2 * linux/fs/ext4/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/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/ext4_fs.h>
25 #include <linux/ext4_jbd2.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
47 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
48 unsigned long journal_devnum);
49 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
51 static void ext4_commit_super (struct super_block * sb,
52 struct ext4_super_block * es,
54 static void ext4_mark_recovery_complete(struct super_block * sb,
55 struct ext4_super_block * es);
56 static void ext4_clear_journal_err(struct super_block * sb,
57 struct ext4_super_block * es);
58 static int ext4_sync_fs(struct super_block *sb, int wait);
59 static const char *ext4_decode_error(struct super_block * sb, int errno,
61 static int ext4_remount (struct super_block * sb, int * flags, char * data);
62 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext4_unlockfs(struct super_block *sb);
64 static void ext4_write_super (struct super_block * sb);
65 static void ext4_write_super_lockfs(struct super_block *sb);
68 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
71 return le32_to_cpu(bg->bg_block_bitmap) |
72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
76 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
79 return le32_to_cpu(bg->bg_inode_bitmap) |
80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
84 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
87 return le32_to_cpu(bg->bg_inode_table) |
88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
92 void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
95 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
100 void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
103 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
108 void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
111 bg->bg_inode_table = cpu_to_le32((u32)blk);
112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
117 * Wrappers for jbd2_journal_start/end.
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
124 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
134 journal = EXT4_SB(sb)->s_journal;
135 if (is_journal_aborted(journal)) {
136 ext4_abort(sb, __FUNCTION__,
137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
141 return jbd2_journal_start(journal, nblocks);
145 * The only special thing we need to do here is to make sure that all
146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
147 * that sync() will call the filesystem's write_super callback if
150 int __ext4_journal_stop(const char *where, handle_t *handle)
152 struct super_block *sb;
156 sb = handle->h_transaction->t_journal->j_private;
158 rc = jbd2_journal_stop(handle);
163 __ext4_std_error(sb, where, err);
167 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
168 struct buffer_head *bh, handle_t *handle, int err)
171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
174 BUFFER_TRACE(bh, "abort");
179 if (is_handle_aborted(handle))
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
185 jbd2_journal_abort_handle(handle);
188 /* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
191 * On ext2, we can store the error state of the filesystem in the
192 * superblock. That is not possible on ext4, because we may have other
193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
198 * We'll just use the jbd2_journal_abort() error code to record an error in
199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
203 static void ext4_handle_error(struct super_block *sb)
205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
210 if (sb->s_flags & MS_RDONLY)
213 if (!test_opt (sb, ERRORS_CONT)) {
214 journal_t *journal = EXT4_SB(sb)->s_journal;
216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
218 jbd2_journal_abort(journal, -EIO);
220 if (test_opt (sb, ERRORS_RO)) {
221 printk (KERN_CRIT "Remounting filesystem read-only\n");
222 sb->s_flags |= MS_RDONLY;
224 ext4_commit_super(sb, es, 1);
225 if (test_opt(sb, ERRORS_PANIC))
226 panic("EXT4-fs (device %s): panic forced after error\n",
230 void ext4_error (struct super_block * sb, const char * function,
231 const char * fmt, ...)
236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
241 ext4_handle_error(sb);
244 static const char *ext4_decode_error(struct super_block * sb, int errno,
251 errstr = "IO failure";
254 errstr = "Out of memory";
257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
258 errstr = "Journal has aborted";
260 errstr = "Readonly filesystem";
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
277 /* __ext4_std_error decodes expected errors from journaling functions
278 * automatically and invokes the appropriate error response. */
280 void __ext4_std_error (struct super_block * sb, const char * function,
286 /* Special case: if the error is EROFS, and we're not already
287 * inside a transaction, then there's really no point in logging
289 if (errno == -EROFS && journal_current_handle() == NULL &&
290 (sb->s_flags & MS_RDONLY))
293 errstr = ext4_decode_error(sb, errno, nbuf);
294 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
295 sb->s_id, function, errstr);
297 ext4_handle_error(sb);
301 * ext4_abort is a much stronger failure handler than ext4_error. The
302 * abort function may be used to deal with unrecoverable failures such
303 * as journal IO errors or ENOMEM at a critical moment in log management.
305 * We unconditionally force the filesystem into an ABORT|READONLY state,
306 * unless the error response on the fs has been set to panic in which
307 * case we take the easy way out and panic immediately.
310 void ext4_abort (struct super_block * sb, const char * function,
311 const char * fmt, ...)
315 printk (KERN_CRIT "ext4_abort called.\n");
318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
323 if (test_opt(sb, ERRORS_PANIC))
324 panic("EXT4-fs panic from previous error\n");
326 if (sb->s_flags & MS_RDONLY)
329 printk(KERN_CRIT "Remounting filesystem read-only\n");
330 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
331 sb->s_flags |= MS_RDONLY;
332 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
333 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
336 void ext4_warning (struct super_block * sb, const char * function,
337 const char * fmt, ...)
342 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
349 void ext4_update_dynamic_rev(struct super_block *sb)
351 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
353 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
356 ext4_warning(sb, __FUNCTION__,
357 "updating to rev %d because of new feature flag, "
358 "running e2fsck is recommended",
361 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
362 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
363 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
364 /* leave es->s_feature_*compat flags alone */
365 /* es->s_uuid will be set by e2fsck if empty */
368 * The rest of the superblock fields should be zero, and if not it
369 * means they are likely already in use, so leave them alone. We
370 * can leave it up to e2fsck to clean up any inconsistencies there.
375 * Open the external journal device
377 static struct block_device *ext4_blkdev_get(dev_t dev)
379 struct block_device *bdev;
380 char b[BDEVNAME_SIZE];
382 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
388 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
389 __bdevname(dev, b), PTR_ERR(bdev));
394 * Release the journal device
396 static int ext4_blkdev_put(struct block_device *bdev)
399 return blkdev_put(bdev);
402 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
404 struct block_device *bdev;
407 bdev = sbi->journal_bdev;
409 ret = ext4_blkdev_put(bdev);
410 sbi->journal_bdev = NULL;
415 static inline struct inode *orphan_list_entry(struct list_head *l)
417 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
420 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
424 printk(KERN_ERR "sb orphan head is %d\n",
425 le32_to_cpu(sbi->s_es->s_last_orphan));
427 printk(KERN_ERR "sb_info orphan list:\n");
428 list_for_each(l, &sbi->s_orphan) {
429 struct inode *inode = orphan_list_entry(l);
431 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
432 inode->i_sb->s_id, inode->i_ino, inode,
433 inode->i_mode, inode->i_nlink,
438 static void ext4_put_super (struct super_block * sb)
440 struct ext4_sb_info *sbi = EXT4_SB(sb);
441 struct ext4_super_block *es = sbi->s_es;
444 ext4_ext_release(sb);
445 ext4_xattr_put_super(sb);
446 jbd2_journal_destroy(sbi->s_journal);
447 if (!(sb->s_flags & MS_RDONLY)) {
448 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
449 es->s_state = cpu_to_le16(sbi->s_mount_state);
450 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
451 mark_buffer_dirty(sbi->s_sbh);
452 ext4_commit_super(sb, es, 1);
455 for (i = 0; i < sbi->s_gdb_count; i++)
456 brelse(sbi->s_group_desc[i]);
457 kfree(sbi->s_group_desc);
458 percpu_counter_destroy(&sbi->s_freeblocks_counter);
459 percpu_counter_destroy(&sbi->s_freeinodes_counter);
460 percpu_counter_destroy(&sbi->s_dirs_counter);
463 for (i = 0; i < MAXQUOTAS; i++)
464 kfree(sbi->s_qf_names[i]);
467 /* Debugging code just in case the in-memory inode orphan list
468 * isn't empty. The on-disk one can be non-empty if we've
469 * detected an error and taken the fs readonly, but the
470 * in-memory list had better be clean by this point. */
471 if (!list_empty(&sbi->s_orphan))
472 dump_orphan_list(sb, sbi);
473 J_ASSERT(list_empty(&sbi->s_orphan));
475 invalidate_bdev(sb->s_bdev);
476 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
478 * Invalidate the journal device's buffers. We don't want them
479 * floating about in memory - the physical journal device may
480 * hotswapped, and it breaks the `ro-after' testing code.
