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 = 0;
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 static void ext3_clear_inode(struct inode *inode)
498 #ifdef CONFIG_EXT3_FS_POSIX_ACL
499 if (EXT3_I(inode)->i_acl &&
500 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
501 posix_acl_release(EXT3_I(inode)->i_acl);
502 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
504 if (EXT3_I(inode)->i_default_acl &&
505 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
506 posix_acl_release(EXT3_I(inode)->i_default_acl);
507 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
510 if (!is_bad_inode(inode))
511 ext3_discard_reservation(inode);
516 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
517 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
519 static int ext3_dquot_initialize(struct inode *inode, int type);
520 static int ext3_dquot_drop(struct inode *inode);
521 static int ext3_write_dquot(struct dquot *dquot);
522 static int ext3_acquire_dquot(struct dquot *dquot);
523 static int ext3_release_dquot(struct dquot *dquot);
524 static int ext3_mark_dquot_dirty(struct dquot *dquot);
525 static int ext3_write_info(struct super_block *sb, int type);
526 static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
527 static int ext3_quota_on_mount(struct super_block *sb, int type);
528 static int ext3_quota_off_mount(struct super_block *sb, int type);
530 static struct dquot_operations ext3_quota_operations = {
531 .initialize = ext3_dquot_initialize,
532 .drop = ext3_dquot_drop,
533 .alloc_space = dquot_alloc_space,
534 .alloc_inode = dquot_alloc_inode,
535 .free_space = dquot_free_space,
536 .free_inode = dquot_free_inode,
537 .transfer = dquot_transfer,
538 .write_dquot = ext3_write_dquot,
539 .acquire_dquot = ext3_acquire_dquot,
540 .release_dquot = ext3_release_dquot,
541 .mark_dirty = ext3_mark_dquot_dirty,
542 .write_info = ext3_write_info
545 static struct quotactl_ops ext3_qctl_operations = {
546 .quota_on = ext3_quota_on,
547 .quota_off = vfs_quota_off,
548 .quota_sync = vfs_quota_sync,
549 .get_info = vfs_get_dqinfo,
550 .set_info = vfs_set_dqinfo,
551 .get_dqblk = vfs_get_dqblk,
552 .set_dqblk = vfs_set_dqblk
556 static struct super_operations ext3_sops = {
557 .alloc_inode = ext3_alloc_inode,
558 .destroy_inode = ext3_destroy_inode,
559 .read_inode = ext3_read_inode,
560 .write_inode = ext3_write_inode,
561 .dirty_inode = ext3_dirty_inode,
562 .delete_inode = ext3_delete_inode,
563 .put_super = ext3_put_super,
564 .write_super = ext3_write_super,
565 .sync_fs = ext3_sync_fs,
566 .write_super_lockfs = ext3_write_super_lockfs,
567 .unlockfs = ext3_unlockfs,
568 .statfs = ext3_statfs,
569 .remount_fs = ext3_remount,
570 .clear_inode = ext3_clear_inode,
573 struct dentry *ext3_get_parent(struct dentry *child);
574 static struct export_operations ext3_export_ops = {
575 .get_parent = ext3_get_parent,
579 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
580 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
581 Opt_nouid32, Opt_check, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
582 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
583 Opt_reservation, Opt_noreservation, Opt_noload,
584 Opt_commit, Opt_journal_update, Opt_journal_inum,
585 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
586 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
587 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0,
588 Opt_ignore, Opt_err, Opt_resize,
591 static match_table_t tokens = {
592 {Opt_bsd_df, "bsddf"},
593 {Opt_minix_df, "minixdf"},
594 {Opt_grpid, "grpid"},
595 {Opt_grpid, "bsdgroups"},
596 {Opt_nogrpid, "nogrpid"},
597 {Opt_nogrpid, "sysvgroups"},
598 {Opt_resgid, "resgid=%u"},
599 {Opt_resuid, "resuid=%u"},
601 {Opt_err_cont, "errors=continue"},
602 {Opt_err_panic, "errors=panic"},
603 {Opt_err_ro, "errors=remount-ro"},
604 {Opt_nouid32, "nouid32"},
605 {Opt_nocheck, "nocheck"},
606 {Opt_nocheck, "check=none"},
607 {Opt_check, "check"},
608 {Opt_debug, "debug"},
609 {Opt_oldalloc, "oldalloc"},
610 {Opt_orlov, "orlov"},
611 {Opt_user_xattr, "user_xattr"},
612 {Opt_nouser_xattr, "nouser_xattr"},
614 {Opt_noacl, "noacl"},
615 {Opt_reservation, "reservation"},
616 {Opt_noreservation, "noreservation"},
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"},
636 {Opt_resize, "resize"}
640 static unsigned long get_sb_block(void **data)
642 unsigned long sb_block;
643 char *options = (char *) *data;
645 if (!options || strncmp(options, "sb=", 3) != 0)
646 return 1; /* Default location */
648 sb_block = simple_strtoul(options, &options, 0);
649 if (*options && *options != ',') {
650 printk("EXT3-fs: Invalid sb specification: %s\n",
656 *data = (void *) options;
660 static int parse_options (char * options, struct super_block *sb,
661 unsigned long * inum, unsigned long *n_blocks_count, int is_remount)
663 struct ext3_sb_info *sbi = EXT3_SB(sb);
665 substring_t args[MAX_OPT_ARGS];
675 while ((p = strsep (&options, ",")) != NULL) {
680 token = match_token(p, tokens, args);
683 clear_opt (sbi->s_mount_opt, MINIX_DF);
686 set_opt (sbi->s_mount_opt, MINIX_DF);
689 set_opt (sbi->s_mount_opt, GRPID);
692 clear_opt (sbi->s_mount_opt, GRPID);
695 if (match_int(&args[0], &option))
697 sbi->s_resuid = option;
700 if (match_int(&args[0], &option))
702 sbi->s_resgid = option;
705 /* handled by get_sb_block() instead of here */
706 /* *sb_block = match_int(&args[0]); */
709 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
