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_lo) |
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_lo) |
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_lo) |
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_lo = 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_lo = 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_lo = 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.
376 int ext4_update_compat_feature(handle_t *handle,
377 struct super_block *sb, __u32 compat)
380 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
381 err = ext4_journal_get_write_access(handle,
385 EXT4_SET_COMPAT_FEATURE(sb, compat);
388 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
389 "call ext4_journal_dirty_met adata");
390 err = ext4_journal_dirty_metadata(handle,
396 int ext4_update_rocompat_feature(handle_t *handle,
397 struct super_block *sb, __u32 rocompat)
400 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
401 err = ext4_journal_get_write_access(handle,
405 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
408 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
409 "call ext4_journal_dirty_met adata");
410 err = ext4_journal_dirty_metadata(handle,
416 int ext4_update_incompat_feature(handle_t *handle,
417 struct super_block *sb, __u32 incompat)
420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
421 err = ext4_journal_get_write_access(handle,
425 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
428 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
429 "call ext4_journal_dirty_met adata");
430 err = ext4_journal_dirty_metadata(handle,
437 * Open the external journal device
439 static struct block_device *ext4_blkdev_get(dev_t dev)
441 struct block_device *bdev;
442 char b[BDEVNAME_SIZE];
444 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
450 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
451 __bdevname(dev, b), PTR_ERR(bdev));
456 * Release the journal device
458 static int ext4_blkdev_put(struct block_device *bdev)
461 return blkdev_put(bdev);
464 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
466 struct block_device *bdev;
469 bdev = sbi->journal_bdev;
471 ret = ext4_blkdev_put(bdev);
472 sbi->journal_bdev = NULL;
477 static inline struct inode *orphan_list_entry(struct list_head *l)
479 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
482 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
486 printk(KERN_ERR "sb orphan head is %d\n",
487 le32_to_cpu(sbi->s_es->s_last_orphan));
489 printk(KERN_ERR "sb_info orphan list:\n");
490 list_for_each(l, &sbi->s_orphan) {
491 struct inode *inode = orphan_list_entry(l);
493 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
494 inode->i_sb->s_id, inode->i_ino, inode,
495 inode->i_mode, inode->i_nlink,
500 static void ext4_put_super (struct super_block * sb)
502 struct ext4_sb_info *sbi = EXT4_SB(sb);
503 struct ext4_super_block *es = sbi->s_es;
506 ext4_ext_release(sb);
507 ext4_xattr_put_super(sb);
508 jbd2_journal_destroy(sbi->s_journal);
509 if (!(sb->s_flags & MS_RDONLY)) {
510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
514 ext4_commit_super(sb, es, 1);
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
537 invalidate_bdev(sb->s_bdev);
538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
544 sync_blockdev(sbi->journal_bdev);
545 invalidate_bdev(sbi->journal_bdev);
546 ext4_blkdev_remove(sbi);
548 sb->s_fs_info = NULL;
553 static struct kmem_cache *ext4_inode_cachep;
556 * Called inside transaction, so use GFP_NOFS
558 static struct inode *ext4_alloc_inode(struct super_block *sb)
560 struct ext4_inode_info *ei;
562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
565 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
572 return &ei->vfs_inode;
575 static void ext4_destroy_inode(struct inode *inode)
577 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
578 printk("EXT4 Inode %p: orphan list check failed!\n",
580 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
581 EXT4_I(inode), sizeof(struct ext4_inode_info),
585 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
588 static void init_once(struct kmem_cache *cachep, void *foo)
590 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
592 INIT_LIST_HEAD(&ei->i_orphan);
593 #ifdef CONFIG_EXT4DEV_FS_XATTR
594 init_rwsem(&ei->xattr_sem);
596 init_rwsem(&ei->i_data_sem);
597 inode_init_once(&ei->vfs_inode);
600 static int init_inodecache(void)
602 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
603 sizeof(struct ext4_inode_info),
604 0, (SLAB_RECLAIM_ACCOUNT|
607 if (ext4_inode_cachep == NULL)
612 static void destroy_inodecache(void)
614 kmem_cache_destroy(ext4_inode_cachep);
617 static void ext4_clear_inode(struct inode *inode)
619 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
620 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
621 if (EXT4_I(inode)->i_acl &&
622 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
623 posix_acl_release(EXT4_I(inode)->i_acl);
624 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
626 if (EXT4_I(inode)->i_default_acl &&
627 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
628 posix_acl_release(EXT4_I(inode)->i_default_acl);
629 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
632 ext4_discard_reservation(inode);
633 EXT4_I(inode)->i_block_alloc_info = NULL;
638 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
640 #if defined(CONFIG_QUOTA)
641 struct ext4_sb_info *sbi = EXT4_SB(sb);
643 if (sbi->s_jquota_fmt)
644 seq_printf(seq, ",jqfmt=%s",
645 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
647 if (sbi->s_qf_names[USRQUOTA])
648 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
650 if (sbi->s_qf_names[GRPQUOTA])
651 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
653 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
654 seq_puts(seq, ",usrquota");
656 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
657 seq_puts(seq, ",grpquota");
663 * - it's set to a non-default value OR
664 * - if the per-sb default is different from the global default
666 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
668 struct super_block *sb = vfs->mnt_sb;
669 struct ext4_sb_info *sbi = EXT4_SB(sb);
670 struct ext4_super_block *es = sbi->s_es;
671 unsigned long def_mount_opts;
673 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
675 if (sbi->s_sb_block != 1)
676 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
677 if (test_opt(sb, MINIX_DF))
678 seq_puts(seq, ",minixdf");
679 if (test_opt(sb, GRPID))
680 seq_puts(seq, ",grpid");
681 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
682 seq_puts(seq, ",nogrpid");
683 if (sbi->s_resuid != EXT4_DEF_RESUID ||
684 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
685 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
687 if (sbi->s_resgid != EXT4_DEF_RESGID ||
688 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
689 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
691 if (test_opt(sb, ERRORS_RO)) {
692 int def_errors = le16_to_cpu(es->s_errors);
694 if (def_errors == EXT4_ERRORS_PANIC ||
695 def_errors == EXT4_ERRORS_CONTINUE) {
696 seq_puts(seq, ",errors=remount-ro");
699 if (test_opt(sb, ERRORS_CONT))
700 seq_puts(seq, ",errors=continue");
701 if (test_opt(sb, ERRORS_PANIC))
702 seq_puts(seq, ",errors=panic");
703 if (test_opt(sb, NO_UID32))
704 seq_puts(seq, ",nouid32");
705 if (test_opt(sb, DEBUG))
706 seq_puts(seq, ",debug");
707 if (test_opt(sb, OLDALLOC))
708 seq_puts(seq, ",oldalloc");
709 #ifdef CONFIG_EXT4DEV_FS_XATTR
710 if (test_opt(sb, XATTR_USER))
711 seq_puts(seq, ",user_xattr");
712 if (!test_opt(sb, XATTR_USER) &&
713 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
714 seq_puts(seq, ",nouser_xattr");
717 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
718 if (test_opt(sb, POSIX_ACL))
719 seq_puts(seq, ",acl");
720 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
721 seq_puts(seq, ",noacl");
723 if (!test_opt(sb, RESERVATION))
724 seq_puts(seq, ",noreservation");
725 if (sbi->s_commit_interval) {
726 seq_printf(seq, ",commit=%u",
727 (unsigned) (sbi->s_commit_interval / HZ));
729 if (test_opt(sb, BARRIER))
730 seq_puts(seq, ",barrier=1");
731 if (test_opt(sb, NOBH))
732 seq_puts(seq, ",nobh");
733 if (!test_opt(sb, EXTENTS))
734 seq_puts(seq, ",noextents");
736 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
737 seq_puts(seq, ",data=journal");
738 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
739 seq_puts(seq, ",data=ordered");
740 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
741 seq_puts(seq, ",data=writeback");
743 ext4_show_quota_options(seq, sb);
749 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
750 u64 ino, u32 generation)
754 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
755 return ERR_PTR(-ESTALE);
756 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
757 return ERR_PTR(-ESTALE);
759 /* iget isn't really right if the inode is currently unallocated!!
761 * ext4_read_inode will return a bad_inode if the inode had been
762 * deleted, so we should be safe.
