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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
195 handle = (handle_t *)ref_cnt;
197 current->journal_info = handle;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
205 unsigned long ref_cnt = (unsigned long)handle;
207 BUG_ON(ref_cnt == 0);
210 handle = (handle_t *)ref_cnt;
212 current->journal_info = handle;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
239 return jbd2_journal_start(journal, nblocks);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where, handle_t *handle)
252 struct super_block *sb;
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
260 sb = handle->h_transaction->t_journal->j_private;
262 rc = jbd2_journal_stop(handle);
267 __ext4_std_error(sb, where, err);
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
277 BUG_ON(!ext4_handle_valid(handle));
280 BUFFER_TRACE(bh, "abort");
285 if (is_handle_aborted(handle))
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
291 jbd2_journal_abort_handle(handle);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block *sb)
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316 if (sb->s_flags & MS_RDONLY)
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
324 jbd2_journal_abort(journal, -EIO);
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
347 ext4_handle_error(sb);
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
357 errstr = "IO failure";
360 errstr = "Out of memory";
363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 errstr = "Journal has aborted";
367 errstr = "Readonly filesystem";
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
399 errstr = ext4_decode_error(sb, errno, nbuf);
400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
403 ext4_handle_error(sb);
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
416 void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
427 if (test_opt(sb, ERRORS_PANIC))
428 panic("EXT4-fs panic from previous error\n");
430 if (sb->s_flags & MS_RDONLY)
433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 sb->s_flags |= MS_RDONLY;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
453 void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 const char *function, const char *fmt, ...)
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 ext4_commit_super(sb, 0);
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
499 ext4_lock_group(sb, grp);
503 void ext4_update_dynamic_rev(struct super_block *sb)
505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
510 ext4_warning(sb, __func__,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
529 * Open the external journal device
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 __bdevname(dev, b), PTR_ERR(bdev));
548 * Release the journal device
550 static int ext4_blkdev_put(struct block_device *bdev)
553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
558 struct block_device *bdev;
561 bdev = sbi->journal_bdev;
563 ret = ext4_blkdev_put(bdev);
564 sbi->journal_bdev = NULL;
569 static inline struct inode *orphan_list_entry(struct list_head *l)
571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
592 static void ext4_put_super(struct super_block *sb)
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
604 ext4_commit_super(sb, 1);
606 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
614 ext4_release_system_zone(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
619 if (!(sb->s_flags & MS_RDONLY)) {
620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
621 es->s_state = cpu_to_le16(sbi->s_mount_state);
622 ext4_commit_super(sb, 1);
625 remove_proc_entry(sb->s_id, ext4_proc_root);
627 kobject_del(&sbi->s_kobj);
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
635 kfree(sbi->s_flex_groups);
636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
654 invalidate_bdev(sb->s_bdev);
655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
661 sync_blockdev(sbi->journal_bdev);
662 invalidate_bdev(sbi->journal_bdev);
663 ext4_blkdev_remove(sbi);
665 sb->s_fs_info = NULL;
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
674 kfree(sbi->s_blockgroup_lock);
678 static struct kmem_cache *ext4_inode_cachep;
681 * Called inside transaction, so use GFP_NOFS
683 static struct inode *ext4_alloc_inode(struct super_block *sb)
685 struct ext4_inode_info *ei;
687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
691 ei->vfs_inode.i_version = 1;
692 ei->vfs_inode.i_data.writeback_index = 0;
693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
705 ei->i_delalloc_reserved_flag = 0;
706 spin_lock_init(&(ei->i_block_reservation_lock));
707 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
708 ei->cur_aio_dio = NULL;
710 ei->i_datasync_tid = 0;
712 return &ei->vfs_inode;
715 static void ext4_destroy_inode(struct inode *inode)
717 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
718 ext4_msg(inode->i_sb, KERN_ERR,
719 "Inode %lu (%p): orphan list check failed!",
720 inode->i_ino, EXT4_I(inode));
721 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
722 EXT4_I(inode), sizeof(struct ext4_inode_info),
726 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
729 static void init_once(void *foo)
731 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
733 INIT_LIST_HEAD(&ei->i_orphan);
734 #ifdef CONFIG_EXT4_FS_XATTR
735 init_rwsem(&ei->xattr_sem);
737 init_rwsem(&ei->i_data_sem);
738 inode_init_once(&ei->vfs_inode);
741 static int init_inodecache(void)
743 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
744 sizeof(struct ext4_inode_info),
745 0, (SLAB_RECLAIM_ACCOUNT|
748 if (ext4_inode_cachep == NULL)
753 static void destroy_inodecache(void)
755 kmem_cache_destroy(ext4_inode_cachep);
758 static void ext4_clear_inode(struct inode *inode)
760 ext4_discard_preallocations(inode);
761 if (EXT4_JOURNAL(inode))
762 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
763 &EXT4_I(inode)->jinode);
766 static inline void ext4_show_quota_options(struct seq_file *seq,
767 struct super_block *sb)
769 #if defined(CONFIG_QUOTA)
770 struct ext4_sb_info *sbi = EXT4_SB(sb);
772 if (sbi->s_jquota_fmt) {
775 switch (sbi->s_jquota_fmt) {
786 seq_printf(seq, ",jqfmt=%s", fmtname);
789 if (sbi->s_qf_names[USRQUOTA])
790 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
792 if (sbi->s_qf_names[GRPQUOTA])
793 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
795 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
796 seq_puts(seq, ",usrquota");
798 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
799 seq_puts(seq, ",grpquota");
805 * - it's set to a non-default value OR
806 * - if the per-sb default is different from the global default
808 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
811 unsigned long def_mount_opts;
812 struct super_block *sb = vfs->mnt_sb;
813 struct ext4_sb_info *sbi = EXT4_SB(sb);
814 struct ext4_super_block *es = sbi->s_es;
816 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
817 def_errors = le16_to_cpu(es->s_errors);
819 if (sbi->s_sb_block != 1)
820 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
821 if (test_opt(sb, MINIX_DF))
822 seq_puts(seq, ",minixdf");
823 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
824 seq_puts(seq, ",grpid");
825 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
826 seq_puts(seq, ",nogrpid");
827 if (sbi->s_resuid != EXT4_DEF_RESUID ||
828 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
829 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
831 if (sbi->s_resgid != EXT4_DEF_RESGID ||
832 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
833 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
835 if (test_opt(sb, ERRORS_RO)) {
836 if (def_errors == EXT4_ERRORS_PANIC ||
837 def_errors == EXT4_ERRORS_CONTINUE) {
838 seq_puts(seq, ",errors=remount-ro");
841 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
842 seq_puts(seq, ",errors=continue");
843 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
844 seq_puts(seq, ",errors=panic");
845 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
846 seq_puts(seq, ",nouid32");
847 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
848 seq_puts(seq, ",debug");
849 if (test_opt(sb, OLDALLOC))
850 seq_puts(seq, ",oldalloc");
851 #ifdef CONFIG_EXT4_FS_XATTR
852 if (test_opt(sb, XATTR_USER) &&
853 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
854 seq_puts(seq, ",user_xattr");
855 if (!test_opt(sb, XATTR_USER) &&
856 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
857 seq_puts(seq, ",nouser_xattr");
860 #ifdef CONFIG_EXT4_FS_POSIX_ACL
861 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
862 seq_puts(seq, ",acl");
863 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
864 seq_puts(seq, ",noacl");
866 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
867 seq_printf(seq, ",commit=%u",
868 (unsigned) (sbi->s_commit_interval / HZ));
870 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
871 seq_printf(seq, ",min_batch_time=%u",
872 (unsigned) sbi->s_min_batch_time);
874 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
875 seq_printf(seq, ",max_batch_time=%u",
876 (unsigned) sbi->s_min_batch_time);
880 * We're changing the default of barrier mount option, so
881 * let's always display its mount state so it's clear what its
884 seq_puts(seq, ",barrier=");
885 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
886 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
887 seq_puts(seq, ",journal_async_commit");
888 if (test_opt(sb, NOBH))
889 seq_puts(seq, ",nobh");
890 if (test_opt(sb, I_VERSION))
891 seq_puts(seq, ",i_version");
892 if (!test_opt(sb, DELALLOC))
893 seq_puts(seq, ",nodelalloc");
897 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
899 * journal mode get enabled in different ways
900 * So just print the value even if we didn't specify it
902 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
903 seq_puts(seq, ",data=journal");
904 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
905 seq_puts(seq, ",data=ordered");
906 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
907 seq_puts(seq, ",data=writeback");
909 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
910 seq_printf(seq, ",inode_readahead_blks=%u",
911 sbi->s_inode_readahead_blks);
913 if (test_opt(sb, DATA_ERR_ABORT))
914 seq_puts(seq, ",data_err=abort");
916 if (test_opt(sb, NO_AUTO_DA_ALLOC))
917 seq_puts(seq, ",noauto_da_alloc");
919 if (test_opt(sb, DISCARD))
920 seq_puts(seq, ",discard");
922 if (test_opt(sb, NOLOAD))
923 seq_puts(seq, ",norecovery");
925 ext4_show_quota_options(seq, sb);
930 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
931 u64 ino, u32 generation)
935 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
936 return ERR_PTR(-ESTALE);
937 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
938 return ERR_PTR(-ESTALE);
940 /* iget isn't really right if the inode is currently unallocated!!
942 * ext4_read_inode will return a bad_inode if the inode had been
943 * deleted, so we should be safe.
945 * Currently we don't know the generation for parent directory, so
946 * a generation of 0 means "accept any"
948 inode = ext4_iget(sb, ino);
950 return ERR_CAST(inode);
951 if (generation && inode->i_generation != generation) {
953 return ERR_PTR(-ESTALE);
959 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
960 int fh_len, int fh_type)
962 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
966 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
967 int fh_len, int fh_type)
969 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
974 * Try to release metadata pages (indirect blocks, directories) which are
975 * mapped via the block device. Since these pages could have journal heads
976 * which would prevent try_to_free_buffers() from freeing them, we must use
977 * jbd2 layer's try_to_free_buffers() function to release them.
