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 mutex_init(&ei->truncate_mutex);
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_CONT)) {
692 int def_errors = le16_to_cpu(es->s_errors);
694 if (def_errors == EXT4_ERRORS_PANIC ||
695 def_errors == EXT4_ERRORS_RO) {
696 seq_puts(seq, ",errors=continue");
699 if (test_opt(sb, ERRORS_RO))
700 seq_puts(seq, ",errors=remount-ro");
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_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
873 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
874 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
875 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
876 Opt_grpquota, Opt_extents, Opt_noextents,
879 static match_table_t tokens = {
880 {Opt_bsd_df, "bsddf"},
881 {Opt_minix_df, "minixdf"},
882 {Opt_grpid, "grpid"},
883 {Opt_grpid, "bsdgroups"},
884 {Opt_nogrpid, "nogrpid"},
885 {Opt_nogrpid, "sysvgroups"},
886 {Opt_resgid, "resgid=%u"},
887 {Opt_resuid, "resuid=%u"},
889 {Opt_err_cont, "errors=continue"},
890 {Opt_err_panic, "errors=panic"},
891 {Opt_err_ro, "errors=remount-ro"},
892 {Opt_nouid32, "nouid32"},
893 {Opt_nocheck, "nocheck"},
894 {Opt_nocheck, "check=none"},
895 {Opt_debug, "debug"},
896 {Opt_oldalloc, "oldalloc"},
897 {Opt_orlov, "orlov"},
898 {Opt_user_xattr, "user_xattr"},
899 {Opt_nouser_xattr, "nouser_xattr"},
901 {Opt_noacl, "noacl"},
902 {Opt_reservation, "reservation"},
903 {Opt_noreservation, "noreservation"},
904 {Opt_noload, "noload"},
907 {Opt_commit, "commit=%u"},
908 {Opt_journal_update, "journal=update"},
909 {Opt_journal_inum, "journal=%u"},
910 {Opt_journal_dev, "journal_dev=%u"},
911 {Opt_abort, "abort"},
912 {Opt_data_journal, "data=journal"},
913 {Opt_data_ordered, "data=ordered"},
914 {Opt_data_writeback, "data=writeback"},
915 {Opt_offusrjquota, "usrjquota="},
916 {Opt_usrjquota, "usrjquota=%s"},
917 {Opt_offgrpjquota, "grpjquota="},
918 {Opt_grpjquota, "grpjquota=%s"},
919 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
920 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
921 {Opt_grpquota, "grpquota"},
922 {Opt_noquota, "noquota"},
923 {Opt_quota, "quota"},
924 {Opt_usrquota, "usrquota"},
925 {Opt_barrier, "barrier=%u"},
926 {Opt_extents, "extents"},
927 {Opt_noextents, "noextents"},
929 {Opt_resize, "resize"},
932 static ext4_fsblk_t get_sb_block(void **data)
934 ext4_fsblk_t sb_block;
935 char *options = (char *) *data;
937 if (!options || strncmp(options, "sb=", 3) != 0)
938 return 1; /* Default location */
940 /*todo: use simple_strtoll with >32bit ext4 */
941 sb_block = simple_strtoul(options, &options, 0);
942 if (*options && *options != ',') {
943 printk("EXT4-fs: Invalid sb specification: %s\n",
949 *data = (void *) options;
953 static int parse_options (char *options, struct super_block *sb,
954 unsigned int *inum, unsigned long *journal_devnum,
955 ext4_fsblk_t *n_blocks_count, int is_remount)
957 struct ext4_sb_info *sbi = EXT4_SB(sb);
959 substring_t args[MAX_OPT_ARGS];
970 while ((p = strsep (&options, ",")) != NULL) {
975 token = match_token(p, tokens, args);
978 clear_opt (sbi->s_mount_opt, MINIX_DF);
981 set_opt (sbi->s_mount_opt, MINIX_DF);
984 set_opt (sbi->s_mount_opt, GRPID);
987 clear_opt (sbi->s_mount_opt, GRPID);
990 if (match_int(&args[0], &option))
992 sbi->s_resuid = option;
995 if (match_int(&args[0], &option))
997 sbi->s_resgid = option;
1000 /* handled by get_sb_block() instead of here */
1001 /* *sb_block = match_int(&args[0]); */
1004 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1005 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1006 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1009 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1010 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1011 set_opt (sbi->s_mount_opt, ERRORS_RO);
1014 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1015 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1016 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1019 set_opt (sbi->s_mount_opt, NO_UID32);
1022 clear_opt (sbi->s_mount_opt, CHECK);
1025 set_opt (sbi->s_mount_opt, DEBUG);
1028 set_opt (sbi->s_mount_opt, OLDALLOC);
1031 clear_opt (sbi->s_mount_opt, OLDALLOC);
1033 #ifdef CONFIG_EXT4DEV_FS_XATTR
1034 case Opt_user_xattr:
1035 set_opt (sbi->s_mount_opt, XATTR_USER);
1037 case Opt_nouser_xattr:
1038 clear_opt (sbi->s_mount_opt, XATTR_USER);
1041 case Opt_user_xattr:
1042 case Opt_nouser_xattr:
1043 printk("EXT4 (no)user_xattr options not supported\n");
1046 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1048 set_opt(sbi->s_mount_opt, POSIX_ACL);
1051 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1056 printk("EXT4 (no)acl options not supported\n");
1059 case Opt_reservation:
1060 set_opt(sbi->s_mount_opt, RESERVATION);
1062 case Opt_noreservation:
1063 clear_opt(sbi->s_mount_opt, RESERVATION);
1065 case Opt_journal_update:
1067 /* Eventually we will want to be able to create
1068 a journal file here. For now, only allow the
1069 user to specify an existing inode to be the
1072 printk(KERN_ERR "EXT4-fs: cannot specify "
1073 "journal on remount\n");
1076 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1078 case Opt_journal_inum:
1080 printk(KERN_ERR "EXT4-fs: cannot specify "
1081 "journal on remount\n");
1084 if (match_int(&args[0], &option))
1088 case Opt_journal_dev:
1090 printk(KERN_ERR "EXT4-fs: cannot specify "
1091 "journal on remount\n");
1094 if (match_int(&args[0], &option))
1096 *journal_devnum = option;
1099 set_opt (sbi->s_mount_opt, NOLOAD);
1102 if (match_int(&args[0], &option))
1107 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1108 sbi->s_commit_interval = HZ * option;
1110 case Opt_data_journal:
1111 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1113 case Opt_data_ordered:
1114 data_opt = EXT4_MOUNT_ORDERED_DATA;
1116 case Opt_data_writeback:
1117 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1120 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1123 "EXT4-fs: cannot change data "
1124 "mode on remount\n");
1128 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1129 sbi->s_mount_opt |= data_opt;
1139 if (sb_any_quota_enabled(sb)) {
1141 "EXT4-fs: Cannot change journalled "
1142 "quota options when quota turned on.\n");
1145 qname = match_strdup(&args[0]);
1148 "EXT4-fs: not enough memory for "
1149 "storing quotafile name.\n");
1152 if (sbi->s_qf_names[qtype] &&
1153 strcmp(sbi->s_qf_names[qtype], qname)) {
1155 "EXT4-fs: %s quota file already "
1156 "specified.\n", QTYPE2NAME(qtype));
1160 sbi->s_qf_names[qtype] = qname;
1161 if (strchr(sbi->s_qf_names[qtype], '/')) {
1163 "EXT4-fs: quotafile must be on "
1164 "filesystem root.\n");
1165 kfree(sbi->s_qf_names[qtype]);
1166 sbi->s_qf_names[qtype] = NULL;
1169 set_opt(sbi->s_mount_opt, QUOTA);
1171 case Opt_offusrjquota:
1174 case Opt_offgrpjquota:
1177 if (sb_any_quota_enabled(sb)) {
1178 printk(KERN_ERR "EXT4-fs: Cannot change "
1179 "journalled quota options when "
1180 "quota turned on.\n");
1184 * The space will be released later when all options
1185 * are confirmed to be correct
1187 sbi->s_qf_names[qtype] = NULL;
1189 case Opt_jqfmt_vfsold:
1190 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1192 case Opt_jqfmt_vfsv0:
1193 sbi->s_jquota_fmt = QFMT_VFS_V0;
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 set_opt(sbi->s_mount_opt, USRQUOTA);
1201 set_opt(sbi->s_mount_opt, QUOTA);
1202 set_opt(sbi->s_mount_opt, GRPQUOTA);
1205 if (sb_any_quota_enabled(sb)) {