764 * Currently we don't know the generation for parent directory, so
765 * a generation of 0 means "accept any"
767 inode = iget(sb, ino);
769 return ERR_PTR(-ENOMEM);
770 if (is_bad_inode(inode) ||
771 (generation && inode->i_generation != generation)) {
773 return ERR_PTR(-ESTALE);
779 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
780 int fh_len, int fh_type)
782 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
786 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
787 int fh_len, int fh_type)
789 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
794 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
795 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
797 static int ext4_dquot_initialize(struct inode *inode, int type);
798 static int ext4_dquot_drop(struct inode *inode);
799 static int ext4_write_dquot(struct dquot *dquot);
800 static int ext4_acquire_dquot(struct dquot *dquot);
801 static int ext4_release_dquot(struct dquot *dquot);
802 static int ext4_mark_dquot_dirty(struct dquot *dquot);
803 static int ext4_write_info(struct super_block *sb, int type);
804 static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
805 static int ext4_quota_on_mount(struct super_block *sb, int type);
806 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
807 size_t len, loff_t off);
808 static ssize_t ext4_quota_write(struct super_block *sb, int type,
809 const char *data, size_t len, loff_t off);
811 static struct dquot_operations ext4_quota_operations = {
812 .initialize = ext4_dquot_initialize,
813 .drop = ext4_dquot_drop,
814 .alloc_space = dquot_alloc_space,
815 .alloc_inode = dquot_alloc_inode,
816 .free_space = dquot_free_space,
817 .free_inode = dquot_free_inode,
818 .transfer = dquot_transfer,
819 .write_dquot = ext4_write_dquot,
820 .acquire_dquot = ext4_acquire_dquot,
821 .release_dquot = ext4_release_dquot,
822 .mark_dirty = ext4_mark_dquot_dirty,
823 .write_info = ext4_write_info
826 static struct quotactl_ops ext4_qctl_operations = {
827 .quota_on = ext4_quota_on,
828 .quota_off = vfs_quota_off,
829 .quota_sync = vfs_quota_sync,
830 .get_info = vfs_get_dqinfo,
831 .set_info = vfs_set_dqinfo,
832 .get_dqblk = vfs_get_dqblk,
833 .set_dqblk = vfs_set_dqblk
837 static const struct super_operations ext4_sops = {
838 .alloc_inode = ext4_alloc_inode,
839 .destroy_inode = ext4_destroy_inode,
840 .read_inode = ext4_read_inode,
841 .write_inode = ext4_write_inode,
842 .dirty_inode = ext4_dirty_inode,
843 .delete_inode = ext4_delete_inode,
844 .put_super = ext4_put_super,
845 .write_super = ext4_write_super,
846 .sync_fs = ext4_sync_fs,
847 .write_super_lockfs = ext4_write_super_lockfs,
848 .unlockfs = ext4_unlockfs,
849 .statfs = ext4_statfs,
850 .remount_fs = ext4_remount,
851 .clear_inode = ext4_clear_inode,
852 .show_options = ext4_show_options,
854 .quota_read = ext4_quota_read,
855 .quota_write = ext4_quota_write,
859 static const struct export_operations ext4_export_ops = {
860 .fh_to_dentry = ext4_fh_to_dentry,
861 .fh_to_parent = ext4_fh_to_parent,
862 .get_parent = ext4_get_parent,
866 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
867 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
868 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
869 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
870 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
871 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
872 Opt_journal_checksum, Opt_journal_async_commit,
873 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
874 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
875 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
876 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
877 Opt_grpquota, Opt_extents, Opt_noextents,
880 static match_table_t tokens = {
881 {Opt_bsd_df, "bsddf"},
882 {Opt_minix_df, "minixdf"},
883 {Opt_grpid, "grpid"},
884 {Opt_grpid, "bsdgroups"},
885 {Opt_nogrpid, "nogrpid"},
886 {Opt_nogrpid, "sysvgroups"},
887 {Opt_resgid, "resgid=%u"},
888 {Opt_resuid, "resuid=%u"},
890 {Opt_err_cont, "errors=continue"},
891 {Opt_err_panic, "errors=panic"},
892 {Opt_err_ro, "errors=remount-ro"},
893 {Opt_nouid32, "nouid32"},
894 {Opt_nocheck, "nocheck"},
895 {Opt_nocheck, "check=none"},
896 {Opt_debug, "debug"},
897 {Opt_oldalloc, "oldalloc"},
898 {Opt_orlov, "orlov"},
899 {Opt_user_xattr, "user_xattr"},
900 {Opt_nouser_xattr, "nouser_xattr"},
902 {Opt_noacl, "noacl"},
903 {Opt_reservation, "reservation"},
904 {Opt_noreservation, "noreservation"},
905 {Opt_noload, "noload"},
908 {Opt_commit, "commit=%u"},
909 {Opt_journal_update, "journal=update"},
910 {Opt_journal_inum, "journal=%u"},
911 {Opt_journal_dev, "journal_dev=%u"},
912 {Opt_journal_checksum, "journal_checksum"},
913 {Opt_journal_async_commit, "journal_async_commit"},
914 {Opt_abort, "abort"},
915 {Opt_data_journal, "data=journal"},
916 {Opt_data_ordered, "data=ordered"},
917 {Opt_data_writeback, "data=writeback"},
918 {Opt_offusrjquota, "usrjquota="},
919 {Opt_usrjquota, "usrjquota=%s"},
920 {Opt_offgrpjquota, "grpjquota="},
921 {Opt_grpjquota, "grpjquota=%s"},
922 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
923 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
924 {Opt_grpquota, "grpquota"},
925 {Opt_noquota, "noquota"},
926 {Opt_quota, "quota"},
927 {Opt_usrquota, "usrquota"},
928 {Opt_barrier, "barrier=%u"},
929 {Opt_extents, "extents"},
930 {Opt_noextents, "noextents"},
932 {Opt_resize, "resize"},
935 static ext4_fsblk_t get_sb_block(void **data)
937 ext4_fsblk_t sb_block;
938 char *options = (char *) *data;
940 if (!options || strncmp(options, "sb=", 3) != 0)
941 return 1; /* Default location */
943 /*todo: use simple_strtoll with >32bit ext4 */
944 sb_block = simple_strtoul(options, &options, 0);
945 if (*options && *options != ',') {
946 printk("EXT4-fs: Invalid sb specification: %s\n",
952 *data = (void *) options;
956 static int parse_options (char *options, struct super_block *sb,
957 unsigned int *inum, unsigned long *journal_devnum,
958 ext4_fsblk_t *n_blocks_count, int is_remount)
960 struct ext4_sb_info *sbi = EXT4_SB(sb);
962 substring_t args[MAX_OPT_ARGS];
973 while ((p = strsep (&options, ",")) != NULL) {
978 token = match_token(p, tokens, args);
981 clear_opt (sbi->s_mount_opt, MINIX_DF);
984 set_opt (sbi->s_mount_opt, MINIX_DF);
987 set_opt (sbi->s_mount_opt, GRPID);
990 clear_opt (sbi->s_mount_opt, GRPID);
993 if (match_int(&args[0], &option))
995 sbi->s_resuid = option;
998 if (match_int(&args[0], &option))
1000 sbi->s_resgid = option;
1003 /* handled by get_sb_block() instead of here */
1004 /* *sb_block = match_int(&args[0]); */
1007 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1008 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1009 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1012 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1013 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1014 set_opt (sbi->s_mount_opt, ERRORS_RO);
1017 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1018 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1019 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1022 set_opt (sbi->s_mount_opt, NO_UID32);
1025 clear_opt (sbi->s_mount_opt, CHECK);
1028 set_opt (sbi->s_mount_opt, DEBUG);
1031 set_opt (sbi->s_mount_opt, OLDALLOC);
1034 clear_opt (sbi->s_mount_opt, OLDALLOC);
1036 #ifdef CONFIG_EXT4DEV_FS_XATTR
1037 case Opt_user_xattr:
1038 set_opt (sbi->s_mount_opt, XATTR_USER);
1040 case Opt_nouser_xattr:
1041 clear_opt (sbi->s_mount_opt, XATTR_USER);
1044 case Opt_user_xattr:
1045 case Opt_nouser_xattr:
1046 printk("EXT4 (no)user_xattr options not supported\n");
1049 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1051 set_opt(sbi->s_mount_opt, POSIX_ACL);
1054 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1059 printk("EXT4 (no)acl options not supported\n");
1062 case Opt_reservation:
1063 set_opt(sbi->s_mount_opt, RESERVATION);
1065 case Opt_noreservation:
1066 clear_opt(sbi->s_mount_opt, RESERVATION);
1068 case Opt_journal_update:
1070 /* Eventually we will want to be able to create
1071 a journal file here. For now, only allow the
1072 user to specify an existing inode to be the
1075 printk(KERN_ERR "EXT4-fs: cannot specify "
1076 "journal on remount\n");
1079 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1081 case Opt_journal_inum:
1083 printk(KERN_ERR "EXT4-fs: cannot specify "
1084 "journal on remount\n");
1087 if (match_int(&args[0], &option))
1091 case Opt_journal_dev:
1093 printk(KERN_ERR "EXT4-fs: cannot specify "
1094 "journal on remount\n");
1097 if (match_int(&args[0], &option))
1099 *journal_devnum = option;
1101 case Opt_journal_checksum:
1102 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1104 case Opt_journal_async_commit:
1105 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1106 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1109 set_opt (sbi->s_mount_opt, NOLOAD);
1112 if (match_int(&args[0], &option))
1117 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1118 sbi->s_commit_interval = HZ * option;
1120 case Opt_data_journal:
1121 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1123 case Opt_data_ordered:
1124 data_opt = EXT4_MOUNT_ORDERED_DATA;
