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1198 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1200 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1202 return (logic_sb_block + nr + 1);
1203 bg = sbi->s_desc_per_block * nr;
1204 if (ext3_bg_has_super(sb, bg))
1206 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
1210 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1212 struct buffer_head * bh;
1213 struct ext3_super_block *es = NULL;
1214 struct ext3_sb_info *sbi;
1215 unsigned long block;
1216 unsigned long sb_block = get_sb_block(&data);
1217 unsigned long logic_sb_block;
1218 unsigned long offset = 0;
1219 unsigned long journal_inum = 0;
1220 unsigned long def_mount_opts;
1228 sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
1231 sb->s_fs_info = sbi;
1232 memset(sbi, 0, sizeof(*sbi));
1233 sbi->s_mount_opt = 0;
1234 sbi->s_resuid = EXT3_DEF_RESUID;
1235 sbi->s_resgid = EXT3_DEF_RESGID;
1237 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1239 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1244 * The ext3 superblock will not be buffer aligned for other than 1kB
1245 * block sizes. We need to calculate the offset from buffer start.
1247 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1248 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1249 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1251 logic_sb_block = sb_block;
1254 if (!(bh = sb_bread(sb, logic_sb_block))) {
1255 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1259 * Note: s_es must be initialized as soon as possible because
1260 * some ext3 macro-instructions depend on its value
1262 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1264 sb->s_magic = le16_to_cpu(es->s_magic);
1265 if (sb->s_magic != EXT3_SUPER_MAGIC) {
1268 "VFS: Can't find ext3 filesystem on dev %s.\n",
1273 /* Set defaults before we parse the mount options */
1274 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1275 if (def_mount_opts & EXT3_DEFM_DEBUG)
1276 set_opt(sbi->s_mount_opt, DEBUG);
1277 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1278 set_opt(sbi->s_mount_opt, GRPID);
1279 if (def_mount_opts & EXT3_DEFM_UID16)
1280 set_opt(sbi->s_mount_opt, NO_UID32);
1281 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1282 set_opt(sbi->s_mount_opt, XATTR_USER);
1283 if (def_mount_opts & EXT3_DEFM_ACL)
1284 set_opt(sbi->s_mount_opt, POSIX_ACL);
1285 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1286 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1287 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1288 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1289 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1290 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1292 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1293 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1294 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1295 set_opt(sbi->s_mount_opt, ERRORS_RO);
1297 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1298 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1300 /* enable barriers by default */
1301 set_opt(sbi->s_mount_opt, BARRIER);
1303 if (!parse_options ((char *) data, sb, &journal_inum, 0))
1306 sb->s_flags |= MS_ONE_SECOND;
1307 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1308 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1310 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1311 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1312 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1313 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1315 "EXT3-fs warning: feature flags set on rev 0 fs, "
1316 "running e2fsck is recommended\n");
1318 * Check feature flags regardless of the revision level, since we
1319 * previously didn't change the revision level when setting the flags,
1320 * so there is a chance incompat flags are set on a rev 0 filesystem.
1322 if ((i = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))) {
1323 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1324 "unsupported optional features (%x).\n",
1328 if (!(sb->s_flags & MS_RDONLY) &&
1329 (i = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))){
1330 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1331 "unsupported optional features (%x).\n",
1335 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1337 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1338 blocksize > EXT3_MAX_BLOCK_SIZE) {
1340 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1341 blocksize, sb->s_id);
1345 hblock = bdev_hardsect_size(sb->s_bdev);
1346 if (sb->s_blocksize != blocksize) {
1348 * Make sure the blocksize for the filesystem is larger
1349 * than the hardware sectorsize for the machine.