482 sync_blockdev(sbi->journal_bdev);
483 invalidate_bdev(sbi->journal_bdev);
484 ext4_blkdev_remove(sbi);
486 sb->s_fs_info = NULL;
491 static struct kmem_cache *ext4_inode_cachep;
494 * Called inside transaction, so use GFP_NOFS
496 static struct inode *ext4_alloc_inode(struct super_block *sb)
498 struct ext4_inode_info *ei;
500 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
503 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
504 ei->i_acl = EXT4_ACL_NOT_CACHED;
505 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
507 ei->i_block_alloc_info = NULL;
508 ei->vfs_inode.i_version = 1;
509 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
510 return &ei->vfs_inode;
513 static void ext4_destroy_inode(struct inode *inode)
515 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
516 printk("EXT4 Inode %p: orphan list check failed!\n",
518 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
519 EXT4_I(inode), sizeof(struct ext4_inode_info),
523 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
526 static void init_once(struct kmem_cache *cachep, void *foo)
528 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
530 INIT_LIST_HEAD(&ei->i_orphan);
531 #ifdef CONFIG_EXT4DEV_FS_XATTR
532 init_rwsem(&ei->xattr_sem);
534 mutex_init(&ei->truncate_mutex);
535 inode_init_once(&ei->vfs_inode);
538 static int init_inodecache(void)
540 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
541 sizeof(struct ext4_inode_info),
542 0, (SLAB_RECLAIM_ACCOUNT|
545 if (ext4_inode_cachep == NULL)
550 static void destroy_inodecache(void)
552 kmem_cache_destroy(ext4_inode_cachep);
555 static void ext4_clear_inode(struct inode *inode)
557 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
558 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
559 if (EXT4_I(inode)->i_acl &&
560 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
561 posix_acl_release(EXT4_I(inode)->i_acl);
562 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
564 if (EXT4_I(inode)->i_default_acl &&
565 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
566 posix_acl_release(EXT4_I(inode)->i_default_acl);
567 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
570 ext4_discard_reservation(inode);
571 EXT4_I(inode)->i_block_alloc_info = NULL;
576 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
578 #if defined(CONFIG_QUOTA)
579 struct ext4_sb_info *sbi = EXT4_SB(sb);
581 if (sbi->s_jquota_fmt)
582 seq_printf(seq, ",jqfmt=%s",
583 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
585 if (sbi->s_qf_names[USRQUOTA])
586 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
588 if (sbi->s_qf_names[GRPQUOTA])
589 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
591 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
592 seq_puts(seq, ",usrquota");
594 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
595 seq_puts(seq, ",grpquota");
601 * - it's set to a non-default value OR
602 * - if the per-sb default is different from the global default
604 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
606 struct super_block *sb = vfs->mnt_sb;
607 struct ext4_sb_info *sbi = EXT4_SB(sb);
608 struct ext4_super_block *es = sbi->s_es;
609 unsigned long def_mount_opts;
611 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
613 if (sbi->s_sb_block != 1)
614 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
615 if (test_opt(sb, MINIX_DF))
616 seq_puts(seq, ",minixdf");
617 if (test_opt(sb, GRPID))
618 seq_puts(seq, ",grpid");
619 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
620 seq_puts(seq, ",nogrpid");
621 if (sbi->s_resuid != EXT4_DEF_RESUID ||
622 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
623 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
625 if (sbi->s_resgid != EXT4_DEF_RESGID ||
626 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
627 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
629 if (test_opt(sb, ERRORS_CONT)) {
630 int def_errors = le16_to_cpu(es->s_errors);
632 if (def_errors == EXT4_ERRORS_PANIC ||
633 def_errors == EXT4_ERRORS_RO) {
634 seq_puts(seq, ",errors=continue");
637 if (test_opt(sb, ERRORS_RO))
638 seq_puts(seq, ",errors=remount-ro");
639 if (test_opt(sb, ERRORS_PANIC))
640 seq_puts(seq, ",errors=panic");
641 if (test_opt(sb, NO_UID32))
642 seq_puts(seq, ",nouid32");
643 if (test_opt(sb, DEBUG))
644 seq_puts(seq, ",debug");
645 if (test_opt(sb, OLDALLOC))
646 seq_puts(seq, ",oldalloc");
647 #ifdef CONFIG_EXT4_FS_XATTR
648 if (test_opt(sb, XATTR_USER))
649 seq_puts(seq, ",user_xattr");
650 if (!test_opt(sb, XATTR_USER) &&
651 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
652 seq_puts(seq, ",nouser_xattr");
655 #ifdef CONFIG_EXT4_FS_POSIX_ACL
656 if (test_opt(sb, POSIX_ACL))
657 seq_puts(seq, ",acl");
658 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
659 seq_puts(seq, ",noacl");
661 if (!test_opt(sb, RESERVATION))
662 seq_puts(seq, ",noreservation");
663 if (sbi->s_commit_interval) {
664 seq_printf(seq, ",commit=%u",
665 (unsigned) (sbi->s_commit_interval / HZ));
667 if (test_opt(sb, BARRIER))
668 seq_puts(seq, ",barrier=1");
669 if (test_opt(sb, NOBH))
670 seq_puts(seq, ",nobh");
671 if (!test_opt(sb, EXTENTS))
672 seq_puts(seq, ",noextents");
674 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
675 seq_puts(seq, ",data=journal");
676 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
677 seq_puts(seq, ",data=ordered");
678 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
679 seq_puts(seq, ",data=writeback");
681 ext4_show_quota_options(seq, sb);
687 static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
690 unsigned long ino = objp[0];
691 __u32 generation = objp[1];
693 struct dentry *result;
695 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
696 return ERR_PTR(-ESTALE);
697 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
698 return ERR_PTR(-ESTALE);
700 /* iget isn't really right if the inode is currently unallocated!!
702 * ext4_read_inode will return a bad_inode if the inode had been
703 * deleted, so we should be safe.
705 * Currently we don't know the generation for parent directory, so
706 * a generation of 0 means "accept any"
708 inode = iget(sb, ino);
710 return ERR_PTR(-ENOMEM);
711 if (is_bad_inode(inode) ||
712 (generation && inode->i_generation != generation)) {
714 return ERR_PTR(-ESTALE);
716 /* now to find a dentry.