979 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
982 journal_t *journal = EXT4_SB(sb)->s_journal;
984 WARN_ON(PageChecked(page));
985 if (!page_has_buffers(page))
988 return jbd2_journal_try_to_free_buffers(journal, page,
990 return try_to_free_buffers(page);
994 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
995 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
997 static int ext4_write_dquot(struct dquot *dquot);
998 static int ext4_acquire_dquot(struct dquot *dquot);
999 static int ext4_release_dquot(struct dquot *dquot);
1000 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1001 static int ext4_write_info(struct super_block *sb, int type);
1002 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1003 char *path, int remount);
1004 static int ext4_quota_on_mount(struct super_block *sb, int type);
1005 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1006 size_t len, loff_t off);
1007 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1008 const char *data, size_t len, loff_t off);
1010 static const struct dquot_operations ext4_quota_operations = {
1011 .initialize = dquot_initialize,
1013 .alloc_space = dquot_alloc_space,
1014 .reserve_space = dquot_reserve_space,
1015 .claim_space = dquot_claim_space,
1016 .release_rsv = dquot_release_reserved_space,
1017 .get_reserved_space = ext4_get_reserved_space,
1018 .alloc_inode = dquot_alloc_inode,
1019 .free_space = dquot_free_space,
1020 .free_inode = dquot_free_inode,
1021 .transfer = dquot_transfer,
1022 .write_dquot = ext4_write_dquot,
1023 .acquire_dquot = ext4_acquire_dquot,
1024 .release_dquot = ext4_release_dquot,
1025 .mark_dirty = ext4_mark_dquot_dirty,
1026 .write_info = ext4_write_info,
1027 .alloc_dquot = dquot_alloc,
1028 .destroy_dquot = dquot_destroy,
1031 static const struct quotactl_ops ext4_qctl_operations = {
1032 .quota_on = ext4_quota_on,
1033 .quota_off = vfs_quota_off,
1034 .quota_sync = vfs_quota_sync,
1035 .get_info = vfs_get_dqinfo,
1036 .set_info = vfs_set_dqinfo,
1037 .get_dqblk = vfs_get_dqblk,
1038 .set_dqblk = vfs_set_dqblk
1042 static const struct super_operations ext4_sops = {
1043 .alloc_inode = ext4_alloc_inode,
1044 .destroy_inode = ext4_destroy_inode,
1045 .write_inode = ext4_write_inode,
1046 .dirty_inode = ext4_dirty_inode,
1047 .delete_inode = ext4_delete_inode,
1048 .put_super = ext4_put_super,
1049 .sync_fs = ext4_sync_fs,
1050 .freeze_fs = ext4_freeze,
1051 .unfreeze_fs = ext4_unfreeze,
1052 .statfs = ext4_statfs,
1053 .remount_fs = ext4_remount,
1054 .clear_inode = ext4_clear_inode,
1055 .show_options = ext4_show_options,
1057 .quota_read = ext4_quota_read,
1058 .quota_write = ext4_quota_write,
1060 .bdev_try_to_free_page = bdev_try_to_free_page,
1063 static const struct super_operations ext4_nojournal_sops = {
1064 .alloc_inode = ext4_alloc_inode,
1065 .destroy_inode = ext4_destroy_inode,
1066 .write_inode = ext4_write_inode,
1067 .dirty_inode = ext4_dirty_inode,
1068 .delete_inode = ext4_delete_inode,
1069 .write_super = ext4_write_super,
1070 .put_super = ext4_put_super,
1071 .statfs = ext4_statfs,
1072 .remount_fs = ext4_remount,
1073 .clear_inode = ext4_clear_inode,
1074 .show_options = ext4_show_options,
1076 .quota_read = ext4_quota_read,
1077 .quota_write = ext4_quota_write,
1079 .bdev_try_to_free_page = bdev_try_to_free_page,
1082 static const struct export_operations ext4_export_ops = {
1083 .fh_to_dentry = ext4_fh_to_dentry,
1084 .fh_to_parent = ext4_fh_to_parent,
1085 .get_parent = ext4_get_parent,
1089 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1090 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1091 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1092 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1093 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1094 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1095 Opt_journal_update, Opt_journal_dev,
1096 Opt_journal_checksum, Opt_journal_async_commit,
1097 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1098 Opt_data_err_abort, Opt_data_err_ignore,
1099 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1100 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1101 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1102 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1103 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1104 Opt_block_validity, Opt_noblock_validity,
1105 Opt_inode_readahead_blks, Opt_journal_ioprio,
1106 Opt_discard, Opt_nodiscard,
1109 static const match_table_t tokens = {
1110 {Opt_bsd_df, "bsddf"},
1111 {Opt_minix_df, "minixdf"},
1112 {Opt_grpid, "grpid"},
1113 {Opt_grpid, "bsdgroups"},
1114 {Opt_nogrpid, "nogrpid"},
1115 {Opt_nogrpid, "sysvgroups"},
1116 {Opt_resgid, "resgid=%u"},
1117 {Opt_resuid, "resuid=%u"},
1119 {Opt_err_cont, "errors=continue"},
1120 {Opt_err_panic, "errors=panic"},
1121 {Opt_err_ro, "errors=remount-ro"},
1122 {Opt_nouid32, "nouid32"},
1123 {Opt_debug, "debug"},
1124 {Opt_oldalloc, "oldalloc"},
1125 {Opt_orlov, "orlov"},
1126 {Opt_user_xattr, "user_xattr"},
1127 {Opt_nouser_xattr, "nouser_xattr"},
1129 {Opt_noacl, "noacl"},
1130 {Opt_noload, "noload"},
1131 {Opt_noload, "norecovery"},
1134 {Opt_commit, "commit=%u"},
1135 {Opt_min_batch_time, "min_batch_time=%u"},
1136 {Opt_max_batch_time, "max_batch_time=%u"},
1137 {Opt_journal_update, "journal=update"},
1138 {Opt_journal_dev, "journal_dev=%u"},
1139 {Opt_journal_checksum, "journal_checksum"},
1140 {Opt_journal_async_commit, "journal_async_commit"},
1141 {Opt_abort, "abort"},
1142 {Opt_data_journal, "data=journal"},
1143 {Opt_data_ordered, "data=ordered"},
1144 {Opt_data_writeback, "data=writeback"},
1145 {Opt_data_err_abort, "data_err=abort"},
1146 {Opt_data_err_ignore, "data_err=ignore"},
1147 {Opt_offusrjquota, "usrjquota="},
1148 {Opt_usrjquota, "usrjquota=%s"},
1149 {Opt_offgrpjquota, "grpjquota="},
1150 {Opt_grpjquota, "grpjquota=%s"},
1151 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1152 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1153 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1154 {Opt_grpquota, "grpquota"},
1155 {Opt_noquota, "noquota"},
1156 {Opt_quota, "quota"},
1157 {Opt_usrquota, "usrquota"},
1158 {Opt_barrier, "barrier=%u"},
1159 {Opt_barrier, "barrier"},
1160 {Opt_nobarrier, "nobarrier"},
1161 {Opt_i_version, "i_version"},
1162 {Opt_stripe, "stripe=%u"},
1163 {Opt_resize, "resize"},
1164 {Opt_delalloc, "delalloc"},
1165 {Opt_nodelalloc, "nodelalloc"},
1166 {Opt_block_validity, "block_validity"},
1167 {Opt_noblock_validity, "noblock_validity"},
1168 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1169 {Opt_journal_ioprio, "journal_ioprio=%u"},
1170 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1171 {Opt_auto_da_alloc, "auto_da_alloc"},
1172 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1173 {Opt_discard, "discard"},
1174 {Opt_nodiscard, "nodiscard"},
1178 static ext4_fsblk_t get_sb_block(void **data)
1180 ext4_fsblk_t sb_block;
1181 char *options = (char *) *data;
1183 if (!options || strncmp(options, "sb=", 3) != 0)
1184 return 1; /* Default location */
1187 /* TODO: use simple_strtoll with >32bit ext4 */
1188 sb_block = simple_strtoul(options, &options, 0);
1189 if (*options && *options != ',') {
1190 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1194 if (*options == ',')
1196 *data = (void *) options;
1201 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1203 static int parse_options(char *options, struct super_block *sb,
1204 unsigned long *journal_devnum,
1205 unsigned int *journal_ioprio,
1206 ext4_fsblk_t *n_blocks_count, int is_remount)
1208 struct ext4_sb_info *sbi = EXT4_SB(sb);
1210 substring_t args[MAX_OPT_ARGS];
1221 while ((p = strsep(&options, ",")) != NULL) {
1226 token = match_token(p, tokens, args);
1229 clear_opt(sbi->s_mount_opt, MINIX_DF);
1232 set_opt(sbi->s_mount_opt, MINIX_DF);
1235 set_opt(sbi->s_mount_opt, GRPID);
1238 clear_opt(sbi->s_mount_opt, GRPID);
1241 if (match_int(&args[0], &option))
1243 sbi->s_resuid = option;
1246 if (match_int(&args[0], &option))
1248 sbi->s_resgid = option;
1251 /* handled by get_sb_block() instead of here */
1252 /* *sb_block = match_int(&args[0]); */
1255 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1256 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1257 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1260 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1261 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1262 set_opt(sbi->s_mount_opt, ERRORS_RO);
1265 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1266 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1267 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1270 set_opt(sbi->s_mount_opt, NO_UID32);
1273 set_opt(sbi->s_mount_opt, DEBUG);
1276 set_opt(sbi->s_mount_opt, OLDALLOC);
1279 clear_opt(sbi->s_mount_opt, OLDALLOC);
1281 #ifdef CONFIG_EXT4_FS_XATTR
1282 case Opt_user_xattr:
1283 set_opt(sbi->s_mount_opt, XATTR_USER);
1285 case Opt_nouser_xattr:
1286 clear_opt(sbi->s_mount_opt, XATTR_USER);
1289 case Opt_user_xattr:
1290 case Opt_nouser_xattr:
1291 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1294 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1296 set_opt(sbi->s_mount_opt, POSIX_ACL);
1299 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1304 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1307 case Opt_journal_update:
1309 /* Eventually we will want to be able to create
1310 a journal file here. For now, only allow the
1311 user to specify an existing inode to be the
1314 ext4_msg(sb, KERN_ERR,
1315 "Cannot specify journal on remount");
1318 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1320 case Opt_journal_dev:
1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot specify journal on remount");
1326 if (match_int(&args[0], &option))
1328 *journal_devnum = option;
1330 case Opt_journal_checksum:
1331 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1333 case Opt_journal_async_commit:
1334 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1335 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1338 set_opt(sbi->s_mount_opt, NOLOAD);
1341 if (match_int(&args[0], &option))
1346 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1347 sbi->s_commit_interval = HZ * option;
1349 case Opt_max_batch_time:
1350 if (match_int(&args[0], &option))
1355 option = EXT4_DEF_MAX_BATCH_TIME;
1356 sbi->s_max_batch_time = option;
1358 case Opt_min_batch_time:
1359 if (match_int(&args[0], &option))
1363 sbi->s_min_batch_time = option;
1365 case Opt_data_journal:
1366 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1368 case Opt_data_ordered:
1369 data_opt = EXT4_MOUNT_ORDERED_DATA;
1371 case Opt_data_writeback:
1372 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1375 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1377 ext4_msg(sb, KERN_ERR,
1378 "Cannot change data mode on remount");
1382 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1383 sbi->s_mount_opt |= data_opt;
1386 case Opt_data_err_abort:
1387 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1389 case Opt_data_err_ignore:
1390 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1399 if (sb_any_quota_loaded(sb) &&
1400 !sbi->s_qf_names[qtype]) {
1401 ext4_msg(sb, KERN_ERR,
1402 "Cannot change journaled "
1403 "quota options when quota turned on");
1406 qname = match_strdup(&args[0]);