1351 if (blocksize < hblock) {
1352 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1353 "device blocksize %d.\n", blocksize, hblock);
1358 sb_set_blocksize(sb, blocksize);
1359 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1360 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1361 bh = sb_bread(sb, logic_sb_block);
1364 "EXT3-fs: Can't read superblock on 2nd try.\n");
1367 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1369 if (es->s_magic != le16_to_cpu(EXT3_SUPER_MAGIC)) {
1371 "EXT3-fs: Magic mismatch, very weird !\n");
1376 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1378 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1379 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1380 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1382 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1383 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1384 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1385 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1386 (sbi->s_inode_size > blocksize)) {
1388 "EXT3-fs: unsupported inode size: %d\n",
1393 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1394 le32_to_cpu(es->s_log_frag_size);
1395 if (blocksize != sbi->s_frag_size) {
1397 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1398 sbi->s_frag_size, blocksize);
1401 sbi->s_frags_per_block = 1;
1402 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1403 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1404 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1405 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1406 sbi->s_itb_per_group = sbi->s_inodes_per_group /sbi->s_inodes_per_block;
1407 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1409 sbi->s_mount_state = le16_to_cpu(es->s_state);
1410 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1411 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1412 for (i=0; i < 4; i++)
1413 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1414 sbi->s_def_hash_version = es->s_def_hash_version;
1416 if (sbi->s_blocks_per_group > blocksize * 8) {
1418 "EXT3-fs: #blocks per group too big: %lu\n",
1419 sbi->s_blocks_per_group);
1422 if (sbi->s_frags_per_group > blocksize * 8) {
1424 "EXT3-fs: #fragments per group too big: %lu\n",
1425 sbi->s_frags_per_group);
1428 if (sbi->s_inodes_per_group > blocksize * 8) {
1430 "EXT3-fs: #inodes per group too big: %lu\n",
1431 sbi->s_inodes_per_group);
1435 sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
1436 le32_to_cpu(es->s_first_data_block) +
1437 EXT3_BLOCKS_PER_GROUP(sb) - 1) /
1438 EXT3_BLOCKS_PER_GROUP(sb);
1439 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1440 EXT3_DESC_PER_BLOCK(sb);
1441 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1443 if (sbi->s_group_desc == NULL) {
1444 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1447 sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(u8),
1449 if (!sbi->s_debts) {
1450 printk("EXT3-fs: not enough memory to allocate s_bgi\n");
1453 memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(u8));
1455 percpu_counter_init(&sbi->s_freeblocks_counter);
1456 percpu_counter_init(&sbi->s_freeinodes_counter);
1457 percpu_counter_init(&sbi->s_dirs_counter);
1458 bgl_lock_init(&sbi->s_blockgroup_lock);
1460 for (i = 0; i < db_count; i++) {
1461 block = descriptor_loc(sb, logic_sb_block, i);
1462 sbi->s_group_desc[i] = sb_bread(sb, block);
1463 if (!sbi->s_group_desc[i]) {
1464 printk (KERN_ERR "EXT3-fs: "
1465 "can't read group descriptor %d\n", i);
1470 if (!ext3_check_descriptors (sb)) {
1471 printk (KERN_ERR "EXT3-fs: group descriptors corrupted !\n");
1474 sbi->s_gdb_count = db_count;
1475 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1476 spin_lock_init(&sbi->s_next_gen_lock);
1478 * set up enough so that it can read an inode
1480 sb->s_op = &ext3_sops;
1481 sb->s_export_op = &ext3_export_ops;
1483 sb->s_qcop = &ext3_qctl_operations;
1484 sb->dq_op = &ext3_quota_operations;
1486 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1490 needs_recovery = (es->s_last_orphan != 0 ||
1491 EXT3_HAS_INCOMPAT_FEATURE(sb,
1492 EXT3_FEATURE_INCOMPAT_RECOVER));
1495 * The first inode we look at is the journal inode. Don't try
1496 * root first: it may be modified in the journal!
