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>
40 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
49 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
50 unsigned long journal_devnum);
51 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
53 static void ext4_commit_super (struct super_block * sb,
54 struct ext4_super_block * es,
56 static void ext4_mark_recovery_complete(struct super_block * sb,
57 struct ext4_super_block * es);
58 static void ext4_clear_journal_err(struct super_block * sb,
59 struct ext4_super_block * es);
60 static int ext4_sync_fs(struct super_block *sb, int wait);
61 static const char *ext4_decode_error(struct super_block * sb, int errno,
63 static int ext4_remount (struct super_block * sb, int * flags, char * data);
64 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
65 static void ext4_unlockfs(struct super_block *sb);
66 static void ext4_write_super (struct super_block * sb);
67 static void ext4_write_super_lockfs(struct super_block *sb);
70 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
73 return le32_to_cpu(bg->bg_block_bitmap) |
74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
81 return le32_to_cpu(bg->bg_inode_bitmap) |
82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
89 return le32_to_cpu(bg->bg_inode_table) |
90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 void ext4_block_bitmap_set(struct super_block *sb,
95 struct ext4_group_desc *bg, ext4_fsblk_t blk)
97 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
98 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
99 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
102 void ext4_inode_bitmap_set(struct super_block *sb,
103 struct ext4_group_desc *bg, ext4_fsblk_t blk)
105 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
106 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
107 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
110 void ext4_inode_table_set(struct super_block *sb,
111 struct ext4_group_desc *bg, ext4_fsblk_t blk)
113 bg->bg_inode_table = cpu_to_le32((u32)blk);
114 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
115 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
119 * Wrappers for jbd2_journal_start/end.
121 * The only special thing we need to do here is to make sure that all
122 * journal_end calls result in the superblock being marked dirty, so
123 * that sync() will call the filesystem's write_super callback if
126 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
130 if (sb->s_flags & MS_RDONLY)
131 return ERR_PTR(-EROFS);
133 /* Special case here: if the journal has aborted behind our
134 * backs (eg. EIO in the commit thread), then we still need to
135 * take the FS itself readonly cleanly. */
136 journal = EXT4_SB(sb)->s_journal;
137 if (is_journal_aborted(journal)) {
138 ext4_abort(sb, __FUNCTION__,
139 "Detected aborted journal");
140 return ERR_PTR(-EROFS);
143 return jbd2_journal_start(journal, nblocks);
147 * The only special thing we need to do here is to make sure that all
148 * jbd2_journal_stop calls result in the superblock being marked dirty, so
149 * that sync() will call the filesystem's write_super callback if
152 int __ext4_journal_stop(const char *where, handle_t *handle)
154 struct super_block *sb;
158 sb = handle->h_transaction->t_journal->j_private;
160 rc = jbd2_journal_stop(handle);
165 __ext4_std_error(sb, where, err);
169 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
170 struct buffer_head *bh, handle_t *handle, int err)
173 const char *errstr = ext4_decode_error(NULL, err, nbuf);
176 BUFFER_TRACE(bh, "abort");
181 if (is_handle_aborted(handle))
184 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
185 caller, errstr, err_fn);
187 jbd2_journal_abort_handle(handle);
190 /* Deal with the reporting of failure conditions on a filesystem such as
191 * inconsistencies detected or read IO failures.
193 * On ext2, we can store the error state of the filesystem in the
194 * superblock. That is not possible on ext4, because we may have other
195 * write ordering constraints on the superblock which prevent us from
196 * writing it out straight away; and given that the journal is about to
197 * be aborted, we can't rely on the current, or future, transactions to
198 * write out the superblock safely.
200 * We'll just use the jbd2_journal_abort() error code to record an error in
201 * the journal instead. On recovery, the journal will compain about
202 * that error until we've noted it down and cleared it.
205 static void ext4_handle_error(struct super_block *sb)
207 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
209 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
210 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
212 if (sb->s_flags & MS_RDONLY)
215 if (!test_opt (sb, ERRORS_CONT)) {
216 journal_t *journal = EXT4_SB(sb)->s_journal;
218 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
220 jbd2_journal_abort(journal, -EIO);
222 if (test_opt (sb, ERRORS_RO)) {
223 printk (KERN_CRIT "Remounting filesystem read-only\n");
224 sb->s_flags |= MS_RDONLY;
226 ext4_commit_super(sb, es, 1);
227 if (test_opt(sb, ERRORS_PANIC))
228 panic("EXT4-fs (device %s): panic forced after error\n",
232 void ext4_error (struct super_block * sb, const char * function,
233 const char * fmt, ...)
238 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
243 ext4_handle_error(sb);
246 static const char *ext4_decode_error(struct super_block * sb, int errno,
253 errstr = "IO failure";
256 errstr = "Out of memory";
259 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
260 errstr = "Journal has aborted";
262 errstr = "Readonly filesystem";
265 /* If the caller passed in an extra buffer for unknown
266 * errors, textualise them now. Else we just return
269 /* Check for truncated error codes... */
270 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
279 /* __ext4_std_error decodes expected errors from journaling functions
280 * automatically and invokes the appropriate error response. */
282 void __ext4_std_error (struct super_block * sb, const char * function,
288 /* Special case: if the error is EROFS, and we're not already
289 * inside a transaction, then there's really no point in logging
291 if (errno == -EROFS && journal_current_handle() == NULL &&
292 (sb->s_flags & MS_RDONLY))
295 errstr = ext4_decode_error(sb, errno, nbuf);
296 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
297 sb->s_id, function, errstr);
299 ext4_handle_error(sb);
303 * ext4_abort is a much stronger failure handler than ext4_error. The
304 * abort function may be used to deal with unrecoverable failures such
305 * as journal IO errors or ENOMEM at a critical moment in log management.
307 * We unconditionally force the filesystem into an ABORT|READONLY state,
308 * unless the error response on the fs has been set to panic in which
309 * case we take the easy way out and panic immediately.
312 void ext4_abort (struct super_block * sb, const char * function,
313 const char * fmt, ...)
317 printk (KERN_CRIT "ext4_abort called.\n");
320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
325 if (test_opt(sb, ERRORS_PANIC))
326 panic("EXT4-fs panic from previous error\n");
328 if (sb->s_flags & MS_RDONLY)
331 printk(KERN_CRIT "Remounting filesystem read-only\n");
332 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
333 sb->s_flags |= MS_RDONLY;
334 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
335 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
338 void ext4_warning (struct super_block * sb, const char * function,
339 const char * fmt, ...)
344 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
351 void ext4_update_dynamic_rev(struct super_block *sb)
353 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
355 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
358 ext4_warning(sb, __FUNCTION__,
359 "updating to rev %d because of new feature flag, "
360 "running e2fsck is recommended",
363 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
366 /* leave es->s_feature_*compat flags alone */
367 /* es->s_uuid will be set by e2fsck if empty */
370 * The rest of the superblock fields should be zero, and if not it
371 * means they are likely already in use, so leave them alone. We
372 * can leave it up to e2fsck to clean up any inconsistencies there.
377 * Open the external journal device
379 static struct block_device *ext4_blkdev_get(dev_t dev)
381 struct block_device *bdev;
382 char b[BDEVNAME_SIZE];
384 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
390 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
391 __bdevname(dev, b), PTR_ERR(bdev));
396 * Release the journal device
398 static int ext4_blkdev_put(struct block_device *bdev)
401 return blkdev_put(bdev);
404 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
406 struct block_device *bdev;
409 bdev = sbi->journal_bdev;
411 ret = ext4_blkdev_put(bdev);
412 sbi->journal_bdev = NULL;
417 static inline struct inode *orphan_list_entry(struct list_head *l)
419 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
422 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
426 printk(KERN_ERR "sb orphan head is %d\n",
427 le32_to_cpu(sbi->s_es->s_last_orphan));
429 printk(KERN_ERR "sb_info orphan list:\n");
430 list_for_each(l, &sbi->s_orphan) {
431 struct inode *inode = orphan_list_entry(l);
433 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
434 inode->i_sb->s_id, inode->i_ino, inode,
435 inode->i_mode, inode->i_nlink,
440 static void ext4_put_super (struct super_block * sb)
442 struct ext4_sb_info *sbi = EXT4_SB(sb);
443 struct ext4_super_block *es = sbi->s_es;
446 ext4_ext_release(sb);
447 ext4_xattr_put_super(sb);
448 jbd2_journal_destroy(sbi->s_journal);
449 if (!(sb->s_flags & MS_RDONLY)) {
450 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
451 es->s_state = cpu_to_le16(sbi->s_mount_state);
452 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
453 mark_buffer_dirty(sbi->s_sbh);
454 ext4_commit_super(sb, es, 1);
457 for (i = 0; i < sbi->s_gdb_count; i++)
458 brelse(sbi->s_group_desc[i]);
459 kfree(sbi->s_group_desc);
460 percpu_counter_destroy(&sbi->s_freeblocks_counter);
461 percpu_counter_destroy(&sbi->s_freeinodes_counter);
462 percpu_counter_destroy(&sbi->s_dirs_counter);
465 for (i = 0; i < MAXQUOTAS; i++)
466 kfree(sbi->s_qf_names[i]);
469 /* Debugging code just in case the in-memory inode orphan list
470 * isn't empty. The on-disk one can be non-empty if we've
471 * detected an error and taken the fs readonly, but the
472 * in-memory list had better be clean by this point. */
473 if (!list_empty(&sbi->s_orphan))
474 dump_orphan_list(sb, sbi);
475 J_ASSERT(list_empty(&sbi->s_orphan));
477 invalidate_bdev(sb->s_bdev);
478 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
480 * Invalidate the journal device's buffers. We don't want them
481 * floating about in memory - the physical journal device may
482 * hotswapped, and it breaks the `ro-after' testing code.
