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_da_metadata_calc_len = 0;
706 ei->i_delalloc_reserved_flag = 0;
707 spin_lock_init(&(ei->i_block_reservation_lock));
709 ei->i_reserved_quota = 0;
711 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
712 ei->cur_aio_dio = NULL;
714 ei->i_datasync_tid = 0;
716 return &ei->vfs_inode;
719 static void ext4_destroy_inode(struct inode *inode)
721 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
722 ext4_msg(inode->i_sb, KERN_ERR,
723 "Inode %lu (%p): orphan list check failed!",
724 inode->i_ino, EXT4_I(inode));
725 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
726 EXT4_I(inode), sizeof(struct ext4_inode_info),
730 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
733 static void init_once(void *foo)
735 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
737 INIT_LIST_HEAD(&ei->i_orphan);
738 #ifdef CONFIG_EXT4_FS_XATTR
739 init_rwsem(&ei->xattr_sem);
741 init_rwsem(&ei->i_data_sem);
742 inode_init_once(&ei->vfs_inode);
745 static int init_inodecache(void)
747 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
748 sizeof(struct ext4_inode_info),
749 0, (SLAB_RECLAIM_ACCOUNT|
752 if (ext4_inode_cachep == NULL)
757 static void destroy_inodecache(void)
759 kmem_cache_destroy(ext4_inode_cachep);
762 static void ext4_clear_inode(struct inode *inode)
764 ext4_discard_preallocations(inode);
765 if (EXT4_JOURNAL(inode))
766 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
767 &EXT4_I(inode)->jinode);
770 static inline void ext4_show_quota_options(struct seq_file *seq,
771 struct super_block *sb)
773 #if defined(CONFIG_QUOTA)
774 struct ext4_sb_info *sbi = EXT4_SB(sb);
776 if (sbi->s_jquota_fmt) {
779 switch (sbi->s_jquota_fmt) {
790 seq_printf(seq, ",jqfmt=%s", fmtname);
793 if (sbi->s_qf_names[USRQUOTA])
794 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
796 if (sbi->s_qf_names[GRPQUOTA])
797 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
799 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
800 seq_puts(seq, ",usrquota");
802 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
803 seq_puts(seq, ",grpquota");
809 * - it's set to a non-default value OR
810 * - if the per-sb default is different from the global default
812 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
815 unsigned long def_mount_opts;
816 struct super_block *sb = vfs->mnt_sb;
817 struct ext4_sb_info *sbi = EXT4_SB(sb);
818 struct ext4_super_block *es = sbi->s_es;
820 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
821 def_errors = le16_to_cpu(es->s_errors);
823 if (sbi->s_sb_block != 1)
824 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
825 if (test_opt(sb, MINIX_DF))
826 seq_puts(seq, ",minixdf");
827 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
828 seq_puts(seq, ",grpid");
829 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
830 seq_puts(seq, ",nogrpid");
831 if (sbi->s_resuid != EXT4_DEF_RESUID ||
832 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
833 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
835 if (sbi->s_resgid != EXT4_DEF_RESGID ||
836 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
837 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
839 if (test_opt(sb, ERRORS_RO)) {
840 if (def_errors == EXT4_ERRORS_PANIC ||
841 def_errors == EXT4_ERRORS_CONTINUE) {
842 seq_puts(seq, ",errors=remount-ro");
845 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
846 seq_puts(seq, ",errors=continue");
847 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
848 seq_puts(seq, ",errors=panic");
849 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
850 seq_puts(seq, ",nouid32");
851 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