1498 if (!test_opt(sb, NOLOAD) &&
1499 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1500 if (ext3_load_journal(sb, es))
1502 } else if (journal_inum) {
1503 if (ext3_create_journal(sb, es, journal_inum))
1508 "ext3: No journal on filesystem on %s\n",
1513 /* We have now updated the journal if required, so we can
1514 * validate the data journaling mode. */
1515 switch (test_opt(sb, DATA_FLAGS)) {
1517 /* No mode set, assume a default based on the journal
1518 capabilities: ORDERED_DATA if the journal can
1519 cope, else JOURNAL_DATA */
1520 if (journal_check_available_features
1521 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1522 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1524 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1527 case EXT3_MOUNT_ORDERED_DATA:
1528 case EXT3_MOUNT_WRITEBACK_DATA:
1529 if (!journal_check_available_features
1530 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1531 printk(KERN_ERR "EXT3-fs: Journal does not support "
1532 "requested data journaling mode\n");
1540 * The journal_load will have done any necessary log recovery,
1541 * so we can safely mount the rest of the filesystem now.
1544 root = iget(sb, EXT3_ROOT_INO);
1545 sb->s_root = d_alloc_root(root);
1547 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1551 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1554 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1558 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1560 * akpm: core read_super() calls in here with the superblock locked.
1561 * That deadlocks, because orphan cleanup needs to lock the superblock
1562 * in numerous places. Here we just pop the lock - it's relatively
1563 * harmless, because we are now ready to accept write_super() requests,
1564 * and aviro says that's the only reason for hanging onto the
1567 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1568 ext3_orphan_cleanup(sb, es);
1569 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1571 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1572 ext3_mark_recovery_complete(sb, es);
1573 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1574 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1575 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1578 percpu_counter_mod(&sbi->s_freeblocks_counter,
1579 ext3_count_free_blocks(sb));
1580 percpu_counter_mod(&sbi->s_freeinodes_counter,
1581 ext3_count_free_inodes(sb));
1582 percpu_counter_mod(&sbi->s_dirs_counter,
1583 ext3_count_dirs(sb));
1588 journal_destroy(sbi->s_journal);
1590 kfree(sbi->s_debts);
1591 for (i = 0; i < db_count; i++)
1592 brelse(sbi->s_group_desc[i]);
1593 kfree(sbi->s_group_desc);
1596 for (i = 0; i < MAXQUOTAS; i++) {
1597 if (sbi->s_qf_names[i])
1598 kfree(sbi->s_qf_names[i]);
1601 ext3_blkdev_remove(sbi);
1604 sb->s_fs_info = NULL;
1610 * Setup any per-fs journal parameters now. We'll do this both on
1611 * initial mount, once the journal has been initialised but before we've
1612 * done any recovery; and again on any subsequent remount.
1614 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1616 struct ext3_sb_info *sbi = EXT3_SB(sb);
1618 if (sbi->s_commit_interval)
1619 journal->j_commit_interval = sbi->s_commit_interval;
1620 /* We could also set up an ext3-specific default for the commit
1621 * interval here, but for now we'll just fall back to the jbd
1624 spin_lock(&journal->j_state_lock);
1625 if (test_opt(sb, BARRIER))
1626 journal->j_flags |= JFS_BARRIER;
1628 journal->j_flags &= ~JFS_BARRIER;
1629 spin_unlock(&journal->j_state_lock);
1632 static journal_t *ext3_get_journal(struct super_block *sb, int journal_inum)
1634 struct inode *journal_inode;
1637 /* First, test for the existence of a valid inode on disk. Bad
1638 * things happen if we iget() an unused inode, as the subsequent
1639 * iput() will try to delete it. */
1641 journal_inode = iget(sb, journal_inum);
1642 if (!journal_inode) {
1643 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1646 if (!journal_inode->i_nlink) {
1647 make_bad_inode(journal_inode);
1648 iput(journal_inode);
1649 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1653 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1654 journal_inode, journal_inode->i_size);
1655 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1656 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1657 iput(journal_inode);
1661 journal = journal_init_inode(journal_inode);
1663 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1664 iput(journal_inode);
1667 journal->j_private = sb;
1668 ext3_init_journal_params(sb, journal);
1672 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1675 struct buffer_head * bh;
1679 int hblock, blocksize;
1680 unsigned long sb_block;
1681 unsigned long offset;
1682 struct ext3_super_block * es;
1683 struct block_device *bdev;
1685 bdev = ext3_blkdev_get(j_dev);