484 sync_blockdev(sbi->journal_bdev);
485 invalidate_bdev(sbi->journal_bdev);
486 ext4_blkdev_remove(sbi);
488 sb->s_fs_info = NULL;
493 static struct kmem_cache *ext4_inode_cachep;
496 * Called inside transaction, so use GFP_NOFS
498 static struct inode *ext4_alloc_inode(struct super_block *sb)
500 struct ext4_inode_info *ei;
502 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
505 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
506 ei->i_acl = EXT4_ACL_NOT_CACHED;
507 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
509 ei->i_block_alloc_info = NULL;
510 ei->vfs_inode.i_version = 1;
511 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
512 return &ei->vfs_inode;
515 static void ext4_destroy_inode(struct inode *inode)
517 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
518 printk("EXT4 Inode %p: orphan list check failed!\n",
520 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
521 EXT4_I(inode), sizeof(struct ext4_inode_info),
525 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
528 static void init_once(struct kmem_cache *cachep, void *foo)
530 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
532 INIT_LIST_HEAD(&ei->i_orphan);
533 #ifdef CONFIG_EXT4DEV_FS_XATTR
534 init_rwsem(&ei->xattr_sem);
536 mutex_init(&ei->truncate_mutex);
537 inode_init_once(&ei->vfs_inode);
540 static int init_inodecache(void)
542 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
543 sizeof(struct ext4_inode_info),
544 0, (SLAB_RECLAIM_ACCOUNT|
547 if (ext4_inode_cachep == NULL)
552 static void destroy_inodecache(void)
554 kmem_cache_destroy(ext4_inode_cachep);
557 static void ext4_clear_inode(struct inode *inode)
559 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
560 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
561 if (EXT4_I(inode)->i_acl &&
562 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
563 posix_acl_release(EXT4_I(inode)->i_acl);
564 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
566 if (EXT4_I(inode)->i_default_acl &&
567 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
568 posix_acl_release(EXT4_I(inode)->i_default_acl);
569 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
572 ext4_discard_reservation(inode);
573 EXT4_I(inode)->i_block_alloc_info = NULL;
578 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
580 #if defined(CONFIG_QUOTA)
581 struct ext4_sb_info *sbi = EXT4_SB(sb);
583 if (sbi->s_jquota_fmt)
584 seq_printf(seq, ",jqfmt=%s",
585 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
587 if (sbi->s_qf_names[USRQUOTA])
588 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
590 if (sbi->s_qf_names[GRPQUOTA])
591 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
593 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
594 seq_puts(seq, ",usrquota");
596 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
597 seq_puts(seq, ",grpquota");
603 * - it's set to a non-default value OR
604 * - if the per-sb default is different from the global default
606 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
608 struct super_block *sb = vfs->mnt_sb;
609 struct ext4_sb_info *sbi = EXT4_SB(sb);
610 struct ext4_super_block *es = sbi->s_es;
611 unsigned long def_mount_opts;
613 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
615 if (sbi->s_sb_block != 1)
616 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
617 if (test_opt(sb, MINIX_DF))
618 seq_puts(seq, ",minixdf");
619 if (test_opt(sb, GRPID))
620 seq_puts(seq, ",grpid");
621 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
622 seq_puts(seq, ",nogrpid");
623 if (sbi->s_resuid != EXT4_DEF_RESUID ||
624 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
625 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
627 if (sbi->s_resgid != EXT4_DEF_RESGID ||
628 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
629 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
631 if (test_opt(sb, ERRORS_CONT)) {
632 int def_errors = le16_to_cpu(es->s_errors);
634 if (def_errors == EXT4_ERRORS_PANIC ||
635 def_errors == EXT4_ERRORS_RO) {
636 seq_puts(seq, ",errors=continue");
639 if (test_opt(sb, ERRORS_RO))
640 seq_puts(seq, ",errors=remount-ro");
641 if (test_opt(sb, ERRORS_PANIC))
642 seq_puts(seq, ",errors=panic");
643 if (test_opt(sb, NO_UID32))
644 seq_puts(seq, ",nouid32");
645 if (test_opt(sb, DEBUG))
646 seq_puts(seq, ",debug");
647 if (test_opt(sb, OLDALLOC))
648 seq_puts(seq, ",oldalloc");
649 #ifdef CONFIG_EXT4_FS_XATTR
650 if (test_opt(sb, XATTR_USER))
651 seq_puts(seq, ",user_xattr");
652 if (!test_opt(sb, XATTR_USER) &&
653 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
654 seq_puts(seq, ",nouser_xattr");
657 #ifdef CONFIG_EXT4_FS_POSIX_ACL
658 if (test_opt(sb, POSIX_ACL))
659 seq_puts(seq, ",acl");
660 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
661 seq_puts(seq, ",noacl");
663 if (!test_opt(sb, RESERVATION))
664 seq_puts(seq, ",noreservation");
665 if (sbi->s_commit_interval) {
666 seq_printf(seq, ",commit=%u",
667 (unsigned) (sbi->s_commit_interval / HZ));
669 if (test_opt(sb, BARRIER))
670 seq_puts(seq, ",barrier=1");
671 if (test_opt(sb, NOBH))
672 seq_puts(seq, ",nobh");
673 if (!test_opt(sb, EXTENTS))
674 seq_puts(seq, ",noextents");
676 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
677 seq_puts(seq, ",data=journal");
678 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
679 seq_puts(seq, ",data=ordered");
680 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
681 seq_puts(seq, ",data=writeback");
683 ext4_show_quota_options(seq, sb);
689 static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
692 unsigned long ino = objp[0];
693 __u32 generation = objp[1];
695 struct dentry *result;
697 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
698 return ERR_PTR(-ESTALE);
699 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
700 return ERR_PTR(-ESTALE);
702 /* iget isn't really right if the inode is currently unallocated!!
704 * ext4_read_inode will return a bad_inode if the inode had been
705 * deleted, so we should be safe.
707 * Currently we don't know the generation for parent directory, so
708 * a generation of 0 means "accept any"
710 inode = iget(sb, ino);
712 return ERR_PTR(-ENOMEM);
713 if (is_bad_inode(inode) ||
714 (generation && inode->i_generation != generation)) {
716 return ERR_PTR(-ESTALE);
718 /* now to find a dentry.