1126 case Opt_data_writeback:
1127 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1130 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1133 "EXT4-fs: cannot change data "
1134 "mode on remount\n");
1138 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1139 sbi->s_mount_opt |= data_opt;
1149 if (sb_any_quota_enabled(sb)) {
1151 "EXT4-fs: Cannot change journalled "
1152 "quota options when quota turned on.\n");
1155 qname = match_strdup(&args[0]);
1158 "EXT4-fs: not enough memory for "
1159 "storing quotafile name.\n");
1162 if (sbi->s_qf_names[qtype] &&
1163 strcmp(sbi->s_qf_names[qtype], qname)) {
1165 "EXT4-fs: %s quota file already "
1166 "specified.\n", QTYPE2NAME(qtype));
1170 sbi->s_qf_names[qtype] = qname;
1171 if (strchr(sbi->s_qf_names[qtype], '/')) {
1173 "EXT4-fs: quotafile must be on "
1174 "filesystem root.\n");
1175 kfree(sbi->s_qf_names[qtype]);
1176 sbi->s_qf_names[qtype] = NULL;
1179 set_opt(sbi->s_mount_opt, QUOTA);
1181 case Opt_offusrjquota:
1184 case Opt_offgrpjquota:
1187 if (sb_any_quota_enabled(sb)) {
1188 printk(KERN_ERR "EXT4-fs: Cannot change "
1189 "journalled quota options when "
1190 "quota turned on.\n");
1194 * The space will be released later when all options
1195 * are confirmed to be correct
1197 sbi->s_qf_names[qtype] = NULL;
1199 case Opt_jqfmt_vfsold:
1200 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1202 case Opt_jqfmt_vfsv0:
1203 sbi->s_jquota_fmt = QFMT_VFS_V0;
1207 set_opt(sbi->s_mount_opt, QUOTA);
1208 set_opt(sbi->s_mount_opt, USRQUOTA);
1211 set_opt(sbi->s_mount_opt, QUOTA);
1212 set_opt(sbi->s_mount_opt, GRPQUOTA);
1215 if (sb_any_quota_enabled(sb)) {
1216 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1217 "options when quota turned on.\n");
1220 clear_opt(sbi->s_mount_opt, QUOTA);
1221 clear_opt(sbi->s_mount_opt, USRQUOTA);
1222 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1230 case Opt_offusrjquota:
1231 case Opt_offgrpjquota:
1232 case Opt_jqfmt_vfsold:
1233 case Opt_jqfmt_vfsv0:
1235 "EXT4-fs: journalled quota options not "
1242 set_opt(sbi->s_mount_opt, ABORT);
1245 if (match_int(&args[0], &option))
1248 set_opt(sbi->s_mount_opt, BARRIER);
1250 clear_opt(sbi->s_mount_opt, BARRIER);
1256 printk("EXT4-fs: resize option only available "
1260 if (match_int(&args[0], &option) != 0)
1262 *n_blocks_count = option;
1265 set_opt(sbi->s_mount_opt, NOBH);
1268 clear_opt(sbi->s_mount_opt, NOBH);
1271 set_opt (sbi->s_mount_opt, EXTENTS);
1274 clear_opt (sbi->s_mount_opt, EXTENTS);
1278 "EXT4-fs: Unrecognized mount option \"%s\" "
1279 "or missing value\n", p);
1284 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1285 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1286 sbi->s_qf_names[USRQUOTA])
1287 clear_opt(sbi->s_mount_opt, USRQUOTA);
1289 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1290 sbi->s_qf_names[GRPQUOTA])
1291 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1293 if ((sbi->s_qf_names[USRQUOTA] &&
1294 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1295 (sbi->s_qf_names[GRPQUOTA] &&
1296 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1297 printk(KERN_ERR "EXT4-fs: old and new quota "
1298 "format mixing.\n");
1302 if (!sbi->s_jquota_fmt) {
1303 printk(KERN_ERR "EXT4-fs: journalled quota format "
1304 "not specified.\n");
1308 if (sbi->s_jquota_fmt) {
1309 printk(KERN_ERR "EXT4-fs: journalled quota format "
1310 "specified with no journalling "
1319 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1322 struct ext4_sb_info *sbi = EXT4_SB(sb);
1325 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1326 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1327 "forcing read-only mode\n");
1332 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1333 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1334 "running e2fsck is recommended\n");
1335 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1336 printk (KERN_WARNING
1337 "EXT4-fs warning: mounting fs with errors, "
1338 "running e2fsck is recommended\n");
1339 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1340 le16_to_cpu(es->s_mnt_count) >=
1341 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1342 printk (KERN_WARNING
1343 "EXT4-fs warning: maximal mount count reached, "
1344 "running e2fsck is recommended\n");
1345 else if (le32_to_cpu(es->s_checkinterval) &&
1346 (le32_to_cpu(es->s_lastcheck) +
1347 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1348 printk (KERN_WARNING
1349 "EXT4-fs warning: checktime reached, "
1350 "running e2fsck is recommended\n");
1352 /* @@@ We _will_ want to clear the valid bit if we find
1353 * inconsistencies, to force a fsck at reboot. But for
1354 * a plain journaled filesystem we can keep it set as
1357 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1359 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1360 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1361 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1362 es->s_mtime = cpu_to_le32(get_seconds());
1363 ext4_update_dynamic_rev(sb);
1364 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1366 ext4_commit_super(sb, es, 1);
1367 if (test_opt(sb, DEBUG))
1368 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1369 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1371 sbi->s_groups_count,
1372 EXT4_BLOCKS_PER_GROUP(sb),
1373 EXT4_INODES_PER_GROUP(sb),
1376 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1377 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1378 char b[BDEVNAME_SIZE];
1380 printk("external journal on %s\n",
1381 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1383 printk("internal journal\n");
1388 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1389 struct ext4_group_desc *gdp)
1393 if (sbi->s_es->s_feature_ro_compat &
1394 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1395 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1396 __le32 le_group = cpu_to_le32(block_group);
1398 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1399 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1400 crc = crc16(crc, (__u8 *)gdp, offset);
1401 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1402 /* for checksum of struct ext4_group_desc do the rest...*/
1403 if ((sbi->s_es->s_feature_incompat &
1404 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1405 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1406 crc = crc16(crc, (__u8 *)gdp + offset,
1407 le16_to_cpu(sbi->s_es->s_desc_size) -
1411 return cpu_to_le16(crc);
1414 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1415 struct ext4_group_desc *gdp)
1417 if ((sbi->s_es->s_feature_ro_compat &
1418 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1419 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1425 /* Called at mount-time, super-block is locked */
1426 static int ext4_check_descriptors (struct super_block * sb)
1428 struct ext4_sb_info *sbi = EXT4_SB(sb);
1429 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1430 ext4_fsblk_t last_block;
1431 ext4_fsblk_t block_bitmap;
1432 ext4_fsblk_t inode_bitmap;
1433 ext4_fsblk_t inode_table;
1434 struct ext4_group_desc * gdp = NULL;
1436 int flexbg_flag = 0;
1439 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1442 ext4_debug ("Checking group descriptors");
1444 for (i = 0; i < sbi->s_groups_count; i++)
1446 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1447 last_block = ext4_blocks_count(sbi->s_es) - 1;
1449 last_block = first_block +
1450 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1452 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1453 gdp = (struct ext4_group_desc *)
1454 sbi->s_group_desc[desc_block++]->b_data;
1455 block_bitmap = ext4_block_bitmap(sb, gdp);
1456 if (block_bitmap < first_block || block_bitmap > last_block)
1458 ext4_error (sb, "ext4_check_descriptors",
1459 "Block bitmap for group %lu"
1460 " not in group (block %llu)!",
1464 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1465 if (inode_bitmap < first_block || inode_bitmap > last_block)
1467 ext4_error (sb, "ext4_check_descriptors",
1468 "Inode bitmap for group %lu"
1469 " not in group (block %llu)!",
1473 inode_table = ext4_inode_table(sb, gdp);
1474 if (inode_table < first_block ||
1475 inode_table + sbi->s_itb_per_group - 1 > last_block)
1477 ext4_error (sb, "ext4_check_descriptors",
1478 "Inode table for group %lu"
1479 " not in group (block %llu)!",
1483 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1484 ext4_error(sb, __FUNCTION__,
1485 "Checksum for group %lu failed (%u!=%u)\n",
1486 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1487 gdp)), le16_to_cpu(gdp->bg_checksum));
1491 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1492 gdp = (struct ext4_group_desc *)
1493 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1496 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1497 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1501 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1502 * the superblock) which were deleted from all directories, but held open by
1503 * a process at the time of a crash. We walk the list and try to delete these
1504 * inodes at recovery time (only with a read-write filesystem).