1206 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1207 "options when quota turned on.\n");
1210 clear_opt(sbi->s_mount_opt, QUOTA);
1211 clear_opt(sbi->s_mount_opt, USRQUOTA);
1212 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1220 case Opt_offusrjquota:
1221 case Opt_offgrpjquota:
1222 case Opt_jqfmt_vfsold:
1223 case Opt_jqfmt_vfsv0:
1225 "EXT4-fs: journalled quota options not "
1232 set_opt(sbi->s_mount_opt, ABORT);
1235 if (match_int(&args[0], &option))
1238 set_opt(sbi->s_mount_opt, BARRIER);
1240 clear_opt(sbi->s_mount_opt, BARRIER);
1246 printk("EXT4-fs: resize option only available "
1250 if (match_int(&args[0], &option) != 0)
1252 *n_blocks_count = option;
1255 set_opt(sbi->s_mount_opt, NOBH);
1258 clear_opt(sbi->s_mount_opt, NOBH);
1261 set_opt (sbi->s_mount_opt, EXTENTS);
1264 clear_opt (sbi->s_mount_opt, EXTENTS);
1268 "EXT4-fs: Unrecognized mount option \"%s\" "
1269 "or missing value\n", p);
1274 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1275 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1276 sbi->s_qf_names[USRQUOTA])
1277 clear_opt(sbi->s_mount_opt, USRQUOTA);
1279 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1280 sbi->s_qf_names[GRPQUOTA])
1281 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1283 if ((sbi->s_qf_names[USRQUOTA] &&
1284 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1285 (sbi->s_qf_names[GRPQUOTA] &&
1286 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1287 printk(KERN_ERR "EXT4-fs: old and new quota "
1288 "format mixing.\n");
1292 if (!sbi->s_jquota_fmt) {
1293 printk(KERN_ERR "EXT4-fs: journalled quota format "
1294 "not specified.\n");
1298 if (sbi->s_jquota_fmt) {
1299 printk(KERN_ERR "EXT4-fs: journalled quota format "
1300 "specified with no journalling "
1309 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1312 struct ext4_sb_info *sbi = EXT4_SB(sb);
1315 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1316 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1317 "forcing read-only mode\n");
1322 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1323 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1324 "running e2fsck is recommended\n");
1325 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1326 printk (KERN_WARNING
1327 "EXT4-fs warning: mounting fs with errors, "
1328 "running e2fsck is recommended\n");
1329 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1330 le16_to_cpu(es->s_mnt_count) >=
1331 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1332 printk (KERN_WARNING
1333 "EXT4-fs warning: maximal mount count reached, "
1334 "running e2fsck is recommended\n");
1335 else if (le32_to_cpu(es->s_checkinterval) &&
1336 (le32_to_cpu(es->s_lastcheck) +
1337 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1338 printk (KERN_WARNING
1339 "EXT4-fs warning: checktime reached, "
1340 "running e2fsck is recommended\n");
1342 /* @@@ We _will_ want to clear the valid bit if we find
1343 * inconsistencies, to force a fsck at reboot. But for
1344 * a plain journaled filesystem we can keep it set as
1347 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
1349 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1350 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1351 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1352 es->s_mtime = cpu_to_le32(get_seconds());
1353 ext4_update_dynamic_rev(sb);
1354 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1356 ext4_commit_super(sb, es, 1);
1357 if (test_opt(sb, DEBUG))
1358 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1359 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1361 sbi->s_groups_count,
1362 EXT4_BLOCKS_PER_GROUP(sb),
1363 EXT4_INODES_PER_GROUP(sb),
1366 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1367 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1368 char b[BDEVNAME_SIZE];
1370 printk("external journal on %s\n",
1371 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1373 printk("internal journal\n");
1378 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1379 struct ext4_group_desc *gdp)
1383 if (sbi->s_es->s_feature_ro_compat &
1384 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1385 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1386 __le32 le_group = cpu_to_le32(block_group);
1388 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1389 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1390 crc = crc16(crc, (__u8 *)gdp, offset);
1391 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1392 /* for checksum of struct ext4_group_desc do the rest...*/
1393 if ((sbi->s_es->s_feature_incompat &
1394 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1395 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1396 crc = crc16(crc, (__u8 *)gdp + offset,
1397 le16_to_cpu(sbi->s_es->s_desc_size) -
1401 return cpu_to_le16(crc);
1404 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1405 struct ext4_group_desc *gdp)
1407 if ((sbi->s_es->s_feature_ro_compat &
1408 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1409 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1415 /* Called at mount-time, super-block is locked */
1416 static int ext4_check_descriptors (struct super_block * sb)
1418 struct ext4_sb_info *sbi = EXT4_SB(sb);
1419 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1420 ext4_fsblk_t last_block;
1421 ext4_fsblk_t block_bitmap;
1422 ext4_fsblk_t inode_bitmap;
1423 ext4_fsblk_t inode_table;
1424 struct ext4_group_desc * gdp = NULL;
1426 int flexbg_flag = 0;
1429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1432 ext4_debug ("Checking group descriptors");
1434 for (i = 0; i < sbi->s_groups_count; i++)
1436 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1437 last_block = ext4_blocks_count(sbi->s_es) - 1;
1439 last_block = first_block +
1440 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1442 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1443 gdp = (struct ext4_group_desc *)
1444 sbi->s_group_desc[desc_block++]->b_data;
1445 block_bitmap = ext4_block_bitmap(sb, gdp);
1446 if (block_bitmap < first_block || block_bitmap > last_block)
1448 ext4_error (sb, "ext4_check_descriptors",
1449 "Block bitmap for group %lu"
1450 " not in group (block %llu)!",
1454 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1455 if (inode_bitmap < first_block || inode_bitmap > last_block)
1457 ext4_error (sb, "ext4_check_descriptors",
1458 "Inode bitmap for group %lu"
1459 " not in group (block %llu)!",
1463 inode_table = ext4_inode_table(sb, gdp);
1464 if (inode_table < first_block ||
1465 inode_table + sbi->s_itb_per_group - 1 > last_block)
1467 ext4_error (sb, "ext4_check_descriptors",
1468 "Inode table for group %lu"
1469 " not in group (block %llu)!",
1473 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1474 ext4_error(sb, __FUNCTION__,
1475 "Checksum for group %lu failed (%u!=%u)\n",
1476 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1477 gdp)), le16_to_cpu(gdp->bg_checksum));
1481 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1482 gdp = (struct ext4_group_desc *)
1483 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
1486 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1487 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1491 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1492 * the superblock) which were deleted from all directories, but held open by
1493 * a process at the time of a crash. We walk the list and try to delete these
1494 * inodes at recovery time (only with a read-write filesystem).