1408 ext4_msg(sb, KERN_ERR,
1409 "Not enough memory for "
1410 "storing quotafile name");
1413 if (sbi->s_qf_names[qtype] &&
1414 strcmp(sbi->s_qf_names[qtype], qname)) {
1415 ext4_msg(sb, KERN_ERR,
1416 "%s quota file already "
1417 "specified", QTYPE2NAME(qtype));
1421 sbi->s_qf_names[qtype] = qname;
1422 if (strchr(sbi->s_qf_names[qtype], '/')) {
1423 ext4_msg(sb, KERN_ERR,
1424 "quotafile must be on "
1426 kfree(sbi->s_qf_names[qtype]);
1427 sbi->s_qf_names[qtype] = NULL;
1430 set_opt(sbi->s_mount_opt, QUOTA);
1432 case Opt_offusrjquota:
1435 case Opt_offgrpjquota:
1438 if (sb_any_quota_loaded(sb) &&
1439 sbi->s_qf_names[qtype]) {
1440 ext4_msg(sb, KERN_ERR, "Cannot change "
1441 "journaled quota options when "
1446 * The space will be released later when all options
1447 * are confirmed to be correct
1449 sbi->s_qf_names[qtype] = NULL;
1451 case Opt_jqfmt_vfsold:
1452 qfmt = QFMT_VFS_OLD;
1454 case Opt_jqfmt_vfsv0:
1457 case Opt_jqfmt_vfsv1:
1460 if (sb_any_quota_loaded(sb) &&
1461 sbi->s_jquota_fmt != qfmt) {
1462 ext4_msg(sb, KERN_ERR, "Cannot change "
1463 "journaled quota options when "
1467 sbi->s_jquota_fmt = qfmt;
1471 set_opt(sbi->s_mount_opt, QUOTA);
1472 set_opt(sbi->s_mount_opt, USRQUOTA);
1475 set_opt(sbi->s_mount_opt, QUOTA);
1476 set_opt(sbi->s_mount_opt, GRPQUOTA);
1479 if (sb_any_quota_loaded(sb)) {
1480 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1481 "options when quota turned on");
1484 clear_opt(sbi->s_mount_opt, QUOTA);
1485 clear_opt(sbi->s_mount_opt, USRQUOTA);
1486 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1492 ext4_msg(sb, KERN_ERR,
1493 "quota options not supported");
1497 case Opt_offusrjquota:
1498 case Opt_offgrpjquota:
1499 case Opt_jqfmt_vfsold:
1500 case Opt_jqfmt_vfsv0:
1501 case Opt_jqfmt_vfsv1:
1502 ext4_msg(sb, KERN_ERR,
1503 "journaled quota options not supported");
1509 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1512 clear_opt(sbi->s_mount_opt, BARRIER);
1515 if (match_int(&args[0], &option)) {
1516 set_opt(sbi->s_mount_opt, BARRIER);
1520 set_opt(sbi->s_mount_opt, BARRIER);
1522 clear_opt(sbi->s_mount_opt, BARRIER);
1528 ext4_msg(sb, KERN_ERR,
1529 "resize option only available "
1533 if (match_int(&args[0], &option) != 0)
1535 *n_blocks_count = option;
1538 set_opt(sbi->s_mount_opt, NOBH);
1541 clear_opt(sbi->s_mount_opt, NOBH);
1544 set_opt(sbi->s_mount_opt, I_VERSION);
1545 sb->s_flags |= MS_I_VERSION;
1547 case Opt_nodelalloc:
1548 clear_opt(sbi->s_mount_opt, DELALLOC);
1551 if (match_int(&args[0], &option))
1555 sbi->s_stripe = option;
1558 set_opt(sbi->s_mount_opt, DELALLOC);
1560 case Opt_block_validity:
1561 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1563 case Opt_noblock_validity:
1564 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1566 case Opt_inode_readahead_blks:
1567 if (match_int(&args[0], &option))
1569 if (option < 0 || option > (1 << 30))
1571 if (!is_power_of_2(option)) {
1572 ext4_msg(sb, KERN_ERR,
1573 "EXT4-fs: inode_readahead_blks"
1574 " must be a power of 2");
1577 sbi->s_inode_readahead_blks = option;
1579 case Opt_journal_ioprio:
1580 if (match_int(&args[0], &option))
1582 if (option < 0 || option > 7)
1584 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1587 case Opt_noauto_da_alloc:
1588 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1590 case Opt_auto_da_alloc:
1591 if (match_int(&args[0], &option)) {
1592 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1596 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1598 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1601 set_opt(sbi->s_mount_opt, DISCARD);
1604 clear_opt(sbi->s_mount_opt, DISCARD);
1607 ext4_msg(sb, KERN_ERR,
1608 "Unrecognized mount option \"%s\" "
1609 "or missing value", p);
1614 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1615 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1616 sbi->s_qf_names[USRQUOTA])
1617 clear_opt(sbi->s_mount_opt, USRQUOTA);
1619 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1620 sbi->s_qf_names[GRPQUOTA])
1621 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1623 if ((sbi->s_qf_names[USRQUOTA] &&
1624 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1625 (sbi->s_qf_names[GRPQUOTA] &&
1626 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1627 ext4_msg(sb, KERN_ERR, "old and new quota "
1632 if (!sbi->s_jquota_fmt) {
1633 ext4_msg(sb, KERN_ERR, "journaled quota format "
1638 if (sbi->s_jquota_fmt) {
1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 "specified with no journaling "
1649 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1652 struct ext4_sb_info *sbi = EXT4_SB(sb);
1655 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1656 ext4_msg(sb, KERN_ERR, "revision level too high, "
1657 "forcing read-only mode");
1662 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1663 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1664 "running e2fsck is recommended");
1665 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1666 ext4_msg(sb, KERN_WARNING,
1667 "warning: mounting fs with errors, "
1668 "running e2fsck is recommended");
1669 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1670 le16_to_cpu(es->s_mnt_count) >=
1671 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1672 ext4_msg(sb, KERN_WARNING,
1673 "warning: maximal mount count reached, "
1674 "running e2fsck is recommended");
1675 else if (le32_to_cpu(es->s_checkinterval) &&
1676 (le32_to_cpu(es->s_lastcheck) +
1677 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1678 ext4_msg(sb, KERN_WARNING,
1679 "warning: checktime reached, "
1680 "running e2fsck is recommended");
1681 if (!sbi->s_journal)
1682 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1683 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1684 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1685 le16_add_cpu(&es->s_mnt_count, 1);
1686 es->s_mtime = cpu_to_le32(get_seconds());
1687 ext4_update_dynamic_rev(sb);
1689 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1691 ext4_commit_super(sb, 1);
1692 if (test_opt(sb, DEBUG))
1693 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1694 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1696 sbi->s_groups_count,
1697 EXT4_BLOCKS_PER_GROUP(sb),
1698 EXT4_INODES_PER_GROUP(sb),
1704 static int ext4_fill_flex_info(struct super_block *sb)
1706 struct ext4_sb_info *sbi = EXT4_SB(sb);
1707 struct ext4_group_desc *gdp = NULL;
1708 ext4_group_t flex_group_count;
1709 ext4_group_t flex_group;
1710 int groups_per_flex = 0;
1714 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1715 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1717 if (groups_per_flex < 2) {
1718 sbi->s_log_groups_per_flex = 0;
1722 /* We allocate both existing and potentially added groups */
1723 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1724 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1725 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1726 size = flex_group_count * sizeof(struct flex_groups);
1727 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1728 if (sbi->s_flex_groups == NULL) {
1729 sbi->s_flex_groups = vmalloc(size);
1730 if (sbi->s_flex_groups)
1731 memset(sbi->s_flex_groups, 0, size);
1733 if (sbi->s_flex_groups == NULL) {
1734 ext4_msg(sb, KERN_ERR, "not enough memory for "
1735 "%u flex groups", flex_group_count);
1739 for (i = 0; i < sbi->s_groups_count; i++) {
1740 gdp = ext4_get_group_desc(sb, i, NULL);
1742 flex_group = ext4_flex_group(sbi, i);
1743 atomic_add(ext4_free_inodes_count(sb, gdp),
1744 &sbi->s_flex_groups[flex_group].free_inodes);
1745 atomic_add(ext4_free_blks_count(sb, gdp),
1746 &sbi->s_flex_groups[flex_group].free_blocks);
1747 atomic_add(ext4_used_dirs_count(sb, gdp),
1748 &sbi->s_flex_groups[flex_group].used_dirs);
1756 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1757 struct ext4_group_desc *gdp)
1761 if (sbi->s_es->s_feature_ro_compat &
1762 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1763 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1764 __le32 le_group = cpu_to_le32(block_group);
1766 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1767 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1768 crc = crc16(crc, (__u8 *)gdp, offset);
1769 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1770 /* for checksum of struct ext4_group_desc do the rest...*/
1771 if ((sbi->s_es->s_feature_incompat &
1772 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1773 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1774 crc = crc16(crc, (__u8 *)gdp + offset,
1775 le16_to_cpu(sbi->s_es->s_desc_size) -
1779 return cpu_to_le16(crc);
1782 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1783 struct ext4_group_desc *gdp)
1785 if ((sbi->s_es->s_feature_ro_compat &
1786 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1787 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1793 /* Called at mount-time, super-block is locked */
1794 static int ext4_check_descriptors(struct super_block *sb)
1796 struct ext4_sb_info *sbi = EXT4_SB(sb);
1797 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1798 ext4_fsblk_t last_block;
1799 ext4_fsblk_t block_bitmap;
1800 ext4_fsblk_t inode_bitmap;
1801 ext4_fsblk_t inode_table;
1802 int flexbg_flag = 0;
1805 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1808 ext4_debug("Checking group descriptors");
1810 for (i = 0; i < sbi->s_groups_count; i++) {
1811 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1813 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1814 last_block = ext4_blocks_count(sbi->s_es) - 1;
1816 last_block = first_block +
1817 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1819 block_bitmap = ext4_block_bitmap(sb, gdp);
1820 if (block_bitmap < first_block || block_bitmap > last_block) {
1821 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1822 "Block bitmap for group %u not in group "
1823 "(block %llu)!", i, block_bitmap);
1826 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1827 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1828 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1829 "Inode bitmap for group %u not in group "
1830 "(block %llu)!", i, inode_bitmap);
1833 inode_table = ext4_inode_table(sb, gdp);
1834 if (inode_table < first_block ||
1835 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1836 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1837 "Inode table for group %u not in group "
1838 "(block %llu)!", i, inode_table);
1841 ext4_lock_group(sb, i);
1842 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1843 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1844 "Checksum for group %u failed (%u!=%u)",
1845 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1846 gdp)), le16_to_cpu(gdp->bg_checksum));
1847 if (!(sb->s_flags & MS_RDONLY)) {
1848 ext4_unlock_group(sb, i);
1852 ext4_unlock_group(sb, i);
1854 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1857 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1858 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1862 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1863 * the superblock) which were deleted from all directories, but held open by
1864 * a process at the time of a crash. We walk the list and try to delete these
1865 * inodes at recovery time (only with a read-write filesystem).