719 * If possible, get a well-connected one
721 result = d_alloc_anon(inode);
724 return ERR_PTR(-ENOMEM);
730 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
731 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
733 static int ext4_dquot_initialize(struct inode *inode, int type);
734 static int ext4_dquot_drop(struct inode *inode);
735 static int ext4_write_dquot(struct dquot *dquot);
736 static int ext4_acquire_dquot(struct dquot *dquot);
737 static int ext4_release_dquot(struct dquot *dquot);
738 static int ext4_mark_dquot_dirty(struct dquot *dquot);
739 static int ext4_write_info(struct super_block *sb, int type);
740 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
741 static int ext4_quota_on_mount(struct super_block *sb, int type);
742 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
743 size_t len, loff_t off);
744 static ssize_t ext4_quota_write(struct super_block *sb, int type,
745 const char *data, size_t len, loff_t off);
747 static struct dquot_operations ext4_quota_operations = {
748 .initialize = ext4_dquot_initialize,
749 .drop = ext4_dquot_drop,
750 .alloc_space = dquot_alloc_space,
751 .alloc_inode = dquot_alloc_inode,
752 .free_space = dquot_free_space,
753 .free_inode = dquot_free_inode,
754 .transfer = dquot_transfer,
755 .write_dquot = ext4_write_dquot,
756 .acquire_dquot = ext4_acquire_dquot,
757 .release_dquot = ext4_release_dquot,
758 .mark_dirty = ext4_mark_dquot_dirty,
759 .write_info = ext4_write_info
762 static struct quotactl_ops ext4_qctl_operations = {
763 .quota_on = ext4_quota_on,
764 .quota_off = vfs_quota_off,
765 .quota_sync = vfs_quota_sync,
766 .get_info = vfs_get_dqinfo,
767 .set_info = vfs_set_dqinfo,
768 .get_dqblk = vfs_get_dqblk,
769 .set_dqblk = vfs_set_dqblk
773 static const struct super_operations ext4_sops = {
774 .alloc_inode = ext4_alloc_inode,
775 .destroy_inode = ext4_destroy_inode,
776 .read_inode = ext4_read_inode,
777 .write_inode = ext4_write_inode,
778 .dirty_inode = ext4_dirty_inode,
779 .delete_inode = ext4_delete_inode,
780 .put_super = ext4_put_super,
781 .write_super = ext4_write_super,
782 .sync_fs = ext4_sync_fs,
783 .write_super_lockfs = ext4_write_super_lockfs,
784 .unlockfs = ext4_unlockfs,
785 .statfs = ext4_statfs,
786 .remount_fs = ext4_remount,
787 .clear_inode = ext4_clear_inode,
788 .show_options = ext4_show_options,
790 .quota_read = ext4_quota_read,
791 .quota_write = ext4_quota_write,
795 static struct export_operations ext4_export_ops = {
796 .get_parent = ext4_get_parent,
797 .get_dentry = ext4_get_dentry,
801 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
802 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
803 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
804 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
805 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
806 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
807 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
808 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
809 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
810 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
811 Opt_grpquota, Opt_extents, Opt_noextents,
814 static match_table_t tokens = {
815 {Opt_bsd_df, "bsddf"},
816 {Opt_minix_df, "minixdf"},
817 {Opt_grpid, "grpid"},
818 {Opt_grpid, "bsdgroups"},
819 {Opt_nogrpid, "nogrpid"},
820 {Opt_nogrpid, "sysvgroups"},
821 {Opt_resgid, "resgid=%u"},
822 {Opt_resuid, "resuid=%u"},
824 {Opt_err_cont, "errors=continue"},
825 {Opt_err_panic, "errors=panic"},
826 {Opt_err_ro, "errors=remount-ro"},
827 {Opt_nouid32, "nouid32"},
828 {Opt_nocheck, "nocheck"},
829 {Opt_nocheck, "check=none"},
830 {Opt_debug, "debug"},
831 {Opt_oldalloc, "oldalloc"},
832 {Opt_orlov, "orlov"},
833 {Opt_user_xattr, "user_xattr"},
834 {Opt_nouser_xattr, "nouser_xattr"},
836 {Opt_noacl, "noacl"},
837 {Opt_reservation, "reservation"},
838 {Opt_noreservation, "noreservation"},
839 {Opt_noload, "noload"},
842 {Opt_commit, "commit=%u"},
843 {Opt_journal_update, "journal=update"},
844 {Opt_journal_inum, "journal=%u"},
845 {Opt_journal_dev, "journal_dev=%u"},
846 {Opt_abort, "abort"},
847 {Opt_data_journal, "data=journal"},
848 {Opt_data_ordered, "data=ordered"},
849 {Opt_data_writeback, "data=writeback"},
850 {Opt_offusrjquota, "usrjquota="},
851 {Opt_usrjquota, "usrjquota=%s"},
852 {Opt_offgrpjquota, "grpjquota="},
853 {Opt_grpjquota, "grpjquota=%s"},
854 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
855 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
856 {Opt_grpquota, "grpquota"},
857 {Opt_noquota, "noquota"},
858 {Opt_quota, "quota"},
859 {Opt_usrquota, "usrquota"},
860 {Opt_barrier, "barrier=%u"},
861 {Opt_extents, "extents"},
862 {Opt_noextents, "noextents"},
864 {Opt_resize, "resize"},
867 static ext4_fsblk_t get_sb_block(void **data)
869 ext4_fsblk_t sb_block;
870 char *options = (char *) *data;
872 if (!options || strncmp(options, "sb=", 3) != 0)
873 return 1; /* Default location */
875 /*todo: use simple_strtoll with >32bit ext4 */
876 sb_block = simple_strtoul(options, &options, 0);
877 if (*options && *options != ',') {
878 printk("EXT4-fs: Invalid sb specification: %s\n",
884 *data = (void *) options;
888 static int parse_options (char *options, struct super_block *sb,
889 unsigned int *inum, unsigned long *journal_devnum,
890 ext4_fsblk_t *n_blocks_count, int is_remount)
892 struct ext4_sb_info *sbi = EXT4_SB(sb);
894 substring_t args[MAX_OPT_ARGS];
905 while ((p = strsep (&options, ",")) != NULL) {
910 token = match_token(p, tokens, args);
913 clear_opt (sbi->s_mount_opt, MINIX_DF);
916 set_opt (sbi->s_mount_opt, MINIX_DF);
919 set_opt (sbi->s_mount_opt, GRPID);
922 clear_opt (sbi->s_mount_opt, GRPID);
925 if (match_int(&args[0], &option))
927 sbi->s_resuid = option;
930 if (match_int(&args[0], &option))
932 sbi->s_resgid = option;
935 /* handled by get_sb_block() instead of here */
936 /* *sb_block = match_int(&args[0]); */
939 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
940 clear_opt (sbi->s_mount_opt, ERRORS_RO);
941 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
944 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
945 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
946 set_opt (sbi->s_mount_opt, ERRORS_RO);
949 clear_opt (sbi->s_mount_opt, ERRORS_RO);
950 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
951 set_opt (sbi->s_mount_opt, ERRORS_CONT);
954 set_opt (sbi->s_mount_opt, NO_UID32);
957 clear_opt (sbi->s_mount_opt, CHECK);
960 set_opt (sbi->s_mount_opt, DEBUG);
963 set_opt (sbi->s_mount_opt, OLDALLOC);
966 clear_opt (sbi->s_mount_opt, OLDALLOC);
968 #ifdef CONFIG_EXT4DEV_FS_XATTR
970 set_opt (sbi->s_mount_opt, XATTR_USER);
972 case Opt_nouser_xattr:
973 clear_opt (sbi->s_mount_opt, XATTR_USER);
977 case Opt_nouser_xattr:
978 printk("EXT4 (no)user_xattr options not supported\n");
981 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
983 set_opt(sbi->s_mount_opt, POSIX_ACL);
986 clear_opt(sbi->s_mount_opt, POSIX_ACL);
991 printk("EXT4 (no)acl options not supported\n");
994 case Opt_reservation:
995 set_opt(sbi->s_mount_opt, RESERVATION);
997 case Opt_noreservation:
998 clear_opt(sbi->s_mount_opt, RESERVATION);
1000 case Opt_journal_update:
1002 /* Eventually we will want to be able to create
1003 a journal file here. For now, only allow the
1004 user to specify an existing inode to be the
1007 printk(KERN_ERR "EXT4-fs: cannot specify "
1008 "journal on remount\n");
1011 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1013 case Opt_journal_inum:
1015 printk(KERN_ERR "EXT4-fs: cannot specify "
1016 "journal on remount\n");
1019 if (match_int(&args[0], &option))
1023 case Opt_journal_dev:
1025 printk(KERN_ERR "EXT4-fs: cannot specify "
1026 "journal on remount\n");
1029 if (match_int(&args[0], &option))
1031 *journal_devnum = option;
1034 set_opt (sbi->s_mount_opt, NOLOAD);
1037 if (match_int(&args[0], &option))
1042 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1043 sbi->s_commit_interval = HZ * option;
1045 case Opt_data_journal:
1046 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1048 case Opt_data_ordered:
1049 data_opt = EXT4_MOUNT_ORDERED_DATA;
1051 case Opt_data_writeback:
1052 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1055 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1058 "EXT4-fs: cannot change data "
1059 "mode on remount\n");
1063 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1064 sbi->s_mount_opt |= data_opt;
1074 if (sb_any_quota_enabled(sb)) {
1076 "EXT4-fs: Cannot change journalled "
1077 "quota options when quota turned on.\n");
1080 qname = match_strdup(&args[0]);
1083 "EXT4-fs: not enough memory for "
1084 "storing quotafile name.\n");
1087 if (sbi->s_qf_names[qtype] &&
1088 strcmp(sbi->s_qf_names[qtype], qname)) {
1090 "EXT4-fs: %s quota file already "
1091 "specified.\n", QTYPE2NAME(qtype));
1095 sbi->s_qf_names[qtype] = qname;
1096 if (strchr(sbi->s_qf_names[qtype], '/')) {
1098 "EXT4-fs: quotafile must be on "