1496 * In order to keep the orphan inode chain consistent during traversal (in
1497 * case of crash during recovery), we link each inode into the superblock
1498 * orphan list_head and handle it the same way as an inode deletion during
1499 * normal operation (which journals the operations for us).
1501 * We only do an iget() and an iput() on each inode, which is very safe if we
1502 * accidentally point at an in-use or already deleted inode. The worst that
1503 * can happen in this case is that we get a "bit already cleared" message from
1504 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1505 * e2fsck was run on this filesystem, and it must have already done the orphan
1506 * inode cleanup for us, so we can safely abort without any further action.
1508 static void ext4_orphan_cleanup (struct super_block * sb,
1509 struct ext4_super_block * es)
1511 unsigned int s_flags = sb->s_flags;
1512 int nr_orphans = 0, nr_truncates = 0;
1516 if (!es->s_last_orphan) {
1517 jbd_debug(4, "no orphan inodes to clean up\n");
1521 if (bdev_read_only(sb->s_bdev)) {
1522 printk(KERN_ERR "EXT4-fs: write access "
1523 "unavailable, skipping orphan cleanup.\n");
1527 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1528 if (es->s_last_orphan)
1529 jbd_debug(1, "Errors on filesystem, "
1530 "clearing orphan list.\n");
1531 es->s_last_orphan = 0;
1532 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1536 if (s_flags & MS_RDONLY) {
1537 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1539 sb->s_flags &= ~MS_RDONLY;
1542 /* Needed for iput() to work correctly and not trash data */
1543 sb->s_flags |= MS_ACTIVE;
1544 /* Turn on quotas so that they are updated correctly */
1545 for (i = 0; i < MAXQUOTAS; i++) {
1546 if (EXT4_SB(sb)->s_qf_names[i]) {
1547 int ret = ext4_quota_on_mount(sb, i);
1550 "EXT4-fs: Cannot turn on journalled "
1551 "quota: error %d\n", ret);
1556 while (es->s_last_orphan) {
1557 struct inode *inode;
1560 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
1561 es->s_last_orphan = 0;
1565 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1567 if (inode->i_nlink) {
1569 "%s: truncating inode %lu to %Ld bytes\n",
1570 __FUNCTION__, inode->i_ino, inode->i_size);
1571 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1572 inode->i_ino, inode->i_size);
1573 ext4_truncate(inode);
1577 "%s: deleting unreferenced inode %lu\n",
1578 __FUNCTION__, inode->i_ino);
1579 jbd_debug(2, "deleting unreferenced inode %lu\n",
1583 iput(inode); /* The delete magic happens here! */
1586 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1589 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1590 sb->s_id, PLURAL(nr_orphans));
1592 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1593 sb->s_id, PLURAL(nr_truncates));
1595 /* Turn quotas off */
1596 for (i = 0; i < MAXQUOTAS; i++) {
1597 if (sb_dqopt(sb)->files[i])
1598 vfs_quota_off(sb, i);
1601 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1604 * Maximal extent format file size.
1605 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1606 * extent format containers, within a sector_t, and within i_blocks
1607 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1608 * so that won't be a limiting factor.
1610 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1612 static loff_t ext4_max_size(int blkbits)
1615 loff_t upper_limit = MAX_LFS_FILESIZE;
1617 /* small i_blocks in vfs inode? */
1618 if (sizeof(blkcnt_t) < sizeof(u64)) {
1620 * CONFIG_LSF is not enabled implies the inode
1621 * i_block represent total blocks in 512 bytes
1622 * 32 == size of vfs inode i_blocks * 8
1624 upper_limit = (1LL << 32) - 1;
1626 /* total blocks in file system block size */
1627 upper_limit >>= (blkbits - 9);
1628 upper_limit <<= blkbits;
1631 /* 32-bit extent-start container, ee_block */
1636 /* Sanity check against vm- & vfs- imposed limits */
1637 if (res > upper_limit)
1644 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1645 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1646 * We need to be 1 filesystem block less than the 2^48 sector limit.