1867 * In order to keep the orphan inode chain consistent during traversal (in
1868 * case of crash during recovery), we link each inode into the superblock
1869 * orphan list_head and handle it the same way as an inode deletion during
1870 * normal operation (which journals the operations for us).
1872 * We only do an iget() and an iput() on each inode, which is very safe if we
1873 * accidentally point at an in-use or already deleted inode. The worst that
1874 * can happen in this case is that we get a "bit already cleared" message from
1875 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1876 * e2fsck was run on this filesystem, and it must have already done the orphan
1877 * inode cleanup for us, so we can safely abort without any further action.
1879 static void ext4_orphan_cleanup(struct super_block *sb,
1880 struct ext4_super_block *es)
1882 unsigned int s_flags = sb->s_flags;
1883 int nr_orphans = 0, nr_truncates = 0;
1887 if (!es->s_last_orphan) {
1888 jbd_debug(4, "no orphan inodes to clean up\n");
1892 if (bdev_read_only(sb->s_bdev)) {
1893 ext4_msg(sb, KERN_ERR, "write access "
1894 "unavailable, skipping orphan cleanup");
1898 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1899 if (es->s_last_orphan)
1900 jbd_debug(1, "Errors on filesystem, "
1901 "clearing orphan list.\n");
1902 es->s_last_orphan = 0;
1903 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1907 if (s_flags & MS_RDONLY) {
1908 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1909 sb->s_flags &= ~MS_RDONLY;
1912 /* Needed for iput() to work correctly and not trash data */
1913 sb->s_flags |= MS_ACTIVE;
1914 /* Turn on quotas so that they are updated correctly */
1915 for (i = 0; i < MAXQUOTAS; i++) {
1916 if (EXT4_SB(sb)->s_qf_names[i]) {
1917 int ret = ext4_quota_on_mount(sb, i);
1919 ext4_msg(sb, KERN_ERR,
1920 "Cannot turn on journaled "
1921 "quota: error %d", ret);
1926 while (es->s_last_orphan) {
1927 struct inode *inode;
1929 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1930 if (IS_ERR(inode)) {
1931 es->s_last_orphan = 0;
1935 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1937 if (inode->i_nlink) {
1938 ext4_msg(sb, KERN_DEBUG,
1939 "%s: truncating inode %lu to %lld bytes",
1940 __func__, inode->i_ino, inode->i_size);
1941 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1942 inode->i_ino, inode->i_size);
1943 ext4_truncate(inode);
1946 ext4_msg(sb, KERN_DEBUG,
1947 "%s: deleting unreferenced inode %lu",
1948 __func__, inode->i_ino);
1949 jbd_debug(2, "deleting unreferenced inode %lu\n",
1953 iput(inode); /* The delete magic happens here! */
1956 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1959 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1960 PLURAL(nr_orphans));
1962 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1963 PLURAL(nr_truncates));
1965 /* Turn quotas off */
1966 for (i = 0; i < MAXQUOTAS; i++) {
1967 if (sb_dqopt(sb)->files[i])
1968 vfs_quota_off(sb, i, 0);
1971 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1975 * Maximal extent format file size.
1976 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1977 * extent format containers, within a sector_t, and within i_blocks
1978 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1979 * so that won't be a limiting factor.
1981 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1983 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1986 loff_t upper_limit = MAX_LFS_FILESIZE;
1988 /* small i_blocks in vfs inode? */
1989 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1991 * CONFIG_LBDAF is not enabled implies the inode
1992 * i_block represent total blocks in 512 bytes
1993 * 32 == size of vfs inode i_blocks * 8
1995 upper_limit = (1LL << 32) - 1;
1997 /* total blocks in file system block size */
1998 upper_limit >>= (blkbits - 9);
1999 upper_limit <<= blkbits;
2002 /* 32-bit extent-start container, ee_block */
2007 /* Sanity check against vm- & vfs- imposed limits */
2008 if (res > upper_limit)
2015 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2016 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2017 * We need to be 1 filesystem block less than the 2^48 sector limit.
2019 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2021 loff_t res = EXT4_NDIR_BLOCKS;
2024 /* This is calculated to be the largest file size for a dense, block
2025 * mapped file such that the file's total number of 512-byte sectors,
2026 * including data and all indirect blocks, does not exceed (2^48 - 1).
2028 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2029 * number of 512-byte sectors of the file.
2032 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2034 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2035 * the inode i_block field represents total file blocks in
2036 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2038 upper_limit = (1LL << 32) - 1;
2040 /* total blocks in file system block size */
2041 upper_limit >>= (bits - 9);
2045 * We use 48 bit ext4_inode i_blocks
2046 * With EXT4_HUGE_FILE_FL set the i_blocks
2047 * represent total number of blocks in
2048 * file system block size
2050 upper_limit = (1LL << 48) - 1;
2054 /* indirect blocks */
2056 /* double indirect blocks */
2057 meta_blocks += 1 + (1LL << (bits-2));
2058 /* tripple indirect blocks */
2059 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2061 upper_limit -= meta_blocks;
2062 upper_limit <<= bits;
2064 res += 1LL << (bits-2);
2065 res += 1LL << (2*(bits-2));
2066 res += 1LL << (3*(bits-2));
2068 if (res > upper_limit)
2071 if (res > MAX_LFS_FILESIZE)
2072 res = MAX_LFS_FILESIZE;
2077 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2078 ext4_fsblk_t logical_sb_block, int nr)
2080 struct ext4_sb_info *sbi = EXT4_SB(sb);
2081 ext4_group_t bg, first_meta_bg;
2084 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2086 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2088 return logical_sb_block + nr + 1;
2089 bg = sbi->s_desc_per_block * nr;
2090 if (ext4_bg_has_super(sb, bg))
2093 return (has_super + ext4_group_first_block_no(sb, bg));
2097 * ext4_get_stripe_size: Get the stripe size.
2098 * @sbi: In memory super block info
2100 * If we have specified it via mount option, then
2101 * use the mount option value. If the value specified at mount time is
2102 * greater than the blocks per group use the super block value.
2103 * If the super block value is greater than blocks per group return 0.
2104 * Allocator needs it be less than blocks per group.
2107 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2109 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2110 unsigned long stripe_width =
2111 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2113 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2114 return sbi->s_stripe;
2116 if (stripe_width <= sbi->s_blocks_per_group)
2117 return stripe_width;
2119 if (stride <= sbi->s_blocks_per_group)
2128 struct attribute attr;
2129 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2130 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2131 const char *, size_t);
2135 static int parse_strtoul(const char *buf,
2136 unsigned long max, unsigned long *value)
2140 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2141 endp = skip_spaces(endp);
2142 if (*endp || *value > max)
2148 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2149 struct ext4_sb_info *sbi,
2152 return snprintf(buf, PAGE_SIZE, "%llu\n",
2153 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2156 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2157 struct ext4_sb_info *sbi, char *buf)
2159 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2161 return snprintf(buf, PAGE_SIZE, "%lu\n",
2162 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2163 sbi->s_sectors_written_start) >> 1);
2166 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2167 struct ext4_sb_info *sbi, char *buf)
2169 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2171 return snprintf(buf, PAGE_SIZE, "%llu\n",
2172 sbi->s_kbytes_written +
2173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2177 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2178 struct ext4_sb_info *sbi,
2179 const char *buf, size_t count)
2183 if (parse_strtoul(buf, 0x40000000, &t))
2186 if (!is_power_of_2(t))
2189 sbi->s_inode_readahead_blks = t;
2193 static ssize_t sbi_ui_show(struct ext4_attr *a,
2194 struct ext4_sb_info *sbi, char *buf)
2196 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2198 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2201 static ssize_t sbi_ui_store(struct ext4_attr *a,
2202 struct ext4_sb_info *sbi,
2203 const char *buf, size_t count)
2205 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2208 if (parse_strtoul(buf, 0xffffffff, &t))
2214 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2215 static struct ext4_attr ext4_attr_##_name = { \
2216 .attr = {.name = __stringify(_name), .mode = _mode }, \
2219 .offset = offsetof(struct ext4_sb_info, _elname), \
2221 #define EXT4_ATTR(name, mode, show, store) \
2222 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2224 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2225 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2226 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2227 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2228 #define ATTR_LIST(name) &ext4_attr_##name.attr
2230 EXT4_RO_ATTR(delayed_allocation_blocks);
2231 EXT4_RO_ATTR(session_write_kbytes);
2232 EXT4_RO_ATTR(lifetime_write_kbytes);
2233 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2234 inode_readahead_blks_store, s_inode_readahead_blks);
2235 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2236 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2237 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2238 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2239 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2240 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2241 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2242 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2244 static struct attribute *ext4_attrs[] = {
2245 ATTR_LIST(delayed_allocation_blocks),
2246 ATTR_LIST(session_write_kbytes),
2247 ATTR_LIST(lifetime_write_kbytes),
2248 ATTR_LIST(inode_readahead_blks),
2249 ATTR_LIST(inode_goal),
2250 ATTR_LIST(mb_stats),
2251 ATTR_LIST(mb_max_to_scan),
2252 ATTR_LIST(mb_min_to_scan),
2253 ATTR_LIST(mb_order2_req),
2254 ATTR_LIST(mb_stream_req),
2255 ATTR_LIST(mb_group_prealloc),
2256 ATTR_LIST(max_writeback_mb_bump),
2260 static ssize_t ext4_attr_show(struct kobject *kobj,
2261 struct attribute *attr, char *buf)
2263 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2265 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2267 return a->show ? a->show(a, sbi, buf) : 0;
2270 static ssize_t ext4_attr_store(struct kobject *kobj,
2271 struct attribute *attr,
2272 const char *buf, size_t len)
2274 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2276 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2278 return a->store ? a->store(a, sbi, buf, len) : 0;
2281 static void ext4_sb_release(struct kobject *kobj)
2283 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2285 complete(&sbi->s_kobj_unregister);
2289 static struct sysfs_ops ext4_attr_ops = {
2290 .show = ext4_attr_show,
2291 .store = ext4_attr_store,
2294 static struct kobj_type ext4_ktype = {
2295 .default_attrs = ext4_attrs,
2296 .sysfs_ops = &ext4_attr_ops,
2297 .release = ext4_sb_release,
2301 * Check whether this filesystem can be mounted based on
2302 * the features present and the RDONLY/RDWR mount requested.