1506 * In order to keep the orphan inode chain consistent during traversal (in
1507 * case of crash during recovery), we link each inode into the superblock
1508 * orphan list_head and handle it the same way as an inode deletion during
1509 * normal operation (which journals the operations for us).
1511 * We only do an iget() and an iput() on each inode, which is very safe if we
1512 * accidentally point at an in-use or already deleted inode. The worst that
1513 * can happen in this case is that we get a "bit already cleared" message from
1514 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1515 * e2fsck was run on this filesystem, and it must have already done the orphan
1516 * inode cleanup for us, so we can safely abort without any further action.
1518 static void ext4_orphan_cleanup (struct super_block * sb,
1519 struct ext4_super_block * es)
1521 unsigned int s_flags = sb->s_flags;
1522 int nr_orphans = 0, nr_truncates = 0;
1526 if (!es->s_last_orphan) {
1527 jbd_debug(4, "no orphan inodes to clean up\n");
1531 if (bdev_read_only(sb->s_bdev)) {
1532 printk(KERN_ERR "EXT4-fs: write access "
1533 "unavailable, skipping orphan cleanup.\n");
1537 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1538 if (es->s_last_orphan)
1539 jbd_debug(1, "Errors on filesystem, "
1540 "clearing orphan list.\n");
1541 es->s_last_orphan = 0;
1542 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1546 if (s_flags & MS_RDONLY) {
1547 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1549 sb->s_flags &= ~MS_RDONLY;
1552 /* Needed for iput() to work correctly and not trash data */
1553 sb->s_flags |= MS_ACTIVE;
1554 /* Turn on quotas so that they are updated correctly */
1555 for (i = 0; i < MAXQUOTAS; i++) {
1556 if (EXT4_SB(sb)->s_qf_names[i]) {
1557 int ret = ext4_quota_on_mount(sb, i);
1560 "EXT4-fs: Cannot turn on journalled "
1561 "quota: error %d\n", ret);
1566 while (es->s_last_orphan) {
1567 struct inode *inode;
1570 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1571 es->s_last_orphan = 0;
1575 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1577 if (inode->i_nlink) {
1579 "%s: truncating inode %lu to %Ld bytes\n",
1580 __FUNCTION__, inode->i_ino, inode->i_size);
1581 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1582 inode->i_ino, inode->i_size);
1583 ext4_truncate(inode);
1587 "%s: deleting unreferenced inode %lu\n",
1588 __FUNCTION__, inode->i_ino);
1589 jbd_debug(2, "deleting unreferenced inode %lu\n",
1593 iput(inode); /* The delete magic happens here! */
1596 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1599 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1600 sb->s_id, PLURAL(nr_orphans));
1602 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1603 sb->s_id, PLURAL(nr_truncates));
1605 /* Turn quotas off */
1606 for (i = 0; i < MAXQUOTAS; i++) {
1607 if (sb_dqopt(sb)->files[i])
1608 vfs_quota_off(sb, i);
1611 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1614 * Maximal extent format file size.
1615 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1616 * extent format containers, within a sector_t, and within i_blocks
1617 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1618 * so that won't be a limiting factor.
1620 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1622 static loff_t ext4_max_size(int blkbits)
1625 loff_t upper_limit = MAX_LFS_FILESIZE;
1627 /* small i_blocks in vfs inode? */
1628 if (sizeof(blkcnt_t) < sizeof(u64)) {
1630 * CONFIG_LSF is not enabled implies the inode
1631 * i_block represent total blocks in 512 bytes
1632 * 32 == size of vfs inode i_blocks * 8
1634 upper_limit = (1LL << 32) - 1;
1636 /* total blocks in file system block size */
1637 upper_limit >>= (blkbits - 9);
1638 upper_limit <<= blkbits;
1641 /* 32-bit extent-start container, ee_block */
1646 /* Sanity check against vm- & vfs- imposed limits */
1647 if (res > upper_limit)
1654 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1655 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1656 * We need to be 1 filesystem block less than the 2^48 sector limit.
1658 static loff_t ext4_max_bitmap_size(int bits)
1660 loff_t res = EXT4_NDIR_BLOCKS;
1663 /* This is calculated to be the largest file size for a
1664 * dense, bitmapped file such that the total number of
1665 * sectors in the file, including data and all indirect blocks,
1666 * does not exceed 2^48 -1
1667 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1668 * total number of 512 bytes blocks of the file
1671 if (sizeof(blkcnt_t) < sizeof(u64)) {
1673 * CONFIG_LSF is not enabled implies the inode
1674 * i_block represent total blocks in 512 bytes
1675 * 32 == size of vfs inode i_blocks * 8
1677 upper_limit = (1LL << 32) - 1;
1679 /* total blocks in file system block size */
1680 upper_limit >>= (bits - 9);
1684 * We use 48 bit ext4_inode i_blocks
1685 * With EXT4_HUGE_FILE_FL set the i_blocks
1686 * represent total number of blocks in
1687 * file system block size
1689 upper_limit = (1LL << 48) - 1;
1693 /* indirect blocks */
1695 /* double indirect blocks */
1696 meta_blocks += 1 + (1LL << (bits-2));
1697 /* tripple indirect blocks */
1698 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1700 upper_limit -= meta_blocks;
1701 upper_limit <<= bits;
1703 res += 1LL << (bits-2);
1704 res += 1LL << (2*(bits-2));
1705 res += 1LL << (3*(bits-2));
1707 if (res > upper_limit)
1710 if (res > MAX_LFS_FILESIZE)
1711 res = MAX_LFS_FILESIZE;
1716 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1717 ext4_fsblk_t logical_sb_block, int nr)
1719 struct ext4_sb_info *sbi = EXT4_SB(sb);
1720 ext4_group_t bg, first_meta_bg;
1723 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1725 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1727 return logical_sb_block + nr + 1;
1728 bg = sbi->s_desc_per_block * nr;
1729 if (ext4_bg_has_super(sb, bg))
1731 return (has_super + ext4_group_first_block_no(sb, bg));
1735 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1736 __releases(kernel_sem)
1737 __acquires(kernel_sem)
1740 struct buffer_head * bh;
1741 struct ext4_super_block *es = NULL;
1742 struct ext4_sb_info *sbi;
1744 ext4_fsblk_t sb_block = get_sb_block(&data);
1745 ext4_fsblk_t logical_sb_block;
1746 unsigned long offset = 0;
1747 unsigned int journal_inum = 0;
1748 unsigned long journal_devnum = 0;
1749 unsigned long def_mount_opts;
1760 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1763 sb->s_fs_info = sbi;
1764 sbi->s_mount_opt = 0;
1765 sbi->s_resuid = EXT4_DEF_RESUID;
1766 sbi->s_resgid = EXT4_DEF_RESGID;
1767 sbi->s_sb_block = sb_block;
1771 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1773 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1777 if (!sb_set_blocksize(sb, blocksize)) {
1778 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1783 * The ext4 superblock will not be buffer aligned for other than 1kB
1784 * block sizes. We need to calculate the offset from buffer start.