1099 "filesystem root.\n");
1100 kfree(sbi->s_qf_names[qtype]);
1101 sbi->s_qf_names[qtype] = NULL;
1104 set_opt(sbi->s_mount_opt, QUOTA);
1106 case Opt_offusrjquota:
1109 case Opt_offgrpjquota:
1112 if (sb_any_quota_enabled(sb)) {
1113 printk(KERN_ERR "EXT4-fs: Cannot change "
1114 "journalled quota options when "
1115 "quota turned on.\n");
1119 * The space will be released later when all options
1120 * are confirmed to be correct
1122 sbi->s_qf_names[qtype] = NULL;
1124 case Opt_jqfmt_vfsold:
1125 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1127 case Opt_jqfmt_vfsv0:
1128 sbi->s_jquota_fmt = QFMT_VFS_V0;
1132 set_opt(sbi->s_mount_opt, QUOTA);
1133 set_opt(sbi->s_mount_opt, USRQUOTA);
1136 set_opt(sbi->s_mount_opt, QUOTA);
1137 set_opt(sbi->s_mount_opt, GRPQUOTA);
1140 if (sb_any_quota_enabled(sb)) {
1141 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1142 "options when quota turned on.\n");
1145 clear_opt(sbi->s_mount_opt, QUOTA);
1146 clear_opt(sbi->s_mount_opt, USRQUOTA);
1147 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1155 case Opt_offusrjquota:
1156 case Opt_offgrpjquota:
1157 case Opt_jqfmt_vfsold:
1158 case Opt_jqfmt_vfsv0:
1160 "EXT4-fs: journalled quota options not "
1167 set_opt(sbi->s_mount_opt, ABORT);
1170 if (match_int(&args[0], &option))
1173 set_opt(sbi->s_mount_opt, BARRIER);
1175 clear_opt(sbi->s_mount_opt, BARRIER);
1181 printk("EXT4-fs: resize option only available "
1185 if (match_int(&args[0], &option) != 0)
1187 *n_blocks_count = option;
1190 set_opt(sbi->s_mount_opt, NOBH);
1193 clear_opt(sbi->s_mount_opt, NOBH);
1196 set_opt (sbi->s_mount_opt, EXTENTS);
1199 clear_opt (sbi->s_mount_opt, EXTENTS);
1203 "EXT4-fs: Unrecognized mount option \"%s\" "
1204 "or missing value\n", p);
1209 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1210 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1211 sbi->s_qf_names[USRQUOTA])
1212 clear_opt(sbi->s_mount_opt, USRQUOTA);
1214 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1215 sbi->s_qf_names[GRPQUOTA])
1216 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1218 if ((sbi->s_qf_names[USRQUOTA] &&
1219 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1220 (sbi->s_qf_names[GRPQUOTA] &&
1221 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1222 printk(KERN_ERR "EXT4-fs: old and new quota "
1223 "format mixing.\n");
1227 if (!sbi->s_jquota_fmt) {
1228 printk(KERN_ERR "EXT4-fs: journalled quota format "
1229 "not specified.\n");
1233 if (sbi->s_jquota_fmt) {
1234 printk(KERN_ERR "EXT4-fs: journalled quota format "
1235 "specified with no journalling "
1244 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1247 struct ext4_sb_info *sbi = EXT4_SB(sb);
1250 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1251 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1252 "forcing read-only mode\n");
1257 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1258 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1259 "running e2fsck is recommended\n");
1260 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1261 printk (KERN_WARNING
1262 "EXT4-fs warning: mounting fs with errors, "
1263 "running e2fsck is recommended\n");
1264 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1265 le16_to_cpu(es->s_mnt_count) >=
1266 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1267 printk (KERN_WARNING
1268 "EXT4-fs warning: maximal mount count reached, "
1269 "running e2fsck is recommended\n");
1270 else if (le32_to_cpu(es->s_checkinterval) &&
1271 (le32_to_cpu(es->s_lastcheck) +
1272 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1273 printk (KERN_WARNING
1274 "EXT4-fs warning: checktime reached, "
1275 "running e2fsck is recommended\n");
1277 /* @@@ We _will_ want to clear the valid bit if we find
1278 * inconsistencies, to force a fsck at reboot. But for
1279 * a plain journaled filesystem we can keep it set as
1282 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1284 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1285 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1286 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1287 es->s_mtime = cpu_to_le32(get_seconds());
1288 ext4_update_dynamic_rev(sb);
1289 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1291 ext4_commit_super(sb, es, 1);
1292 if (test_opt(sb, DEBUG))
1293 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1294 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1296 sbi->s_groups_count,
1297 EXT4_BLOCKS_PER_GROUP(sb),
1298 EXT4_INODES_PER_GROUP(sb),
1301 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1302 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1303 char b[BDEVNAME_SIZE];
1305 printk("external journal on %s\n",
1306 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1308 printk("internal journal\n");
1313 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1314 struct ext4_group_desc *gdp)
1318 if (sbi->s_es->s_feature_ro_compat &
1319 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1320 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1321 __le32 le_group = cpu_to_le32(block_group);
1323 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1324 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1325 crc = crc16(crc, (__u8 *)gdp, offset);
1326 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1327 /* for checksum of struct ext4_group_desc do the rest...*/
1328 if ((sbi->s_es->s_feature_incompat &
1329 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1330 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1331 crc = crc16(crc, (__u8 *)gdp + offset,
1332 le16_to_cpu(sbi->s_es->s_desc_size) -
1336 return cpu_to_le16(crc);
1339 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1340 struct ext4_group_desc *gdp)
1342 if ((sbi->s_es->s_feature_ro_compat &
1343 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1344 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1350 /* Called at mount-time, super-block is locked */
1351 static int ext4_check_descriptors (struct super_block * sb)
1353 struct ext4_sb_info *sbi = EXT4_SB(sb);
1354 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1355 ext4_fsblk_t last_block;
1356 ext4_fsblk_t block_bitmap;
1357 ext4_fsblk_t inode_bitmap;
1358 ext4_fsblk_t inode_table;
1359 struct ext4_group_desc * gdp = NULL;
1363 ext4_debug ("Checking group descriptors");
1365 for (i = 0; i < sbi->s_groups_count; i++)
1367 if (i == sbi->s_groups_count - 1)
1368 last_block = ext4_blocks_count(sbi->s_es) - 1;
1370 last_block = first_block +
1371 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1373 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1374 gdp = (struct ext4_group_desc *)
1375 sbi->s_group_desc[desc_block++]->b_data;
1376 block_bitmap = ext4_block_bitmap(sb, gdp);
1377 if (block_bitmap < first_block || block_bitmap > last_block)
1379 ext4_error (sb, "ext4_check_descriptors",
1380 "Block bitmap for group %d"
1381 " not in group (block %llu)!",
1385 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1386 if (inode_bitmap < first_block || inode_bitmap > last_block)
1388 ext4_error (sb, "ext4_check_descriptors",
1389 "Inode bitmap for group %d"
1390 " not in group (block %llu)!",
1394 inode_table = ext4_inode_table(sb, gdp);
1395 if (inode_table < first_block ||
1396 inode_table + sbi->s_itb_per_group - 1 > last_block)
1398 ext4_error (sb, "ext4_check_descriptors",
1399 "Inode table for group %d"
1400 " not in group (block %llu)!",
1404 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1405 ext4_error(sb, __FUNCTION__,
1406 "Checksum for group %d failed (%u!=%u)\n", i,
1407 le16_to_cpu(ext4_group_desc_csum(sbi, i,
1409 le16_to_cpu(gdp->bg_checksum));
1412 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1413 gdp = (struct ext4_group_desc *)
1414 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1417 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1418 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1423 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1424 * the superblock) which were deleted from all directories, but held open by
1425 * a process at the time of a crash. We walk the list and try to delete these
1426 * inodes at recovery time (only with a read-write filesystem).
1428 * In order to keep the orphan inode chain consistent during traversal (in
1429 * case of crash during recovery), we link each inode into the superblock
1430 * orphan list_head and handle it the same way as an inode deletion during
1431 * normal operation (which journals the operations for us).
1433 * We only do an iget() and an iput() on each inode, which is very safe if we
1434 * accidentally point at an in-use or already deleted inode. The worst that
1435 * can happen in this case is that we get a "bit already cleared" message from
1436 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1437 * e2fsck was run on this filesystem, and it must have already done the orphan
1438 * inode cleanup for us, so we can safely abort without any further action.