717 * If possible, get a well-connected one
719 result = d_alloc_anon(inode);
722 return ERR_PTR(-ENOMEM);
728 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
729 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
731 static int ext4_dquot_initialize(struct inode *inode, int type);
732 static int ext4_dquot_drop(struct inode *inode);
733 static int ext4_write_dquot(struct dquot *dquot);
734 static int ext4_acquire_dquot(struct dquot *dquot);
735 static int ext4_release_dquot(struct dquot *dquot);
736 static int ext4_mark_dquot_dirty(struct dquot *dquot);
737 static int ext4_write_info(struct super_block *sb, int type);
738 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
739 static int ext4_quota_on_mount(struct super_block *sb, int type);
740 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
741 size_t len, loff_t off);
742 static ssize_t ext4_quota_write(struct super_block *sb, int type,
743 const char *data, size_t len, loff_t off);
745 static struct dquot_operations ext4_quota_operations = {
746 .initialize = ext4_dquot_initialize,
747 .drop = ext4_dquot_drop,
748 .alloc_space = dquot_alloc_space,
749 .alloc_inode = dquot_alloc_inode,
750 .free_space = dquot_free_space,
751 .free_inode = dquot_free_inode,
752 .transfer = dquot_transfer,
753 .write_dquot = ext4_write_dquot,
754 .acquire_dquot = ext4_acquire_dquot,
755 .release_dquot = ext4_release_dquot,
756 .mark_dirty = ext4_mark_dquot_dirty,
757 .write_info = ext4_write_info
760 static struct quotactl_ops ext4_qctl_operations = {
761 .quota_on = ext4_quota_on,
762 .quota_off = vfs_quota_off,
763 .quota_sync = vfs_quota_sync,
764 .get_info = vfs_get_dqinfo,
765 .set_info = vfs_set_dqinfo,
766 .get_dqblk = vfs_get_dqblk,
767 .set_dqblk = vfs_set_dqblk
771 static const struct super_operations ext4_sops = {
772 .alloc_inode = ext4_alloc_inode,
773 .destroy_inode = ext4_destroy_inode,
774 .read_inode = ext4_read_inode,
775 .write_inode = ext4_write_inode,
776 .dirty_inode = ext4_dirty_inode,
777 .delete_inode = ext4_delete_inode,
778 .put_super = ext4_put_super,
779 .write_super = ext4_write_super,
780 .sync_fs = ext4_sync_fs,
781 .write_super_lockfs = ext4_write_super_lockfs,
782 .unlockfs = ext4_unlockfs,
783 .statfs = ext4_statfs,
784 .remount_fs = ext4_remount,
785 .clear_inode = ext4_clear_inode,
786 .show_options = ext4_show_options,
788 .quota_read = ext4_quota_read,
789 .quota_write = ext4_quota_write,
793 static struct export_operations ext4_export_ops = {
794 .get_parent = ext4_get_parent,
795 .get_dentry = ext4_get_dentry,
799 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
800 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
801 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
802 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
803 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
804 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
805 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
806 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
807 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
808 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
809 Opt_grpquota, Opt_extents, Opt_noextents,
812 static match_table_t tokens = {
813 {Opt_bsd_df, "bsddf"},
814 {Opt_minix_df, "minixdf"},
815 {Opt_grpid, "grpid"},
816 {Opt_grpid, "bsdgroups"},
817 {Opt_nogrpid, "nogrpid"},
818 {Opt_nogrpid, "sysvgroups"},
819 {Opt_resgid, "resgid=%u"},
820 {Opt_resuid, "resuid=%u"},
822 {Opt_err_cont, "errors=continue"},
823 {Opt_err_panic, "errors=panic"},
824 {Opt_err_ro, "errors=remount-ro"},
825 {Opt_nouid32, "nouid32"},
826 {Opt_nocheck, "nocheck"},
827 {Opt_nocheck, "check=none"},
828 {Opt_debug, "debug"},
829 {Opt_oldalloc, "oldalloc"},
830 {Opt_orlov, "orlov"},
831 {Opt_user_xattr, "user_xattr"},
832 {Opt_nouser_xattr, "nouser_xattr"},
834 {Opt_noacl, "noacl"},
835 {Opt_reservation, "reservation"},
836 {Opt_noreservation, "noreservation"},
837 {Opt_noload, "noload"},
840 {Opt_commit, "commit=%u"},
841 {Opt_journal_update, "journal=update"},
842 {Opt_journal_inum, "journal=%u"},
843 {Opt_journal_dev, "journal_dev=%u"},
844 {Opt_abort, "abort"},
845 {Opt_data_journal, "data=journal"},
846 {Opt_data_ordered, "data=ordered"},
847 {Opt_data_writeback, "data=writeback"},
848 {Opt_offusrjquota, "usrjquota="},
849 {Opt_usrjquota, "usrjquota=%s"},
850 {Opt_offgrpjquota, "grpjquota="},
851 {Opt_grpjquota, "grpjquota=%s"},
852 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
853 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
854 {Opt_grpquota, "grpquota"},
855 {Opt_noquota, "noquota"},
856 {Opt_quota, "quota"},
857 {Opt_usrquota, "usrquota"},
858 {Opt_barrier, "barrier=%u"},
859 {Opt_extents, "extents"},
860 {Opt_noextents, "noextents"},
862 {Opt_resize, "resize"},
865 static ext4_fsblk_t get_sb_block(void **data)
867 ext4_fsblk_t sb_block;
868 char *options = (char *) *data;
870 if (!options || strncmp(options, "sb=", 3) != 0)
871 return 1; /* Default location */
873 /*todo: use simple_strtoll with >32bit ext4 */
874 sb_block = simple_strtoul(options, &options, 0);
875 if (*options && *options != ',') {
876 printk("EXT4-fs: Invalid sb specification: %s\n",
882 *data = (void *) options;
886 static int parse_options (char *options, struct super_block *sb,
887 unsigned int *inum, unsigned long *journal_devnum,
888 ext4_fsblk_t *n_blocks_count, int is_remount)
890 struct ext4_sb_info *sbi = EXT4_SB(sb);
892 substring_t args[MAX_OPT_ARGS];
903 while ((p = strsep (&options, ",")) != NULL) {
908 token = match_token(p, tokens, args);
911 clear_opt (sbi->s_mount_opt, MINIX_DF);
914 set_opt (sbi->s_mount_opt, MINIX_DF);
917 set_opt (sbi->s_mount_opt, GRPID);
920 clear_opt (sbi->s_mount_opt, GRPID);
923 if (match_int(&args[0], &option))
925 sbi->s_resuid = option;
928 if (match_int(&args[0], &option))
930 sbi->s_resgid = option;
933 /* handled by get_sb_block() instead of here */
934 /* *sb_block = match_int(&args[0]); */
937 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
938 clear_opt (sbi->s_mount_opt, ERRORS_RO);
939 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
942 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
943 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
944 set_opt (sbi->s_mount_opt, ERRORS_RO);
947 clear_opt (sbi->s_mount_opt, ERRORS_RO);
948 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
949 set_opt (sbi->s_mount_opt, ERRORS_CONT);
952 set_opt (sbi->s_mount_opt, NO_UID32);
955 clear_opt (sbi->s_mount_opt, CHECK);
958 set_opt (sbi->s_mount_opt, DEBUG);
961 set_opt (sbi->s_mount_opt, OLDALLOC);
964 clear_opt (sbi->s_mount_opt, OLDALLOC);
966 #ifdef CONFIG_EXT4DEV_FS_XATTR
968 set_opt (sbi->s_mount_opt, XATTR_USER);
970 case Opt_nouser_xattr:
971 clear_opt (sbi->s_mount_opt, XATTR_USER);
975 case Opt_nouser_xattr:
976 printk("EXT4 (no)user_xattr options not supported\n");
979 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
981 set_opt(sbi->s_mount_opt, POSIX_ACL);
984 clear_opt(sbi->s_mount_opt, POSIX_ACL);
989 printk("EXT4 (no)acl options not supported\n");
992 case Opt_reservation:
993 set_opt(sbi->s_mount_opt, RESERVATION);
995 case Opt_noreservation:
996 clear_opt(sbi->s_mount_opt, RESERVATION);
998 case Opt_journal_update:
1000 /* Eventually we will want to be able to create
1001 a journal file here. For now, only allow the
1002 user to specify an existing inode to be the
1005 printk(KERN_ERR "EXT4-fs: cannot specify "
1006 "journal on remount\n");
1009 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1011 case Opt_journal_inum:
1013 printk(KERN_ERR "EXT4-fs: cannot specify "
1014 "journal on remount\n");
1017 if (match_int(&args[0], &option))
1021 case Opt_journal_dev:
1023 printk(KERN_ERR "EXT4-fs: cannot specify "
1024 "journal on remount\n");
1027 if (match_int(&args[0], &option))
1029 *journal_devnum = option;
1032 set_opt (sbi->s_mount_opt, NOLOAD);
1035 if (match_int(&args[0], &option))
1040 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1041 sbi->s_commit_interval = HZ * option;
1043 case Opt_data_journal:
1044 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1046 case Opt_data_ordered:
1047 data_opt = EXT4_MOUNT_ORDERED_DATA;
1049 case Opt_data_writeback:
1050 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1053 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1056 "EXT4-fs: cannot change data "
1057 "mode on remount\n");
1061 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1062 sbi->s_mount_opt |= data_opt;
1072 if (sb_any_quota_enabled(sb)) {
1074 "EXT4-fs: Cannot change journalled "
1075 "quota options when quota turned on.\n");
1078 qname = match_strdup(&args[0]);
1081 "EXT4-fs: not enough memory for "
1082 "storing quotafile name.\n");
1085 if (sbi->s_qf_names[qtype] &&
1086 strcmp(sbi->s_qf_names[qtype], qname)) {
1088 "EXT4-fs: %s quota file already "
1089 "specified.\n", QTYPE2NAME(qtype));
1093 sbi->s_qf_names[qtype] = qname;
1094 if (strchr(sbi->s_qf_names[qtype], '/')) {
1096 "EXT4-fs: quotafile must be on "
1097 "filesystem root.\n");
1098 kfree(sbi->s_qf_names[qtype]);
1099 sbi->s_qf_names[qtype] = NULL;
1102 set_opt(sbi->s_mount_opt, QUOTA);
1104 case Opt_offusrjquota:
1107 case Opt_offgrpjquota:
1110 if (sb_any_quota_enabled(sb)) {
1111 printk(KERN_ERR "EXT4-fs: Cannot change "
1112 "journalled quota options when "
1113 "quota turned on.\n");
1117 * The space will be released later when all options
1118 * are confirmed to be correct
1120 sbi->s_qf_names[qtype] = NULL;
1122 case Opt_jqfmt_vfsold:
1123 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1125 case Opt_jqfmt_vfsv0:
1126 sbi->s_jquota_fmt = QFMT_VFS_V0;
1130 set_opt(sbi->s_mount_opt, QUOTA);
1131 set_opt(sbi->s_mount_opt, USRQUOTA);
1134 set_opt(sbi->s_mount_opt, QUOTA);
1135 set_opt(sbi->s_mount_opt, GRPQUOTA);
1138 if (sb_any_quota_enabled(sb)) {
1139 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1140 "options when quota turned on.\n");
1143 clear_opt(sbi->s_mount_opt, QUOTA);
1144 clear_opt(sbi->s_mount_opt, USRQUOTA);
1145 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1153 case Opt_offusrjquota:
1154 case Opt_offgrpjquota:
1155 case Opt_jqfmt_vfsold:
1156 case Opt_jqfmt_vfsv0:
1158 "EXT4-fs: journalled quota options not "
1165 set_opt(sbi->s_mount_opt, ABORT);
1168 if (match_int(&args[0], &option))
1171 set_opt(sbi->s_mount_opt, BARRIER);
1173 clear_opt(sbi->s_mount_opt, BARRIER);
1179 printk("EXT4-fs: resize option only available "
1183 if (match_int(&args[0], &option) != 0)
1185 *n_blocks_count = option;
1188 set_opt(sbi->s_mount_opt, NOBH);
1191 clear_opt(sbi->s_mount_opt, NOBH);
1194 set_opt (sbi->s_mount_opt, EXTENTS);
1197 clear_opt (sbi->s_mount_opt, EXTENTS);
1201 "EXT4-fs: Unrecognized mount option \"%s\" "
1202 "or missing value\n", p);
1207 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1208 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1209 sbi->s_qf_names[USRQUOTA])
1210 clear_opt(sbi->s_mount_opt, USRQUOTA);
1212 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1213 sbi->s_qf_names[GRPQUOTA])
1214 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1216 if ((sbi->s_qf_names[USRQUOTA] &&
1217 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1218 (sbi->s_qf_names[GRPQUOTA] &&
1219 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1220 printk(KERN_ERR "EXT4-fs: old and new quota "
1221 "format mixing.\n");
1225 if (!sbi->s_jquota_fmt) {
1226 printk(KERN_ERR "EXT4-fs: journalled quota format "
1227 "not specified.\n");
1231 if (sbi->s_jquota_fmt) {
1232 printk(KERN_ERR "EXT4-fs: journalled quota format "
1233 "specified with no journalling "
1242 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1245 struct ext4_sb_info *sbi = EXT4_SB(sb);
1248 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1249 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1250 "forcing read-only mode\n");
1255 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1256 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1257 "running e2fsck is recommended\n");
1258 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1259 printk (KERN_WARNING
1260 "EXT4-fs warning: mounting fs with errors, "
1261 "running e2fsck is recommended\n");
1262 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1263 le16_to_cpu(es->s_mnt_count) >=
1264 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1265 printk (KERN_WARNING
1266 "EXT4-fs warning: maximal mount count reached, "
1267 "running e2fsck is recommended\n");
1268 else if (le32_to_cpu(es->s_checkinterval) &&
1269 (le32_to_cpu(es->s_lastcheck) +
1270 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1271 printk (KERN_WARNING
1272 "EXT4-fs warning: checktime reached, "
1273 "running e2fsck is recommended\n");
1275 /* @@@ We _will_ want to clear the valid bit if we find
1276 * inconsistencies, to force a fsck at reboot. But for
1277 * a plain journaled filesystem we can keep it set as
1280 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1282 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1283 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1284 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1285 es->s_mtime = cpu_to_le32(get_seconds());
1286 ext4_update_dynamic_rev(sb);
1287 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1289 ext4_commit_super(sb, es, 1);
1290 if (test_opt(sb, DEBUG))
1291 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1292 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1294 sbi->s_groups_count,
1295 EXT4_BLOCKS_PER_GROUP(sb),
1296 EXT4_INODES_PER_GROUP(sb),
1299 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1300 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1301 char b[BDEVNAME_SIZE];
1303 printk("external journal on %s\n",
1304 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1306 printk("internal journal\n");
1311 /* Called at mount-time, super-block is locked */
1312 static int ext4_check_descriptors (struct super_block * sb)
1314 struct ext4_sb_info *sbi = EXT4_SB(sb);
1315 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1316 ext4_fsblk_t last_block;
1317 ext4_fsblk_t block_bitmap;
1318 ext4_fsblk_t inode_bitmap;
1319 ext4_fsblk_t inode_table;
1320 struct ext4_group_desc * gdp = NULL;
1324 ext4_debug ("Checking group descriptors");
1326 for (i = 0; i < sbi->s_groups_count; i++)
1328 if (i == sbi->s_groups_count - 1)
1329 last_block = ext4_blocks_count(sbi->s_es) - 1;
1331 last_block = first_block +
1332 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1334 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1335 gdp = (struct ext4_group_desc *)
1336 sbi->s_group_desc[desc_block++]->b_data;
1337 block_bitmap = ext4_block_bitmap(sb, gdp);
1338 if (block_bitmap < first_block || block_bitmap > last_block)
1340 ext4_error (sb, "ext4_check_descriptors",
1341 "Block bitmap for group %d"
1342 " not in group (block %llu)!",
1346 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1347 if (inode_bitmap < first_block || inode_bitmap > last_block)
1349 ext4_error (sb, "ext4_check_descriptors",
1350 "Inode bitmap for group %d"
1351 " not in group (block %llu)!",
1355 inode_table = ext4_inode_table(sb, gdp);
1356 if (inode_table < first_block ||
1357 inode_table + sbi->s_itb_per_group - 1 > last_block)
1359 ext4_error (sb, "ext4_check_descriptors",
1360 "Inode table for group %d"
1361 " not in group (block %llu)!",
1365 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1366 gdp = (struct ext4_group_desc *)
1367 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1370 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1371 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1376 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1377 * the superblock) which were deleted from all directories, but held open by
1378 * a process at the time of a crash. We walk the list and try to delete these
1379 * inodes at recovery time (only with a read-write filesystem).
1381 * In order to keep the orphan inode chain consistent during traversal (in
1382 * case of crash during recovery), we link each inode into the superblock
1383 * orphan list_head and handle it the same way as an inode deletion during
1384 * normal operation (which journals the operations for us).
1386 * We only do an iget() and an iput() on each inode, which is very safe if we
1387 * accidentally point at an in-use or already deleted inode. The worst that
1388 * can happen in this case is that we get a "bit already cleared" message from
1389 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1390 * e2fsck was run on this filesystem, and it must have already done the orphan
1391 * inode cleanup for us, so we can safely abort without any further action.
1393 static void ext4_orphan_cleanup (struct super_block * sb,
1394 struct ext4_super_block * es)
1396 unsigned int s_flags = sb->s_flags;
1397 int nr_orphans = 0, nr_truncates = 0;
1401 if (!es->s_last_orphan) {
1402 jbd_debug(4, "no orphan inodes to clean up\n");
1406 if (bdev_read_only(sb->s_bdev)) {
1407 printk(KERN_ERR "EXT4-fs: write access "
1408 "unavailable, skipping orphan cleanup.\n");
1412 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1413 if (es->s_last_orphan)
1414 jbd_debug(1, "Errors on filesystem, "
1415 "clearing orphan list.\n");
1416 es->s_last_orphan = 0;
1417 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1421 if (s_flags & MS_RDONLY) {
1422 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1424 sb->s_flags &= ~MS_RDONLY;
1427 /* Needed for iput() to work correctly and not trash data */
1428 sb->s_flags |= MS_ACTIVE;
1429 /* Turn on quotas so that they are updated correctly */
1430 for (i = 0; i < MAXQUOTAS; i++) {
1431 if (EXT4_SB(sb)->s_qf_names[i]) {
1432 int ret = ext4_quota_on_mount(sb, i);
1435 "EXT4-fs: Cannot turn on journalled "
1436 "quota: error %d\n", ret);
1441 while (es->s_last_orphan) {
1442 struct inode *inode;
1445 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1446 es->s_last_orphan = 0;
1450 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1452 if (inode->i_nlink) {
1454 "%s: truncating inode %lu to %Ld bytes\n",
1455 __FUNCTION__, inode->i_ino, inode->i_size);
1456 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1457 inode->i_ino, inode->i_size);
1458 ext4_truncate(inode);
1462 "%s: deleting unreferenced inode %lu\n",
1463 __FUNCTION__, inode->i_ino);
1464 jbd_debug(2, "deleting unreferenced inode %lu\n",
1468 iput(inode); /* The delete magic happens here! */
1471 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1474 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1475 sb->s_id, PLURAL(nr_orphans));
1477 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1478 sb->s_id, PLURAL(nr_truncates));
1480 /* Turn quotas off */
1481 for (i = 0; i < MAXQUOTAS; i++) {
1482 if (sb_dqopt(sb)->files[i])
1483 vfs_quota_off(sb, i);
1486 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1490 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1491 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1492 * We need to be 1 filesystem block less than the 2^32 sector limit.