710 clear_opt (sbi->s_mount_opt, ERRORS_RO);
711 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
714 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
715 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
716 set_opt (sbi->s_mount_opt, ERRORS_RO);
719 clear_opt (sbi->s_mount_opt, ERRORS_RO);
720 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
721 set_opt (sbi->s_mount_opt, ERRORS_CONT);
724 set_opt (sbi->s_mount_opt, NO_UID32);
727 #ifdef CONFIG_EXT3_CHECK
728 set_opt (sbi->s_mount_opt, CHECK);
731 "EXT3 Check option not supported\n");
735 clear_opt (sbi->s_mount_opt, CHECK);
738 set_opt (sbi->s_mount_opt, DEBUG);
741 set_opt (sbi->s_mount_opt, OLDALLOC);
744 clear_opt (sbi->s_mount_opt, OLDALLOC);
746 #ifdef CONFIG_EXT3_FS_XATTR
748 set_opt (sbi->s_mount_opt, XATTR_USER);
750 case Opt_nouser_xattr:
751 clear_opt (sbi->s_mount_opt, XATTR_USER);
755 case Opt_nouser_xattr:
756 printk("EXT3 (no)user_xattr options not supported\n");
759 #ifdef CONFIG_EXT3_FS_POSIX_ACL
761 set_opt(sbi->s_mount_opt, POSIX_ACL);
764 clear_opt(sbi->s_mount_opt, POSIX_ACL);
769 printk("EXT3 (no)acl options not supported\n");
772 case Opt_reservation:
773 set_opt(sbi->s_mount_opt, RESERVATION);
775 case Opt_noreservation:
776 clear_opt(sbi->s_mount_opt, RESERVATION);
778 case Opt_journal_update:
780 /* Eventually we will want to be able to create
781 a journal file here. For now, only allow the
782 user to specify an existing inode to be the
785 printk(KERN_ERR "EXT3-fs: cannot specify "
786 "journal on remount\n");
789 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
791 case Opt_journal_inum:
793 printk(KERN_ERR "EXT3-fs: cannot specify "
794 "journal on remount\n");
797 if (match_int(&args[0], &option))
802 set_opt (sbi->s_mount_opt, NOLOAD);
805 if (match_int(&args[0], &option))
810 option = JBD_DEFAULT_MAX_COMMIT_AGE;
811 sbi->s_commit_interval = HZ * option;
813 case Opt_data_journal:
814 data_opt = EXT3_MOUNT_JOURNAL_DATA;
816 case Opt_data_ordered:
817 data_opt = EXT3_MOUNT_ORDERED_DATA;
819 case Opt_data_writeback:
820 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
823 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
826 "EXT3-fs: cannot change data "
827 "mode on remount\n");
831 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
832 sbi->s_mount_opt |= data_opt;
842 if (sb_any_quota_enabled(sb)) {
844 "EXT3-fs: Cannot change journalled "
845 "quota options when quota turned on.\n");
848 if (sbi->s_qf_names[qtype]) {
850 "EXT3-fs: %s quota file already "
851 "specified.\n", QTYPE2NAME(qtype));
854 sbi->s_qf_names[qtype] = match_strdup(&args[0]);
855 if (!sbi->s_qf_names[qtype]) {
857 "EXT3-fs: not enough memory for "
858 "storing quotafile name.\n");
861 if (strchr(sbi->s_qf_names[qtype], '/')) {
863 "EXT3-fs: quotafile must be on "
864 "filesystem root.\n");
865 kfree(sbi->s_qf_names[qtype]);
866 sbi->s_qf_names[qtype] = NULL;
870 case Opt_offusrjquota:
873 case Opt_offgrpjquota:
876 if (sb_any_quota_enabled(sb)) {
877 printk(KERN_ERR "EXT3-fs: Cannot change "
878 "journalled quota options when "
879 "quota turned on.\n");
882 if (sbi->s_qf_names[qtype]) {
883 kfree(sbi->s_qf_names[qtype]);
884 sbi->s_qf_names[qtype] = NULL;
887 case Opt_jqfmt_vfsold:
888 sbi->s_jquota_fmt = QFMT_VFS_OLD;
890 case Opt_jqfmt_vfsv0:
891 sbi->s_jquota_fmt = QFMT_VFS_V0;
896 case Opt_offusrjquota:
897 case Opt_offgrpjquota:
898 case Opt_jqfmt_vfsold:
899 case Opt_jqfmt_vfsv0:
901 "EXT3-fs: journalled quota options not "
906 set_opt(sbi->s_mount_opt, ABORT);
911 if (!n_blocks_count) {
912 printk("EXT3-fs: resize option only available "
916 match_int(&args[0], &option);
917 *n_blocks_count = option;
921 "EXT3-fs: Unrecognized mount option \"%s\" "
922 "or missing value\n", p);
927 if (!sbi->s_jquota_fmt && (sbi->s_qf_names[USRQUOTA] ||
928 sbi->s_qf_names[GRPQUOTA])) {
930 "EXT3-fs: journalled quota format not specified.\n");
938 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
941 struct ext3_sb_info *sbi = EXT3_SB(sb);
944 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
945 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
946 "forcing read-only mode\n");
951 if (!(sbi->s_mount_state & EXT3_VALID_FS))
952 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
953 "running e2fsck is recommended\n");
954 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
956 "EXT3-fs warning: mounting fs with errors, "
957 "running e2fsck is recommended\n");
958 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
959 le16_to_cpu(es->s_mnt_count) >=
960 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
962 "EXT3-fs warning: maximal mount count reached, "
963 "running e2fsck is recommended\n");
964 else if (le32_to_cpu(es->s_checkinterval) &&
965 (le32_to_cpu(es->s_lastcheck) +
966 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
968 "EXT3-fs warning: checktime reached, "
969 "running e2fsck is recommended\n");
971 /* @@@ We _will_ want to clear the valid bit if we find
972 inconsistencies, to force a fsck at reboot. But for
973 a plain journaled filesystem we can keep it set as
975 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
977 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
978 es->s_max_mnt_count =
979 (__s16) cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
980 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
981 es->s_mtime = cpu_to_le32(get_seconds());
982 ext3_update_dynamic_rev(sb);
983 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
985 ext3_commit_super(sb, es, 1);