852 seq_puts(seq, ",debug");
853 if (test_opt(sb, OLDALLOC))
854 seq_puts(seq, ",oldalloc");
855 #ifdef CONFIG_EXT4_FS_XATTR
856 if (test_opt(sb, XATTR_USER) &&
857 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
858 seq_puts(seq, ",user_xattr");
859 if (!test_opt(sb, XATTR_USER) &&
860 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
861 seq_puts(seq, ",nouser_xattr");
864 #ifdef CONFIG_EXT4_FS_POSIX_ACL
865 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
866 seq_puts(seq, ",acl");
867 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
868 seq_puts(seq, ",noacl");
870 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
871 seq_printf(seq, ",commit=%u",
872 (unsigned) (sbi->s_commit_interval / HZ));
874 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
875 seq_printf(seq, ",min_batch_time=%u",
876 (unsigned) sbi->s_min_batch_time);
878 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
879 seq_printf(seq, ",max_batch_time=%u",
880 (unsigned) sbi->s_min_batch_time);
884 * We're changing the default of barrier mount option, so
885 * let's always display its mount state so it's clear what its
888 seq_puts(seq, ",barrier=");
889 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
890 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
891 seq_puts(seq, ",journal_async_commit");
892 if (test_opt(sb, NOBH))
893 seq_puts(seq, ",nobh");
894 if (test_opt(sb, I_VERSION))
895 seq_puts(seq, ",i_version");
896 if (!test_opt(sb, DELALLOC))
897 seq_puts(seq, ",nodelalloc");
901 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
903 * journal mode get enabled in different ways
904 * So just print the value even if we didn't specify it
906 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
907 seq_puts(seq, ",data=journal");
908 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
909 seq_puts(seq, ",data=ordered");
910 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
911 seq_puts(seq, ",data=writeback");
913 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
914 seq_printf(seq, ",inode_readahead_blks=%u",
915 sbi->s_inode_readahead_blks);
917 if (test_opt(sb, DATA_ERR_ABORT))
918 seq_puts(seq, ",data_err=abort");
920 if (test_opt(sb, NO_AUTO_DA_ALLOC))
921 seq_puts(seq, ",noauto_da_alloc");
923 if (test_opt(sb, DISCARD))
924 seq_puts(seq, ",discard");
926 if (test_opt(sb, NOLOAD))
927 seq_puts(seq, ",norecovery");
929 ext4_show_quota_options(seq, sb);
934 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
935 u64 ino, u32 generation)
939 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
940 return ERR_PTR(-ESTALE);
941 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
942 return ERR_PTR(-ESTALE);
944 /* iget isn't really right if the inode is currently unallocated!!
946 * ext4_read_inode will return a bad_inode if the inode had been
947 * deleted, so we should be safe.
949 * Currently we don't know the generation for parent directory, so
950 * a generation of 0 means "accept any"
952 inode = ext4_iget(sb, ino);
954 return ERR_CAST(inode);
955 if (generation && inode->i_generation != generation) {
957 return ERR_PTR(-ESTALE);
963 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
964 int fh_len, int fh_type)
966 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
970 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
971 int fh_len, int fh_type)
973 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
978 * Try to release metadata pages (indirect blocks, directories) which are
979 * mapped via the block device. Since these pages could have journal heads
980 * which would prevent try_to_free_buffers() from freeing them, we must use
981 * jbd2 layer's try_to_free_buffers() function to release them.