1648 static loff_t ext4_max_bitmap_size(int bits)
1650 loff_t res = EXT4_NDIR_BLOCKS;
1653 /* This is calculated to be the largest file size for a
1654 * dense, bitmapped file such that the total number of
1655 * sectors in the file, including data and all indirect blocks,
1656 * does not exceed 2^48 -1
1657 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1658 * total number of 512 bytes blocks of the file
1661 if (sizeof(blkcnt_t) < sizeof(u64)) {
1663 * CONFIG_LSF is not enabled implies the inode
1664 * i_block represent total blocks in 512 bytes
1665 * 32 == size of vfs inode i_blocks * 8
1667 upper_limit = (1LL << 32) - 1;
1669 /* total blocks in file system block size */
1670 upper_limit >>= (bits - 9);
1674 * We use 48 bit ext4_inode i_blocks
1675 * With EXT4_HUGE_FILE_FL set the i_blocks
1676 * represent total number of blocks in
1677 * file system block size
1679 upper_limit = (1LL << 48) - 1;
1683 /* indirect blocks */
1685 /* double indirect blocks */
1686 meta_blocks += 1 + (1LL << (bits-2));
1687 /* tripple indirect blocks */
1688 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1690 upper_limit -= meta_blocks;
1691 upper_limit <<= bits;
1693 res += 1LL << (bits-2);
1694 res += 1LL << (2*(bits-2));
1695 res += 1LL << (3*(bits-2));
1697 if (res > upper_limit)
1700 if (res > MAX_LFS_FILESIZE)
1701 res = MAX_LFS_FILESIZE;
1706 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1707 ext4_fsblk_t logical_sb_block, int nr)
1709 struct ext4_sb_info *sbi = EXT4_SB(sb);
1710 ext4_group_t bg, first_meta_bg;
1713 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1715 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1717 return logical_sb_block + nr + 1;
1718 bg = sbi->s_desc_per_block * nr;
1719 if (ext4_bg_has_super(sb, bg))
1721 return (has_super + ext4_group_first_block_no(sb, bg));
1725 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1726 __releases(kernel_sem)
1727 __acquires(kernel_sem)
1730 struct buffer_head * bh;
1731 struct ext4_super_block *es = NULL;
1732 struct ext4_sb_info *sbi;
1734 ext4_fsblk_t sb_block = get_sb_block(&data);
1735 ext4_fsblk_t logical_sb_block;
1736 unsigned long offset = 0;
1737 unsigned int journal_inum = 0;
1738 unsigned long journal_devnum = 0;
1739 unsigned long def_mount_opts;
1750 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1753 sb->s_fs_info = sbi;
1754 sbi->s_mount_opt = 0;
1755 sbi->s_resuid = EXT4_DEF_RESUID;
1756 sbi->s_resgid = EXT4_DEF_RESGID;
1757 sbi->s_sb_block = sb_block;
1761 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1763 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1767 if (!sb_set_blocksize(sb, blocksize)) {
1768 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1773 * The ext4 superblock will not be buffer aligned for other than 1kB
1774 * block sizes. We need to calculate the offset from buffer start.
1776 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1777 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1778 offset = do_div(logical_sb_block, blocksize);
1780 logical_sb_block = sb_block;
1783 if (!(bh = sb_bread(sb, logical_sb_block))) {
1784 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1788 * Note: s_es must be initialized as soon as possible because
1789 * some ext4 macro-instructions depend on its value
1791 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1793 sb->s_magic = le16_to_cpu(es->s_magic);
1794 if (sb->s_magic != EXT4_SUPER_MAGIC)
1797 /* Set defaults before we parse the mount options */
1798 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1799 if (def_mount_opts & EXT4_DEFM_DEBUG)
1800 set_opt(sbi->s_mount_opt, DEBUG);
1801 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1802 set_opt(sbi->s_mount_opt, GRPID);
1803 if (def_mount_opts & EXT4_DEFM_UID16)
1804 set_opt(sbi->s_mount_opt, NO_UID32);
1805 #ifdef CONFIG_EXT4DEV_FS_XATTR
1806 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1807 set_opt(sbi->s_mount_opt, XATTR_USER);
1809 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1810 if (def_mount_opts & EXT4_DEFM_ACL)
1811 set_opt(sbi->s_mount_opt, POSIX_ACL);
1813 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1814 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1815 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1816 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1817 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1818 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1820 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1821 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1822 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
1823 set_opt(sbi->s_mount_opt, ERRORS_RO);
1825 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1827 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1828 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1830 set_opt(sbi->s_mount_opt, RESERVATION);
1833 * turn on extents feature by default in ext4 filesystem
1834 * User -o noextents to turn it off
1836 set_opt(sbi->s_mount_opt, EXTENTS);
1838 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1842 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1843 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1845 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1846 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1847 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1848 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1850 "EXT4-fs warning: feature flags set on rev 0 fs, "
1851 "running e2fsck is recommended\n");
1853 * Check feature flags regardless of the revision level, since we
1854 * previously didn't change the revision level when setting the flags,
1855 * so there is a chance incompat flags are set on a rev 0 filesystem.
1857 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1859 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1860 "unsupported optional features (%x).\n",
1861 sb->s_id, le32_to_cpu(features));
1864 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1865 if (!(sb->s_flags & MS_RDONLY) && features) {
1866 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1867 "unsupported optional features (%x).\n",
1868 sb->s_id, le32_to_cpu(features));
1871 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1873 * Large file size enabled file system can only be
1874 * mount if kernel is build with CONFIG_LSF
1876 if (sizeof(root->i_blocks) < sizeof(u64) &&
1877 !(sb->s_flags & MS_RDONLY)) {
1878 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1879 "files cannot be mounted read-write "
1880 "without CONFIG_LSF.\n", sb->s_id);
1884 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1886 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1887 blocksize > EXT4_MAX_BLOCK_SIZE) {
1889 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1890 blocksize, sb->s_id);
1894 hblock = bdev_hardsect_size(sb->s_bdev);
1895 if (sb->s_blocksize != blocksize) {
1897 * Make sure the blocksize for the filesystem is larger
1898 * than the hardware sectorsize for the machine.