986 if (test_opt(sb, DEBUG))
987 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
988 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
991 EXT3_BLOCKS_PER_GROUP(sb),
992 EXT3_INODES_PER_GROUP(sb),
995 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
996 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
997 char b[BDEVNAME_SIZE];
999 printk("external journal on %s\n",
1000 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1002 printk("internal journal\n");
1004 #ifdef CONFIG_EXT3_CHECK
1005 if (test_opt (sb, CHECK)) {
1006 ext3_check_blocks_bitmap (sb);
1007 ext3_check_inodes_bitmap (sb);
1013 static int ext3_check_descriptors (struct super_block * sb)
1015 struct ext3_sb_info *sbi = EXT3_SB(sb);
1016 unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
1017 struct ext3_group_desc * gdp = NULL;
1021 ext3_debug ("Checking group descriptors");
1023 for (i = 0; i < sbi->s_groups_count; i++)
1025 if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
1026 gdp = (struct ext3_group_desc *)
1027 sbi->s_group_desc[desc_block++]->b_data;
1028 if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
1029 le32_to_cpu(gdp->bg_block_bitmap) >=
1030 block + EXT3_BLOCKS_PER_GROUP(sb))
1032 ext3_error (sb, "ext3_check_descriptors",
1033 "Block bitmap for group %d"
1034 " not in group (block %lu)!",
1036 le32_to_cpu(gdp->bg_block_bitmap));
1039 if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
1040 le32_to_cpu(gdp->bg_inode_bitmap) >=
1041 block + EXT3_BLOCKS_PER_GROUP(sb))
1043 ext3_error (sb, "ext3_check_descriptors",
1044 "Inode bitmap for group %d"
1045 " not in group (block %lu)!",
1047 le32_to_cpu(gdp->bg_inode_bitmap));
1050 if (le32_to_cpu(gdp->bg_inode_table) < block ||
1051 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
1052 block + EXT3_BLOCKS_PER_GROUP(sb))
1054 ext3_error (sb, "ext3_check_descriptors",
1055 "Inode table for group %d"
1056 " not in group (block %lu)!",
1058 le32_to_cpu(gdp->bg_inode_table));
1061 block += EXT3_BLOCKS_PER_GROUP(sb);
1065 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1066 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1071 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1072 * the superblock) which were deleted from all directories, but held open by
1073 * a process at the time of a crash. We walk the list and try to delete these
1074 * inodes at recovery time (only with a read-write filesystem).
1076 * In order to keep the orphan inode chain consistent during traversal (in
1077 * case of crash during recovery), we link each inode into the superblock
1078 * orphan list_head and handle it the same way as an inode deletion during
1079 * normal operation (which journals the operations for us).
1081 * We only do an iget() and an iput() on each inode, which is very safe if we
1082 * accidentally point at an in-use or already deleted inode. The worst that
1083 * can happen in this case is that we get a "bit already cleared" message from
1084 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1085 * e2fsck was run on this filesystem, and it must have already done the orphan
1086 * inode cleanup for us, so we can safely abort without any further action.
1088 static void ext3_orphan_cleanup (struct super_block * sb,
1089 struct ext3_super_block * es)
1091 unsigned int s_flags = sb->s_flags;
1092 int nr_orphans = 0, nr_truncates = 0;
1096 if (!es->s_last_orphan) {
1097 jbd_debug(4, "no orphan inodes to clean up\n");
1101 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1102 if (es->s_last_orphan)
1103 jbd_debug(1, "Errors on filesystem, "
1104 "clearing orphan list.\n");
1105 es->s_last_orphan = 0;
1106 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1110 if (s_flags & MS_RDONLY) {
1111 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1113 sb->s_flags &= ~MS_RDONLY;
1116 /* Needed for iput() to work correctly and not trash data */
1117 sb->s_flags |= MS_ACTIVE;
1118 /* Turn on quotas so that they are updated correctly */
1119 for (i = 0; i < MAXQUOTAS; i++) {
1120 if (EXT3_SB(sb)->s_qf_names[i]) {
1121 int ret = ext3_quota_on_mount(sb, i);
1124 "EXT3-fs: Cannot turn on journalled "
1125 "quota: error %d\n", ret);
1130 while (es->s_last_orphan) {
1131 struct inode *inode;
1134 ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1135 es->s_last_orphan = 0;
1139 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1141 if (inode->i_nlink) {
1143 "%s: truncating inode %ld to %Ld bytes\n",
1144 __FUNCTION__, inode->i_ino, inode->i_size);
1145 jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
1146 inode->i_ino, inode->i_size);
1147 ext3_truncate(inode);
1151 "%s: deleting unreferenced inode %ld\n",
1152 __FUNCTION__, inode->i_ino);
1153 jbd_debug(2, "deleting unreferenced inode %ld\n",
1157 iput(inode); /* The delete magic happens here! */
1160 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1163 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1164 sb->s_id, PLURAL(nr_orphans));
1166 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1167 sb->s_id, PLURAL(nr_truncates));
1169 /* Turn quotas off */
1170 for (i = 0; i < MAXQUOTAS; i++) {
1171 if (sb_dqopt(sb)->files[i])
1172 ext3_quota_off_mount(sb, i);
1175 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1178 #define log2(n) ffz(~(n))
1181 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1182 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1183 * We need to be 1 filesystem block less than the 2^32 sector limit.