1440 static void ext4_orphan_cleanup (struct super_block * sb,
1441 struct ext4_super_block * es)
1443 unsigned int s_flags = sb->s_flags;
1444 int nr_orphans = 0, nr_truncates = 0;
1448 if (!es->s_last_orphan) {
1449 jbd_debug(4, "no orphan inodes to clean up\n");
1453 if (bdev_read_only(sb->s_bdev)) {
1454 printk(KERN_ERR "EXT4-fs: write access "
1455 "unavailable, skipping orphan cleanup.\n");
1459 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1460 if (es->s_last_orphan)
1461 jbd_debug(1, "Errors on filesystem, "
1462 "clearing orphan list.\n");
1463 es->s_last_orphan = 0;
1464 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1468 if (s_flags & MS_RDONLY) {
1469 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1471 sb->s_flags &= ~MS_RDONLY;
1474 /* Needed for iput() to work correctly and not trash data */
1475 sb->s_flags |= MS_ACTIVE;
1476 /* Turn on quotas so that they are updated correctly */
1477 for (i = 0; i < MAXQUOTAS; i++) {
1478 if (EXT4_SB(sb)->s_qf_names[i]) {
1479 int ret = ext4_quota_on_mount(sb, i);
1482 "EXT4-fs: Cannot turn on journalled "
1483 "quota: error %d\n", ret);
1488 while (es->s_last_orphan) {
1489 struct inode *inode;
1492 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1493 es->s_last_orphan = 0;
1497 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1499 if (inode->i_nlink) {
1501 "%s: truncating inode %lu to %Ld bytes\n",
1502 __FUNCTION__, inode->i_ino, inode->i_size);
1503 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1504 inode->i_ino, inode->i_size);
1505 ext4_truncate(inode);
1509 "%s: deleting unreferenced inode %lu\n",
1510 __FUNCTION__, inode->i_ino);
1511 jbd_debug(2, "deleting unreferenced inode %lu\n",
1515 iput(inode); /* The delete magic happens here! */
1518 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1521 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1522 sb->s_id, PLURAL(nr_orphans));
1524 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1525 sb->s_id, PLURAL(nr_truncates));
1527 /* Turn quotas off */
1528 for (i = 0; i < MAXQUOTAS; i++) {
1529 if (sb_dqopt(sb)->files[i])
1530 vfs_quota_off(sb, i);
1533 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1537 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1538 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1539 * We need to be 1 filesystem block less than the 2^32 sector limit.
1541 static loff_t ext4_max_size(int bits)
1543 loff_t res = EXT4_NDIR_BLOCKS;
1544 /* This constant is calculated to be the largest file size for a
1545 * dense, 4k-blocksize file such that the total number of
1546 * sectors in the file, including data and all indirect blocks,
1547 * does not exceed 2^32. */
1548 const loff_t upper_limit = 0x1ff7fffd000LL;
1550 res += 1LL << (bits-2);
1551 res += 1LL << (2*(bits-2));
1552 res += 1LL << (3*(bits-2));
1554 if (res > upper_limit)
1559 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1560 ext4_fsblk_t logical_sb_block, int nr)
1562 struct ext4_sb_info *sbi = EXT4_SB(sb);
1563 unsigned long bg, first_meta_bg;
1566 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1568 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1570 return logical_sb_block + nr + 1;
1571 bg = sbi->s_desc_per_block * nr;
1572 if (ext4_bg_has_super(sb, bg))
1574 return (has_super + ext4_group_first_block_no(sb, bg));
1578 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1580 struct buffer_head * bh;
1581 struct ext4_super_block *es = NULL;
1582 struct ext4_sb_info *sbi;
1584 ext4_fsblk_t sb_block = get_sb_block(&data);
1585 ext4_fsblk_t logical_sb_block;
1586 unsigned long offset = 0;
1587 unsigned int journal_inum = 0;
1588 unsigned long journal_devnum = 0;
1589 unsigned long def_mount_opts;
1600 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1603 sb->s_fs_info = sbi;
1604 sbi->s_mount_opt = 0;
1605 sbi->s_resuid = EXT4_DEF_RESUID;
1606 sbi->s_resgid = EXT4_DEF_RESGID;
1607 sbi->s_sb_block = sb_block;
1611 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1613 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1618 * The ext4 superblock will not be buffer aligned for other than 1kB
1619 * block sizes. We need to calculate the offset from buffer start.
1621 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1622 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1623 offset = do_div(logical_sb_block, blocksize);
1625 logical_sb_block = sb_block;
1628 if (!(bh = sb_bread(sb, logical_sb_block))) {
1629 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1633 * Note: s_es must be initialized as soon as possible because
1634 * some ext4 macro-instructions depend on its value
1636 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1638 sb->s_magic = le16_to_cpu(es->s_magic);
1639 if (sb->s_magic != EXT4_SUPER_MAGIC)
1642 /* Set defaults before we parse the mount options */
1643 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1644 if (def_mount_opts & EXT4_DEFM_DEBUG)
1645 set_opt(sbi->s_mount_opt, DEBUG);
1646 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1647 set_opt(sbi->s_mount_opt, GRPID);
1648 if (def_mount_opts & EXT4_DEFM_UID16)
1649 set_opt(sbi->s_mount_opt, NO_UID32);
1650 #ifdef CONFIG_EXT4DEV_FS_XATTR
1651 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1652 set_opt(sbi->s_mount_opt, XATTR_USER);
1654 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1655 if (def_mount_opts & EXT4_DEFM_ACL)
1656 set_opt(sbi->s_mount_opt, POSIX_ACL);
1658 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1659 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1660 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1661 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1662 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1663 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1665 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1666 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1667 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1668 set_opt(sbi->s_mount_opt, ERRORS_RO);
1670 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1672 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1673 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1675 set_opt(sbi->s_mount_opt, RESERVATION);
1678 * turn on extents feature by default in ext4 filesystem
1679 * User -o noextents to turn it off
1681 set_opt(sbi->s_mount_opt, EXTENTS);
1683 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1687 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1688 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1690 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1691 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1692 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1693 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1695 "EXT4-fs warning: feature flags set on rev 0 fs, "
1696 "running e2fsck is recommended\n");
1698 * Check feature flags regardless of the revision level, since we
1699 * previously didn't change the revision level when setting the flags,
1700 * so there is a chance incompat flags are set on a rev 0 filesystem.
1702 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1704 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1705 "unsupported optional features (%x).\n",
1706 sb->s_id, le32_to_cpu(features));
1709 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1710 if (!(sb->s_flags & MS_RDONLY) && features) {
1711 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1712 "unsupported optional features (%x).\n",
1713 sb->s_id, le32_to_cpu(features));
1716 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1718 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1719 blocksize > EXT4_MAX_BLOCK_SIZE) {
1721 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1722 blocksize, sb->s_id);
1726 hblock = bdev_hardsect_size(sb->s_bdev);
1727 if (sb->s_blocksize != blocksize) {
1729 * Make sure the blocksize for the filesystem is larger
1730 * than the hardware sectorsize for the machine.
1732 if (blocksize < hblock) {
1733 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1734 "device blocksize %d.\n", blocksize, hblock);
1739 sb_set_blocksize(sb, blocksize);
1740 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1741 offset = do_div(logical_sb_block, blocksize);
1742 bh = sb_bread(sb, logical_sb_block);
1745 "EXT4-fs: Can't read superblock on 2nd try.\n");
1748 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1750 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1752 "EXT4-fs: Magic mismatch, very weird !\n");
1757 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1759 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1760 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1761 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1763 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1764 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1765 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1766 (!is_power_of_2(sbi->s_inode_size)) ||
1767 (sbi->s_inode_size > blocksize)) {
1769 "EXT4-fs: unsupported inode size: %d\n",
1773 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1774 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1776 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1777 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1778 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1779 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1780 !is_power_of_2(sbi->s_desc_size)) {
1782 "EXT4-fs: unsupported descriptor size %lu\n",
1787 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1788 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1789 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1790 if (EXT4_INODE_SIZE(sb) == 0)
1792 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1793 if (sbi->s_inodes_per_block == 0)
1795 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1796 sbi->s_inodes_per_block;
1797 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1799 sbi->s_mount_state = le16_to_cpu(es->s_state);
1800 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1801 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1802 for (i=0; i < 4; i++)
1803 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1804 sbi->s_def_hash_version = es->s_def_hash_version;
1806 if (sbi->s_blocks_per_group > blocksize * 8) {
1808 "EXT4-fs: #blocks per group too big: %lu\n",
1809 sbi->s_blocks_per_group);
1812 if (sbi->s_inodes_per_group > blocksize * 8) {
1814 "EXT4-fs: #inodes per group too big: %lu\n",
1815 sbi->s_inodes_per_group);
1819 if (ext4_blocks_count(es) >
1820 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1821 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1822 " too large to mount safely\n", sb->s_id);
1823 if (sizeof(sector_t) < 8)
1824 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1829 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1831 blocks_count = (ext4_blocks_count(es) -
1832 le32_to_cpu(es->s_first_data_block) +
1833 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1834 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1835 sbi->s_groups_count = blocks_count;
1836 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1837 EXT4_DESC_PER_BLOCK(sb);
1838 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1840 if (sbi->s_group_desc == NULL) {
1841 printk (KERN_ERR "EXT4-fs: not enough memory\n");
1845 bgl_lock_init(&sbi->s_blockgroup_lock);
1847 for (i = 0; i < db_count; i++) {
1848 block = descriptor_loc(sb, logical_sb_block, i);
1849 sbi->s_group_desc[i] = sb_bread(sb, block);
1850 if (!sbi->s_group_desc[i]) {
1851 printk (KERN_ERR "EXT4-fs: "
1852 "can't read group descriptor %d\n", i);
1857 if (!ext4_check_descriptors (sb)) {
1858 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
1861 sbi->s_gdb_count = db_count;
1862 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1863 spin_lock_init(&sbi->s_next_gen_lock);
1865 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1866 ext4_count_free_blocks(sb));
1868 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1869 ext4_count_free_inodes(sb));
1872 err = percpu_counter_init(&sbi->s_dirs_counter,
1873 ext4_count_dirs(sb));
1876 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
1880 /* per fileystem reservation list head & lock */
1881 spin_lock_init(&sbi->s_rsv_window_lock);
1882 sbi->s_rsv_window_root = RB_ROOT;
1883 /* Add a single, static dummy reservation to the start of the
1884 * reservation window list --- it gives us a placeholder for
1885 * append-at-start-of-list which makes the allocation logic
1886 * _much_ simpler. */
1887 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1888 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1889 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1890 sbi->s_rsv_window_head.rsv_goal_size = 0;
1891 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
1894 * set up enough so that it can read an inode
1896 sb->s_op = &ext4_sops;
1897 sb->s_export_op = &ext4_export_ops;
1898 sb->s_xattr = ext4_xattr_handlers;
1900 sb->s_qcop = &ext4_qctl_operations;
1901 sb->dq_op = &ext4_quota_operations;
1903 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1907 needs_recovery = (es->s_last_orphan != 0 ||
1908 EXT4_HAS_INCOMPAT_FEATURE(sb,
1909 EXT4_FEATURE_INCOMPAT_RECOVER));
1912 * The first inode we look at is the journal inode. Don't try
1913 * root first: it may be modified in the journal!