1900 if (blocksize < hblock) {
1901 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
1902 "device blocksize %d.\n", blocksize, hblock);
1907 sb_set_blocksize(sb, blocksize);
1908 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1909 offset = do_div(logical_sb_block, blocksize);
1910 bh = sb_bread(sb, logical_sb_block);
1913 "EXT4-fs: Can't read superblock on 2nd try.\n");
1916 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
1918 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
1920 "EXT4-fs: Magic mismatch, very weird !\n");
1925 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
1926 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
1928 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1929 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1930 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
1932 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1933 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1934 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1935 (!is_power_of_2(sbi->s_inode_size)) ||
1936 (sbi->s_inode_size > blocksize)) {
1938 "EXT4-fs: unsupported inode size: %d\n",
1942 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1943 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1945 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1946 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
1947 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
1948 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1949 !is_power_of_2(sbi->s_desc_size)) {
1951 "EXT4-fs: unsupported descriptor size %lu\n",
1956 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
1957 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1958 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1959 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
1961 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
1962 if (sbi->s_inodes_per_block == 0)
1964 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1965 sbi->s_inodes_per_block;
1966 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
1968 sbi->s_mount_state = le16_to_cpu(es->s_state);
1969 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1970 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
1971 for (i=0; i < 4; i++)
1972 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1973 sbi->s_def_hash_version = es->s_def_hash_version;
1975 if (sbi->s_blocks_per_group > blocksize * 8) {
1977 "EXT4-fs: #blocks per group too big: %lu\n",
1978 sbi->s_blocks_per_group);
1981 if (sbi->s_inodes_per_group > blocksize * 8) {
1983 "EXT4-fs: #inodes per group too big: %lu\n",
1984 sbi->s_inodes_per_group);
1988 if (ext4_blocks_count(es) >
1989 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1990 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
1991 " too large to mount safely\n", sb->s_id);
1992 if (sizeof(sector_t) < 8)
1993 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
1998 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2000 blocks_count = (ext4_blocks_count(es) -
2001 le32_to_cpu(es->s_first_data_block) +
2002 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2003 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2004 sbi->s_groups_count = blocks_count;
2005 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2006 EXT4_DESC_PER_BLOCK(sb);
2007 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2009 if (sbi->s_group_desc == NULL) {
2010 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2014 bgl_lock_init(&sbi->s_blockgroup_lock);
2016 for (i = 0; i < db_count; i++) {
2017 block = descriptor_loc(sb, logical_sb_block, i);
2018 sbi->s_group_desc[i] = sb_bread(sb, block);
2019 if (!sbi->s_group_desc[i]) {
2020 printk (KERN_ERR "EXT4-fs: "
2021 "can't read group descriptor %d\n", i);
2026 if (!ext4_check_descriptors (sb)) {
2027 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2030 sbi->s_gdb_count = db_count;
2031 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2032 spin_lock_init(&sbi->s_next_gen_lock);
2034 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2035 ext4_count_free_blocks(sb));
2037 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2038 ext4_count_free_inodes(sb));
2041 err = percpu_counter_init(&sbi->s_dirs_counter,
2042 ext4_count_dirs(sb));
2045 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2049 /* per fileystem reservation list head & lock */
2050 spin_lock_init(&sbi->s_rsv_window_lock);
2051 sbi->s_rsv_window_root = RB_ROOT;
2052 /* Add a single, static dummy reservation to the start of the
2053 * reservation window list --- it gives us a placeholder for
2054 * append-at-start-of-list which makes the allocation logic
2055 * _much_ simpler. */
2056 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2057 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2058 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2059 sbi->s_rsv_window_head.rsv_goal_size = 0;
2060 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2063 * set up enough so that it can read an inode
2065 sb->s_op = &ext4_sops;
2066 sb->s_export_op = &ext4_export_ops;
2067 sb->s_xattr = ext4_xattr_handlers;
2069 sb->s_qcop = &ext4_qctl_operations;
2070 sb->dq_op = &ext4_quota_operations;
2072 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2076 needs_recovery = (es->s_last_orphan != 0 ||
2077 EXT4_HAS_INCOMPAT_FEATURE(sb,
2078 EXT4_FEATURE_INCOMPAT_RECOVER));
2081 * The first inode we look at is the journal inode. Don't try
2082 * root first: it may be modified in the journal!
2084 if (!test_opt(sb, NOLOAD) &&
2085 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2086 if (ext4_load_journal(sb, es, journal_devnum))
2088 } else if (journal_inum) {
2089 if (ext4_create_journal(sb, es, journal_inum))
2094 "ext4: No journal on filesystem on %s\n",
2099 if (ext4_blocks_count(es) > 0xffffffffULL &&
2100 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2101 JBD2_FEATURE_INCOMPAT_64BIT)) {
2102 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2106 /* We have now updated the journal if required, so we can
2107 * validate the data journaling mode. */
2108 switch (test_opt(sb, DATA_FLAGS)) {
2110 /* No mode set, assume a default based on the journal
2111 * capabilities: ORDERED_DATA if the journal can
2112 * cope, else JOURNAL_DATA
2114 if (jbd2_journal_check_available_features
2115 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2116 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2118 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2121 case EXT4_MOUNT_ORDERED_DATA:
2122 case EXT4_MOUNT_WRITEBACK_DATA:
2123 if (!jbd2_journal_check_available_features
2124 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2125 printk(KERN_ERR "EXT4-fs: Journal does not support "
2126 "requested data journaling mode\n");
2133 if (test_opt(sb, NOBH)) {
2134 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2135 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2136 "its supported only with writeback mode\n");
2137 clear_opt(sbi->s_mount_opt, NOBH);
2141 * The jbd2_journal_load will have done any necessary log recovery,
2142 * so we can safely mount the rest of the filesystem now.
2145 root = iget(sb, EXT4_ROOT_INO);
2146 sb->s_root = d_alloc_root(root);
2148 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2152 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2155 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2159 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2161 /* determine the minimum size of new large inodes, if present */
2162 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2163 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2164 EXT4_GOOD_OLD_INODE_SIZE;
2165 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2166 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2167 if (sbi->s_want_extra_isize <
2168 le16_to_cpu(es->s_want_extra_isize))
2169 sbi->s_want_extra_isize =
2170 le16_to_cpu(es->s_want_extra_isize);
2171 if (sbi->s_want_extra_isize <
2172 le16_to_cpu(es->s_min_extra_isize))
2173 sbi->s_want_extra_isize =
2174 le16_to_cpu(es->s_min_extra_isize);
2177 /* Check if enough inode space is available */
2178 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2179 sbi->s_inode_size) {
2180 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2181 EXT4_GOOD_OLD_INODE_SIZE;
2182 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2187 * akpm: core read_super() calls in here with the superblock locked.
2188 * That deadlocks, because orphan cleanup needs to lock the superblock
2189 * in numerous places. Here we just pop the lock - it's relatively
2190 * harmless, because we are now ready to accept write_super() requests,
2191 * and aviro says that's the only reason for hanging onto the
2194 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2195 ext4_orphan_cleanup(sb, es);
2196 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2198 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2199 ext4_mark_recovery_complete(sb, es);
2200 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2201 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2202 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2212 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2217 jbd2_journal_destroy(sbi->s_journal);
2219 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2220 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2221 percpu_counter_destroy(&sbi->s_dirs_counter);
2223 for (i = 0; i < db_count; i++)
2224 brelse(sbi->s_group_desc[i]);
2225 kfree(sbi->s_group_desc);
2228 for (i = 0; i < MAXQUOTAS; i++)
2229 kfree(sbi->s_qf_names[i]);
2231 ext4_blkdev_remove(sbi);
2234 sb->s_fs_info = NULL;
2241 * Setup any per-fs journal parameters now. We'll do this both on
2242 * initial mount, once the journal has been initialised but before we've
2243 * done any recovery; and again on any subsequent remount.