1185 static loff_t ext3_max_size(int bits)
1187 loff_t res = EXT3_NDIR_BLOCKS;
1188 res += 1LL << (bits-2);
1189 res += 1LL << (2*(bits-2));
1190 res += 1LL << (3*(bits-2));
1192 if (res > (512LL << 32) - (1 << bits))
1193 res = (512LL << 32) - (1 << bits);
1197 static unsigned long descriptor_loc(struct super_block *sb,
1198 unsigned long logic_sb_block,
1201 struct ext3_sb_info *sbi = EXT3_SB(sb);
1202 unsigned long bg, first_data_block, first_meta_bg;
1205 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1206 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1208 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1210 return (logic_sb_block + nr + 1);
1211 bg = sbi->s_desc_per_block * nr;
1212 if (ext3_bg_has_super(sb, bg))
1214 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
1218 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1220 struct buffer_head * bh;
1221 struct ext3_super_block *es = 0;
1222 struct ext3_sb_info *sbi;
1223 unsigned long block;
1224 unsigned long sb_block = get_sb_block(&data);
1225 unsigned long logic_sb_block;
1226 unsigned long offset = 0;
1227 unsigned long journal_inum = 0;
1228 unsigned long def_mount_opts;
1236 sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
1239 sb->s_fs_info = sbi;
1240 memset(sbi, 0, sizeof(*sbi));
1241 sbi->s_mount_opt = 0;
1242 sbi->s_resuid = EXT3_DEF_RESUID;
1243 sbi->s_resgid = EXT3_DEF_RESGID;
1245 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1247 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1252 * The ext3 superblock will not be buffer aligned for other than 1kB
1253 * block sizes. We need to calculate the offset from buffer start.
1255 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1256 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1257 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1259 logic_sb_block = sb_block;
1262 if (!(bh = sb_bread(sb, logic_sb_block))) {
1263 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1267 * Note: s_es must be initialized as soon as possible because
1268 * some ext3 macro-instructions depend on its value
1270 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1272 sb->s_magic = le16_to_cpu(es->s_magic);
1273 if (sb->s_magic != EXT3_SUPER_MAGIC) {
1276 "VFS: Can't find ext3 filesystem on dev %s.\n",
1281 /* Set defaults before we parse the mount options */
1282 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1283 if (def_mount_opts & EXT3_DEFM_DEBUG)
1284 set_opt(sbi->s_mount_opt, DEBUG);
1285 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1286 set_opt(sbi->s_mount_opt, GRPID);
1287 if (def_mount_opts & EXT3_DEFM_UID16)
1288 set_opt(sbi->s_mount_opt, NO_UID32);
1289 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1290 set_opt(sbi->s_mount_opt, XATTR_USER);
1291 if (def_mount_opts & EXT3_DEFM_ACL)
1292 set_opt(sbi->s_mount_opt, POSIX_ACL);
1293 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1294 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1295 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1296 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1297 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1298 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1300 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1301 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1302 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
1303 set_opt(sbi->s_mount_opt, ERRORS_RO);
1305 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1306 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1308 set_opt(sbi->s_mount_opt, RESERVATION);
1310 if (!parse_options ((char *) data, sb, &journal_inum, NULL, 0))
1313 sb->s_flags |= MS_ONE_SECOND;
1314 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1315 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1317 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1318 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1319 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1320 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1322 "EXT3-fs warning: feature flags set on rev 0 fs, "
1323 "running e2fsck is recommended\n");
1325 * Check feature flags regardless of the revision level, since we
1326 * previously didn't change the revision level when setting the flags,
1327 * so there is a chance incompat flags are set on a rev 0 filesystem.
1329 if ((i = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))) {
1330 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1331 "unsupported optional features (%x).\n",
1335 if (!(sb->s_flags & MS_RDONLY) &&
1336 (i = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))){
1337 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1338 "unsupported optional features (%x).\n",
1342 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1344 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1345 blocksize > EXT3_MAX_BLOCK_SIZE) {
1347 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1348 blocksize, sb->s_id);
1352 hblock = bdev_hardsect_size(sb->s_bdev);
1353 if (sb->s_blocksize != blocksize) {
1355 * Make sure the blocksize for the filesystem is larger
1356 * than the hardware sectorsize for the machine.