1786 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1787 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1788 offset = do_div(logical_sb_block, blocksize);
1790 logical_sb_block = sb_block;
1793 if (!(bh = sb_bread(sb, logical_sb_block))) {
1794 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1798 * Note: s_es must be initialized as soon as possible because
1799 * some ext4 macro-instructions depend on its value
1801 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1803 sb->s_magic = le16_to_cpu(es->s_magic);
1804 if (sb->s_magic != EXT4_SUPER_MAGIC)
1807 /* Set defaults before we parse the mount options */
1808 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1809 if (def_mount_opts & EXT4_DEFM_DEBUG)
1810 set_opt(sbi->s_mount_opt, DEBUG);
1811 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1812 set_opt(sbi->s_mount_opt, GRPID);
1813 if (def_mount_opts & EXT4_DEFM_UID16)
1814 set_opt(sbi->s_mount_opt, NO_UID32);
1815 #ifdef CONFIG_EXT4DEV_FS_XATTR
1816 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1817 set_opt(sbi->s_mount_opt, XATTR_USER);
1819 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1820 if (def_mount_opts & EXT4_DEFM_ACL)
1821 set_opt(sbi->s_mount_opt, POSIX_ACL);
1823 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1824 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1825 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1826 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1827 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1828 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1830 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1831 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1832 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1833 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1835 set_opt(sbi->s_mount_opt, ERRORS_RO);
1837 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1838 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1840 set_opt(sbi->s_mount_opt, RESERVATION);
1843 * turn on extents feature by default in ext4 filesystem
1844 * User -o noextents to turn it off
1846 set_opt(sbi->s_mount_opt, EXTENTS);
1848 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1852 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1853 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1855 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1856 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1857 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1858 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1860 "EXT4-fs warning: feature flags set on rev 0 fs, "
1861 "running e2fsck is recommended\n");
1863 * Check feature flags regardless of the revision level, since we
1864 * previously didn't change the revision level when setting the flags,
1865 * so there is a chance incompat flags are set on a rev 0 filesystem.
1867 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1869 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1870 "unsupported optional features (%x).\n",
1871 sb->s_id, le32_to_cpu(features));
1874 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1875 if (!(sb->s_flags & MS_RDONLY) && features) {
1876 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1877 "unsupported optional features (%x).\n",
1878 sb->s_id, le32_to_cpu(features));
1881 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1883 * Large file size enabled file system can only be
1884 * mount if kernel is build with CONFIG_LSF
1886 if (sizeof(root->i_blocks) < sizeof(u64) &&
1887 !(sb->s_flags & MS_RDONLY)) {
1888 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1889 "files cannot be mounted read-write "
1890 "without CONFIG_LSF.\n", sb->s_id);
1894 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1896 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1897 blocksize > EXT4_MAX_BLOCK_SIZE) {
1899 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1900 blocksize, sb->s_id);
1904 hblock = bdev_hardsect_size(sb->s_bdev);
1905 if (sb->s_blocksize != blocksize) {
1907 * Make sure the blocksize for the filesystem is larger
1908 * than the hardware sectorsize for the machine.
1910 if (blocksize < hblock) {
1911 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1912 "device blocksize %d.\n", blocksize, hblock);
1917 sb_set_blocksize(sb, blocksize);
1918 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1919 offset = do_div(logical_sb_block, blocksize);
1920 bh = sb_bread(sb, logical_sb_block);
1923 "EXT4-fs: Can't read superblock on 2nd try.\n");
1926 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1928 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1930 "EXT4-fs: Magic mismatch, very weird !\n");
1935 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
1936 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1938 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1939 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1940 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1942 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1943 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1944 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1945 (!is_power_of_2(sbi->s_inode_size)) ||
1946 (sbi->s_inode_size > blocksize)) {
1948 "EXT4-fs: unsupported inode size: %d\n",
1952 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1953 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1955 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1956 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1957 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1958 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1959 !is_power_of_2(sbi->s_desc_size)) {
1961 "EXT4-fs: unsupported descriptor size %lu\n",
1966 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1967 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1968 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1969 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
1971 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1972 if (sbi->s_inodes_per_block == 0)
1974 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1975 sbi->s_inodes_per_block;
1976 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1978 sbi->s_mount_state = le16_to_cpu(es->s_state);
1979 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1980 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1981 for (i=0; i < 4; i++)
1982 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1983 sbi->s_def_hash_version = es->s_def_hash_version;
1985 if (sbi->s_blocks_per_group > blocksize * 8) {
1987 "EXT4-fs: #blocks per group too big: %lu\n",
1988 sbi->s_blocks_per_group);
1991 if (sbi->s_inodes_per_group > blocksize * 8) {
1993 "EXT4-fs: #inodes per group too big: %lu\n",
1994 sbi->s_inodes_per_group);
1998 if (ext4_blocks_count(es) >
1999 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2000 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2001 " too large to mount safely\n", sb->s_id);
2002 if (sizeof(sector_t) < 8)
2003 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2008 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2011 /* ensure blocks_count calculation below doesn't sign-extend */
2012 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2013 le32_to_cpu(es->s_first_data_block) + 1) {
2014 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2015 "first data block %u, blocks per group %lu\n",
2016 ext4_blocks_count(es),
2017 le32_to_cpu(es->s_first_data_block),
2018 EXT4_BLOCKS_PER_GROUP(sb));
2021 blocks_count = (ext4_blocks_count(es) -
2022 le32_to_cpu(es->s_first_data_block) +
2023 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2024 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2025 sbi->s_groups_count = blocks_count;
2026 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2027 EXT4_DESC_PER_BLOCK(sb);
2028 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2030 if (sbi->s_group_desc == NULL) {
2031 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2035 bgl_lock_init(&sbi->s_blockgroup_lock);
2037 for (i = 0; i < db_count; i++) {
2038 block = descriptor_loc(sb, logical_sb_block, i);
2039 sbi->s_group_desc[i] = sb_bread(sb, block);
2040 if (!sbi->s_group_desc[i]) {
2041 printk (KERN_ERR "EXT4-fs: "
2042 "can't read group descriptor %d\n", i);
2047 if (!ext4_check_descriptors (sb)) {
2048 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2051 sbi->s_gdb_count = db_count;
2052 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2053 spin_lock_init(&sbi->s_next_gen_lock);
2055 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2056 ext4_count_free_blocks(sb));
2058 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2059 ext4_count_free_inodes(sb));
2062 err = percpu_counter_init(&sbi->s_dirs_counter,
2063 ext4_count_dirs(sb));
2066 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2070 /* per fileystem reservation list head & lock */
2071 spin_lock_init(&sbi->s_rsv_window_lock);
2072 sbi->s_rsv_window_root = RB_ROOT;
2073 /* Add a single, static dummy reservation to the start of the
2074 * reservation window list --- it gives us a placeholder for
2075 * append-at-start-of-list which makes the allocation logic
2076 * _much_ simpler. */
2077 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2078 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2079 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2080 sbi->s_rsv_window_head.rsv_goal_size = 0;
2081 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2084 * set up enough so that it can read an inode
2086 sb->s_op = &ext4_sops;
2087 sb->s_export_op = &ext4_export_ops;
2088 sb->s_xattr = ext4_xattr_handlers;
2090 sb->s_qcop = &ext4_qctl_operations;
2091 sb->dq_op = &ext4_quota_operations;
2093 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2097 needs_recovery = (es->s_last_orphan != 0 ||
2098 EXT4_HAS_INCOMPAT_FEATURE(sb,
2099 EXT4_FEATURE_INCOMPAT_RECOVER));
2102 * The first inode we look at is the journal inode. Don't try
2103 * root first: it may be modified in the journal!