2303 * Returns 1 if this filesystem can be mounted as requested,
2304 * 0 if it cannot be.
2306 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2308 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2309 ext4_msg(sb, KERN_ERR,
2310 "Couldn't mount because of "
2311 "unsupported optional features (%x)",
2312 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2313 ~EXT4_FEATURE_INCOMPAT_SUPP));
2320 /* Check that feature set is OK for a read-write mount */
2321 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2322 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2323 "unsupported optional features (%x)",
2324 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2325 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2329 * Large file size enabled file system can only be mounted
2330 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2332 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2333 if (sizeof(blkcnt_t) < sizeof(u64)) {
2334 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2335 "cannot be mounted RDWR without "
2343 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2344 __releases(kernel_lock)
2345 __acquires(kernel_lock)
2347 struct buffer_head *bh;
2348 struct ext4_super_block *es = NULL;
2349 struct ext4_sb_info *sbi;
2351 ext4_fsblk_t sb_block = get_sb_block(&data);
2352 ext4_fsblk_t logical_sb_block;
2353 unsigned long offset = 0;
2354 unsigned long journal_devnum = 0;
2355 unsigned long def_mount_opts;
2361 unsigned int db_count;
2363 int needs_recovery, has_huge_files;
2366 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2368 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2372 sbi->s_blockgroup_lock =
2373 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2374 if (!sbi->s_blockgroup_lock) {
2378 sb->s_fs_info = sbi;
2379 sbi->s_mount_opt = 0;
2380 sbi->s_resuid = EXT4_DEF_RESUID;
2381 sbi->s_resgid = EXT4_DEF_RESGID;
2382 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2383 sbi->s_sb_block = sb_block;
2384 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2389 /* Cleanup superblock name */
2390 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2393 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2395 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2400 * The ext4 superblock will not be buffer aligned for other than 1kB
2401 * block sizes. We need to calculate the offset from buffer start.
2403 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2404 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2405 offset = do_div(logical_sb_block, blocksize);
2407 logical_sb_block = sb_block;
2410 if (!(bh = sb_bread(sb, logical_sb_block))) {
2411 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2415 * Note: s_es must be initialized as soon as possible because
2416 * some ext4 macro-instructions depend on its value
2418 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2420 sb->s_magic = le16_to_cpu(es->s_magic);
2421 if (sb->s_magic != EXT4_SUPER_MAGIC)
2423 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2425 /* Set defaults before we parse the mount options */
2426 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2427 if (def_mount_opts & EXT4_DEFM_DEBUG)
2428 set_opt(sbi->s_mount_opt, DEBUG);
2429 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2430 set_opt(sbi->s_mount_opt, GRPID);
2431 if (def_mount_opts & EXT4_DEFM_UID16)
2432 set_opt(sbi->s_mount_opt, NO_UID32);
2433 #ifdef CONFIG_EXT4_FS_XATTR
2434 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2435 set_opt(sbi->s_mount_opt, XATTR_USER);
2437 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2438 if (def_mount_opts & EXT4_DEFM_ACL)
2439 set_opt(sbi->s_mount_opt, POSIX_ACL);
2441 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2442 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2443 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2444 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2445 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2446 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2448 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2449 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2450 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2451 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2453 set_opt(sbi->s_mount_opt, ERRORS_RO);
2455 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2456 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2461 set_opt(sbi->s_mount_opt, BARRIER);
2464 * enable delayed allocation by default
2465 * Use -o nodelalloc to turn it off
2467 set_opt(sbi->s_mount_opt, DELALLOC);
2469 if (!parse_options((char *) data, sb, &journal_devnum,
2470 &journal_ioprio, NULL, 0))
2473 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2474 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2477 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2478 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2479 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2480 ext4_msg(sb, KERN_WARNING,
2481 "feature flags set on rev 0 fs, "
2482 "running e2fsck is recommended");
2485 * Check feature flags regardless of the revision level, since we
2486 * previously didn't change the revision level when setting the flags,
2487 * so there is a chance incompat flags are set on a rev 0 filesystem.
2489 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2492 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2494 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2495 blocksize > EXT4_MAX_BLOCK_SIZE) {
2496 ext4_msg(sb, KERN_ERR,
2497 "Unsupported filesystem blocksize %d", blocksize);
2501 if (sb->s_blocksize != blocksize) {
2502 /* Validate the filesystem blocksize */
2503 if (!sb_set_blocksize(sb, blocksize)) {
2504 ext4_msg(sb, KERN_ERR, "bad block size %d",
2510 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2511 offset = do_div(logical_sb_block, blocksize);
2512 bh = sb_bread(sb, logical_sb_block);
2514 ext4_msg(sb, KERN_ERR,
2515 "Can't read superblock on 2nd try");
2518 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2520 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2521 ext4_msg(sb, KERN_ERR,
2522 "Magic mismatch, very weird!");
2527 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2528 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2529 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2531 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2533 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2534 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2535 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2537 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2538 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2539 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2540 (!is_power_of_2(sbi->s_inode_size)) ||
2541 (sbi->s_inode_size > blocksize)) {
2542 ext4_msg(sb, KERN_ERR,
2543 "unsupported inode size: %d",
2547 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2548 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2551 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2552 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2553 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2554 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2555 !is_power_of_2(sbi->s_desc_size)) {
2556 ext4_msg(sb, KERN_ERR,
2557 "unsupported descriptor size %lu",
2562 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2564 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2565 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2566 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2569 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2570 if (sbi->s_inodes_per_block == 0)
2572 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2573 sbi->s_inodes_per_block;
2574 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2576 sbi->s_mount_state = le16_to_cpu(es->s_state);
2577 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2578 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2580 for (i = 0; i < 4; i++)
2581 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2582 sbi->s_def_hash_version = es->s_def_hash_version;
2583 i = le32_to_cpu(es->s_flags);
2584 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2585 sbi->s_hash_unsigned = 3;
2586 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2587 #ifdef __CHAR_UNSIGNED__
2588 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2589 sbi->s_hash_unsigned = 3;
2591 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2596 if (sbi->s_blocks_per_group > blocksize * 8) {
2597 ext4_msg(sb, KERN_ERR,
2598 "#blocks per group too big: %lu",
2599 sbi->s_blocks_per_group);
2602 if (sbi->s_inodes_per_group > blocksize * 8) {
2603 ext4_msg(sb, KERN_ERR,
2604 "#inodes per group too big: %lu",
2605 sbi->s_inodes_per_group);
2610 * Test whether we have more sectors than will fit in sector_t,
2611 * and whether the max offset is addressable by the page cache.
2613 if ((ext4_blocks_count(es) >
2614 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2615 (ext4_blocks_count(es) >
2616 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2617 ext4_msg(sb, KERN_ERR, "filesystem"
2618 " too large to mount safely on this system");
2619 if (sizeof(sector_t) < 8)
2620 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2625 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2628 /* check blocks count against device size */
2629 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2630 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2631 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2632 "exceeds size of device (%llu blocks)",
2633 ext4_blocks_count(es), blocks_count);
2638 * It makes no sense for the first data block to be beyond the end
2639 * of the filesystem.
2641 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2642 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2643 "block %u is beyond end of filesystem (%llu)",
2644 le32_to_cpu(es->s_first_data_block),
2645 ext4_blocks_count(es));
2648 blocks_count = (ext4_blocks_count(es) -
2649 le32_to_cpu(es->s_first_data_block) +
2650 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2651 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2652 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2653 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2654 "(block count %llu, first data block %u, "
2655 "blocks per group %lu)", sbi->s_groups_count,
2656 ext4_blocks_count(es),
2657 le32_to_cpu(es->s_first_data_block),
2658 EXT4_BLOCKS_PER_GROUP(sb));
2661 sbi->s_groups_count = blocks_count;
2662 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2663 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2664 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2665 EXT4_DESC_PER_BLOCK(sb);
2666 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2668 if (sbi->s_group_desc == NULL) {
2669 ext4_msg(sb, KERN_ERR, "not enough memory");
2673 #ifdef CONFIG_PROC_FS
2675 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2678 bgl_lock_init(sbi->s_blockgroup_lock);
2680 for (i = 0; i < db_count; i++) {
2681 block = descriptor_loc(sb, logical_sb_block, i);
2682 sbi->s_group_desc[i] = sb_bread(sb, block);
2683 if (!sbi->s_group_desc[i]) {
2684 ext4_msg(sb, KERN_ERR,
2685 "can't read group descriptor %d", i);
2690 if (!ext4_check_descriptors(sb)) {
2691 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2694 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2695 if (!ext4_fill_flex_info(sb)) {
2696 ext4_msg(sb, KERN_ERR,
2697 "unable to initialize "
2698 "flex_bg meta info!");
2702 sbi->s_gdb_count = db_count;
2703 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2704 spin_lock_init(&sbi->s_next_gen_lock);
2706 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2707 ext4_count_free_blocks(sb));
2709 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2710 ext4_count_free_inodes(sb));
2713 err = percpu_counter_init(&sbi->s_dirs_counter,
2714 ext4_count_dirs(sb));
2717 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2720 ext4_msg(sb, KERN_ERR, "insufficient memory");
2724 sbi->s_stripe = ext4_get_stripe_size(sbi);
2725 sbi->s_max_writeback_mb_bump = 128;
2728 * set up enough so that it can read an inode
2730 if (!test_opt(sb, NOLOAD) &&
2731 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2732 sb->s_op = &ext4_sops;
2734 sb->s_op = &ext4_nojournal_sops;
2735 sb->s_export_op = &ext4_export_ops;
2736 sb->s_xattr = ext4_xattr_handlers;
2738 sb->s_qcop = &ext4_qctl_operations;
2739 sb->dq_op = &ext4_quota_operations;
2741 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2742 mutex_init(&sbi->s_orphan_lock);
2743 mutex_init(&sbi->s_resize_lock);
2747 needs_recovery = (es->s_last_orphan != 0 ||
2748 EXT4_HAS_INCOMPAT_FEATURE(sb,
2749 EXT4_FEATURE_INCOMPAT_RECOVER));
2752 * The first inode we look at is the journal inode. Don't try
2753 * root first: it may be modified in the journal!