1689 if (bd_claim(bdev, sb)) {
1691 "EXT3: failed to claim external journal device.\n");
1696 blocksize = sb->s_blocksize;
1697 hblock = bdev_hardsect_size(bdev);
1698 if (blocksize < hblock) {
1700 "EXT3-fs: blocksize too small for journal device.\n");
1704 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1705 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1706 set_blocksize(bdev, blocksize);
1707 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1708 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1709 "external journal\n");
1713 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1714 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1715 !(le32_to_cpu(es->s_feature_incompat) &
1716 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1717 printk(KERN_ERR "EXT3-fs: external journal has "
1718 "bad superblock\n");
1723 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1724 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1729 len = le32_to_cpu(es->s_blocks_count);
1730 start = sb_block + 1;
1731 brelse(bh); /* we're done with the superblock */
1733 journal = journal_init_dev(bdev, sb->s_bdev,
1734 start, len, blocksize);
1736 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1739 journal->j_private = sb;
1740 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1741 wait_on_buffer(journal->j_sb_buffer);
1742 if (!buffer_uptodate(journal->j_sb_buffer)) {
1743 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1746 if (ntohl(journal->j_superblock->s_nr_users) != 1) {
1747 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1748 "user (unsupported) - %d\n",
1749 ntohl(journal->j_superblock->s_nr_users));
1752 EXT3_SB(sb)->journal_bdev = bdev;
1753 ext3_init_journal_params(sb, journal);
1756 journal_destroy(journal);
1758 ext3_blkdev_put(bdev);
1762 static int ext3_load_journal(struct super_block * sb,
1763 struct ext3_super_block * es)
1766 int journal_inum = le32_to_cpu(es->s_journal_inum);
1767 dev_t journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
1769 int really_read_only;
1771 really_read_only = bdev_read_only(sb->s_bdev);
1774 * Are we loading a blank journal or performing recovery after a
1775 * crash? For recovery, we need to check in advance whether we
1776 * can get read-write access to the device.
1779 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
1780 if (sb->s_flags & MS_RDONLY) {
1781 printk(KERN_INFO "EXT3-fs: INFO: recovery "
1782 "required on readonly filesystem.\n");
1783 if (really_read_only) {
1784 printk(KERN_ERR "EXT3-fs: write access "
1785 "unavailable, cannot proceed.\n");
1788 printk (KERN_INFO "EXT3-fs: write access will "
1789 "be enabled during recovery.\n");
1793 if (journal_inum && journal_dev) {
1794 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
1795 "and inode journals!\n");
1800 if (!(journal = ext3_get_journal(sb, journal_inum)))
1803 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
1807 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
1808 err = journal_update_format(journal);
1810 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
1811 journal_destroy(journal);
1816 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
1817 err = journal_wipe(journal, !really_read_only);
1819 err = journal_load(journal);
1822 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
1823 journal_destroy(journal);
1827 EXT3_SB(sb)->s_journal = journal;
1828 ext3_clear_journal_err(sb, es);
1832 static int ext3_create_journal(struct super_block * sb,
1833 struct ext3_super_block * es,
1838 if (sb->s_flags & MS_RDONLY) {
1839 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
1840 "create journal.\n");
1844 if (!(journal = ext3_get_journal(sb, journal_inum)))
1847 printk(KERN_INFO "EXT3-fs: creating new journal on inode %d\n",
1850 if (journal_create(journal)) {
1851 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
1852 journal_destroy(journal);
1856 EXT3_SB(sb)->s_journal = journal;
1858 ext3_update_dynamic_rev(sb);
1859 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1860 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
1862 es->s_journal_inum = cpu_to_le32(journal_inum);
1865 /* Make sure we flush the recovery flag to disk. */
1866 ext3_commit_super(sb, es, 1);
1871 static void ext3_commit_super (struct super_block * sb,
1872 struct ext3_super_block * es,
1875 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
1879 es->s_wtime = cpu_to_le32(get_seconds());
1880 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
1881 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
1882 BUFFER_TRACE(sbh, "marking dirty");
1883 mark_buffer_dirty(sbh);
1885 sync_dirty_buffer(sbh);
1890 * Have we just finished recovery? If so, and if we are mounting (or
1891 * remounting) the filesystem readonly, then we will end up with a
1892 * consistent fs on disk. Record that fact.