1358 if (blocksize < hblock) {
1359 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1360 "device blocksize %d.\n", blocksize, hblock);
1365 sb_set_blocksize(sb, blocksize);
1366 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1367 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1368 bh = sb_bread(sb, logic_sb_block);
1371 "EXT3-fs: Can't read superblock on 2nd try.\n");
1374 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1376 if (es->s_magic != le16_to_cpu(EXT3_SUPER_MAGIC)) {
1378 "EXT3-fs: Magic mismatch, very weird !\n");
1383 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1385 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1386 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1387 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1389 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1390 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1391 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1392 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1393 (sbi->s_inode_size > blocksize)) {
1395 "EXT3-fs: unsupported inode size: %d\n",
1400 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1401 le32_to_cpu(es->s_log_frag_size);
1402 if (blocksize != sbi->s_frag_size) {
1404 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1405 sbi->s_frag_size, blocksize);
1408 sbi->s_frags_per_block = 1;
1409 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1410 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1411 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1412 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1413 sbi->s_itb_per_group = sbi->s_inodes_per_group /sbi->s_inodes_per_block;
1414 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1416 sbi->s_mount_state = le16_to_cpu(es->s_state);
1417 sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
1418 sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
1419 for (i=0; i < 4; i++)
1420 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1421 sbi->s_def_hash_version = es->s_def_hash_version;
1423 if (sbi->s_blocks_per_group > blocksize * 8) {
1425 "EXT3-fs: #blocks per group too big: %lu\n",
1426 sbi->s_blocks_per_group);
1429 if (sbi->s_frags_per_group > blocksize * 8) {
1431 "EXT3-fs: #fragments per group too big: %lu\n",
1432 sbi->s_frags_per_group);
1435 if (sbi->s_inodes_per_group > blocksize * 8) {
1437 "EXT3-fs: #inodes per group too big: %lu\n",
1438 sbi->s_inodes_per_group);
1442 sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
1443 le32_to_cpu(es->s_first_data_block) +
1444 EXT3_BLOCKS_PER_GROUP(sb) - 1) /
1445 EXT3_BLOCKS_PER_GROUP(sb);
1446 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1447 EXT3_DESC_PER_BLOCK(sb);
1448 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1450 if (sbi->s_group_desc == NULL) {
1451 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1454 sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(u8),
1456 if (!sbi->s_debts) {
1457 printk("EXT3-fs: not enough memory to allocate s_bgi\n");
1460 memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(u8));
1462 percpu_counter_init(&sbi->s_freeblocks_counter);
1463 percpu_counter_init(&sbi->s_freeinodes_counter);
1464 percpu_counter_init(&sbi->s_dirs_counter);
1465 bgl_lock_init(&sbi->s_blockgroup_lock);
1467 for (i = 0; i < db_count; i++) {
1468 block = descriptor_loc(sb, logic_sb_block, i);
1469 sbi->s_group_desc[i] = sb_bread(sb, block);
1470 if (!sbi->s_group_desc[i]) {
1471 printk (KERN_ERR "EXT3-fs: "
1472 "can't read group descriptor %d\n", i);
1477 if (!ext3_check_descriptors (sb)) {
1478 printk (KERN_ERR "EXT3-fs: group descriptors corrupted !\n");
1481 sbi->s_gdb_count = db_count;
1482 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1483 spin_lock_init(&sbi->s_next_gen_lock);
1484 /* per fileystem reservation list head & lock */
1485 spin_lock_init(&sbi->s_rsv_window_lock);
1486 INIT_LIST_HEAD(&sbi->s_rsv_window_head.rsv_list);
1487 sbi->s_rsv_window_head.rsv_start = 0;
1488 sbi->s_rsv_window_head.rsv_end = 0;
1489 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1490 atomic_set(&sbi->s_rsv_window_head.rsv_goal_size, 0);
1493 * set up enough so that it can read an inode
1495 sb->s_op = &ext3_sops;
1496 sb->s_export_op = &ext3_export_ops;
1498 sb->s_qcop = &ext3_qctl_operations;
1499 sb->dq_op = &ext3_quota_operations;
1501 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1505 needs_recovery = (es->s_last_orphan != 0 ||
1506 EXT3_HAS_INCOMPAT_FEATURE(sb,
1507 EXT3_FEATURE_INCOMPAT_RECOVER));
1510 * The first inode we look at is the journal inode. Don't try
1511 * root first: it may be modified in the journal!
1513 if (!test_opt(sb, NOLOAD) &&
1514 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1515 if (ext3_load_journal(sb, es))
1517 } else if (journal_inum) {
1518 if (ext3_create_journal(sb, es, journal_inum))
1523 "ext3: No journal on filesystem on %s\n",
1528 /* We have now updated the journal if required, so we can
1529 * validate the data journaling mode. */
1530 switch (test_opt(sb, DATA_FLAGS)) {
1532 /* No mode set, assume a default based on the journal
1533 capabilities: ORDERED_DATA if the journal can
1534 cope, else JOURNAL_DATA */
1535 if (journal_check_available_features
1536 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1537 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1539 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1542 case EXT3_MOUNT_ORDERED_DATA:
1543 case EXT3_MOUNT_WRITEBACK_DATA:
1544 if (!journal_check_available_features
1545 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1546 printk(KERN_ERR "EXT3-fs: Journal does not support "
1547 "requested data journaling mode\n");
1555 * The journal_load will have done any necessary log recovery,
1556 * so we can safely mount the rest of the filesystem now.
1559 root = iget(sb, EXT3_ROOT_INO);
1560 sb->s_root = d_alloc_root(root);
1562 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1566 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1569 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1573 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1575 * akpm: core read_super() calls in here with the superblock locked.
1576 * That deadlocks, because orphan cleanup needs to lock the superblock
1577 * in numerous places. Here we just pop the lock - it's relatively
1578 * harmless, because we are now ready to accept write_super() requests,
1579 * and aviro says that's the only reason for hanging onto the
1582 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1583 ext3_orphan_cleanup(sb, es);
1584 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1586 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1587 ext3_mark_recovery_complete(sb, es);
1588 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1589 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1590 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1593 percpu_counter_mod(&sbi->s_freeblocks_counter,
1594 ext3_count_free_blocks(sb));
1595 percpu_counter_mod(&sbi->s_freeinodes_counter,
1596 ext3_count_free_inodes(sb));
1597 percpu_counter_mod(&sbi->s_dirs_counter,
1598 ext3_count_dirs(sb));
1603 journal_destroy(sbi->s_journal);
1605 kfree(sbi->s_debts);
1606 for (i = 0; i < db_count; i++)
1607 brelse(sbi->s_group_desc[i]);
1608 kfree(sbi->s_group_desc);
1611 for (i = 0; i < MAXQUOTAS; i++) {
1612 if (sbi->s_qf_names[i])
1613 kfree(sbi->s_qf_names[i]);
1616 ext3_blkdev_remove(sbi);
1619 sb->s_fs_info = NULL;
1625 * Setup any per-fs journal parameters now. We'll do this both on
1626 * initial mount, once the journal has been initialised but before we've
1627 * done any recovery; and again on any subsequent remount.