2755 if (!test_opt(sb, NOLOAD) &&
2756 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2757 if (ext4_load_journal(sb, es, journal_devnum))
2759 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2760 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2761 ext4_msg(sb, KERN_ERR, "required journal recovery "
2762 "suppressed and not mounted read-only");
2765 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2766 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2767 sbi->s_journal = NULL;
2772 if (ext4_blocks_count(es) > 0xffffffffULL &&
2773 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2774 JBD2_FEATURE_INCOMPAT_64BIT)) {
2775 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2779 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2780 jbd2_journal_set_features(sbi->s_journal,
2781 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2782 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2783 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2784 jbd2_journal_set_features(sbi->s_journal,
2785 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2786 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2787 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2789 jbd2_journal_clear_features(sbi->s_journal,
2790 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2791 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2794 /* We have now updated the journal if required, so we can
2795 * validate the data journaling mode. */
2796 switch (test_opt(sb, DATA_FLAGS)) {
2798 /* No mode set, assume a default based on the journal
2799 * capabilities: ORDERED_DATA if the journal can
2800 * cope, else JOURNAL_DATA
2802 if (jbd2_journal_check_available_features
2803 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2804 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2806 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2809 case EXT4_MOUNT_ORDERED_DATA:
2810 case EXT4_MOUNT_WRITEBACK_DATA:
2811 if (!jbd2_journal_check_available_features
2812 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2813 ext4_msg(sb, KERN_ERR, "Journal does not support "
2814 "requested data journaling mode");
2820 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2824 if (test_opt(sb, NOBH)) {
2825 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2826 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2827 "its supported only with writeback mode");
2828 clear_opt(sbi->s_mount_opt, NOBH);
2831 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2832 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2833 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2834 goto failed_mount_wq;
2838 * The jbd2_journal_load will have done any necessary log recovery,
2839 * so we can safely mount the rest of the filesystem now.
2842 root = ext4_iget(sb, EXT4_ROOT_INO);
2844 ext4_msg(sb, KERN_ERR, "get root inode failed");
2845 ret = PTR_ERR(root);
2848 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2850 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2853 sb->s_root = d_alloc_root(root);
2855 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2861 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2863 /* determine the minimum size of new large inodes, if present */
2864 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2865 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2866 EXT4_GOOD_OLD_INODE_SIZE;
2867 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2868 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2869 if (sbi->s_want_extra_isize <
2870 le16_to_cpu(es->s_want_extra_isize))
2871 sbi->s_want_extra_isize =
2872 le16_to_cpu(es->s_want_extra_isize);
2873 if (sbi->s_want_extra_isize <
2874 le16_to_cpu(es->s_min_extra_isize))
2875 sbi->s_want_extra_isize =
2876 le16_to_cpu(es->s_min_extra_isize);
2879 /* Check if enough inode space is available */
2880 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2881 sbi->s_inode_size) {
2882 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2883 EXT4_GOOD_OLD_INODE_SIZE;
2884 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2888 if (test_opt(sb, DELALLOC) &&
2889 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2890 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2891 "requested data journaling mode");
2892 clear_opt(sbi->s_mount_opt, DELALLOC);
2895 err = ext4_setup_system_zone(sb);
2897 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2898 "zone (%d)\n", err);
2903 err = ext4_mb_init(sb, needs_recovery);
2905 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2910 sbi->s_kobj.kset = ext4_kset;
2911 init_completion(&sbi->s_kobj_unregister);
2912 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2915 ext4_mb_release(sb);
2916 ext4_ext_release(sb);
2920 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2921 ext4_orphan_cleanup(sb, es);
2922 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2923 if (needs_recovery) {
2924 ext4_msg(sb, KERN_INFO, "recovery complete");
2925 ext4_mark_recovery_complete(sb, es);
2927 if (EXT4_SB(sb)->s_journal) {
2928 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2929 descr = " journalled data mode";
2930 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2931 descr = " ordered data mode";
2933 descr = " writeback data mode";
2935 descr = "out journal";
2937 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2944 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2948 ext4_msg(sb, KERN_ERR, "mount failed");
2949 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2951 ext4_release_system_zone(sb);
2952 if (sbi->s_journal) {
2953 jbd2_journal_destroy(sbi->s_journal);
2954 sbi->s_journal = NULL;
2957 if (sbi->s_flex_groups) {
2958 if (is_vmalloc_addr(sbi->s_flex_groups))
2959 vfree(sbi->s_flex_groups);
2961 kfree(sbi->s_flex_groups);
2963 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2964 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2965 percpu_counter_destroy(&sbi->s_dirs_counter);
2966 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2968 for (i = 0; i < db_count; i++)
2969 brelse(sbi->s_group_desc[i]);
2970 kfree(sbi->s_group_desc);
2973 remove_proc_entry(sb->s_id, ext4_proc_root);
2976 for (i = 0; i < MAXQUOTAS; i++)
2977 kfree(sbi->s_qf_names[i]);
2979 ext4_blkdev_remove(sbi);
2982 sb->s_fs_info = NULL;
2983 kfree(sbi->s_blockgroup_lock);
2990 * Setup any per-fs journal parameters now. We'll do this both on
2991 * initial mount, once the journal has been initialised but before we've
2992 * done any recovery; and again on any subsequent remount.
2994 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
2998 journal->j_commit_interval = sbi->s_commit_interval;
2999 journal->j_min_batch_time = sbi->s_min_batch_time;
3000 journal->j_max_batch_time = sbi->s_max_batch_time;
3002 spin_lock(&journal->j_state_lock);
3003 if (test_opt(sb, BARRIER))
3004 journal->j_flags |= JBD2_BARRIER;
3006 journal->j_flags &= ~JBD2_BARRIER;
3007 if (test_opt(sb, DATA_ERR_ABORT))
3008 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3010 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3011 spin_unlock(&journal->j_state_lock);
3014 static journal_t *ext4_get_journal(struct super_block *sb,
3015 unsigned int journal_inum)
3017 struct inode *journal_inode;
3020 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3022 /* First, test for the existence of a valid inode on disk. Bad
3023 * things happen if we iget() an unused inode, as the subsequent
3024 * iput() will try to delete it. */
3026 journal_inode = ext4_iget(sb, journal_inum);
3027 if (IS_ERR(journal_inode)) {
3028 ext4_msg(sb, KERN_ERR, "no journal found");
3031 if (!journal_inode->i_nlink) {
3032 make_bad_inode(journal_inode);
3033 iput(journal_inode);
3034 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3038 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3039 journal_inode, journal_inode->i_size);
3040 if (!S_ISREG(journal_inode->i_mode)) {
3041 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3042 iput(journal_inode);
3046 journal = jbd2_journal_init_inode(journal_inode);
3048 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3049 iput(journal_inode);
3052 journal->j_private = sb;
3053 ext4_init_journal_params(sb, journal);
3057 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3060 struct buffer_head *bh;
3064 int hblock, blocksize;
3065 ext4_fsblk_t sb_block;
3066 unsigned long offset;
3067 struct ext4_super_block *es;
3068 struct block_device *bdev;
3070 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3072 bdev = ext4_blkdev_get(j_dev, sb);
3076 if (bd_claim(bdev, sb)) {
3077 ext4_msg(sb, KERN_ERR,
3078 "failed to claim external journal device");
3079 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3083 blocksize = sb->s_blocksize;
3084 hblock = bdev_logical_block_size(bdev);
3085 if (blocksize < hblock) {
3086 ext4_msg(sb, KERN_ERR,
3087 "blocksize too small for journal device");
3091 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3092 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3093 set_blocksize(bdev, blocksize);
3094 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3095 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3096 "external journal");
3100 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3101 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3102 !(le32_to_cpu(es->s_feature_incompat) &
3103 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3104 ext4_msg(sb, KERN_ERR, "external journal has "
3110 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3111 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3116 len = ext4_blocks_count(es);
3117 start = sb_block + 1;
3118 brelse(bh); /* we're done with the superblock */
3120 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3121 start, len, blocksize);
3123 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3126 journal->j_private = sb;
3127 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3128 wait_on_buffer(journal->j_sb_buffer);
3129 if (!buffer_uptodate(journal->j_sb_buffer)) {
3130 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3133 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3134 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3135 "user (unsupported) - %d",
3136 be32_to_cpu(journal->j_superblock->s_nr_users));
3139 EXT4_SB(sb)->journal_bdev = bdev;
3140 ext4_init_journal_params(sb, journal);
3144 jbd2_journal_destroy(journal);
3146 ext4_blkdev_put(bdev);
3150 static int ext4_load_journal(struct super_block *sb,
3151 struct ext4_super_block *es,
3152 unsigned long journal_devnum)
3155 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3158 int really_read_only;
3160 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3162 if (journal_devnum &&
3163 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3164 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3165 "numbers have changed");
3166 journal_dev = new_decode_dev(journal_devnum);
3168 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3170 really_read_only = bdev_read_only(sb->s_bdev);
3173 * Are we loading a blank journal or performing recovery after a
3174 * crash? For recovery, we need to check in advance whether we
3175 * can get read-write access to the device.