1494 static loff_t ext4_max_size(int bits)
1496 loff_t res = EXT4_NDIR_BLOCKS;
1497 /* This constant is calculated to be the largest file size for a
1498 * dense, 4k-blocksize file such that the total number of
1499 * sectors in the file, including data and all indirect blocks,
1500 * does not exceed 2^32. */
1501 const loff_t upper_limit = 0x1ff7fffd000LL;
1503 res += 1LL << (bits-2);
1504 res += 1LL << (2*(bits-2));
1505 res += 1LL << (3*(bits-2));
1507 if (res > upper_limit)
1512 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1513 ext4_fsblk_t logical_sb_block, int nr)
1515 struct ext4_sb_info *sbi = EXT4_SB(sb);
1516 unsigned long bg, first_meta_bg;
1519 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1521 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1523 return logical_sb_block + nr + 1;
1524 bg = sbi->s_desc_per_block * nr;
1525 if (ext4_bg_has_super(sb, bg))
1527 return (has_super + ext4_group_first_block_no(sb, bg));
1531 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1533 struct buffer_head * bh;
1534 struct ext4_super_block *es = NULL;
1535 struct ext4_sb_info *sbi;
1537 ext4_fsblk_t sb_block = get_sb_block(&data);
1538 ext4_fsblk_t logical_sb_block;
1539 unsigned long offset = 0;
1540 unsigned int journal_inum = 0;
1541 unsigned long journal_devnum = 0;
1542 unsigned long def_mount_opts;
1553 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1556 sb->s_fs_info = sbi;
1557 sbi->s_mount_opt = 0;
1558 sbi->s_resuid = EXT4_DEF_RESUID;
1559 sbi->s_resgid = EXT4_DEF_RESGID;
1560 sbi->s_sb_block = sb_block;
1564 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1566 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1571 * The ext4 superblock will not be buffer aligned for other than 1kB
1572 * block sizes. We need to calculate the offset from buffer start.
1574 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1575 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1576 offset = do_div(logical_sb_block, blocksize);
1578 logical_sb_block = sb_block;
1581 if (!(bh = sb_bread(sb, logical_sb_block))) {
1582 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1586 * Note: s_es must be initialized as soon as possible because
1587 * some ext4 macro-instructions depend on its value
1589 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1591 sb->s_magic = le16_to_cpu(es->s_magic);
1592 if (sb->s_magic != EXT4_SUPER_MAGIC)
1595 /* Set defaults before we parse the mount options */
1596 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1597 if (def_mount_opts & EXT4_DEFM_DEBUG)
1598 set_opt(sbi->s_mount_opt, DEBUG);
1599 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1600 set_opt(sbi->s_mount_opt, GRPID);
1601 if (def_mount_opts & EXT4_DEFM_UID16)
1602 set_opt(sbi->s_mount_opt, NO_UID32);
1603 #ifdef CONFIG_EXT4DEV_FS_XATTR
1604 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1605 set_opt(sbi->s_mount_opt, XATTR_USER);
1607 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1608 if (def_mount_opts & EXT4_DEFM_ACL)
1609 set_opt(sbi->s_mount_opt, POSIX_ACL);
1611 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1612 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1613 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1614 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1615 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1616 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1618 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1619 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1620 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1621 set_opt(sbi->s_mount_opt, ERRORS_RO);
1623 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1625 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1626 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1628 set_opt(sbi->s_mount_opt, RESERVATION);
1631 * turn on extents feature by default in ext4 filesystem
1632 * User -o noextents to turn it off
1634 set_opt(sbi->s_mount_opt, EXTENTS);
1636 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1640 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1641 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1643 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1644 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1645 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1646 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1648 "EXT4-fs warning: feature flags set on rev 0 fs, "
1649 "running e2fsck is recommended\n");
1651 * Check feature flags regardless of the revision level, since we
1652 * previously didn't change the revision level when setting the flags,
1653 * so there is a chance incompat flags are set on a rev 0 filesystem.
1655 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1657 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1658 "unsupported optional features (%x).\n",
1659 sb->s_id, le32_to_cpu(features));
1662 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1663 if (!(sb->s_flags & MS_RDONLY) && features) {
1664 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1665 "unsupported optional features (%x).\n",
1666 sb->s_id, le32_to_cpu(features));
1669 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1671 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1672 blocksize > EXT4_MAX_BLOCK_SIZE) {
1674 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1675 blocksize, sb->s_id);
1679 hblock = bdev_hardsect_size(sb->s_bdev);
1680 if (sb->s_blocksize != blocksize) {
1682 * Make sure the blocksize for the filesystem is larger
1683 * than the hardware sectorsize for the machine.
1685 if (blocksize < hblock) {
1686 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1687 "device blocksize %d.\n", blocksize, hblock);
1692 sb_set_blocksize(sb, blocksize);
1693 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1694 offset = do_div(logical_sb_block, blocksize);
1695 bh = sb_bread(sb, logical_sb_block);
1698 "EXT4-fs: Can't read superblock on 2nd try.\n");
1701 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1703 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1705 "EXT4-fs: Magic mismatch, very weird !\n");
1710 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1712 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1713 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1714 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1716 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1717 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1718 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1719 (!is_power_of_2(sbi->s_inode_size)) ||
1720 (sbi->s_inode_size > blocksize)) {
1722 "EXT4-fs: unsupported inode size: %d\n",
1726 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1727 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1729 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1730 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1731 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1732 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1733 !is_power_of_2(sbi->s_desc_size)) {
1735 "EXT4-fs: unsupported descriptor size %lu\n",
1740 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1741 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1742 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1743 if (EXT4_INODE_SIZE(sb) == 0)
1745 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1746 if (sbi->s_inodes_per_block == 0)
1748 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1749 sbi->s_inodes_per_block;
1750 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1752 sbi->s_mount_state = le16_to_cpu(es->s_state);
1753 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1754 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1755 for (i=0; i < 4; i++)
1756 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1757 sbi->s_def_hash_version = es->s_def_hash_version;
1759 if (sbi->s_blocks_per_group > blocksize * 8) {
1761 "EXT4-fs: #blocks per group too big: %lu\n",
1762 sbi->s_blocks_per_group);
1765 if (sbi->s_inodes_per_group > blocksize * 8) {
1767 "EXT4-fs: #inodes per group too big: %lu\n",
1768 sbi->s_inodes_per_group);
1772 if (ext4_blocks_count(es) >
1773 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1774 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1775 " too large to mount safely\n", sb->s_id);
1776 if (sizeof(sector_t) < 8)
1777 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1782 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1784 blocks_count = (ext4_blocks_count(es) -
1785 le32_to_cpu(es->s_first_data_block) +
1786 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1787 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1788 sbi->s_groups_count = blocks_count;
1789 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1790 EXT4_DESC_PER_BLOCK(sb);
1791 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1793 if (sbi->s_group_desc == NULL) {
1794 printk (KERN_ERR "EXT4-fs: not enough memory\n");
1798 bgl_lock_init(&sbi->s_blockgroup_lock);
1800 for (i = 0; i < db_count; i++) {
1801 block = descriptor_loc(sb, logical_sb_block, i);
1802 sbi->s_group_desc[i] = sb_bread(sb, block);
1803 if (!sbi->s_group_desc[i]) {
1804 printk (KERN_ERR "EXT4-fs: "
1805 "can't read group descriptor %d\n", i);
1810 if (!ext4_check_descriptors (sb)) {
1811 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1814 sbi->s_gdb_count = db_count;
1815 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1816 spin_lock_init(&sbi->s_next_gen_lock);
1818 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1819 ext4_count_free_blocks(sb));
1821 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1822 ext4_count_free_inodes(sb));
1825 err = percpu_counter_init(&sbi->s_dirs_counter,
1826 ext4_count_dirs(sb));
1829 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
1833 /* per fileystem reservation list head & lock */
1834 spin_lock_init(&sbi->s_rsv_window_lock);
1835 sbi->s_rsv_window_root = RB_ROOT;
1836 /* Add a single, static dummy reservation to the start of the
1837 * reservation window list --- it gives us a placeholder for
1838 * append-at-start-of-list which makes the allocation logic
1839 * _much_ simpler. */
1840 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1841 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1842 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1843 sbi->s_rsv_window_head.rsv_goal_size = 0;
1844 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1847 * set up enough so that it can read an inode
1849 sb->s_op = &ext4_sops;
1850 sb->s_export_op = &ext4_export_ops;
1851 sb->s_xattr = ext4_xattr_handlers;
1853 sb->s_qcop = &ext4_qctl_operations;
1854 sb->dq_op = &ext4_quota_operations;
1856 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1860 needs_recovery = (es->s_last_orphan != 0 ||
1861 EXT4_HAS_INCOMPAT_FEATURE(sb,
1862 EXT4_FEATURE_INCOMPAT_RECOVER));
1865 * The first inode we look at is the journal inode. Don't try
1866 * root first: it may be modified in the journal!