1894 static void ext3_mark_recovery_complete(struct super_block * sb,
1895 struct ext3_super_block * es)
1897 journal_t *journal = EXT3_SB(sb)->s_journal;
1899 journal_lock_updates(journal);
1900 journal_flush(journal);
1901 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
1902 sb->s_flags & MS_RDONLY) {
1903 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1905 ext3_commit_super(sb, es, 1);
1907 journal_unlock_updates(journal);
1911 * If we are mounting (or read-write remounting) a filesystem whose journal
1912 * has recorded an error from a previous lifetime, move that error to the
1913 * main filesystem now.
1915 static void ext3_clear_journal_err(struct super_block * sb,
1916 struct ext3_super_block * es)
1922 journal = EXT3_SB(sb)->s_journal;
1925 * Now check for any error status which may have been recorded in the
1926 * journal by a prior ext3_error() or ext3_abort()
1929 j_errno = journal_errno(journal);
1933 errstr = ext3_decode_error(sb, j_errno, nbuf);
1934 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
1935 "from previous mount: %s", errstr);
1936 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
1937 "filesystem check.");
1939 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
1940 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
1941 ext3_commit_super (sb, es, 1);
1943 journal_clear_err(journal);
1948 * Force the running and committing transactions to commit,
1949 * and wait on the commit.
1951 int ext3_force_commit(struct super_block *sb)
1956 if (sb->s_flags & MS_RDONLY)
1959 journal = EXT3_SB(sb)->s_journal;
1961 ret = ext3_journal_force_commit(journal);
1966 * Ext3 always journals updates to the superblock itself, so we don't
1967 * have to propagate any other updates to the superblock on disk at this
1968 * point. Just start an async writeback to get the buffers on their way
1971 * This implicitly triggers the writebehind on sync().
1974 void ext3_write_super (struct super_block * sb)
1976 if (down_trylock(&sb->s_lock) == 0)
1981 static int ext3_sync_fs(struct super_block *sb, int wait)
1986 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
1988 log_wait_commit(EXT3_SB(sb)->s_journal, target);
1994 * LVM calls this function before a (read-only) snapshot is created. This
1995 * gives us a chance to flush the journal completely and mark the fs clean.
1997 void ext3_write_super_lockfs(struct super_block *sb)
2001 if (!(sb->s_flags & MS_RDONLY)) {
2002 journal_t *journal = EXT3_SB(sb)->s_journal;
2004 /* Now we set up the journal barrier. */
2005 journal_lock_updates(journal);
2006 journal_flush(journal);
2008 /* Journal blocked and flushed, clear needs_recovery flag. */
2009 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2010 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2015 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2016 * flag here, even though the filesystem is not technically dirty yet.
2018 void ext3_unlockfs(struct super_block *sb)
2020 if (!(sb->s_flags & MS_RDONLY)) {
2022 /* Reser the needs_recovery flag before the fs is unlocked. */
2023 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2024 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2026 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2030 int ext3_remount (struct super_block * sb, int * flags, char * data)
2032 struct ext3_super_block * es;
2033 struct ext3_sb_info *sbi = EXT3_SB(sb);
2037 * Allow the "check" option to be passed as a remount option.