2245 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2247 struct ext4_sb_info *sbi = EXT4_SB(sb);
2249 if (sbi->s_commit_interval)
2250 journal->j_commit_interval = sbi->s_commit_interval;
2251 /* We could also set up an ext4-specific default for the commit
2252 * interval here, but for now we'll just fall back to the jbd
2255 spin_lock(&journal->j_state_lock);
2256 if (test_opt(sb, BARRIER))
2257 journal->j_flags |= JBD2_BARRIER;
2259 journal->j_flags &= ~JBD2_BARRIER;
2260 spin_unlock(&journal->j_state_lock);
2263 static journal_t *ext4_get_journal(struct super_block *sb,
2264 unsigned int journal_inum)
2266 struct inode *journal_inode;
2269 /* First, test for the existence of a valid inode on disk. Bad
2270 * things happen if we iget() an unused inode, as the subsequent
2271 * iput() will try to delete it. */
2273 journal_inode = iget(sb, journal_inum);
2274 if (!journal_inode) {
2275 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2278 if (!journal_inode->i_nlink) {
2279 make_bad_inode(journal_inode);
2280 iput(journal_inode);
2281 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2285 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2286 journal_inode, journal_inode->i_size);
2287 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
2288 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2289 iput(journal_inode);
2293 journal = jbd2_journal_init_inode(journal_inode);
2295 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2296 iput(journal_inode);
2299 journal->j_private = sb;
2300 ext4_init_journal_params(sb, journal);
2304 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2307 struct buffer_head * bh;
2311 int hblock, blocksize;
2312 ext4_fsblk_t sb_block;
2313 unsigned long offset;
2314 struct ext4_super_block * es;
2315 struct block_device *bdev;
2317 bdev = ext4_blkdev_get(j_dev);
2321 if (bd_claim(bdev, sb)) {
2323 "EXT4: failed to claim external journal device.\n");
2328 blocksize = sb->s_blocksize;
2329 hblock = bdev_hardsect_size(bdev);
2330 if (blocksize < hblock) {
2332 "EXT4-fs: blocksize too small for journal device.\n");
2336 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2337 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2338 set_blocksize(bdev, blocksize);
2339 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2340 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2341 "external journal\n");
2345 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2346 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2347 !(le32_to_cpu(es->s_feature_incompat) &
2348 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2349 printk(KERN_ERR "EXT4-fs: external journal has "
2350 "bad superblock\n");
2355 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2356 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2361 len = ext4_blocks_count(es);
2362 start = sb_block + 1;
2363 brelse(bh); /* we're done with the superblock */
2365 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2366 start, len, blocksize);
2368 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2371 journal->j_private = sb;
2372 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2373 wait_on_buffer(journal->j_sb_buffer);
2374 if (!buffer_uptodate(journal->j_sb_buffer)) {
2375 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2378 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2379 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2380 "user (unsupported) - %d\n",
2381 be32_to_cpu(journal->j_superblock->s_nr_users));
2384 EXT4_SB(sb)->journal_bdev = bdev;
2385 ext4_init_journal_params(sb, journal);
2388 jbd2_journal_destroy(journal);
2390 ext4_blkdev_put(bdev);
2394 static int ext4_load_journal(struct super_block *sb,
2395 struct ext4_super_block *es,
2396 unsigned long journal_devnum)
2399 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2402 int really_read_only;
2404 if (journal_devnum &&
2405 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2406 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2407 "numbers have changed\n");
2408 journal_dev = new_decode_dev(journal_devnum);
2410 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2412 really_read_only = bdev_read_only(sb->s_bdev);
2415 * Are we loading a blank journal or performing recovery after a
2416 * crash? For recovery, we need to check in advance whether we
2417 * can get read-write access to the device.
2420 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2421 if (sb->s_flags & MS_RDONLY) {
2422 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2423 "required on readonly filesystem.\n");
2424 if (really_read_only) {
2425 printk(KERN_ERR "EXT4-fs: write access "
2426 "unavailable, cannot proceed.\n");
2429 printk (KERN_INFO "EXT4-fs: write access will "
2430 "be enabled during recovery.\n");
2434 if (journal_inum && journal_dev) {
2435 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2436 "and inode journals!\n");
2441 if (!(journal = ext4_get_journal(sb, journal_inum)))
2444 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2448 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2449 err = jbd2_journal_update_format(journal);
2451 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2452 jbd2_journal_destroy(journal);
2457 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2458 err = jbd2_journal_wipe(journal, !really_read_only);
2460 err = jbd2_journal_load(journal);
2463 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2464 jbd2_journal_destroy(journal);
2468 EXT4_SB(sb)->s_journal = journal;
2469 ext4_clear_journal_err(sb, es);
2471 if (journal_devnum &&
2472 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2473 es->s_journal_dev = cpu_to_le32(journal_devnum);
2476 /* Make sure we flush the recovery flag to disk. */
2477 ext4_commit_super(sb, es, 1);
2483 static int ext4_create_journal(struct super_block * sb,
2484 struct ext4_super_block * es,
2485 unsigned int journal_inum)
2490 if (sb->s_flags & MS_RDONLY) {
2491 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2492 "create journal.\n");
2496 journal = ext4_get_journal(sb, journal_inum);
2500 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2503 err = jbd2_journal_create(journal);
2505 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2506 jbd2_journal_destroy(journal);
2510 EXT4_SB(sb)->s_journal = journal;
2512 ext4_update_dynamic_rev(sb);
2513 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2514 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2516 es->s_journal_inum = cpu_to_le32(journal_inum);
2519 /* Make sure we flush the recovery flag to disk. */
2520 ext4_commit_super(sb, es, 1);
2525 static void ext4_commit_super (struct super_block * sb,
2526 struct ext4_super_block * es,
2529 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2533 es->s_wtime = cpu_to_le32(get_seconds());
2534 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2535 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2536 BUFFER_TRACE(sbh, "marking dirty");
2537 mark_buffer_dirty(sbh);
2539 sync_dirty_buffer(sbh);
2544 * Have we just finished recovery? If so, and if we are mounting (or
2545 * remounting) the filesystem readonly, then we will end up with a
2546 * consistent fs on disk. Record that fact.
2548 static void ext4_mark_recovery_complete(struct super_block * sb,
2549 struct ext4_super_block * es)
2551 journal_t *journal = EXT4_SB(sb)->s_journal;
2553 jbd2_journal_lock_updates(journal);
2554 jbd2_journal_flush(journal);
2556 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2557 sb->s_flags & MS_RDONLY) {
2558 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2560 ext4_commit_super(sb, es, 1);
2563 jbd2_journal_unlock_updates(journal);
2567 * If we are mounting (or read-write remounting) a filesystem whose journal
2568 * has recorded an error from a previous lifetime, move that error to the
2569 * main filesystem now.
2571 static void ext4_clear_journal_err(struct super_block * sb,
2572 struct ext4_super_block * es)
2578 journal = EXT4_SB(sb)->s_journal;
2581 * Now check for any error status which may have been recorded in the
2582 * journal by a prior ext4_error() or ext4_abort()
2585 j_errno = jbd2_journal_errno(journal);
2589 errstr = ext4_decode_error(sb, j_errno, nbuf);
2590 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
2591 "from previous mount: %s", errstr);
2592 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
2593 "filesystem check.");
2595 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2596 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2597 ext4_commit_super (sb, es, 1);
2599 jbd2_journal_clear_err(journal);
2604 * Force the running and committing transactions to commit,
2605 * and wait on the commit.