1629 static void ext3_init_journal_params(struct ext3_sb_info *sbi,
1632 if (sbi->s_commit_interval)
1633 journal->j_commit_interval = sbi->s_commit_interval;
1634 /* We could also set up an ext3-specific default for the commit
1635 * interval here, but for now we'll just fall back to the jbd
1640 static journal_t *ext3_get_journal(struct super_block *sb, int journal_inum)
1642 struct inode *journal_inode;
1645 /* First, test for the existence of a valid inode on disk. Bad
1646 * things happen if we iget() an unused inode, as the subsequent
1647 * iput() will try to delete it. */
1649 journal_inode = iget(sb, journal_inum);
1650 if (!journal_inode) {
1651 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1654 if (!journal_inode->i_nlink) {
1655 make_bad_inode(journal_inode);
1656 iput(journal_inode);
1657 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1661 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1662 journal_inode, journal_inode->i_size);
1663 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
1664 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
1665 iput(journal_inode);
1669 journal = journal_init_inode(journal_inode);
1671 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
1672 iput(journal_inode);
1674 journal->j_private = sb;
1675 ext3_init_journal_params(EXT3_SB(sb), journal);
1679 static journal_t *ext3_get_dev_journal(struct super_block *sb,
1682 struct buffer_head * bh;
1686 int hblock, blocksize;
1687 unsigned long sb_block;
1688 unsigned long offset;
1689 struct ext3_super_block * es;
1690 struct block_device *bdev;
1692 bdev = ext3_blkdev_get(j_dev);
1696 if (bd_claim(bdev, sb)) {
1698 "EXT3: failed to claim external journal device.\n");
1703 blocksize = sb->s_blocksize;
1704 hblock = bdev_hardsect_size(bdev);
1705 if (blocksize < hblock) {
1707 "EXT3-fs: blocksize too small for journal device.\n");
1711 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
1712 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
1713 set_blocksize(bdev, blocksize);
1714 if (!(bh = __bread(bdev, sb_block, blocksize))) {
1715 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
1716 "external journal\n");
1720 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1721 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
1722 !(le32_to_cpu(es->s_feature_incompat) &
1723 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
1724 printk(KERN_ERR "EXT3-fs: external journal has "
1725 "bad superblock\n");
1730 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
1731 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
1736 len = le32_to_cpu(es->s_blocks_count);
1737 start = sb_block + 1;
1738 brelse(bh); /* we're done with the superblock */
1740 journal = journal_init_dev(bdev, sb->s_bdev,
1741 start, len, blocksize);
1743 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
1746 journal->j_private = sb;
1747 ll_rw_block(READ, 1, &journal->j_sb_buffer);
1748 wait_on_buffer(journal->j_sb_buffer);
1749 if (!buffer_uptodate(journal->j_sb_buffer)) {
1750 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
1753 if (ntohl(journal->j_superblock->s_nr_users) != 1) {
1754 printk(KERN_ERR "EXT3-fs: External journal has more than one "
1755 "user (unsupported) - %d\n",
1756 ntohl(journal->j_superblock->s_nr_users));
1759 EXT3_SB(sb)->journal_bdev = bdev;
1760 ext3_init_journal_params(EXT3_SB(sb), journal);
1763 journal_destroy(journal);
1765 ext3_blkdev_put(bdev);
1769 static int ext3_load_journal(struct super_block * sb,
1770 struct ext3_super_block * es)
1773 int journal_inum = le32_to_cpu(es->s_journal_inum);
1774 dev_t journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
1776 int really_read_only;
1778 really_read_only = bdev_read_only(sb->s_bdev);
1781 * Are we loading a blank journal or performing recovery after a
1782 * crash? For recovery, we need to check in advance whether we
1783 * can get read-write access to the device.
1786 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
1787 if (sb->s_flags & MS_RDONLY) {
1788 printk(KERN_INFO "EXT3-fs: INFO: recovery "
1789 "required on readonly filesystem.\n");
1790 if (really_read_only) {
1791 printk(KERN_ERR "EXT3-fs: write access "
1792 "unavailable, cannot proceed.\n");
1795 printk (KERN_INFO "EXT3-fs: write access will "
1796 "be enabled during recovery.\n");
1800 if (journal_inum && journal_dev) {
1801 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
1802 "and inode journals!\n");
1807 if (!(journal = ext3_get_journal(sb, journal_inum)))
1810 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
1814 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
1815 err = journal_update_format(journal);
1817 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
1818 journal_destroy(journal);
1823 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
1824 err = journal_wipe(journal, !really_read_only);
1826 err = journal_load(journal);
1829 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
1830 journal_destroy(journal);
1834 EXT3_SB(sb)->s_journal = journal;
1835 ext3_clear_journal_err(sb, es);
1839 static int ext3_create_journal(struct super_block * sb,
1840 struct ext3_super_block * es,
1845 if (sb->s_flags & MS_RDONLY) {
1846 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
1847 "create journal.\n");
1851 if (!(journal = ext3_get_journal(sb, journal_inum)))
1854 printk(KERN_INFO "EXT3-fs: creating new journal on inode %d\n",
1857 if (journal_create(journal)) {
1858 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
1859 journal_destroy(journal);
1863 EXT3_SB(sb)->s_journal = journal;
1865 ext3_update_dynamic_rev(sb);
1866 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1867 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
1869 es->s_journal_inum = cpu_to_le32(journal_inum);
1872 /* Make sure we flush the recovery flag to disk. */
1873 ext3_commit_super(sb, es, 1);
1878 static void ext3_commit_super (struct super_block * sb,
1879 struct ext3_super_block * es,
1882 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
1886 es->s_wtime = cpu_to_le32(get_seconds());
1887 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
1888 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
1889 BUFFER_TRACE(sbh, "marking dirty");
1890 mark_buffer_dirty(sbh);
1892 sync_dirty_buffer(sbh);
1897 * Have we just finished recovery? If so, and if we are mounting (or
1898 * remounting) the filesystem readonly, then we will end up with a
1899 * consistent fs on disk. Record that fact.