2039 if (!parse_options(data, sb, &tmp, 1))
2042 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2043 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2045 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2046 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2050 ext3_init_journal_params(sb, sbi->s_journal);
2052 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
2053 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2056 if (*flags & MS_RDONLY) {
2058 * First of all, the unconditional stuff we have to do
2059 * to disable replay of the journal when we next remount
2061 sb->s_flags |= MS_RDONLY;
2064 * OK, test if we are remounting a valid rw partition
2065 * readonly, and if so set the rdonly flag and then
2066 * mark the partition as valid again.
2068 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2069 (sbi->s_mount_state & EXT3_VALID_FS))
2070 es->s_state = cpu_to_le16(sbi->s_mount_state);
2072 ext3_mark_recovery_complete(sb, es);
2075 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2076 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2077 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2078 "remount RDWR because of unsupported "
2079 "optional features (%x).\n",
2084 * Mounting a RDONLY partition read-write, so reread
2085 * and store the current valid flag. (It may have
2086 * been changed by e2fsck since we originally mounted
2089 ext3_clear_journal_err(sb, es);
2090 sbi->s_mount_state = le16_to_cpu(es->s_state);
2091 if (!ext3_setup_super (sb, es, 0))
2092 sb->s_flags &= ~MS_RDONLY;
2098 int ext3_statfs (struct super_block * sb, struct kstatfs * buf)
2100 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
2101 unsigned long overhead;
2104 if (test_opt (sb, MINIX_DF))
2108 * Compute the overhead (FS structures)
2112 * All of the blocks before first_data_block are
2115 overhead = le32_to_cpu(es->s_first_data_block);
2118 * Add the overhead attributed to the superblock and
2119 * block group descriptors. If the sparse superblocks
2120 * feature is turned on, then not all groups have this.
2122 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++)
2123 overhead += ext3_bg_has_super(sb, i) +
2124 ext3_bg_num_gdb(sb, i);
2127 * Every block group has an inode bitmap, a block
2128 * bitmap, and an inode table.
2130 overhead += (EXT3_SB(sb)->s_groups_count *
2131 (2 + EXT3_SB(sb)->s_itb_per_group));
2134 buf->f_type = EXT3_SUPER_MAGIC;
2135 buf->f_bsize = sb->s_blocksize;
2136 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2137 buf->f_bfree = ext3_count_free_blocks (sb);
2138 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2139 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2141 buf->f_files = le32_to_cpu(es->s_inodes_count);
2142 buf->f_ffree = ext3_count_free_inodes (sb);
2143 buf->f_namelen = EXT3_NAME_LEN;
2147 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2148 * is locked for write. Otherwise the are possible deadlocks:
2149 * Process 1 Process 2
2150 * ext3_create() quota_sync()
2151 * journal_start() write_dquot()
2152 * DQUOT_INIT() down(dqio_sem)
2153 * down(dqio_sem) journal_start()
2159 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2161 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]->f_dentry->d_inode;
2164 static int ext3_dquot_initialize(struct inode *inode, int type)
2169 /* We may create quota structure so we need to reserve enough blocks */
2170 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS);
2172 return PTR_ERR(handle);
2173 ret = dquot_initialize(inode, type);
2174 err = ext3_journal_stop(handle);
2180 static int ext3_dquot_drop(struct inode *inode)
2185 /* We may delete quota structure so we need to reserve enough blocks */
2186 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS);
2188 return PTR_ERR(handle);
2189 ret = dquot_drop(inode);
2190 err = ext3_journal_stop(handle);
2196 static int ext3_write_dquot(struct dquot *dquot)