1868 if (!test_opt(sb, NOLOAD) &&
1869 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1870 if (ext4_load_journal(sb, es, journal_devnum))
1872 } else if (journal_inum) {
1873 if (ext4_create_journal(sb, es, journal_inum))
1878 "ext4: No journal on filesystem on %s\n",
1883 if (ext4_blocks_count(es) > 0xffffffffULL &&
1884 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1885 JBD2_FEATURE_INCOMPAT_64BIT)) {
1886 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1890 /* We have now updated the journal if required, so we can
1891 * validate the data journaling mode. */
1892 switch (test_opt(sb, DATA_FLAGS)) {
1894 /* No mode set, assume a default based on the journal
1895 * capabilities: ORDERED_DATA if the journal can
1896 * cope, else JOURNAL_DATA
1898 if (jbd2_journal_check_available_features
1899 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1900 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1902 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1905 case EXT4_MOUNT_ORDERED_DATA:
1906 case EXT4_MOUNT_WRITEBACK_DATA:
1907 if (!jbd2_journal_check_available_features
1908 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1909 printk(KERN_ERR "EXT4-fs: Journal does not support "
1910 "requested data journaling mode\n");
1917 if (test_opt(sb, NOBH)) {
1918 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1919 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1920 "its supported only with writeback mode\n");
1921 clear_opt(sbi->s_mount_opt, NOBH);
1925 * The jbd2_journal_load will have done any necessary log recovery,
1926 * so we can safely mount the rest of the filesystem now.
1929 root = iget(sb, EXT4_ROOT_INO);
1930 sb->s_root = d_alloc_root(root);
1932 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1936 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1939 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1943 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1945 /* determine the minimum size of new large inodes, if present */
1946 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1947 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1948 EXT4_GOOD_OLD_INODE_SIZE;
1949 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1950 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
1951 if (sbi->s_want_extra_isize <
1952 le16_to_cpu(es->s_want_extra_isize))
1953 sbi->s_want_extra_isize =
1954 le16_to_cpu(es->s_want_extra_isize);
1955 if (sbi->s_want_extra_isize <
1956 le16_to_cpu(es->s_min_extra_isize))
1957 sbi->s_want_extra_isize =
1958 le16_to_cpu(es->s_min_extra_isize);
1961 /* Check if enough inode space is available */
1962 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
1963 sbi->s_inode_size) {
1964 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1965 EXT4_GOOD_OLD_INODE_SIZE;
1966 printk(KERN_INFO "EXT4-fs: required extra inode space not"
1971 * akpm: core read_super() calls in here with the superblock locked.
1972 * That deadlocks, because orphan cleanup needs to lock the superblock
1973 * in numerous places. Here we just pop the lock - it's relatively
1974 * harmless, because we are now ready to accept write_super() requests,
1975 * and aviro says that's the only reason for hanging onto the
1978 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1979 ext4_orphan_cleanup(sb, es);
1980 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
1982 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1983 ext4_mark_recovery_complete(sb, es);
1984 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1985 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1986 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
1996 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2001 jbd2_journal_destroy(sbi->s_journal);
2003 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2004 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2005 percpu_counter_destroy(&sbi->s_dirs_counter);
2007 for (i = 0; i < db_count; i++)
2008 brelse(sbi->s_group_desc[i]);
2009 kfree(sbi->s_group_desc);
2012 for (i = 0; i < MAXQUOTAS; i++)
2013 kfree(sbi->s_qf_names[i]);
2015 ext4_blkdev_remove(sbi);
2018 sb->s_fs_info = NULL;
2025 * Setup any per-fs journal parameters now. We'll do this both on
2026 * initial mount, once the journal has been initialised but before we've
2027 * done any recovery; and again on any subsequent remount.
2029 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2031 struct ext4_sb_info *sbi = EXT4_SB(sb);
2033 if (sbi->s_commit_interval)
2034 journal->j_commit_interval = sbi->s_commit_interval;
2035 /* We could also set up an ext4-specific default for the commit
2036 * interval here, but for now we'll just fall back to the jbd
2039 spin_lock(&journal->j_state_lock);
2040 if (test_opt(sb, BARRIER))
2041 journal->j_flags |= JBD2_BARRIER;
2043 journal->j_flags &= ~JBD2_BARRIER;
2044 spin_unlock(&journal->j_state_lock);
2047 static journal_t *ext4_get_journal(struct super_block *sb,
2048 unsigned int journal_inum)
2050 struct inode *journal_inode;
2053 /* First, test for the existence of a valid inode on disk. Bad
2054 * things happen if we iget() an unused inode, as the subsequent
2055 * iput() will try to delete it. */
2057 journal_inode = iget(sb, journal_inum);
2058 if (!journal_inode) {
2059 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2062 if (!journal_inode->i_nlink) {
2063 make_bad_inode(journal_inode);
2064 iput(journal_inode);
2065 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2069 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2070 journal_inode, journal_inode->i_size);
2071 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2072 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2073 iput(journal_inode);
2077 journal = jbd2_journal_init_inode(journal_inode);
2079 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2080 iput(journal_inode);
2083 journal->j_private = sb;
2084 ext4_init_journal_params(sb, journal);
2088 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2091 struct buffer_head * bh;
2095 int hblock, blocksize;
2096 ext4_fsblk_t sb_block;
2097 unsigned long offset;
2098 struct ext4_super_block * es;
2099 struct block_device *bdev;
2101 bdev = ext4_blkdev_get(j_dev);
2105 if (bd_claim(bdev, sb)) {
2107 "EXT4: failed to claim external journal device.\n");
2112 blocksize = sb->s_blocksize;
2113 hblock = bdev_hardsect_size(bdev);
2114 if (blocksize < hblock) {
2116 "EXT4-fs: blocksize too small for journal device.\n");
2120 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2121 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2122 set_blocksize(bdev, blocksize);
2123 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2124 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2125 "external journal\n");
2129 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2130 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2131 !(le32_to_cpu(es->s_feature_incompat) &
2132 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2133 printk(KERN_ERR "EXT4-fs: external journal has "
2134 "bad superblock\n");
2139 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2140 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2145 len = ext4_blocks_count(es);
2146 start = sb_block + 1;
2147 brelse(bh); /* we're done with the superblock */
2149 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2150 start, len, blocksize);
2152 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2155 journal->j_private = sb;
2156 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2157 wait_on_buffer(journal->j_sb_buffer);
2158 if (!buffer_uptodate(journal->j_sb_buffer)) {
2159 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2162 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2163 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2164 "user (unsupported) - %d\n",
2165 be32_to_cpu(journal->j_superblock->s_nr_users));
2168 EXT4_SB(sb)->journal_bdev = bdev;
2169 ext4_init_journal_params(sb, journal);
2172 jbd2_journal_destroy(journal);
2174 ext4_blkdev_put(bdev);
2178 static int ext4_load_journal(struct super_block *sb,
2179 struct ext4_super_block *es,
2180 unsigned long journal_devnum)
2183 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2186 int really_read_only;
2188 if (journal_devnum &&
2189 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2190 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2191 "numbers have changed\n");
2192 journal_dev = new_decode_dev(journal_devnum);
2194 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2196 really_read_only = bdev_read_only(sb->s_bdev);
2199 * Are we loading a blank journal or performing recovery after a
2200 * crash? For recovery, we need to check in advance whether we
2201 * can get read-write access to the device.