1915 if (!test_opt(sb, NOLOAD) &&
1916 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1917 if (ext4_load_journal(sb, es, journal_devnum))
1919 } else if (journal_inum) {
1920 if (ext4_create_journal(sb, es, journal_inum))
1925 "ext4: No journal on filesystem on %s\n",
1930 if (ext4_blocks_count(es) > 0xffffffffULL &&
1931 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1932 JBD2_FEATURE_INCOMPAT_64BIT)) {
1933 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1937 /* We have now updated the journal if required, so we can
1938 * validate the data journaling mode. */
1939 switch (test_opt(sb, DATA_FLAGS)) {
1941 /* No mode set, assume a default based on the journal
1942 * capabilities: ORDERED_DATA if the journal can
1943 * cope, else JOURNAL_DATA
1945 if (jbd2_journal_check_available_features
1946 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
1947 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1949 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1952 case EXT4_MOUNT_ORDERED_DATA:
1953 case EXT4_MOUNT_WRITEBACK_DATA:
1954 if (!jbd2_journal_check_available_features
1955 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
1956 printk(KERN_ERR "EXT4-fs: Journal does not support "
1957 "requested data journaling mode\n");
1964 if (test_opt(sb, NOBH)) {
1965 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1966 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
1967 "its supported only with writeback mode\n");
1968 clear_opt(sbi->s_mount_opt, NOBH);
1972 * The jbd2_journal_load will have done any necessary log recovery,
1973 * so we can safely mount the rest of the filesystem now.
1976 root = iget(sb, EXT4_ROOT_INO);
1977 sb->s_root = d_alloc_root(root);
1979 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1983 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1986 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
1990 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1992 /* determine the minimum size of new large inodes, if present */
1993 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1994 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1995 EXT4_GOOD_OLD_INODE_SIZE;
1996 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1997 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
1998 if (sbi->s_want_extra_isize <
1999 le16_to_cpu(es->s_want_extra_isize))
2000 sbi->s_want_extra_isize =
2001 le16_to_cpu(es->s_want_extra_isize);
2002 if (sbi->s_want_extra_isize <
2003 le16_to_cpu(es->s_min_extra_isize))
2004 sbi->s_want_extra_isize =
2005 le16_to_cpu(es->s_min_extra_isize);
2008 /* Check if enough inode space is available */
2009 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2010 sbi->s_inode_size) {
2011 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2012 EXT4_GOOD_OLD_INODE_SIZE;
2013 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2018 * akpm: core read_super() calls in here with the superblock locked.
2019 * That deadlocks, because orphan cleanup needs to lock the superblock
2020 * in numerous places. Here we just pop the lock - it's relatively
2021 * harmless, because we are now ready to accept write_super() requests,
2022 * and aviro says that's the only reason for hanging onto the
2025 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2026 ext4_orphan_cleanup(sb, es);
2027 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2029 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2030 ext4_mark_recovery_complete(sb, es);
2031 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2032 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2033 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2043 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2048 jbd2_journal_destroy(sbi->s_journal);
2050 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2051 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2052 percpu_counter_destroy(&sbi->s_dirs_counter);
2054 for (i = 0; i < db_count; i++)
2055 brelse(sbi->s_group_desc[i]);
2056 kfree(sbi->s_group_desc);
2059 for (i = 0; i < MAXQUOTAS; i++)
2060 kfree(sbi->s_qf_names[i]);
2062 ext4_blkdev_remove(sbi);
2065 sb->s_fs_info = NULL;
2072 * Setup any per-fs journal parameters now. We'll do this both on
2073 * initial mount, once the journal has been initialised but before we've
2074 * done any recovery; and again on any subsequent remount.
2076 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2078 struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 if (sbi->s_commit_interval)
2081 journal->j_commit_interval = sbi->s_commit_interval;
2082 /* We could also set up an ext4-specific default for the commit
2083 * interval here, but for now we'll just fall back to the jbd
2086 spin_lock(&journal->j_state_lock);
2087 if (test_opt(sb, BARRIER))
2088 journal->j_flags |= JBD2_BARRIER;
2090 journal->j_flags &= ~JBD2_BARRIER;
2091 spin_unlock(&journal->j_state_lock);
2094 static journal_t *ext4_get_journal(struct super_block *sb,
2095 unsigned int journal_inum)
2097 struct inode *journal_inode;
2100 /* First, test for the existence of a valid inode on disk. Bad
2101 * things happen if we iget() an unused inode, as the subsequent
2102 * iput() will try to delete it. */
2104 journal_inode = iget(sb, journal_inum);
2105 if (!journal_inode) {
2106 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2109 if (!journal_inode->i_nlink) {
2110 make_bad_inode(journal_inode);
2111 iput(journal_inode);
2112 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2116 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2117 journal_inode, journal_inode->i_size);
2118 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2119 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2120 iput(journal_inode);
2124 journal = jbd2_journal_init_inode(journal_inode);
2126 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2127 iput(journal_inode);
2130 journal->j_private = sb;
2131 ext4_init_journal_params(sb, journal);
2135 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2138 struct buffer_head * bh;
2142 int hblock, blocksize;
2143 ext4_fsblk_t sb_block;
2144 unsigned long offset;
2145 struct ext4_super_block * es;
2146 struct block_device *bdev;
2148 bdev = ext4_blkdev_get(j_dev);
2152 if (bd_claim(bdev, sb)) {
2154 "EXT4: failed to claim external journal device.\n");
2159 blocksize = sb->s_blocksize;
2160 hblock = bdev_hardsect_size(bdev);
2161 if (blocksize < hblock) {
2163 "EXT4-fs: blocksize too small for journal device.\n");
2167 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2168 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2169 set_blocksize(bdev, blocksize);
2170 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2171 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2172 "external journal\n");
2176 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2177 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2178 !(le32_to_cpu(es->s_feature_incompat) &
2179 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2180 printk(KERN_ERR "EXT4-fs: external journal has "
2181 "bad superblock\n");
2186 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2187 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2192 len = ext4_blocks_count(es);
2193 start = sb_block + 1;
2194 brelse(bh); /* we're done with the superblock */
2196 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2197 start, len, blocksize);
2199 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2202 journal->j_private = sb;
2203 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2204 wait_on_buffer(journal->j_sb_buffer);
2205 if (!buffer_uptodate(journal->j_sb_buffer)) {
2206 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2209 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2210 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2211 "user (unsupported) - %d\n",
2212 be32_to_cpu(journal->j_superblock->s_nr_users));
2215 EXT4_SB(sb)->journal_bdev = bdev;
2216 ext4_init_journal_params(sb, journal);
2219 jbd2_journal_destroy(journal);
2221 ext4_blkdev_put(bdev);
2225 static int ext4_load_journal(struct super_block *sb,
2226 struct ext4_super_block *es,
2227 unsigned long journal_devnum)
2230 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2233 int really_read_only;
2235 if (journal_devnum &&
2236 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2237 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2238 "numbers have changed\n");
2239 journal_dev = new_decode_dev(journal_devnum);
2241 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2243 really_read_only = bdev_read_only(sb->s_bdev);
2246 * Are we loading a blank journal or performing recovery after a
2247 * crash? For recovery, we need to check in advance whether we
2248 * can get read-write access to the device.