3177 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3178 if (sb->s_flags & MS_RDONLY) {
3179 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3180 "required on readonly filesystem");
3181 if (really_read_only) {
3182 ext4_msg(sb, KERN_ERR, "write access "
3183 "unavailable, cannot proceed");
3186 ext4_msg(sb, KERN_INFO, "write access will "
3187 "be enabled during recovery");
3191 if (journal_inum && journal_dev) {
3192 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3193 "and inode journals!");
3198 if (!(journal = ext4_get_journal(sb, journal_inum)))
3201 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3205 if (!(journal->j_flags & JBD2_BARRIER))
3206 ext4_msg(sb, KERN_INFO, "barriers disabled");
3208 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3209 err = jbd2_journal_update_format(journal);
3211 ext4_msg(sb, KERN_ERR, "error updating journal");
3212 jbd2_journal_destroy(journal);
3217 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3218 err = jbd2_journal_wipe(journal, !really_read_only);
3220 err = jbd2_journal_load(journal);
3223 ext4_msg(sb, KERN_ERR, "error loading journal");
3224 jbd2_journal_destroy(journal);
3228 EXT4_SB(sb)->s_journal = journal;
3229 ext4_clear_journal_err(sb, es);
3231 if (journal_devnum &&
3232 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3233 es->s_journal_dev = cpu_to_le32(journal_devnum);
3235 /* Make sure we flush the recovery flag to disk. */
3236 ext4_commit_super(sb, 1);
3242 static int ext4_commit_super(struct super_block *sb, int sync)
3244 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3245 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3250 if (buffer_write_io_error(sbh)) {
3252 * Oh, dear. A previous attempt to write the
3253 * superblock failed. This could happen because the
3254 * USB device was yanked out. Or it could happen to
3255 * be a transient write error and maybe the block will
3256 * be remapped. Nothing we can do but to retry the
3257 * write and hope for the best.
3259 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3260 "superblock detected");
3261 clear_buffer_write_io_error(sbh);
3262 set_buffer_uptodate(sbh);
3265 * If the file system is mounted read-only, don't update the
3266 * superblock write time. This avoids updating the superblock
3267 * write time when we are mounting the root file system
3268 * read/only but we need to replay the journal; at that point,
3269 * for people who are east of GMT and who make their clock
3270 * tick in localtime for Windows bug-for-bug compatibility,
3271 * the clock is set in the future, and this will cause e2fsck
3272 * to complain and force a full file system check.
3274 if (!(sb->s_flags & MS_RDONLY))
3275 es->s_wtime = cpu_to_le32(get_seconds());
3276 es->s_kbytes_written =
3277 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3278 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3279 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3280 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3281 &EXT4_SB(sb)->s_freeblocks_counter));
3282 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3283 &EXT4_SB(sb)->s_freeinodes_counter));
3285 BUFFER_TRACE(sbh, "marking dirty");
3286 mark_buffer_dirty(sbh);
3288 error = sync_dirty_buffer(sbh);
3292 error = buffer_write_io_error(sbh);
3294 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3296 clear_buffer_write_io_error(sbh);
3297 set_buffer_uptodate(sbh);
3304 * Have we just finished recovery? If so, and if we are mounting (or
3305 * remounting) the filesystem readonly, then we will end up with a
3306 * consistent fs on disk. Record that fact.
3308 static void ext4_mark_recovery_complete(struct super_block *sb,
3309 struct ext4_super_block *es)
3311 journal_t *journal = EXT4_SB(sb)->s_journal;
3313 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3314 BUG_ON(journal != NULL);
3317 jbd2_journal_lock_updates(journal);
3318 if (jbd2_journal_flush(journal) < 0)
3321 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3322 sb->s_flags & MS_RDONLY) {
3323 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3324 ext4_commit_super(sb, 1);
3328 jbd2_journal_unlock_updates(journal);
3332 * If we are mounting (or read-write remounting) a filesystem whose journal
3333 * has recorded an error from a previous lifetime, move that error to the
3334 * main filesystem now.
3336 static void ext4_clear_journal_err(struct super_block *sb,
3337 struct ext4_super_block *es)
3343 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3345 journal = EXT4_SB(sb)->s_journal;
3348 * Now check for any error status which may have been recorded in the
3349 * journal by a prior ext4_error() or ext4_abort()
3352 j_errno = jbd2_journal_errno(journal);
3356 errstr = ext4_decode_error(sb, j_errno, nbuf);
3357 ext4_warning(sb, __func__, "Filesystem error recorded "
3358 "from previous mount: %s", errstr);
3359 ext4_warning(sb, __func__, "Marking fs in need of "
3360 "filesystem check.");
3362 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3363 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3364 ext4_commit_super(sb, 1);
3366 jbd2_journal_clear_err(journal);
3371 * Force the running and committing transactions to commit,
3372 * and wait on the commit.
3374 int ext4_force_commit(struct super_block *sb)
3379 if (sb->s_flags & MS_RDONLY)
3382 journal = EXT4_SB(sb)->s_journal;
3384 ret = ext4_journal_force_commit(journal);
3389 static void ext4_write_super(struct super_block *sb)
3392 ext4_commit_super(sb, 1);
3396 static int ext4_sync_fs(struct super_block *sb, int wait)
3400 struct ext4_sb_info *sbi = EXT4_SB(sb);
3402 trace_ext4_sync_fs(sb, wait);
3403 flush_workqueue(sbi->dio_unwritten_wq);
3404 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3406 jbd2_log_wait_commit(sbi->s_journal, target);
3412 * LVM calls this function before a (read-only) snapshot is created. This
3413 * gives us a chance to flush the journal completely and mark the fs clean.
3415 static int ext4_freeze(struct super_block *sb)
3420 if (sb->s_flags & MS_RDONLY)
3423 journal = EXT4_SB(sb)->s_journal;
3425 /* Now we set up the journal barrier. */
3426 jbd2_journal_lock_updates(journal);
3429 * Don't clear the needs_recovery flag if we failed to flush
3432 error = jbd2_journal_flush(journal);
3435 jbd2_journal_unlock_updates(journal);
3439 /* Journal blocked and flushed, clear needs_recovery flag. */
3440 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3441 error = ext4_commit_super(sb, 1);
3448 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3449 * flag here, even though the filesystem is not technically dirty yet.
3451 static int ext4_unfreeze(struct super_block *sb)
3453 if (sb->s_flags & MS_RDONLY)
3457 /* Reset the needs_recovery flag before the fs is unlocked. */
3458 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3459 ext4_commit_super(sb, 1);
3461 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3465 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3467 struct ext4_super_block *es;
3468 struct ext4_sb_info *sbi = EXT4_SB(sb);
3469 ext4_fsblk_t n_blocks_count = 0;
3470 unsigned long old_sb_flags;
3471 struct ext4_mount_options old_opts;
3473 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3481 /* Store the original options */
3483 old_sb_flags = sb->s_flags;
3484 old_opts.s_mount_opt = sbi->s_mount_opt;
3485 old_opts.s_resuid = sbi->s_resuid;
3486 old_opts.s_resgid = sbi->s_resgid;
3487 old_opts.s_commit_interval = sbi->s_commit_interval;
3488 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3489 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3491 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3492 for (i = 0; i < MAXQUOTAS; i++)
3493 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3495 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3496 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3499 * Allow the "check" option to be passed as a remount option.
3501 if (!parse_options(data, sb, NULL, &journal_ioprio,
3502 &n_blocks_count, 1)) {
3507 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3508 ext4_abort(sb, __func__, "Abort forced by user");
3510 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3511 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3515 if (sbi->s_journal) {
3516 ext4_init_journal_params(sb, sbi->s_journal);
3517 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3520 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3521 n_blocks_count > ext4_blocks_count(es)) {
3522 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3527 if (*flags & MS_RDONLY) {
3529 * First of all, the unconditional stuff we have to do
3530 * to disable replay of the journal when we next remount
3532 sb->s_flags |= MS_RDONLY;
3535 * OK, test if we are remounting a valid rw partition
3536 * readonly, and if so set the rdonly flag and then
3537 * mark the partition as valid again.
3539 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3540 (sbi->s_mount_state & EXT4_VALID_FS))
3541 es->s_state = cpu_to_le16(sbi->s_mount_state);
3544 ext4_mark_recovery_complete(sb, es);
3546 /* Make sure we can mount this feature set readwrite */
3547 if (!ext4_feature_set_ok(sb, 0)) {
3552 * Make sure the group descriptor checksums
3553 * are sane. If they aren't, refuse to remount r/w.
3555 for (g = 0; g < sbi->s_groups_count; g++) {
3556 struct ext4_group_desc *gdp =
3557 ext4_get_group_desc(sb, g, NULL);
3559 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3560 ext4_msg(sb, KERN_ERR,
3561 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3562 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3563 le16_to_cpu(gdp->bg_checksum));
3570 * If we have an unprocessed orphan list hanging
3571 * around from a previously readonly bdev mount,
3572 * require a full umount/remount for now.
3574 if (es->s_last_orphan) {
3575 ext4_msg(sb, KERN_WARNING, "Couldn't "
3576 "remount RDWR because of unprocessed "
3577 "orphan inode list. Please "
3578 "umount/remount instead");
3584 * Mounting a RDONLY partition read-write, so reread
3585 * and store the current valid flag. (It may have
3586 * been changed by e2fsck since we originally mounted
3590 ext4_clear_journal_err(sb, es);
3591 sbi->s_mount_state = le16_to_cpu(es->s_state);
3592 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3594 if (!ext4_setup_super(sb, es, 0))
3595 sb->s_flags &= ~MS_RDONLY;
3598 ext4_setup_system_zone(sb);
3599 if (sbi->s_journal == NULL)
3600 ext4_commit_super(sb, 1);
3603 /* Release old quota file names */
3604 for (i = 0; i < MAXQUOTAS; i++)
3605 if (old_opts.s_qf_names[i] &&
3606 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3607 kfree(old_opts.s_qf_names[i]);
3614 sb->s_flags = old_sb_flags;
3615 sbi->s_mount_opt = old_opts.s_mount_opt;
3616 sbi->s_resuid = old_opts.s_resuid;
3617 sbi->s_resgid = old_opts.s_resgid;
3618 sbi->s_commit_interval = old_opts.s_commit_interval;
3619 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3620 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3622 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3623 for (i = 0; i < MAXQUOTAS; i++) {
3624 if (sbi->s_qf_names[i] &&
3625 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3626 kfree(sbi->s_qf_names[i]);
3627 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3635 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3637 struct super_block *sb = dentry->d_sb;
3638 struct ext4_sb_info *sbi = EXT4_SB(sb);
3639 struct ext4_super_block *es = sbi->s_es;
3642 if (test_opt(sb, MINIX_DF)) {
3643 sbi->s_overhead_last = 0;
3644 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3645 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3646 ext4_fsblk_t overhead = 0;
3649 * Compute the overhead (FS structures). This is constant
3650 * for a given filesystem unless the number of block groups
3651 * changes so we cache the previous value until it does.