1901 static void ext3_mark_recovery_complete(struct super_block * sb,
1902 struct ext3_super_block * es)
1904 journal_flush(EXT3_SB(sb)->s_journal);
1905 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
1906 sb->s_flags & MS_RDONLY) {
1907 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1909 ext3_commit_super(sb, es, 1);
1914 * If we are mounting (or read-write remounting) a filesystem whose journal
1915 * has recorded an error from a previous lifetime, move that error to the
1916 * main filesystem now.
1918 static void ext3_clear_journal_err(struct super_block * sb,
1919 struct ext3_super_block * es)
1925 journal = EXT3_SB(sb)->s_journal;
1928 * Now check for any error status which may have been recorded in the
1929 * journal by a prior ext3_error() or ext3_abort()
1932 j_errno = journal_errno(journal);
1936 errstr = ext3_decode_error(sb, j_errno, nbuf);
1937 ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
1938 "from previous mount: %s", errstr);
1939 ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
1940 "filesystem check.");
1942 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
1943 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
1944 ext3_commit_super (sb, es, 1);
1946 journal_clear_err(journal);
1951 * Force the running and committing transactions to commit,
1952 * and wait on the commit.
1954 int ext3_force_commit(struct super_block *sb)
1959 if (sb->s_flags & MS_RDONLY)
1962 journal = EXT3_SB(sb)->s_journal;
1964 ret = ext3_journal_force_commit(journal);
1969 * Ext3 always journals updates to the superblock itself, so we don't
1970 * have to propagate any other updates to the superblock on disk at this
1971 * point. Just start an async writeback to get the buffers on their way
1974 * This implicitly triggers the writebehind on sync().
1977 void ext3_write_super (struct super_block * sb)
1979 if (down_trylock(&sb->s_lock) == 0)
1984 static int ext3_sync_fs(struct super_block *sb, int wait)
1989 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
1991 log_wait_commit(EXT3_SB(sb)->s_journal, target);
1997 * LVM calls this function before a (read-only) snapshot is created. This
1998 * gives us a chance to flush the journal completely and mark the fs clean.
2000 void ext3_write_super_lockfs(struct super_block *sb)
2004 if (!(sb->s_flags & MS_RDONLY)) {
2005 journal_t *journal = EXT3_SB(sb)->s_journal;
2007 /* Now we set up the journal barrier. */
2008 journal_lock_updates(journal);
2009 journal_flush(journal);
2011 /* Journal blocked and flushed, clear needs_recovery flag. */
2012 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2013 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2018 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2019 * flag here, even though the filesystem is not technically dirty yet.
2021 void ext3_unlockfs(struct super_block *sb)
2023 if (!(sb->s_flags & MS_RDONLY)) {
2025 /* Reser the needs_recovery flag before the fs is unlocked. */
2026 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2027 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2029 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2033 int ext3_remount (struct super_block * sb, int * flags, char * data)
2035 struct ext3_super_block * es;
2036 struct ext3_sb_info *sbi = EXT3_SB(sb);
2038 unsigned long n_blocks_count = 0;
2041 * Allow the "check" option to be passed as a remount option.
2043 if (!parse_options(data, sb, &tmp, &n_blocks_count, 1))
2046 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2047 ext3_abort(sb, __FUNCTION__, "Abort forced by user");
2049 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2050 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2054 ext3_init_journal_params(sbi, sbi->s_journal);
2056 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2057 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2058 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2061 if (*flags & MS_RDONLY) {
2063 * First of all, the unconditional stuff we have to do
2064 * to disable replay of the journal when we next remount
2066 sb->s_flags |= MS_RDONLY;
2069 * OK, test if we are remounting a valid rw partition
2070 * readonly, and if so set the rdonly flag and then
2071 * mark the partition as valid again.
2073 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2074 (sbi->s_mount_state & EXT3_VALID_FS))
2075 es->s_state = cpu_to_le16(sbi->s_mount_state);
2077 ext3_mark_recovery_complete(sb, es);
2080 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2081 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2082 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2083 "remount RDWR because of unsupported "
2084 "optional features (%x).\n",
2089 * Mounting a RDONLY partition read-write, so reread
2090 * and store the current valid flag. (It may have
2091 * been changed by e2fsck since we originally mounted
2094 ext3_clear_journal_err(sb, es);
2095 sbi->s_mount_state = le16_to_cpu(es->s_state);
2096 if ((ret = ext3_group_extend(sb, es, n_blocks_count)))
2098 if (!ext3_setup_super (sb, es, 0))
2099 sb->s_flags &= ~MS_RDONLY;
2105 int ext3_statfs (struct super_block * sb, struct kstatfs * buf)
2107 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
2108 unsigned long overhead;
2111 if (test_opt (sb, MINIX_DF))
2115 * Compute the overhead (FS structures)
2119 * All of the blocks before first_data_block are
2122 overhead = le32_to_cpu(es->s_first_data_block);
2125 * Add the overhead attributed to the superblock and
2126 * block group descriptors. If the sparse superblocks
2127 * feature is turned on, then not all groups have this.
2129 for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++)
2130 overhead += ext3_bg_has_super(sb, i) +
2131 ext3_bg_num_gdb(sb, i);
2134 * Every block group has an inode bitmap, a block
2135 * bitmap, and an inode table.