2251 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2252 if (sb->s_flags & MS_RDONLY) {
2253 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2254 "required on readonly filesystem.\n");
2255 if (really_read_only) {
2256 printk(KERN_ERR "EXT4-fs: write access "
2257 "unavailable, cannot proceed.\n");
2260 printk (KERN_INFO "EXT4-fs: write access will "
2261 "be enabled during recovery.\n");
2265 if (journal_inum && journal_dev) {
2266 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2267 "and inode journals!\n");
2272 if (!(journal = ext4_get_journal(sb, journal_inum)))
2275 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2279 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2280 err = jbd2_journal_update_format(journal);
2282 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2283 jbd2_journal_destroy(journal);
2288 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2289 err = jbd2_journal_wipe(journal, !really_read_only);
2291 err = jbd2_journal_load(journal);
2294 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2295 jbd2_journal_destroy(journal);
2299 EXT4_SB(sb)->s_journal = journal;
2300 ext4_clear_journal_err(sb, es);
2302 if (journal_devnum &&
2303 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2304 es->s_journal_dev = cpu_to_le32(journal_devnum);
2307 /* Make sure we flush the recovery flag to disk. */
2308 ext4_commit_super(sb, es, 1);
2314 static int ext4_create_journal(struct super_block * sb,
2315 struct ext4_super_block * es,
2316 unsigned int journal_inum)
2321 if (sb->s_flags & MS_RDONLY) {
2322 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2323 "create journal.\n");
2327 journal = ext4_get_journal(sb, journal_inum);
2331 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2334 err = jbd2_journal_create(journal);
2336 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2337 jbd2_journal_destroy(journal);
2341 EXT4_SB(sb)->s_journal = journal;
2343 ext4_update_dynamic_rev(sb);
2344 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2345 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2347 es->s_journal_inum = cpu_to_le32(journal_inum);
2350 /* Make sure we flush the recovery flag to disk. */
2351 ext4_commit_super(sb, es, 1);
2356 static void ext4_commit_super (struct super_block * sb,
2357 struct ext4_super_block * es,
2360 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2364 es->s_wtime = cpu_to_le32(get_seconds());
2365 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2366 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2367 BUFFER_TRACE(sbh, "marking dirty");
2368 mark_buffer_dirty(sbh);
2370 sync_dirty_buffer(sbh);
2375 * Have we just finished recovery? If so, and if we are mounting (or
2376 * remounting) the filesystem readonly, then we will end up with a
2377 * consistent fs on disk. Record that fact.
2379 static void ext4_mark_recovery_complete(struct super_block * sb,
2380 struct ext4_super_block * es)
2382 journal_t *journal = EXT4_SB(sb)->s_journal;
2384 jbd2_journal_lock_updates(journal);
2385 jbd2_journal_flush(journal);
2387 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2388 sb->s_flags & MS_RDONLY) {
2389 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2391 ext4_commit_super(sb, es, 1);
2394 jbd2_journal_unlock_updates(journal);
2398 * If we are mounting (or read-write remounting) a filesystem whose journal
2399 * has recorded an error from a previous lifetime, move that error to the
2400 * main filesystem now.
2402 static void ext4_clear_journal_err(struct super_block * sb,
2403 struct ext4_super_block * es)
2409 journal = EXT4_SB(sb)->s_journal;
2412 * Now check for any error status which may have been recorded in the
2413 * journal by a prior ext4_error() or ext4_abort()
2416 j_errno = jbd2_journal_errno(journal);
2420 errstr = ext4_decode_error(sb, j_errno, nbuf);
2421 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2422 "from previous mount: %s", errstr);
2423 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2424 "filesystem check.");
2426 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2427 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2428 ext4_commit_super (sb, es, 1);
2430 jbd2_journal_clear_err(journal);
2435 * Force the running and committing transactions to commit,
2436 * and wait on the commit.
2438 int ext4_force_commit(struct super_block *sb)
2443 if (sb->s_flags & MS_RDONLY)
2446 journal = EXT4_SB(sb)->s_journal;
2448 ret = ext4_journal_force_commit(journal);
2453 * Ext4 always journals updates to the superblock itself, so we don't
2454 * have to propagate any other updates to the superblock on disk at this
2455 * point. Just start an async writeback to get the buffers on their way
2458 * This implicitly triggers the writebehind on sync().
2461 static void ext4_write_super (struct super_block * sb)
2463 if (mutex_trylock(&sb->s_lock) != 0)
2468 static int ext4_sync_fs(struct super_block *sb, int wait)
2473 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2475 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2481 * LVM calls this function before a (read-only) snapshot is created. This
2482 * gives us a chance to flush the journal completely and mark the fs clean.
2484 static void ext4_write_super_lockfs(struct super_block *sb)
2488 if (!(sb->s_flags & MS_RDONLY)) {
2489 journal_t *journal = EXT4_SB(sb)->s_journal;
2491 /* Now we set up the journal barrier. */
2492 jbd2_journal_lock_updates(journal);
2493 jbd2_journal_flush(journal);
2495 /* Journal blocked and flushed, clear needs_recovery flag. */
2496 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2497 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2502 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2503 * flag here, even though the filesystem is not technically dirty yet.
2505 static void ext4_unlockfs(struct super_block *sb)
2507 if (!(sb->s_flags & MS_RDONLY)) {
2509 /* Reser the needs_recovery flag before the fs is unlocked. */
2510 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2511 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2513 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2517 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2519 struct ext4_super_block * es;
2520 struct ext4_sb_info *sbi = EXT4_SB(sb);
2521 ext4_fsblk_t n_blocks_count = 0;
2522 unsigned long old_sb_flags;
2523 struct ext4_mount_options old_opts;
2529 /* Store the original options */
2530 old_sb_flags = sb->s_flags;
2531 old_opts.s_mount_opt = sbi->s_mount_opt;
2532 old_opts.s_resuid = sbi->s_resuid;
2533 old_opts.s_resgid = sbi->s_resgid;
2534 old_opts.s_commit_interval = sbi->s_commit_interval;
2536 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2537 for (i = 0; i < MAXQUOTAS; i++)
2538 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2542 * Allow the "check" option to be passed as a remount option.
2544 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2549 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2550 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2552 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2553 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2557 ext4_init_journal_params(sb, sbi->s_journal);
2559 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2560 n_blocks_count > ext4_blocks_count(es)) {
2561 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2566 if (*flags & MS_RDONLY) {
2568 * First of all, the unconditional stuff we have to do
2569 * to disable replay of the journal when we next remount
2571 sb->s_flags |= MS_RDONLY;
2574 * OK, test if we are remounting a valid rw partition
2575 * readonly, and if so set the rdonly flag and then
2576 * mark the partition as valid again.
2578 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2579 (sbi->s_mount_state & EXT4_VALID_FS))
2580 es->s_state = cpu_to_le16(sbi->s_mount_state);
2583 * We have to unlock super so that we can wait for
2587 ext4_mark_recovery_complete(sb, es);
2591 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2592 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2593 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2594 "remount RDWR because of unsupported "
2595 "optional features (%x).\n",
2596 sb->s_id, le32_to_cpu(ret));
2602 * If we have an unprocessed orphan list hanging
2603 * around from a previously readonly bdev mount,
2604 * require a full umount/remount for now.
2606 if (es->s_last_orphan) {
2607 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2608 "remount RDWR because of unprocessed "
2609 "orphan inode list. Please "
2610 "umount/remount instead.\n",
2617 * Mounting a RDONLY partition read-write, so reread
2618 * and store the current valid flag. (It may have
2619 * been changed by e2fsck since we originally mounted
2622 ext4_clear_journal_err(sb, es);
2623 sbi->s_mount_state = le16_to_cpu(es->s_state);
2624 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2626 if (!ext4_setup_super (sb, es, 0))
2627 sb->s_flags &= ~MS_RDONLY;
2631 /* Release old quota file names */
2632 for (i = 0; i < MAXQUOTAS; i++)
2633 if (old_opts.s_qf_names[i] &&
2634 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2635 kfree(old_opts.s_qf_names[i]);
2639 sb->s_flags = old_sb_flags;
2640 sbi->s_mount_opt = old_opts.s_mount_opt;
2641 sbi->s_resuid = old_opts.s_resuid;
2642 sbi->s_resgid = old_opts.s_resgid;
2643 sbi->s_commit_interval = old_opts.s_commit_interval;
2645 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2646 for (i = 0; i < MAXQUOTAS; i++) {
2647 if (sbi->s_qf_names[i] &&
2648 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2649 kfree(sbi->s_qf_names[i]);
2650 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2656 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2658 struct super_block *sb = dentry->d_sb;
2659 struct ext4_sb_info *sbi = EXT4_SB(sb);
2660 struct ext4_super_block *es = sbi->s_es;
2663 if (test_opt(sb, MINIX_DF)) {
2664 sbi->s_overhead_last = 0;
2665 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2666 unsigned long ngroups = sbi->s_groups_count, i;
2667 ext4_fsblk_t overhead = 0;
2671 * Compute the overhead (FS structures). This is constant
2672 * for a given filesystem unless the number of block groups
2673 * changes so we cache the previous value until it does.