2105 if (!test_opt(sb, NOLOAD) &&
2106 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2107 if (ext4_load_journal(sb, es, journal_devnum))
2109 } else if (journal_inum) {
2110 if (ext4_create_journal(sb, es, journal_inum))
2115 "ext4: No journal on filesystem on %s\n",
2120 if (ext4_blocks_count(es) > 0xffffffffULL &&
2121 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2122 JBD2_FEATURE_INCOMPAT_64BIT)) {
2123 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2127 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2128 jbd2_journal_set_features(sbi->s_journal,
2129 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2131 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2132 jbd2_journal_set_features(sbi->s_journal,
2133 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2134 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2135 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2137 jbd2_journal_clear_features(sbi->s_journal,
2138 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2139 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2142 /* We have now updated the journal if required, so we can
2143 * validate the data journaling mode. */
2144 switch (test_opt(sb, DATA_FLAGS)) {
2146 /* No mode set, assume a default based on the journal
2147 * capabilities: ORDERED_DATA if the journal can
2148 * cope, else JOURNAL_DATA
2150 if (jbd2_journal_check_available_features
2151 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2152 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2154 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2157 case EXT4_MOUNT_ORDERED_DATA:
2158 case EXT4_MOUNT_WRITEBACK_DATA:
2159 if (!jbd2_journal_check_available_features
2160 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2161 printk(KERN_ERR "EXT4-fs: Journal does not support "
2162 "requested data journaling mode\n");
2169 if (test_opt(sb, NOBH)) {
2170 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2171 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2172 "its supported only with writeback mode\n");
2173 clear_opt(sbi->s_mount_opt, NOBH);
2177 * The jbd2_journal_load will have done any necessary log recovery,
2178 * so we can safely mount the rest of the filesystem now.
2181 root = iget(sb, EXT4_ROOT_INO);
2182 sb->s_root = d_alloc_root(root);
2184 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2188 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2191 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2195 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2197 /* determine the minimum size of new large inodes, if present */
2198 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2199 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2200 EXT4_GOOD_OLD_INODE_SIZE;
2201 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2202 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2203 if (sbi->s_want_extra_isize <
2204 le16_to_cpu(es->s_want_extra_isize))
2205 sbi->s_want_extra_isize =
2206 le16_to_cpu(es->s_want_extra_isize);
2207 if (sbi->s_want_extra_isize <
2208 le16_to_cpu(es->s_min_extra_isize))
2209 sbi->s_want_extra_isize =
2210 le16_to_cpu(es->s_min_extra_isize);
2213 /* Check if enough inode space is available */
2214 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2215 sbi->s_inode_size) {
2216 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2217 EXT4_GOOD_OLD_INODE_SIZE;
2218 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2223 * akpm: core read_super() calls in here with the superblock locked.
2224 * That deadlocks, because orphan cleanup needs to lock the superblock
2225 * in numerous places. Here we just pop the lock - it's relatively
2226 * harmless, because we are now ready to accept write_super() requests,
2227 * and aviro says that's the only reason for hanging onto the
2230 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2231 ext4_orphan_cleanup(sb, es);
2232 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2234 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2235 ext4_mark_recovery_complete(sb, es);
2236 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2237 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2238 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2248 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2253 jbd2_journal_destroy(sbi->s_journal);
2255 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2256 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2257 percpu_counter_destroy(&sbi->s_dirs_counter);
2259 for (i = 0; i < db_count; i++)
2260 brelse(sbi->s_group_desc[i]);
2261 kfree(sbi->s_group_desc);
2264 for (i = 0; i < MAXQUOTAS; i++)
2265 kfree(sbi->s_qf_names[i]);
2267 ext4_blkdev_remove(sbi);
2270 sb->s_fs_info = NULL;
2277 * Setup any per-fs journal parameters now. We'll do this both on
2278 * initial mount, once the journal has been initialised but before we've
2279 * done any recovery; and again on any subsequent remount.
2281 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2283 struct ext4_sb_info *sbi = EXT4_SB(sb);
2285 if (sbi->s_commit_interval)
2286 journal->j_commit_interval = sbi->s_commit_interval;
2287 /* We could also set up an ext4-specific default for the commit
2288 * interval here, but for now we'll just fall back to the jbd
2291 spin_lock(&journal->j_state_lock);
2292 if (test_opt(sb, BARRIER))
2293 journal->j_flags |= JBD2_BARRIER;
2295 journal->j_flags &= ~JBD2_BARRIER;
2296 spin_unlock(&journal->j_state_lock);
2299 static journal_t *ext4_get_journal(struct super_block *sb,
2300 unsigned int journal_inum)
2302 struct inode *journal_inode;
2305 /* First, test for the existence of a valid inode on disk. Bad
2306 * things happen if we iget() an unused inode, as the subsequent
2307 * iput() will try to delete it. */
2309 journal_inode = iget(sb, journal_inum);
2310 if (!journal_inode) {
2311 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2314 if (!journal_inode->i_nlink) {
2315 make_bad_inode(journal_inode);
2316 iput(journal_inode);
2317 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2321 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2322 journal_inode, journal_inode->i_size);
2323 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2324 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2325 iput(journal_inode);
2329 journal = jbd2_journal_init_inode(journal_inode);
2331 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2332 iput(journal_inode);
2335 journal->j_private = sb;
2336 ext4_init_journal_params(sb, journal);
2340 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2343 struct buffer_head * bh;
2347 int hblock, blocksize;
2348 ext4_fsblk_t sb_block;
2349 unsigned long offset;
2350 struct ext4_super_block * es;
2351 struct block_device *bdev;
2353 bdev = ext4_blkdev_get(j_dev);
2357 if (bd_claim(bdev, sb)) {
2359 "EXT4: failed to claim external journal device.\n");
2364 blocksize = sb->s_blocksize;
2365 hblock = bdev_hardsect_size(bdev);
2366 if (blocksize < hblock) {
2368 "EXT4-fs: blocksize too small for journal device.\n");
2372 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2373 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2374 set_blocksize(bdev, blocksize);
2375 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2376 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2377 "external journal\n");
2381 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2382 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2383 !(le32_to_cpu(es->s_feature_incompat) &
2384 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2385 printk(KERN_ERR "EXT4-fs: external journal has "
2386 "bad superblock\n");
2391 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2392 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2397 len = ext4_blocks_count(es);
2398 start = sb_block + 1;
2399 brelse(bh); /* we're done with the superblock */
2401 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2402 start, len, blocksize);
2404 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2407 journal->j_private = sb;
2408 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2409 wait_on_buffer(journal->j_sb_buffer);
2410 if (!buffer_uptodate(journal->j_sb_buffer)) {
2411 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2414 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2415 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2416 "user (unsupported) - %d\n",
2417 be32_to_cpu(journal->j_superblock->s_nr_users));
2420 EXT4_SB(sb)->journal_bdev = bdev;
2421 ext4_init_journal_params(sb, journal);
2424 jbd2_journal_destroy(journal);
2426 ext4_blkdev_put(bdev);
2430 static int ext4_load_journal(struct super_block *sb,
2431 struct ext4_super_block *es,
2432 unsigned long journal_devnum)
2435 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2438 int really_read_only;
2440 if (journal_devnum &&
2441 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2442 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2443 "numbers have changed\n");
2444 journal_dev = new_decode_dev(journal_devnum);
2446 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2448 really_read_only = bdev_read_only(sb->s_bdev);
2451 * Are we loading a blank journal or performing recovery after a
2452 * crash? For recovery, we need to check in advance whether we
2453 * can get read-write access to the device.