983 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
986 journal_t *journal = EXT4_SB(sb)->s_journal;
988 WARN_ON(PageChecked(page));
989 if (!page_has_buffers(page))
992 return jbd2_journal_try_to_free_buffers(journal, page,
994 return try_to_free_buffers(page);
998 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
999 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1001 static int ext4_write_dquot(struct dquot *dquot);
1002 static int ext4_acquire_dquot(struct dquot *dquot);
1003 static int ext4_release_dquot(struct dquot *dquot);
1004 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1005 static int ext4_write_info(struct super_block *sb, int type);
1006 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1007 char *path, int remount);
1008 static int ext4_quota_on_mount(struct super_block *sb, int type);
1009 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1010 size_t len, loff_t off);
1011 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1012 const char *data, size_t len, loff_t off);
1014 static const struct dquot_operations ext4_quota_operations = {
1015 .initialize = dquot_initialize,
1018 .get_reserved_space = ext4_get_reserved_space,
1020 .transfer = dquot_transfer,
1021 .write_dquot = ext4_write_dquot,
1022 .acquire_dquot = ext4_acquire_dquot,
1023 .release_dquot = ext4_release_dquot,
1024 .mark_dirty = ext4_mark_dquot_dirty,
1025 .write_info = ext4_write_info,
1026 .alloc_dquot = dquot_alloc,
1027 .destroy_dquot = dquot_destroy,
1030 static const struct quotactl_ops ext4_qctl_operations = {
1031 .quota_on = ext4_quota_on,
1032 .quota_off = vfs_quota_off,
1033 .quota_sync = vfs_quota_sync,
1034 .get_info = vfs_get_dqinfo,
1035 .set_info = vfs_set_dqinfo,
1036 .get_dqblk = vfs_get_dqblk,
1037 .set_dqblk = vfs_set_dqblk
1041 static const struct super_operations ext4_sops = {
1042 .alloc_inode = ext4_alloc_inode,
1043 .destroy_inode = ext4_destroy_inode,
1044 .write_inode = ext4_write_inode,
1045 .dirty_inode = ext4_dirty_inode,
1046 .delete_inode = ext4_delete_inode,
1047 .put_super = ext4_put_super,
1048 .sync_fs = ext4_sync_fs,
1049 .freeze_fs = ext4_freeze,
1050 .unfreeze_fs = ext4_unfreeze,
1051 .statfs = ext4_statfs,
1052 .remount_fs = ext4_remount,
1053 .clear_inode = ext4_clear_inode,
1054 .show_options = ext4_show_options,
1056 .quota_read = ext4_quota_read,
1057 .quota_write = ext4_quota_write,
1059 .bdev_try_to_free_page = bdev_try_to_free_page,
1062 static const struct super_operations ext4_nojournal_sops = {
1063 .alloc_inode = ext4_alloc_inode,
1064 .destroy_inode = ext4_destroy_inode,
1065 .write_inode = ext4_write_inode,
1066 .dirty_inode = ext4_dirty_inode,
1067 .delete_inode = ext4_delete_inode,
1068 .write_super = ext4_write_super,
1069 .put_super = ext4_put_super,
1070 .statfs = ext4_statfs,
1071 .remount_fs = ext4_remount,
1072 .clear_inode = ext4_clear_inode,
1073 .show_options = ext4_show_options,
1075 .quota_read = ext4_quota_read,
1076 .quota_write = ext4_quota_write,
1078 .bdev_try_to_free_page = bdev_try_to_free_page,
1081 static const struct export_operations ext4_export_ops = {
1082 .fh_to_dentry = ext4_fh_to_dentry,
1083 .fh_to_parent = ext4_fh_to_parent,
1084 .get_parent = ext4_get_parent,
1088 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1089 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1090 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1091 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1092 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1093 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1094 Opt_journal_update, Opt_journal_dev,
1095 Opt_journal_checksum, Opt_journal_async_commit,
1096 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1097 Opt_data_err_abort, Opt_data_err_ignore,
1098 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1099 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1100 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1101 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1102 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1103 Opt_block_validity, Opt_noblock_validity,