2204 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2205 if (sb->s_flags & MS_RDONLY) {
2206 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2207 "required on readonly filesystem.\n");
2208 if (really_read_only) {
2209 printk(KERN_ERR "EXT4-fs: write access "
2210 "unavailable, cannot proceed.\n");
2213 printk (KERN_INFO "EXT4-fs: write access will "
2214 "be enabled during recovery.\n");
2218 if (journal_inum && journal_dev) {
2219 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2220 "and inode journals!\n");
2225 if (!(journal = ext4_get_journal(sb, journal_inum)))
2228 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2232 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2233 err = jbd2_journal_update_format(journal);
2235 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2236 jbd2_journal_destroy(journal);
2241 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2242 err = jbd2_journal_wipe(journal, !really_read_only);
2244 err = jbd2_journal_load(journal);
2247 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2248 jbd2_journal_destroy(journal);
2252 EXT4_SB(sb)->s_journal = journal;
2253 ext4_clear_journal_err(sb, es);
2255 if (journal_devnum &&
2256 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2257 es->s_journal_dev = cpu_to_le32(journal_devnum);
2260 /* Make sure we flush the recovery flag to disk. */
2261 ext4_commit_super(sb, es, 1);
2267 static int ext4_create_journal(struct super_block * sb,
2268 struct ext4_super_block * es,
2269 unsigned int journal_inum)
2274 if (sb->s_flags & MS_RDONLY) {
2275 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2276 "create journal.\n");
2280 journal = ext4_get_journal(sb, journal_inum);
2284 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2287 err = jbd2_journal_create(journal);
2289 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2290 jbd2_journal_destroy(journal);
2294 EXT4_SB(sb)->s_journal = journal;
2296 ext4_update_dynamic_rev(sb);
2297 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2298 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2300 es->s_journal_inum = cpu_to_le32(journal_inum);
2303 /* Make sure we flush the recovery flag to disk. */
2304 ext4_commit_super(sb, es, 1);
2309 static void ext4_commit_super (struct super_block * sb,
2310 struct ext4_super_block * es,
2313 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2317 es->s_wtime = cpu_to_le32(get_seconds());
2318 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2319 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2320 BUFFER_TRACE(sbh, "marking dirty");
2321 mark_buffer_dirty(sbh);
2323 sync_dirty_buffer(sbh);
2328 * Have we just finished recovery? If so, and if we are mounting (or
2329 * remounting) the filesystem readonly, then we will end up with a
2330 * consistent fs on disk. Record that fact.
2332 static void ext4_mark_recovery_complete(struct super_block * sb,
2333 struct ext4_super_block * es)
2335 journal_t *journal = EXT4_SB(sb)->s_journal;
2337 jbd2_journal_lock_updates(journal);
2338 jbd2_journal_flush(journal);
2340 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2341 sb->s_flags & MS_RDONLY) {
2342 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2344 ext4_commit_super(sb, es, 1);
2347 jbd2_journal_unlock_updates(journal);
2351 * If we are mounting (or read-write remounting) a filesystem whose journal
2352 * has recorded an error from a previous lifetime, move that error to the
2353 * main filesystem now.
2355 static void ext4_clear_journal_err(struct super_block * sb,
2356 struct ext4_super_block * es)
2362 journal = EXT4_SB(sb)->s_journal;
2365 * Now check for any error status which may have been recorded in the
2366 * journal by a prior ext4_error() or ext4_abort()
2369 j_errno = jbd2_journal_errno(journal);
2373 errstr = ext4_decode_error(sb, j_errno, nbuf);
2374 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2375 "from previous mount: %s", errstr);
2376 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2377 "filesystem check.");
2379 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2380 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2381 ext4_commit_super (sb, es, 1);
2383 jbd2_journal_clear_err(journal);
2388 * Force the running and committing transactions to commit,
2389 * and wait on the commit.
2391 int ext4_force_commit(struct super_block *sb)
2396 if (sb->s_flags & MS_RDONLY)
2399 journal = EXT4_SB(sb)->s_journal;
2401 ret = ext4_journal_force_commit(journal);
2406 * Ext4 always journals updates to the superblock itself, so we don't
2407 * have to propagate any other updates to the superblock on disk at this
2408 * point. Just start an async writeback to get the buffers on their way
2411 * This implicitly triggers the writebehind on sync().
2414 static void ext4_write_super (struct super_block * sb)
2416 if (mutex_trylock(&sb->s_lock) != 0)
2421 static int ext4_sync_fs(struct super_block *sb, int wait)
2426 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2428 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2434 * LVM calls this function before a (read-only) snapshot is created. This
2435 * gives us a chance to flush the journal completely and mark the fs clean.
2437 static void ext4_write_super_lockfs(struct super_block *sb)
2441 if (!(sb->s_flags & MS_RDONLY)) {
2442 journal_t *journal = EXT4_SB(sb)->s_journal;
2444 /* Now we set up the journal barrier. */
2445 jbd2_journal_lock_updates(journal);
2446 jbd2_journal_flush(journal);
2448 /* Journal blocked and flushed, clear needs_recovery flag. */
2449 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2450 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2455 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2456 * flag here, even though the filesystem is not technically dirty yet.
2458 static void ext4_unlockfs(struct super_block *sb)
2460 if (!(sb->s_flags & MS_RDONLY)) {
2462 /* Reser the needs_recovery flag before the fs is unlocked. */
2463 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2464 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2466 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2470 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2472 struct ext4_super_block * es;
2473 struct ext4_sb_info *sbi = EXT4_SB(sb);
2474 ext4_fsblk_t n_blocks_count = 0;
2475 unsigned long old_sb_flags;
2476 struct ext4_mount_options old_opts;
2482 /* Store the original options */
2483 old_sb_flags = sb->s_flags;
2484 old_opts.s_mount_opt = sbi->s_mount_opt;
2485 old_opts.s_resuid = sbi->s_resuid;
2486 old_opts.s_resgid = sbi->s_resgid;
2487 old_opts.s_commit_interval = sbi->s_commit_interval;
2489 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2490 for (i = 0; i < MAXQUOTAS; i++)
2491 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2495 * Allow the "check" option to be passed as a remount option.
2497 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2502 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2503 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2505 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2506 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2510 ext4_init_journal_params(sb, sbi->s_journal);
2512 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2513 n_blocks_count > ext4_blocks_count(es)) {
2514 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2519 if (*flags & MS_RDONLY) {
2521 * First of all, the unconditional stuff we have to do
2522 * to disable replay of the journal when we next remount
2524 sb->s_flags |= MS_RDONLY;
2527 * OK, test if we are remounting a valid rw partition
2528 * readonly, and if so set the rdonly flag and then
2529 * mark the partition as valid again.
2531 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2532 (sbi->s_mount_state & EXT4_VALID_FS))
2533 es->s_state = cpu_to_le16(sbi->s_mount_state);
2536 * We have to unlock super so that we can wait for
2540 ext4_mark_recovery_complete(sb, es);
2544 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2545 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2546 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2547 "remount RDWR because of unsupported "
2548 "optional features (%x).\n",
2549 sb->s_id, le32_to_cpu(ret));
2555 * If we have an unprocessed orphan list hanging
2556 * around from a previously readonly bdev mount,
2557 * require a full umount/remount for now.
2559 if (es->s_last_orphan) {
2560 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2561 "remount RDWR because of unprocessed "
2562 "orphan inode list. Please "
2563 "umount/remount instead.\n",
2570 * Mounting a RDONLY partition read-write, so reread
2571 * and store the current valid flag. (It may have
2572 * been changed by e2fsck since we originally mounted
2575 ext4_clear_journal_err(sb, es);
2576 sbi->s_mount_state = le16_to_cpu(es->s_state);
2577 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2579 if (!ext4_setup_super (sb, es, 0))
2580 sb->s_flags &= ~MS_RDONLY;
2584 /* Release old quota file names */
2585 for (i = 0; i < MAXQUOTAS; i++)
2586 if (old_opts.s_qf_names[i] &&
2587 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2588 kfree(old_opts.s_qf_names[i]);
2592 sb->s_flags = old_sb_flags;
2593 sbi->s_mount_opt = old_opts.s_mount_opt;
2594 sbi->s_resuid = old_opts.s_resuid;
2595 sbi->s_resgid = old_opts.s_resgid;
2596 sbi->s_commit_interval = old_opts.s_commit_interval;
2598 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2599 for (i = 0; i < MAXQUOTAS; i++) {
2600 if (sbi->s_qf_names[i] &&
2601 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2602 kfree(sbi->s_qf_names[i]);
2603 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2609 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2611 struct super_block *sb = dentry->d_sb;
2612 struct ext4_sb_info *sbi = EXT4_SB(sb);
2613 struct ext4_super_block *es = sbi->s_es;
2616 if (test_opt(sb, MINIX_DF)) {
2617 sbi->s_overhead_last = 0;
2618 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2619 unsigned long ngroups = sbi->s_groups_count, i;
2620 ext4_fsblk_t overhead = 0;
2624 * Compute the overhead (FS structures). This is constant
2625 * for a given filesystem unless the number of block groups
2626 * changes so we cache the previous value until it does.