2201 handle = ext3_journal_start(dquot_to_inode(dquot),
2202 EXT3_QUOTA_TRANS_BLOCKS);
2204 return PTR_ERR(handle);
2205 ret = dquot_commit(dquot);
2206 err = ext3_journal_stop(handle);
2212 static int ext3_acquire_dquot(struct dquot *dquot)
2217 handle = ext3_journal_start(dquot_to_inode(dquot),
2218 EXT3_QUOTA_INIT_BLOCKS);
2220 return PTR_ERR(handle);
2221 ret = dquot_acquire(dquot);
2222 err = ext3_journal_stop(handle);
2228 static int ext3_release_dquot(struct dquot *dquot)
2233 handle = ext3_journal_start(dquot_to_inode(dquot),
2234 EXT3_QUOTA_INIT_BLOCKS);
2236 return PTR_ERR(handle);
2237 ret = dquot_release(dquot);
2238 err = ext3_journal_stop(handle);
2244 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2246 /* Are we journalling quotas? */
2247 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2248 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2249 dquot_mark_dquot_dirty(dquot);
2250 return ext3_write_dquot(dquot);
2252 return dquot_mark_dquot_dirty(dquot);
2256 static int ext3_write_info(struct super_block *sb, int type)
2261 /* Data block + inode block */
2262 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2264 return PTR_ERR(handle);
2265 ret = dquot_commit_info(sb, type);
2266 err = ext3_journal_stop(handle);
2273 * Turn on quotas during mount time - we need to find
2274 * the quota file and such...
2276 static int ext3_quota_on_mount(struct super_block *sb, int type)
2279 struct dentry *dentry;
2280 struct qstr name = { .name = EXT3_SB(sb)->s_qf_names[type],
2282 .len = strlen(EXT3_SB(sb)->s_qf_names[type])};
2284 dentry = lookup_hash(&name, sb->s_root);
2286 return PTR_ERR(dentry);
2287 err = vfs_quota_on_mount(type, EXT3_SB(sb)->s_jquota_fmt, dentry);
2290 /* We keep the dentry reference if everything went ok - we drop it
2291 * on quota_off time */
2295 /* Turn quotas off during mount time */
2296 static int ext3_quota_off_mount(struct super_block *sb, int type)
2299 struct dentry *dentry;
2301 dentry = sb_dqopt(sb)->files[type]->f_dentry;
2302 err = vfs_quota_off_mount(sb, type);
2303 /* We invalidate dentry - it has at least wrong hash... */
2304 d_invalidate(dentry);
2310 * Standard function to be called on quota_on
2312 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2316 struct nameidata nd;
2318 /* Not journalling quota? */
2319 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2320 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2321 return vfs_quota_on(sb, type, format_id, path);
2322 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2325 /* Quotafile not on the same filesystem? */
2326 if (nd.mnt->mnt_sb != sb)
2328 /* Quotafile not of fs root? */
2329 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2331 "EXT3-fs: Quota file not on filesystem root. "
2332 "Journalled quota will not work.\n");
2333 if (!ext3_should_journal_data(nd.dentry->d_inode))
2334 printk(KERN_WARNING "EXT3-fs: Quota file does not have "
2335 "data-journalling. Journalled quota will not work.\n");
2337 return vfs_quota_on(sb, type, format_id, path);
2342 static struct super_block *ext3_get_sb(struct file_system_type *fs_type,
2343 int flags, const char *dev_name, void *data)
2345 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
2348 static struct file_system_type ext3_fs_type = {
2349 .owner = THIS_MODULE,
2351 .get_sb = ext3_get_sb,
2352 .kill_sb = kill_block_super,
2353 .fs_flags = FS_REQUIRES_DEV,
2356 static int __init init_ext3_fs(void)
2358 int err = init_ext3_xattr();
2361 err = init_inodecache();
2364 err = register_filesystem_lifo(&ext3_fs_type);
2369 destroy_inodecache();
2375 static void __exit exit_ext3_fs(void)
2377 unregister_filesystem(&ext3_fs_type);
2378 destroy_inodecache();
2382 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2383 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2384 MODULE_LICENSE("GPL");
2385 module_init(init_ext3_fs)
2386 module_exit(exit_ext3_fs)