2677 * All of the blocks before first_data_block are
2680 overhead = le32_to_cpu(es->s_first_data_block);
2683 * Add the overhead attributed to the superblock and
2684 * block group descriptors. If the sparse superblocks
2685 * feature is turned on, then not all groups have this.
2687 for (i = 0; i < ngroups; i++) {
2688 overhead += ext4_bg_has_super(sb, i) +
2689 ext4_bg_num_gdb(sb, i);
2694 * Every block group has an inode bitmap, a block
2695 * bitmap, and an inode table.
2697 overhead += ngroups * (2 + sbi->s_itb_per_group);
2698 sbi->s_overhead_last = overhead;
2700 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2703 buf->f_type = EXT4_SUPER_MAGIC;
2704 buf->f_bsize = sb->s_blocksize;
2705 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2706 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2707 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2708 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2709 if (buf->f_bfree < ext4_r_blocks_count(es))
2711 buf->f_files = le32_to_cpu(es->s_inodes_count);
2712 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2713 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2714 buf->f_namelen = EXT4_NAME_LEN;
2715 fsid = le64_to_cpup((void *)es->s_uuid) ^
2716 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2717 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2718 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2722 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2723 * is locked for write. Otherwise the are possible deadlocks:
2724 * Process 1 Process 2
2725 * ext4_create() quota_sync()
2726 * jbd2_journal_start() write_dquot()
2727 * DQUOT_INIT() down(dqio_mutex)
2728 * down(dqio_mutex) jbd2_journal_start()
2734 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2736 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2739 static int ext4_dquot_initialize(struct inode *inode, int type)
2744 /* We may create quota structure so we need to reserve enough blocks */
2745 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2747 return PTR_ERR(handle);
2748 ret = dquot_initialize(inode, type);
2749 err = ext4_journal_stop(handle);
2755 static int ext4_dquot_drop(struct inode *inode)
2760 /* We may delete quota structure so we need to reserve enough blocks */
2761 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2763 return PTR_ERR(handle);
2764 ret = dquot_drop(inode);
2765 err = ext4_journal_stop(handle);
2771 static int ext4_write_dquot(struct dquot *dquot)
2775 struct inode *inode;
2777 inode = dquot_to_inode(dquot);
2778 handle = ext4_journal_start(inode,
2779 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2781 return PTR_ERR(handle);
2782 ret = dquot_commit(dquot);
2783 err = ext4_journal_stop(handle);
2789 static int ext4_acquire_dquot(struct dquot *dquot)
2794 handle = ext4_journal_start(dquot_to_inode(dquot),
2795 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2797 return PTR_ERR(handle);
2798 ret = dquot_acquire(dquot);
2799 err = ext4_journal_stop(handle);
2805 static int ext4_release_dquot(struct dquot *dquot)
2810 handle = ext4_journal_start(dquot_to_inode(dquot),
2811 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2812 if (IS_ERR(handle)) {
2813 /* Release dquot anyway to avoid endless cycle in dqput() */
2814 dquot_release(dquot);
2815 return PTR_ERR(handle);
2817 ret = dquot_release(dquot);
2818 err = ext4_journal_stop(handle);
2824 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2826 /* Are we journalling quotas? */
2827 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2828 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2829 dquot_mark_dquot_dirty(dquot);
2830 return ext4_write_dquot(dquot);
2832 return dquot_mark_dquot_dirty(dquot);
2836 static int ext4_write_info(struct super_block *sb, int type)
2841 /* Data block + inode block */
2842 handle = ext4_journal_start(sb->s_root->d_inode, 2);
2844 return PTR_ERR(handle);
2845 ret = dquot_commit_info(sb, type);
2846 err = ext4_journal_stop(handle);
2853 * Turn on quotas during mount time - we need to find
2854 * the quota file and such...
2856 static int ext4_quota_on_mount(struct super_block *sb, int type)
2858 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2859 EXT4_SB(sb)->s_jquota_fmt, type);
2863 * Standard function to be called on quota_on
2865 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
2869 struct nameidata nd;
2871 if (!test_opt(sb, QUOTA))
2873 /* Not journalling quota? */
2874 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2875 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
2876 return vfs_quota_on(sb, type, format_id, path);
2877 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2880 /* Quotafile not on the same filesystem? */
2881 if (nd.mnt->mnt_sb != sb) {
2885 /* Quotafile not of fs root? */
2886 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2888 "EXT4-fs: Quota file not on filesystem root. "
2889 "Journalled quota will not work.\n");
2891 return vfs_quota_on(sb, type, format_id, path);
2894 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2895 * acquiring the locks... As quota files are never truncated and quota code
2896 * itself serializes the operations (and noone else should touch the files)
2897 * we don't have to be afraid of races */
2898 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
2899 size_t len, loff_t off)
2901 struct inode *inode = sb_dqopt(sb)->files[type];
2902 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2904 int offset = off & (sb->s_blocksize - 1);
2907 struct buffer_head *bh;
2908 loff_t i_size = i_size_read(inode);
2912 if (off+len > i_size)
2915 while (toread > 0) {
2916 tocopy = sb->s_blocksize - offset < toread ?
2917 sb->s_blocksize - offset : toread;
2918 bh = ext4_bread(NULL, inode, blk, 0, &err);
2921 if (!bh) /* A hole? */
2922 memset(data, 0, tocopy);
2924 memcpy(data, bh->b_data+offset, tocopy);
2934 /* Write to quotafile (we know the transaction is already started and has
2935 * enough credits) */
2936 static ssize_t ext4_quota_write(struct super_block *sb, int type,
2937 const char *data, size_t len, loff_t off)
2939 struct inode *inode = sb_dqopt(sb)->files[type];
2940 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
2942 int offset = off & (sb->s_blocksize - 1);
2944 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
2945 size_t towrite = len;
2946 struct buffer_head *bh;
2947 handle_t *handle = journal_current_handle();
2950 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2951 " cancelled because transaction is not started.\n",
2952 (unsigned long long)off, (unsigned long long)len);
2955 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2956 while (towrite > 0) {
2957 tocopy = sb->s_blocksize - offset < towrite ?
2958 sb->s_blocksize - offset : towrite;
2959 bh = ext4_bread(handle, inode, blk, 1, &err);
2962 if (journal_quota) {
2963 err = ext4_journal_get_write_access(handle, bh);
2970 memcpy(bh->b_data+offset, data, tocopy);
2971 flush_dcache_page(bh->b_page);
2974 err = ext4_journal_dirty_metadata(handle, bh);
2976 /* Always do at least ordered writes for quotas */
2977 err = ext4_journal_dirty_data(handle, bh);
2978 mark_buffer_dirty(bh);
2991 if (inode->i_size < off+len-towrite) {
2992 i_size_write(inode, off+len-towrite);
2993 EXT4_I(inode)->i_disksize = inode->i_size;
2996 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2997 ext4_mark_inode_dirty(handle, inode);
2998 mutex_unlock(&inode->i_mutex);
2999 return len - towrite;
3004 static int ext4_get_sb(struct file_system_type *fs_type,
3005 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3007 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3010 static struct file_system_type ext4dev_fs_type = {
3011 .owner = THIS_MODULE,
3013 .get_sb = ext4_get_sb,
3014 .kill_sb = kill_block_super,
3015 .fs_flags = FS_REQUIRES_DEV,
3018 static int __init init_ext4_fs(void)
3020 int err = init_ext4_xattr();
3023 err = init_inodecache();
3026 err = register_filesystem(&ext4dev_fs_type);
3031 destroy_inodecache();
3037 static void __exit exit_ext4_fs(void)
3039 unregister_filesystem(&ext4dev_fs_type);
3040 destroy_inodecache();
3044 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3045 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3046 MODULE_LICENSE("GPL");
3047 module_init(init_ext4_fs)
3048 module_exit(exit_ext4_fs)