2630 * All of the blocks before first_data_block are
2633 overhead = le32_to_cpu(es->s_first_data_block);
2636 * Add the overhead attributed to the superblock and
2637 * block group descriptors. If the sparse superblocks
2638 * feature is turned on, then not all groups have this.
2640 for (i = 0; i < ngroups; i++) {
2641 overhead += ext4_bg_has_super(sb, i) +
2642 ext4_bg_num_gdb(sb, i);
2647 * Every block group has an inode bitmap, a block
2648 * bitmap, and an inode table.
2650 overhead += ngroups * (2 + sbi->s_itb_per_group);
2651 sbi->s_overhead_last = overhead;
2653 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2656 buf->f_type = EXT4_SUPER_MAGIC;
2657 buf->f_bsize = sb->s_blocksize;
2658 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2659 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2660 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2661 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2662 if (buf->f_bfree < ext4_r_blocks_count(es))
2664 buf->f_files = le32_to_cpu(es->s_inodes_count);
2665 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2666 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2667 buf->f_namelen = EXT4_NAME_LEN;
2668 fsid = le64_to_cpup((void *)es->s_uuid) ^
2669 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2670 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2671 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2675 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2676 * is locked for write. Otherwise the are possible deadlocks:
2677 * Process 1 Process 2
2678 * ext4_create() quota_sync()
2679 * jbd2_journal_start() write_dquot()
2680 * DQUOT_INIT() down(dqio_mutex)
2681 * down(dqio_mutex) jbd2_journal_start()
2687 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2689 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2692 static int ext4_dquot_initialize(struct inode *inode, int type)
2697 /* We may create quota structure so we need to reserve enough blocks */
2698 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2700 return PTR_ERR(handle);
2701 ret = dquot_initialize(inode, type);
2702 err = ext4_journal_stop(handle);
2708 static int ext4_dquot_drop(struct inode *inode)
2713 /* We may delete quota structure so we need to reserve enough blocks */
2714 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2716 return PTR_ERR(handle);
2717 ret = dquot_drop(inode);
2718 err = ext4_journal_stop(handle);
2724 static int ext4_write_dquot(struct dquot *dquot)
2728 struct inode *inode;
2730 inode = dquot_to_inode(dquot);
2731 handle = ext4_journal_start(inode,
2732 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2734 return PTR_ERR(handle);
2735 ret = dquot_commit(dquot);
2736 err = ext4_journal_stop(handle);
2742 static int ext4_acquire_dquot(struct dquot *dquot)
2747 handle = ext4_journal_start(dquot_to_inode(dquot),
2748 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2750 return PTR_ERR(handle);
2751 ret = dquot_acquire(dquot);
2752 err = ext4_journal_stop(handle);
2758 static int ext4_release_dquot(struct dquot *dquot)
2763 handle = ext4_journal_start(dquot_to_inode(dquot),
2764 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2765 if (IS_ERR(handle)) {
2766 /* Release dquot anyway to avoid endless cycle in dqput() */
2767 dquot_release(dquot);
2768 return PTR_ERR(handle);
2770 ret = dquot_release(dquot);
2771 err = ext4_journal_stop(handle);
2777 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2779 /* Are we journalling quotas? */
2780 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2781 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2782 dquot_mark_dquot_dirty(dquot);
2783 return ext4_write_dquot(dquot);
2785 return dquot_mark_dquot_dirty(dquot);
2789 static int ext4_write_info(struct super_block *sb, int type)
2794 /* Data block + inode block */
2795 handle = ext4_journal_start(sb->s_root->d_inode, 2);
2797 return PTR_ERR(handle);
2798 ret = dquot_commit_info(sb, type);
2799 err = ext4_journal_stop(handle);
2806 * Turn on quotas during mount time - we need to find
2807 * the quota file and such...
2809 static int ext4_quota_on_mount(struct super_block *sb, int type)
2811 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2812 EXT4_SB(sb)->s_jquota_fmt, type);
2816 * Standard function to be called on quota_on
2818 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2822 struct nameidata nd;
2824 if (!test_opt(sb, QUOTA))
2826 /* Not journalling quota? */
2827 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2828 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2829 return vfs_quota_on(sb, type, format_id, path);
2830 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2833 /* Quotafile not on the same filesystem? */
2834 if (nd.mnt->mnt_sb != sb) {
2838 /* Quotafile not of fs root? */
2839 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2841 "EXT4-fs: Quota file not on filesystem root. "
2842 "Journalled quota will not work.\n");
2844 return vfs_quota_on(sb, type, format_id, path);
2847 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2848 * acquiring the locks... As quota files are never truncated and quota code
2849 * itself serializes the operations (and noone else should touch the files)
2850 * we don't have to be afraid of races */
2851 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2852 size_t len, loff_t off)
2854 struct inode *inode = sb_dqopt(sb)->files[type];
2855 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2857 int offset = off & (sb->s_blocksize - 1);
2860 struct buffer_head *bh;
2861 loff_t i_size = i_size_read(inode);
2865 if (off+len > i_size)
2868 while (toread > 0) {
2869 tocopy = sb->s_blocksize - offset < toread ?
2870 sb->s_blocksize - offset : toread;
2871 bh = ext4_bread(NULL, inode, blk, 0, &err);
2874 if (!bh) /* A hole? */
2875 memset(data, 0, tocopy);
2877 memcpy(data, bh->b_data+offset, tocopy);
2887 /* Write to quotafile (we know the transaction is already started and has
2888 * enough credits) */
2889 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2890 const char *data, size_t len, loff_t off)
2892 struct inode *inode = sb_dqopt(sb)->files[type];
2893 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2895 int offset = off & (sb->s_blocksize - 1);
2897 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2898 size_t towrite = len;
2899 struct buffer_head *bh;
2900 handle_t *handle = journal_current_handle();
2903 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2904 " cancelled because transaction is not started.\n",
2905 (unsigned long long)off, (unsigned long long)len);
2908 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2909 while (towrite > 0) {
2910 tocopy = sb->s_blocksize - offset < towrite ?
2911 sb->s_blocksize - offset : towrite;
2912 bh = ext4_bread(handle, inode, blk, 1, &err);
2915 if (journal_quota) {
2916 err = ext4_journal_get_write_access(handle, bh);
2923 memcpy(bh->b_data+offset, data, tocopy);
2924 flush_dcache_page(bh->b_page);
2927 err = ext4_journal_dirty_metadata(handle, bh);
2929 /* Always do at least ordered writes for quotas */
2930 err = ext4_journal_dirty_data(handle, bh);
2931 mark_buffer_dirty(bh);
2944 if (inode->i_size < off+len-towrite) {
2945 i_size_write(inode, off+len-towrite);
2946 EXT4_I(inode)->i_disksize = inode->i_size;
2949 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2950 ext4_mark_inode_dirty(handle, inode);
2951 mutex_unlock(&inode->i_mutex);
2952 return len - towrite;
2957 static int ext4_get_sb(struct file_system_type *fs_type,
2958 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2960 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
2963 static struct file_system_type ext4dev_fs_type = {
2964 .owner = THIS_MODULE,
2966 .get_sb = ext4_get_sb,
2967 .kill_sb = kill_block_super,
2968 .fs_flags = FS_REQUIRES_DEV,
2971 static int __init init_ext4_fs(void)
2973 int err = init_ext4_xattr();
2976 err = init_inodecache();
2979 err = register_filesystem(&ext4dev_fs_type);
2984 destroy_inodecache();
2990 static void __exit exit_ext4_fs(void)
2992 unregister_filesystem(&ext4dev_fs_type);
2993 destroy_inodecache();
2997 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2998 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
2999 MODULE_LICENSE("GPL");
3000 module_init(init_ext4_fs)
3001 module_exit(exit_ext4_fs)