3655 * All of the blocks before first_data_block are
3658 overhead = le32_to_cpu(es->s_first_data_block);
3661 * Add the overhead attributed to the superblock and
3662 * block group descriptors. If the sparse superblocks
3663 * feature is turned on, then not all groups have this.
3665 for (i = 0; i < ngroups; i++) {
3666 overhead += ext4_bg_has_super(sb, i) +
3667 ext4_bg_num_gdb(sb, i);
3672 * Every block group has an inode bitmap, a block
3673 * bitmap, and an inode table.
3675 overhead += ngroups * (2 + sbi->s_itb_per_group);
3676 sbi->s_overhead_last = overhead;
3678 sbi->s_blocks_last = ext4_blocks_count(es);
3681 buf->f_type = EXT4_SUPER_MAGIC;
3682 buf->f_bsize = sb->s_blocksize;
3683 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3684 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3685 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3686 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3687 if (buf->f_bfree < ext4_r_blocks_count(es))
3689 buf->f_files = le32_to_cpu(es->s_inodes_count);
3690 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3691 buf->f_namelen = EXT4_NAME_LEN;
3692 fsid = le64_to_cpup((void *)es->s_uuid) ^
3693 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3694 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3695 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3700 /* Helper function for writing quotas on sync - we need to start transaction
3701 * before quota file is locked for write. Otherwise the are possible deadlocks:
3702 * Process 1 Process 2
3703 * ext4_create() quota_sync()
3704 * jbd2_journal_start() write_dquot()
3705 * vfs_dq_init() down(dqio_mutex)
3706 * down(dqio_mutex) jbd2_journal_start()
3712 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3714 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3717 static int ext4_write_dquot(struct dquot *dquot)
3721 struct inode *inode;
3723 inode = dquot_to_inode(dquot);
3724 handle = ext4_journal_start(inode,
3725 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3727 return PTR_ERR(handle);
3728 ret = dquot_commit(dquot);
3729 err = ext4_journal_stop(handle);
3735 static int ext4_acquire_dquot(struct dquot *dquot)
3740 handle = ext4_journal_start(dquot_to_inode(dquot),
3741 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3743 return PTR_ERR(handle);
3744 ret = dquot_acquire(dquot);
3745 err = ext4_journal_stop(handle);
3751 static int ext4_release_dquot(struct dquot *dquot)
3756 handle = ext4_journal_start(dquot_to_inode(dquot),
3757 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3758 if (IS_ERR(handle)) {
3759 /* Release dquot anyway to avoid endless cycle in dqput() */
3760 dquot_release(dquot);
3761 return PTR_ERR(handle);
3763 ret = dquot_release(dquot);
3764 err = ext4_journal_stop(handle);
3770 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3772 /* Are we journaling quotas? */
3773 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3774 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3775 dquot_mark_dquot_dirty(dquot);
3776 return ext4_write_dquot(dquot);
3778 return dquot_mark_dquot_dirty(dquot);
3782 static int ext4_write_info(struct super_block *sb, int type)
3787 /* Data block + inode block */
3788 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3790 return PTR_ERR(handle);
3791 ret = dquot_commit_info(sb, type);
3792 err = ext4_journal_stop(handle);
3799 * Turn on quotas during mount time - we need to find
3800 * the quota file and such...
3802 static int ext4_quota_on_mount(struct super_block *sb, int type)
3804 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3805 EXT4_SB(sb)->s_jquota_fmt, type);
3809 * Standard function to be called on quota_on
3811 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3812 char *name, int remount)
3817 if (!test_opt(sb, QUOTA))
3819 /* When remounting, no checks are needed and in fact, name is NULL */
3821 return vfs_quota_on(sb, type, format_id, name, remount);
3823 err = kern_path(name, LOOKUP_FOLLOW, &path);
3827 /* Quotafile not on the same filesystem? */
3828 if (path.mnt->mnt_sb != sb) {
3832 /* Journaling quota? */
3833 if (EXT4_SB(sb)->s_qf_names[type]) {
3834 /* Quotafile not in fs root? */
3835 if (path.dentry->d_parent != sb->s_root)
3836 ext4_msg(sb, KERN_WARNING,
3837 "Quota file not on filesystem root. "
3838 "Journaled quota will not work");
3842 * When we journal data on quota file, we have to flush journal to see
3843 * all updates to the file when we bypass pagecache...
3845 if (EXT4_SB(sb)->s_journal &&
3846 ext4_should_journal_data(path.dentry->d_inode)) {
3848 * We don't need to lock updates but journal_flush() could
3849 * otherwise be livelocked...
3851 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3852 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3853 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3860 err = vfs_quota_on_path(sb, type, format_id, &path);
3865 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3866 * acquiring the locks... As quota files are never truncated and quota code
3867 * itself serializes the operations (and noone else should touch the files)
3868 * we don't have to be afraid of races */
3869 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3870 size_t len, loff_t off)
3872 struct inode *inode = sb_dqopt(sb)->files[type];
3873 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3875 int offset = off & (sb->s_blocksize - 1);
3878 struct buffer_head *bh;
3879 loff_t i_size = i_size_read(inode);
3883 if (off+len > i_size)
3886 while (toread > 0) {
3887 tocopy = sb->s_blocksize - offset < toread ?
3888 sb->s_blocksize - offset : toread;
3889 bh = ext4_bread(NULL, inode, blk, 0, &err);
3892 if (!bh) /* A hole? */
3893 memset(data, 0, tocopy);
3895 memcpy(data, bh->b_data+offset, tocopy);
3905 /* Write to quotafile (we know the transaction is already started and has
3906 * enough credits) */
3907 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3908 const char *data, size_t len, loff_t off)
3910 struct inode *inode = sb_dqopt(sb)->files[type];
3911 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3913 int offset = off & (sb->s_blocksize - 1);
3915 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3916 size_t towrite = len;
3917 struct buffer_head *bh;
3918 handle_t *handle = journal_current_handle();
3920 if (EXT4_SB(sb)->s_journal && !handle) {
3921 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3922 " cancelled because transaction is not started",
3923 (unsigned long long)off, (unsigned long long)len);
3926 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3927 while (towrite > 0) {
3928 tocopy = sb->s_blocksize - offset < towrite ?
3929 sb->s_blocksize - offset : towrite;
3930 bh = ext4_bread(handle, inode, blk, 1, &err);
3933 if (journal_quota) {
3934 err = ext4_journal_get_write_access(handle, bh);
3941 memcpy(bh->b_data+offset, data, tocopy);
3942 flush_dcache_page(bh->b_page);
3945 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3947 /* Always do at least ordered writes for quotas */
3948 err = ext4_jbd2_file_inode(handle, inode);
3949 mark_buffer_dirty(bh);
3960 if (len == towrite) {
3961 mutex_unlock(&inode->i_mutex);
3964 if (inode->i_size < off+len-towrite) {
3965 i_size_write(inode, off+len-towrite);
3966 EXT4_I(inode)->i_disksize = inode->i_size;
3968 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3969 ext4_mark_inode_dirty(handle, inode);
3970 mutex_unlock(&inode->i_mutex);
3971 return len - towrite;
3976 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3977 const char *dev_name, void *data, struct vfsmount *mnt)
3979 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3982 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3983 static struct file_system_type ext2_fs_type = {
3984 .owner = THIS_MODULE,
3986 .get_sb = ext4_get_sb,
3987 .kill_sb = kill_block_super,
3988 .fs_flags = FS_REQUIRES_DEV,
3991 static inline void register_as_ext2(void)
3993 int err = register_filesystem(&ext2_fs_type);
3996 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
3999 static inline void unregister_as_ext2(void)
4001 unregister_filesystem(&ext2_fs_type);
4004 static inline void register_as_ext2(void) { }
4005 static inline void unregister_as_ext2(void) { }
4008 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4009 static struct file_system_type ext3_fs_type = {
4010 .owner = THIS_MODULE,
4012 .get_sb = ext4_get_sb,
4013 .kill_sb = kill_block_super,
4014 .fs_flags = FS_REQUIRES_DEV,
4017 static inline void register_as_ext3(void)
4019 int err = register_filesystem(&ext3_fs_type);
4022 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4025 static inline void unregister_as_ext3(void)
4027 unregister_filesystem(&ext3_fs_type);
4030 static inline void register_as_ext3(void) { }
4031 static inline void unregister_as_ext3(void) { }
4034 static struct file_system_type ext4_fs_type = {
4035 .owner = THIS_MODULE,
4037 .get_sb = ext4_get_sb,
4038 .kill_sb = kill_block_super,
4039 .fs_flags = FS_REQUIRES_DEV,
4042 static int __init init_ext4_fs(void)
4046 err = init_ext4_system_zone();
4049 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4052 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4053 err = init_ext4_mballoc();
4057 err = init_ext4_xattr();
4060 err = init_inodecache();
4065 err = register_filesystem(&ext4_fs_type);
4070 unregister_as_ext2();
4071 unregister_as_ext3();
4072 destroy_inodecache();
4076 exit_ext4_mballoc();
4078 remove_proc_entry("fs/ext4", NULL);
4079 kset_unregister(ext4_kset);
4081 exit_ext4_system_zone();
4085 static void __exit exit_ext4_fs(void)
4087 unregister_as_ext2();
4088 unregister_as_ext3();
4089 unregister_filesystem(&ext4_fs_type);
4090 destroy_inodecache();
4092 exit_ext4_mballoc();
4093 remove_proc_entry("fs/ext4", NULL);
4094 kset_unregister(ext4_kset);
4095 exit_ext4_system_zone();
4098 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4099 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4100 MODULE_LICENSE("GPL");
4101 module_init(init_ext4_fs)
4102 module_exit(exit_ext4_fs)