1104 Opt_inode_readahead_blks, Opt_journal_ioprio,
1105 Opt_discard, Opt_nodiscard,
1108 static const match_table_t tokens = {
1109 {Opt_bsd_df, "bsddf"},
1110 {Opt_minix_df, "minixdf"},
1111 {Opt_grpid, "grpid"},
1112 {Opt_grpid, "bsdgroups"},
1113 {Opt_nogrpid, "nogrpid"},
1114 {Opt_nogrpid, "sysvgroups"},
1115 {Opt_resgid, "resgid=%u"},
1116 {Opt_resuid, "resuid=%u"},
1118 {Opt_err_cont, "errors=continue"},
1119 {Opt_err_panic, "errors=panic"},
1120 {Opt_err_ro, "errors=remount-ro"},
1121 {Opt_nouid32, "nouid32"},
1122 {Opt_debug, "debug"},
1123 {Opt_oldalloc, "oldalloc"},
1124 {Opt_orlov, "orlov"},
1125 {Opt_user_xattr, "user_xattr"},
1126 {Opt_nouser_xattr, "nouser_xattr"},
1128 {Opt_noacl, "noacl"},
1129 {Opt_noload, "noload"},
1130 {Opt_noload, "norecovery"},
1133 {Opt_commit, "commit=%u"},
1134 {Opt_min_batch_time, "min_batch_time=%u"},
1135 {Opt_max_batch_time, "max_batch_time=%u"},
1136 {Opt_journal_update, "journal=update"},
1137 {Opt_journal_dev, "journal_dev=%u"},
1138 {Opt_journal_checksum, "journal_checksum"},
1139 {Opt_journal_async_commit, "journal_async_commit"},
1140 {Opt_abort, "abort"},
1141 {Opt_data_journal, "data=journal"},
1142 {Opt_data_ordered, "data=ordered"},
1143 {Opt_data_writeback, "data=writeback"},
1144 {Opt_data_err_abort, "data_err=abort"},
1145 {Opt_data_err_ignore, "data_err=ignore"},
1146 {Opt_offusrjquota, "usrjquota="},
1147 {Opt_usrjquota, "usrjquota=%s"},
1148 {Opt_offgrpjquota, "grpjquota="},
1149 {Opt_grpjquota, "grpjquota=%s"},
1150 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1151 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1152 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1153 {Opt_grpquota, "grpquota"},
1154 {Opt_noquota, "noquota"},
1155 {Opt_quota, "quota"},
1156 {Opt_usrquota, "usrquota"},
1157 {Opt_barrier, "barrier=%u"},
1158 {Opt_barrier, "barrier"},
1159 {Opt_nobarrier, "nobarrier"},
1160 {Opt_i_version, "i_version"},
1161 {Opt_stripe, "stripe=%u"},
1162 {Opt_resize, "resize"},
1163 {Opt_delalloc, "delalloc"},
1164 {Opt_nodelalloc, "nodelalloc"},
1165 {Opt_block_validity, "block_validity"},
1166 {Opt_noblock_validity, "noblock_validity"},
1167 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1168 {Opt_journal_ioprio, "journal_ioprio=%u"},
1169 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1170 {Opt_auto_da_alloc, "auto_da_alloc"},
1171 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1172 {Opt_discard, "discard"},
1173 {Opt_nodiscard, "nodiscard"},
1177 static ext4_fsblk_t get_sb_block(void **data)
1179 ext4_fsblk_t sb_block;
1180 char *options = (char *) *data;
1182 if (!options || strncmp(options, "sb=", 3) != 0)
1183 return 1; /* Default location */
1186 /* TODO: use simple_strtoll with >32bit ext4 */
1187 sb_block = simple_strtoul(options, &options, 0);
1188 if (*options && *options != ',') {
1189 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1193 if (*options == ',')
1195 *data = (void *) options;
1200 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1202 static int parse_options(char *options, struct super_block *sb,
1203 unsigned long *journal_devnum,
1204 unsigned int *journal_ioprio,
1205 ext4_fsblk_t *n_blocks_count, int is_remount)
1207 struct ext4_sb_info *sbi = EXT4_SB(sb);
1209 substring_t args[MAX_OPT_ARGS];
1220 while ((p = strsep(&options, ",")) != NULL) {
1225 token = match_token(p, tokens, args);
1228 clear_opt(sbi->s_mount_opt, MINIX_DF);
1231 set_opt(sbi->s_mount_opt, MINIX_DF);
1234 set_opt(sbi->s_mount_opt, GRPID);
1237 clear_opt(sbi->s_mount_opt, GRPID);
1240 if (match_int(&args[0], &option))
1242 sbi->s_resuid = option;
1245 if (match_int(&args[0], &option))
1247 sbi->s_resgid = option;
1250 /* handled by get_sb_block() instead of here */
1251 /* *sb_block = match_int(&args[0]); */