2607 int ext4_force_commit(struct super_block *sb)
2612 if (sb->s_flags & MS_RDONLY)
2615 journal = EXT4_SB(sb)->s_journal;
2617 ret = ext4_journal_force_commit(journal);
2622 * Ext4 always journals updates to the superblock itself, so we don't
2623 * have to propagate any other updates to the superblock on disk at this
2624 * point. Just start an async writeback to get the buffers on their way
2627 * This implicitly triggers the writebehind on sync().
2630 static void ext4_write_super (struct super_block * sb)
2632 if (mutex_trylock(&sb->s_lock) != 0)
2637 static int ext4_sync_fs(struct super_block *sb, int wait)
2642 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2644 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2650 * LVM calls this function before a (read-only) snapshot is created. This
2651 * gives us a chance to flush the journal completely and mark the fs clean.
2653 static void ext4_write_super_lockfs(struct super_block *sb)
2657 if (!(sb->s_flags & MS_RDONLY)) {
2658 journal_t *journal = EXT4_SB(sb)->s_journal;
2660 /* Now we set up the journal barrier. */
2661 jbd2_journal_lock_updates(journal);
2662 jbd2_journal_flush(journal);
2664 /* Journal blocked and flushed, clear needs_recovery flag. */
2665 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2666 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2671 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2672 * flag here, even though the filesystem is not technically dirty yet.
2674 static void ext4_unlockfs(struct super_block *sb)
2676 if (!(sb->s_flags & MS_RDONLY)) {
2678 /* Reser the needs_recovery flag before the fs is unlocked. */
2679 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2680 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2682 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2686 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2688 struct ext4_super_block * es;
2689 struct ext4_sb_info *sbi = EXT4_SB(sb);
2690 ext4_fsblk_t n_blocks_count = 0;
2691 unsigned long old_sb_flags;
2692 struct ext4_mount_options old_opts;
2698 /* Store the original options */
2699 old_sb_flags = sb->s_flags;
2700 old_opts.s_mount_opt = sbi->s_mount_opt;
2701 old_opts.s_resuid = sbi->s_resuid;
2702 old_opts.s_resgid = sbi->s_resgid;
2703 old_opts.s_commit_interval = sbi->s_commit_interval;
2705 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2706 for (i = 0; i < MAXQUOTAS; i++)
2707 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2711 * Allow the "check" option to be passed as a remount option.
2713 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2718 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2719 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
2721 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2722 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2726 ext4_init_journal_params(sb, sbi->s_journal);
2728 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2729 n_blocks_count > ext4_blocks_count(es)) {
2730 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2735 if (*flags & MS_RDONLY) {
2737 * First of all, the unconditional stuff we have to do
2738 * to disable replay of the journal when we next remount
2740 sb->s_flags |= MS_RDONLY;
2743 * OK, test if we are remounting a valid rw partition
2744 * readonly, and if so set the rdonly flag and then
2745 * mark the partition as valid again.
2747 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2748 (sbi->s_mount_state & EXT4_VALID_FS))
2749 es->s_state = cpu_to_le16(sbi->s_mount_state);
2752 * We have to unlock super so that we can wait for
2756 ext4_mark_recovery_complete(sb, es);
2760 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2761 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2762 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2763 "remount RDWR because of unsupported "
2764 "optional features (%x).\n",
2765 sb->s_id, le32_to_cpu(ret));
2771 * If we have an unprocessed orphan list hanging
2772 * around from a previously readonly bdev mount,
2773 * require a full umount/remount for now.
2775 if (es->s_last_orphan) {
2776 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2777 "remount RDWR because of unprocessed "
2778 "orphan inode list. Please "
2779 "umount/remount instead.\n",
2786 * Mounting a RDONLY partition read-write, so reread
2787 * and store the current valid flag. (It may have
2788 * been changed by e2fsck since we originally mounted
2791 ext4_clear_journal_err(sb, es);
2792 sbi->s_mount_state = le16_to_cpu(es->s_state);
2793 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2795 if (!ext4_setup_super (sb, es, 0))
2796 sb->s_flags &= ~MS_RDONLY;
2800 /* Release old quota file names */
2801 for (i = 0; i < MAXQUOTAS; i++)
2802 if (old_opts.s_qf_names[i] &&
2803 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2804 kfree(old_opts.s_qf_names[i]);
2808 sb->s_flags = old_sb_flags;
2809 sbi->s_mount_opt = old_opts.s_mount_opt;
2810 sbi->s_resuid = old_opts.s_resuid;
2811 sbi->s_resgid = old_opts.s_resgid;
2812 sbi->s_commit_interval = old_opts.s_commit_interval;
2814 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2815 for (i = 0; i < MAXQUOTAS; i++) {
2816 if (sbi->s_qf_names[i] &&
2817 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2818 kfree(sbi->s_qf_names[i]);
2819 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2825 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2827 struct super_block *sb = dentry->d_sb;
2828 struct ext4_sb_info *sbi = EXT4_SB(sb);
2829 struct ext4_super_block *es = sbi->s_es;
2832 if (test_opt(sb, MINIX_DF)) {
2833 sbi->s_overhead_last = 0;
2834 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2835 ext4_group_t ngroups = sbi->s_groups_count, i;
2836 ext4_fsblk_t overhead = 0;
2840 * Compute the overhead (FS structures). This is constant
2841 * for a given filesystem unless the number of block groups
2842 * changes so we cache the previous value until it does.
2846 * All of the blocks before first_data_block are
2849 overhead = le32_to_cpu(es->s_first_data_block);
2852 * Add the overhead attributed to the superblock and
2853 * block group descriptors. If the sparse superblocks
2854 * feature is turned on, then not all groups have this.
2856 for (i = 0; i < ngroups; i++) {
2857 overhead += ext4_bg_has_super(sb, i) +
2858 ext4_bg_num_gdb(sb, i);
2863 * Every block group has an inode bitmap, a block
2864 * bitmap, and an inode table.