2456 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2457 if (sb->s_flags & MS_RDONLY) {
2458 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2459 "required on readonly filesystem.\n");
2460 if (really_read_only) {
2461 printk(KERN_ERR "EXT4-fs: write access "
2462 "unavailable, cannot proceed.\n");
2465 printk (KERN_INFO "EXT4-fs: write access will "
2466 "be enabled during recovery.\n");
2470 if (journal_inum && journal_dev) {
2471 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2472 "and inode journals!\n");
2477 if (!(journal = ext4_get_journal(sb, journal_inum)))
2480 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2484 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2485 err = jbd2_journal_update_format(journal);
2487 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2488 jbd2_journal_destroy(journal);
2493 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2494 err = jbd2_journal_wipe(journal, !really_read_only);
2496 err = jbd2_journal_load(journal);
2499 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2500 jbd2_journal_destroy(journal);
2504 EXT4_SB(sb)->s_journal = journal;
2505 ext4_clear_journal_err(sb, es);
2507 if (journal_devnum &&
2508 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2509 es->s_journal_dev = cpu_to_le32(journal_devnum);
2512 /* Make sure we flush the recovery flag to disk. */
2513 ext4_commit_super(sb, es, 1);
2519 static int ext4_create_journal(struct super_block * sb,
2520 struct ext4_super_block * es,
2521 unsigned int journal_inum)
2526 if (sb->s_flags & MS_RDONLY) {
2527 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2528 "create journal.\n");
2532 journal = ext4_get_journal(sb, journal_inum);
2536 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2539 err = jbd2_journal_create(journal);
2541 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2542 jbd2_journal_destroy(journal);
2546 EXT4_SB(sb)->s_journal = journal;
2548 ext4_update_dynamic_rev(sb);
2549 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2550 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2552 es->s_journal_inum = cpu_to_le32(journal_inum);
2555 /* Make sure we flush the recovery flag to disk. */
2556 ext4_commit_super(sb, es, 1);
2561 static void ext4_commit_super (struct super_block * sb,
2562 struct ext4_super_block * es,
2565 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2569 es->s_wtime = cpu_to_le32(get_seconds());
2570 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2571 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2572 BUFFER_TRACE(sbh, "marking dirty");
2573 mark_buffer_dirty(sbh);
2575 sync_dirty_buffer(sbh);
2580 * Have we just finished recovery? If so, and if we are mounting (or
2581 * remounting) the filesystem readonly, then we will end up with a
2582 * consistent fs on disk. Record that fact.
2584 static void ext4_mark_recovery_complete(struct super_block * sb,
2585 struct ext4_super_block * es)
2587 journal_t *journal = EXT4_SB(sb)->s_journal;
2589 jbd2_journal_lock_updates(journal);
2590 jbd2_journal_flush(journal);
2592 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2593 sb->s_flags & MS_RDONLY) {
2594 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2596 ext4_commit_super(sb, es, 1);
2599 jbd2_journal_unlock_updates(journal);
2603 * If we are mounting (or read-write remounting) a filesystem whose journal
2604 * has recorded an error from a previous lifetime, move that error to the
2605 * main filesystem now.
2607 static void ext4_clear_journal_err(struct super_block * sb,
2608 struct ext4_super_block * es)
2614 journal = EXT4_SB(sb)->s_journal;
2617 * Now check for any error status which may have been recorded in the
2618 * journal by a prior ext4_error() or ext4_abort()
2621 j_errno = jbd2_journal_errno(journal);
2625 errstr = ext4_decode_error(sb, j_errno, nbuf);
2626 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2627 "from previous mount: %s", errstr);
2628 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2629 "filesystem check.");
2631 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2632 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2633 ext4_commit_super (sb, es, 1);
2635 jbd2_journal_clear_err(journal);
2640 * Force the running and committing transactions to commit,
2641 * and wait on the commit.
2643 int ext4_force_commit(struct super_block *sb)
2648 if (sb->s_flags & MS_RDONLY)
2651 journal = EXT4_SB(sb)->s_journal;
2653 ret = ext4_journal_force_commit(journal);
2658 * Ext4 always journals updates to the superblock itself, so we don't
2659 * have to propagate any other updates to the superblock on disk at this
2660 * point. Just start an async writeback to get the buffers on their way
2663 * This implicitly triggers the writebehind on sync().
2666 static void ext4_write_super (struct super_block * sb)
2668 if (mutex_trylock(&sb->s_lock) != 0)
2673 static int ext4_sync_fs(struct super_block *sb, int wait)
2678 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2680 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2686 * LVM calls this function before a (read-only) snapshot is created. This
2687 * gives us a chance to flush the journal completely and mark the fs clean.
2689 static void ext4_write_super_lockfs(struct super_block *sb)
2693 if (!(sb->s_flags & MS_RDONLY)) {
2694 journal_t *journal = EXT4_SB(sb)->s_journal;
2696 /* Now we set up the journal barrier. */
2697 jbd2_journal_lock_updates(journal);
2698 jbd2_journal_flush(journal);
2700 /* Journal blocked and flushed, clear needs_recovery flag. */
2701 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2702 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2707 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2708 * flag here, even though the filesystem is not technically dirty yet.
2710 static void ext4_unlockfs(struct super_block *sb)
2712 if (!(sb->s_flags & MS_RDONLY)) {
2714 /* Reser the needs_recovery flag before the fs is unlocked. */
2715 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2716 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2718 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2722 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2724 struct ext4_super_block * es;
2725 struct ext4_sb_info *sbi = EXT4_SB(sb);
2726 ext4_fsblk_t n_blocks_count = 0;
2727 unsigned long old_sb_flags;
2728 struct ext4_mount_options old_opts;
2734 /* Store the original options */
2735 old_sb_flags = sb->s_flags;
2736 old_opts.s_mount_opt = sbi->s_mount_opt;
2737 old_opts.s_resuid = sbi->s_resuid;
2738 old_opts.s_resgid = sbi->s_resgid;
2739 old_opts.s_commit_interval = sbi->s_commit_interval;
2741 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2742 for (i = 0; i < MAXQUOTAS; i++)
2743 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2747 * Allow the "check" option to be passed as a remount option.
2749 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2754 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2755 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2757 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2758 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2762 ext4_init_journal_params(sb, sbi->s_journal);
2764 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2765 n_blocks_count > ext4_blocks_count(es)) {
2766 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2771 if (*flags & MS_RDONLY) {
2773 * First of all, the unconditional stuff we have to do
2774 * to disable replay of the journal when we next remount
2776 sb->s_flags |= MS_RDONLY;
2779 * OK, test if we are remounting a valid rw partition
2780 * readonly, and if so set the rdonly flag and then
2781 * mark the partition as valid again.
2783 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2784 (sbi->s_mount_state & EXT4_VALID_FS))
2785 es->s_state = cpu_to_le16(sbi->s_mount_state);
2788 * We have to unlock super so that we can wait for
2792 ext4_mark_recovery_complete(sb, es);
2796 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2797 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2798 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2799 "remount RDWR because of unsupported "
2800 "optional features (%x).\n",
2801 sb->s_id, le32_to_cpu(ret));
2807 * If we have an unprocessed orphan list hanging
2808 * around from a previously readonly bdev mount,
2809 * require a full umount/remount for now.
2811 if (es->s_last_orphan) {
2812 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2813 "remount RDWR because of unprocessed "
2814 "orphan inode list. Please "
2815 "umount/remount instead.\n",
2822 * Mounting a RDONLY partition read-write, so reread
2823 * and store the current valid flag. (It may have
2824 * been changed by e2fsck since we originally mounted
2827 ext4_clear_journal_err(sb, es);
2828 sbi->s_mount_state = le16_to_cpu(es->s_state);
2829 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2831 if (!ext4_setup_super (sb, es, 0))
2832 sb->s_flags &= ~MS_RDONLY;
2836 /* Release old quota file names */
2837 for (i = 0; i < MAXQUOTAS; i++)
2838 if (old_opts.s_qf_names[i] &&
2839 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2840 kfree(old_opts.s_qf_names[i]);
2844 sb->s_flags = old_sb_flags;
2845 sbi->s_mount_opt = old_opts.s_mount_opt;
2846 sbi->s_resuid = old_opts.s_resuid;
2847 sbi->s_resgid = old_opts.s_resgid;
2848 sbi->s_commit_interval = old_opts.s_commit_interval;
2850 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2851 for (i = 0; i < MAXQUOTAS; i++) {
2852 if (sbi->s_qf_names[i] &&
2853 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2854 kfree(sbi->s_qf_names[i]);
2855 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2861 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2863 struct super_block *sb = dentry->d_sb;
2864 struct ext4_sb_info *sbi = EXT4_SB(sb);
2865 struct ext4_super_block *es = sbi->s_es;
2868 if (test_opt(sb, MINIX_DF)) {
2869 sbi->s_overhead_last = 0;
2870 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2871 ext4_group_t ngroups = sbi->s_groups_count, i;
2872 ext4_fsblk_t overhead = 0;
2876 * Compute the overhead (FS structures). This is constant
2877 * for a given filesystem unless the number of block groups
2878 * changes so we cache the previous value until it does.