2137 overhead += (EXT3_SB(sb)->s_groups_count *
2138 (2 + EXT3_SB(sb)->s_itb_per_group));
2141 buf->f_type = EXT3_SUPER_MAGIC;
2142 buf->f_bsize = sb->s_blocksize;
2143 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
2144 buf->f_bfree = ext3_count_free_blocks (sb);
2145 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2146 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2148 buf->f_files = le32_to_cpu(es->s_inodes_count);
2149 buf->f_ffree = ext3_count_free_inodes (sb);
2150 buf->f_namelen = EXT3_NAME_LEN;
2154 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2155 * is locked for write. Otherwise the are possible deadlocks:
2156 * Process 1 Process 2
2157 * ext3_create() quota_sync()
2158 * journal_start() write_dquot()
2159 * DQUOT_INIT() down(dqio_sem)
2160 * down(dqio_sem) journal_start()
2166 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2168 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type]->f_dentry->d_inode;
2171 static int ext3_dquot_initialize(struct inode *inode, int type)
2176 /* We may create quota structure so we need to reserve enough blocks */
2177 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS);
2179 return PTR_ERR(handle);
2180 ret = dquot_initialize(inode, type);
2181 err = ext3_journal_stop(handle);
2187 static int ext3_dquot_drop(struct inode *inode)
2192 /* We may delete quota structure so we need to reserve enough blocks */
2193 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS);
2195 return PTR_ERR(handle);
2196 ret = dquot_drop(inode);
2197 err = ext3_journal_stop(handle);
2203 static int ext3_write_dquot(struct dquot *dquot)
2208 handle = ext3_journal_start(dquot_to_inode(dquot),
2209 EXT3_QUOTA_TRANS_BLOCKS);
2211 return PTR_ERR(handle);
2212 ret = dquot_commit(dquot);
2213 err = ext3_journal_stop(handle);
2219 static int ext3_acquire_dquot(struct dquot *dquot)
2224 handle = ext3_journal_start(dquot_to_inode(dquot),
2225 EXT3_QUOTA_INIT_BLOCKS);
2227 return PTR_ERR(handle);
2228 ret = dquot_acquire(dquot);
2229 err = ext3_journal_stop(handle);
2235 static int ext3_release_dquot(struct dquot *dquot)
2240 handle = ext3_journal_start(dquot_to_inode(dquot),
2241 EXT3_QUOTA_INIT_BLOCKS);
2243 return PTR_ERR(handle);
2244 ret = dquot_release(dquot);
2245 err = ext3_journal_stop(handle);
2251 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2253 /* Are we journalling quotas? */
2254 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2255 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2256 dquot_mark_dquot_dirty(dquot);
2257 return ext3_write_dquot(dquot);
2259 return dquot_mark_dquot_dirty(dquot);
2263 static int ext3_write_info(struct super_block *sb, int type)
2268 /* Data block + inode block */
2269 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2271 return PTR_ERR(handle);
2272 ret = dquot_commit_info(sb, type);
2273 err = ext3_journal_stop(handle);
2280 * Turn on quotas during mount time - we need to find
2281 * the quota file and such...
2283 static int ext3_quota_on_mount(struct super_block *sb, int type)
2286 struct dentry *dentry;
2287 struct qstr name = { .name = EXT3_SB(sb)->s_qf_names[type],
2289 .len = strlen(EXT3_SB(sb)->s_qf_names[type])};
2291 dentry = lookup_hash(&name, sb->s_root);
2293 return PTR_ERR(dentry);
2294 err = vfs_quota_on_mount(type, EXT3_SB(sb)->s_jquota_fmt, dentry);
2297 /* We keep the dentry reference if everything went ok - we drop it
2298 * on quota_off time */
2302 /* Turn quotas off during mount time */
2303 static int ext3_quota_off_mount(struct super_block *sb, int type)
2306 struct dentry *dentry;
2308 dentry = sb_dqopt(sb)->files[type]->f_dentry;
2309 err = vfs_quota_off_mount(sb, type);
2310 /* We invalidate dentry - it has at least wrong hash... */
2311 d_invalidate(dentry);
2317 * Standard function to be called on quota_on
2319 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2323 struct nameidata nd;
2325 /* Not journalling quota? */
2326 if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2327 !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
2328 return vfs_quota_on(sb, type, format_id, path);
2329 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2332 /* Quotafile not on the same filesystem? */
2333 if (nd.mnt->mnt_sb != sb)
2335 /* Quotafile not of fs root? */
2336 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2338 "EXT3-fs: Quota file not on filesystem root. "
2339 "Journalled quota will not work.\n");
2340 if (!ext3_should_journal_data(nd.dentry->d_inode))
2341 printk(KERN_WARNING "EXT3-fs: Quota file does not have "
2342 "data-journalling. Journalled quota will not work.\n");
2344 return vfs_quota_on(sb, type, format_id, path);
2349 static struct super_block *ext3_get_sb(struct file_system_type *fs_type,
2350 int flags, const char *dev_name, void *data)
2352 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
2355 static struct file_system_type ext3_fs_type = {
2356 .owner = THIS_MODULE,
2358 .get_sb = ext3_get_sb,
2359 .kill_sb = kill_block_super,
2360 .fs_flags = FS_REQUIRES_DEV,
2363 static int __init init_ext3_fs(void)
2365 int err = init_ext3_xattr();
2368 err = init_inodecache();
2371 err = register_filesystem(&ext3_fs_type);
2376 destroy_inodecache();
2382 static void __exit exit_ext3_fs(void)
2384 unregister_filesystem(&ext3_fs_type);
2385 destroy_inodecache();
2389 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2390 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2391 MODULE_LICENSE("GPL");
2392 module_init(init_ext3_fs)
2393 module_exit(exit_ext3_fs)