2866 overhead += ngroups * (2 + sbi->s_itb_per_group);
2867 sbi->s_overhead_last = overhead;
2869 sbi->s_blocks_last = ext4_blocks_count(es);
2872 buf->f_type = EXT4_SUPER_MAGIC;
2873 buf->f_bsize = sb->s_blocksize;
2874 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
2875 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2876 ext4_free_blocks_count_set(es, buf->f_bfree);
2877 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2878 if (buf->f_bfree < ext4_r_blocks_count(es))
2880 buf->f_files = le32_to_cpu(es->s_inodes_count);
2881 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2882 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2883 buf->f_namelen = EXT4_NAME_LEN;
2884 fsid = le64_to_cpup((void *)es->s_uuid) ^
2885 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2886 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2887 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2891 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2892 * is locked for write. Otherwise the are possible deadlocks:
2893 * Process 1 Process 2
2894 * ext4_create() quota_sync()
2895 * jbd2_journal_start() write_dquot()
2896 * DQUOT_INIT() down(dqio_mutex)
2897 * down(dqio_mutex) jbd2_journal_start()
2903 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2905 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2908 static int ext4_dquot_initialize(struct inode *inode, int type)
2913 /* We may create quota structure so we need to reserve enough blocks */
2914 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
2916 return PTR_ERR(handle);
2917 ret = dquot_initialize(inode, type);
2918 err = ext4_journal_stop(handle);
2924 static int ext4_dquot_drop(struct inode *inode)
2929 /* We may delete quota structure so we need to reserve enough blocks */
2930 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2932 return PTR_ERR(handle);
2933 ret = dquot_drop(inode);
2934 err = ext4_journal_stop(handle);
2940 static int ext4_write_dquot(struct dquot *dquot)
2944 struct inode *inode;
2946 inode = dquot_to_inode(dquot);
2947 handle = ext4_journal_start(inode,
2948 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2950 return PTR_ERR(handle);
2951 ret = dquot_commit(dquot);
2952 err = ext4_journal_stop(handle);
2958 static int ext4_acquire_dquot(struct dquot *dquot)
2963 handle = ext4_journal_start(dquot_to_inode(dquot),
2964 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2966 return PTR_ERR(handle);
2967 ret = dquot_acquire(dquot);
2968 err = ext4_journal_stop(handle);
2974 static int ext4_release_dquot(struct dquot *dquot)
2979 handle = ext4_journal_start(dquot_to_inode(dquot),
2980 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2981 if (IS_ERR(handle)) {
2982 /* Release dquot anyway to avoid endless cycle in dqput() */
2983 dquot_release(dquot);
2984 return PTR_ERR(handle);
2986 ret = dquot_release(dquot);
2987 err = ext4_journal_stop(handle);
2993 static int ext4_mark_dquot_dirty(struct dquot *dquot)
2995 /* Are we journalling quotas? */
2996 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2997 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2998 dquot_mark_dquot_dirty(dquot);
2999 return ext4_write_dquot(dquot);
3001 return dquot_mark_dquot_dirty(dquot);
3005 static int ext4_write_info(struct super_block *sb, int type)
3010 /* Data block + inode block */
3011 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3013 return PTR_ERR(handle);
3014 ret = dquot_commit_info(sb, type);
3015 err = ext4_journal_stop(handle);
3022 * Turn on quotas during mount time - we need to find
3023 * the quota file and such...
3025 static int ext4_quota_on_mount(struct super_block *sb, int type)
3027 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3028 EXT4_SB(sb)->s_jquota_fmt, type);
3032 * Standard function to be called on quota_on
3034 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3038 struct nameidata nd;
3040 if (!test_opt(sb, QUOTA))
3042 /* Not journalling quota? */
3043 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3044 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
3045 return vfs_quota_on(sb, type, format_id, path);
3046 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3049 /* Quotafile not on the same filesystem? */
3050 if (nd.mnt->mnt_sb != sb) {
3054 /* Quotafile not of fs root? */
3055 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3057 "EXT4-fs: Quota file not on filesystem root. "
3058 "Journalled quota will not work.\n");
3060 return vfs_quota_on(sb, type, format_id, path);
3063 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3064 * acquiring the locks... As quota files are never truncated and quota code
3065 * itself serializes the operations (and noone else should touch the files)
3066 * we don't have to be afraid of races */
3067 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3068 size_t len, loff_t off)
3070 struct inode *inode = sb_dqopt(sb)->files[type];
3071 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3073 int offset = off & (sb->s_blocksize - 1);
3076 struct buffer_head *bh;
3077 loff_t i_size = i_size_read(inode);
3081 if (off+len > i_size)
3084 while (toread > 0) {
3085 tocopy = sb->s_blocksize - offset < toread ?
3086 sb->s_blocksize - offset : toread;
3087 bh = ext4_bread(NULL, inode, blk, 0, &err);
3090 if (!bh) /* A hole? */
3091 memset(data, 0, tocopy);
3093 memcpy(data, bh->b_data+offset, tocopy);
3103 /* Write to quotafile (we know the transaction is already started and has
3104 * enough credits) */
3105 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3106 const char *data, size_t len, loff_t off)
3108 struct inode *inode = sb_dqopt(sb)->files[type];
3109 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3111 int offset = off & (sb->s_blocksize - 1);
3113 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3114 size_t towrite = len;
3115 struct buffer_head *bh;
3116 handle_t *handle = journal_current_handle();
3119 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3120 " cancelled because transaction is not started.\n",
3121 (unsigned long long)off, (unsigned long long)len);
3124 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3125 while (towrite > 0) {
3126 tocopy = sb->s_blocksize - offset < towrite ?
3127 sb->s_blocksize - offset : towrite;
3128 bh = ext4_bread(handle, inode, blk, 1, &err);
3131 if (journal_quota) {
3132 err = ext4_journal_get_write_access(handle, bh);
3139 memcpy(bh->b_data+offset, data, tocopy);
3140 flush_dcache_page(bh->b_page);
3143 err = ext4_journal_dirty_metadata(handle, bh);
3145 /* Always do at least ordered writes for quotas */
3146 err = ext4_journal_dirty_data(handle, bh);
3147 mark_buffer_dirty(bh);
3160 if (inode->i_size < off+len-towrite) {
3161 i_size_write(inode, off+len-towrite);
3162 EXT4_I(inode)->i_disksize = inode->i_size;
3165 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3166 ext4_mark_inode_dirty(handle, inode);
3167 mutex_unlock(&inode->i_mutex);
3168 return len - towrite;
3173 static int ext4_get_sb(struct file_system_type *fs_type,
3174 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3176 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3179 static struct file_system_type ext4dev_fs_type = {
3180 .owner = THIS_MODULE,
3182 .get_sb = ext4_get_sb,
3183 .kill_sb = kill_block_super,
3184 .fs_flags = FS_REQUIRES_DEV,
3187 static int __init init_ext4_fs(void)
3189 int err = init_ext4_xattr();
3192 err = init_inodecache();
3195 err = register_filesystem(&ext4dev_fs_type);
3200 destroy_inodecache();
3206 static void __exit exit_ext4_fs(void)
3208 unregister_filesystem(&ext4dev_fs_type);
3209 destroy_inodecache();
3213 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3214 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3215 MODULE_LICENSE("GPL");
3216 module_init(init_ext4_fs)
3217 module_exit(exit_ext4_fs)