2882 * All of the blocks before first_data_block are
2885 overhead = le32_to_cpu(es->s_first_data_block);
2888 * Add the overhead attributed to the superblock and
2889 * block group descriptors. If the sparse superblocks
2890 * feature is turned on, then not all groups have this.
2892 for (i = 0; i < ngroups; i++) {
2893 overhead += ext4_bg_has_super(sb, i) +
2894 ext4_bg_num_gdb(sb, i);
2899 * Every block group has an inode bitmap, a block
2900 * bitmap, and an inode table.
2902 overhead += ngroups * (2 + sbi->s_itb_per_group);
2903 sbi->s_overhead_last = overhead;
2905 sbi->s_blocks_last = ext4_blocks_count(es);
2908 buf->f_type = EXT4_SUPER_MAGIC;
2909 buf->f_bsize = sb->s_blocksize;
2910 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2911 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2912 ext4_free_blocks_count_set(es, buf->f_bfree);
2913 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2914 if (buf->f_bfree < ext4_r_blocks_count(es))
2916 buf->f_files = le32_to_cpu(es->s_inodes_count);
2917 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2918 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2919 buf->f_namelen = EXT4_NAME_LEN;
2920 fsid = le64_to_cpup((void *)es->s_uuid) ^
2921 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2922 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2923 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2927 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2928 * is locked for write. Otherwise the are possible deadlocks:
2929 * Process 1 Process 2
2930 * ext4_create() quota_sync()
2931 * jbd2_journal_start() write_dquot()
2932 * DQUOT_INIT() down(dqio_mutex)
2933 * down(dqio_mutex) jbd2_journal_start()
2939 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2941 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2944 static int ext4_dquot_initialize(struct inode *inode, int type)
2949 /* We may create quota structure so we need to reserve enough blocks */
2950 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2952 return PTR_ERR(handle);
2953 ret = dquot_initialize(inode, type);
2954 err = ext4_journal_stop(handle);
2960 static int ext4_dquot_drop(struct inode *inode)
2965 /* We may delete quota structure so we need to reserve enough blocks */
2966 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2968 return PTR_ERR(handle);
2969 ret = dquot_drop(inode);
2970 err = ext4_journal_stop(handle);
2976 static int ext4_write_dquot(struct dquot *dquot)
2980 struct inode *inode;
2982 inode = dquot_to_inode(dquot);
2983 handle = ext4_journal_start(inode,
2984 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2986 return PTR_ERR(handle);
2987 ret = dquot_commit(dquot);
2988 err = ext4_journal_stop(handle);
2994 static int ext4_acquire_dquot(struct dquot *dquot)
2999 handle = ext4_journal_start(dquot_to_inode(dquot),
3000 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3002 return PTR_ERR(handle);
3003 ret = dquot_acquire(dquot);
3004 err = ext4_journal_stop(handle);
3010 static int ext4_release_dquot(struct dquot *dquot)
3015 handle = ext4_journal_start(dquot_to_inode(dquot),
3016 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3017 if (IS_ERR(handle)) {
3018 /* Release dquot anyway to avoid endless cycle in dqput() */
3019 dquot_release(dquot);
3020 return PTR_ERR(handle);
3022 ret = dquot_release(dquot);
3023 err = ext4_journal_stop(handle);
3029 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3031 /* Are we journalling quotas? */
3032 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3033 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3034 dquot_mark_dquot_dirty(dquot);
3035 return ext4_write_dquot(dquot);
3037 return dquot_mark_dquot_dirty(dquot);
3041 static int ext4_write_info(struct super_block *sb, int type)
3046 /* Data block + inode block */
3047 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3049 return PTR_ERR(handle);
3050 ret = dquot_commit_info(sb, type);
3051 err = ext4_journal_stop(handle);
3058 * Turn on quotas during mount time - we need to find
3059 * the quota file and such...
3061 static int ext4_quota_on_mount(struct super_block *sb, int type)
3063 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3064 EXT4_SB(sb)->s_jquota_fmt, type);
3068 * Standard function to be called on quota_on
3070 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3074 struct nameidata nd;
3076 if (!test_opt(sb, QUOTA))
3078 /* Not journalling quota? */
3079 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3080 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
3081 return vfs_quota_on(sb, type, format_id, path);
3082 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3085 /* Quotafile not on the same filesystem? */
3086 if (nd.mnt->mnt_sb != sb) {
3090 /* Quotafile not of fs root? */
3091 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3093 "EXT4-fs: Quota file not on filesystem root. "
3094 "Journalled quota will not work.\n");
3096 return vfs_quota_on(sb, type, format_id, path);
3099 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3100 * acquiring the locks... As quota files are never truncated and quota code
3101 * itself serializes the operations (and noone else should touch the files)
3102 * we don't have to be afraid of races */
3103 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3104 size_t len, loff_t off)
3106 struct inode *inode = sb_dqopt(sb)->files[type];
3107 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3109 int offset = off & (sb->s_blocksize - 1);
3112 struct buffer_head *bh;
3113 loff_t i_size = i_size_read(inode);
3117 if (off+len > i_size)
3120 while (toread > 0) {
3121 tocopy = sb->s_blocksize - offset < toread ?
3122 sb->s_blocksize - offset : toread;
3123 bh = ext4_bread(NULL, inode, blk, 0, &err);
3126 if (!bh) /* A hole? */
3127 memset(data, 0, tocopy);
3129 memcpy(data, bh->b_data+offset, tocopy);
3139 /* Write to quotafile (we know the transaction is already started and has
3140 * enough credits) */
3141 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3142 const char *data, size_t len, loff_t off)
3144 struct inode *inode = sb_dqopt(sb)->files[type];
3145 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3147 int offset = off & (sb->s_blocksize - 1);
3149 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3150 size_t towrite = len;
3151 struct buffer_head *bh;
3152 handle_t *handle = journal_current_handle();
3155 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3156 " cancelled because transaction is not started.\n",
3157 (unsigned long long)off, (unsigned long long)len);
3160 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3161 while (towrite > 0) {
3162 tocopy = sb->s_blocksize - offset < towrite ?
3163 sb->s_blocksize - offset : towrite;
3164 bh = ext4_bread(handle, inode, blk, 1, &err);
3167 if (journal_quota) {
3168 err = ext4_journal_get_write_access(handle, bh);
3175 memcpy(bh->b_data+offset, data, tocopy);
3176 flush_dcache_page(bh->b_page);
3179 err = ext4_journal_dirty_metadata(handle, bh);
3181 /* Always do at least ordered writes for quotas */
3182 err = ext4_journal_dirty_data(handle, bh);
3183 mark_buffer_dirty(bh);
3196 if (inode->i_size < off+len-towrite) {
3197 i_size_write(inode, off+len-towrite);
3198 EXT4_I(inode)->i_disksize = inode->i_size;
3201 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3202 ext4_mark_inode_dirty(handle, inode);
3203 mutex_unlock(&inode->i_mutex);
3204 return len - towrite;
3209 static int ext4_get_sb(struct file_system_type *fs_type,
3210 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3212 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3215 static struct file_system_type ext4dev_fs_type = {
3216 .owner = THIS_MODULE,
3218 .get_sb = ext4_get_sb,
3219 .kill_sb = kill_block_super,
3220 .fs_flags = FS_REQUIRES_DEV,
3223 static int __init init_ext4_fs(void)
3225 int err = init_ext4_xattr();
3228 err = init_inodecache();
3231 err = register_filesystem(&ext4dev_fs_type);
3236 destroy_inodecache();
3242 static void __exit exit_ext4_fs(void)
3244 unregister_filesystem(&ext4dev_fs_type);
3245 destroy_inodecache();
3249 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3250 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3251 MODULE_LICENSE("GPL");
3252 module_init(init_ext4_fs)
3253 module_exit(exit_ext4_fs)