1254 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1255 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1256 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1259 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1260 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1261 set_opt(sbi->s_mount_opt, ERRORS_RO);
1264 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1265 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1266 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1269 set_opt(sbi->s_mount_opt, NO_UID32);
1272 set_opt(sbi->s_mount_opt, DEBUG);
1275 set_opt(sbi->s_mount_opt, OLDALLOC);
1278 clear_opt(sbi->s_mount_opt, OLDALLOC);
1280 #ifdef CONFIG_EXT4_FS_XATTR
1281 case Opt_user_xattr:
1282 set_opt(sbi->s_mount_opt, XATTR_USER);
1284 case Opt_nouser_xattr:
1285 clear_opt(sbi->s_mount_opt, XATTR_USER);
1288 case Opt_user_xattr:
1289 case Opt_nouser_xattr:
1290 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1293 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1295 set_opt(sbi->s_mount_opt, POSIX_ACL);
1298 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1303 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1306 case Opt_journal_update:
1308 /* Eventually we will want to be able to create
1309 a journal file here. For now, only allow the
1310 user to specify an existing inode to be the
1313 ext4_msg(sb, KERN_ERR,
1314 "Cannot specify journal on remount");
1317 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1319 case Opt_journal_dev:
1321 ext4_msg(sb, KERN_ERR,
1322 "Cannot specify journal on remount");
1325 if (match_int(&args[0], &option))
1327 *journal_devnum = option;
1329 case Opt_journal_checksum:
1330 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1332 case Opt_journal_async_commit:
1333 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1334 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1337 set_opt(sbi->s_mount_opt, NOLOAD);
1340 if (match_int(&args[0], &option))
1345 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1346 sbi->s_commit_interval = HZ * option;
1348 case Opt_max_batch_time:
1349 if (match_int(&args[0], &option))
1354 option = EXT4_DEF_MAX_BATCH_TIME;
1355 sbi->s_max_batch_time = option;
1357 case Opt_min_batch_time:
1358 if (match_int(&args[0], &option))
1362 sbi->s_min_batch_time = option;
1364 case Opt_data_journal:
1365 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1367 case Opt_data_ordered:
1368 data_opt = EXT4_MOUNT_ORDERED_DATA;
1370 case Opt_data_writeback:
1371 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1374 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1376 ext4_msg(sb, KERN_ERR,
1377 "Cannot change data mode on remount");
1381 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1382 sbi->s_mount_opt |= data_opt;
1385 case Opt_data_err_abort:
1386 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1388 case Opt_data_err_ignore:
1389 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1398 if (sb_any_quota_loaded(sb) &&
1399 !sbi->s_qf_names[qtype]) {
1400 ext4_msg(sb, KERN_ERR,
1401 "Cannot change journaled "
1402 "quota options when quota turned on");
1405 qname = match_strdup(&args[0]);
1407 ext4_msg(sb, KERN_ERR,
1408 "Not enough memory for "
1409 "storing quotafile name");
1412 if (sbi->s_qf_names[qtype] &&
1413 strcmp(sbi->s_qf_names[qtype], qname)) {
1414 ext4_msg(sb, KERN_ERR,
1415 "%s quota file already "
1416 "specified", QTYPE2NAME(qtype));
1420 sbi->s_qf_names[qtype] = qname;
1421 if (strchr(sbi->s_qf_names[qtype], '/')) {
1422 ext4_msg(sb, KERN_ERR,
1423 "quotafile must be on "
1425 kfree(sbi->s_qf_names[qtype]);
1426 sbi->s_qf_names[qtype] = NULL;
1429 set_opt(sbi->s_mount_opt, QUOTA);
1431 case Opt_offusrjquota:
1434 case Opt_offgrpjquota:
1437 if (sb_any_quota_loaded(sb) &&
1438 sbi->s_qf_names[qtype]) {
1439 ext4_msg(sb, KERN_ERR, "Cannot change "
1440 "journaled quota options when "
1445 * The space will be released later when all options
1446 * are confirmed to be correct
1448 sbi->s_qf_names[qtype] = NULL;
1450 case Opt_jqfmt_vfsold:
1451 qfmt = QFMT_VFS_OLD;
1453 case Opt_jqfmt_vfsv0:
1456 case Opt_jqfmt_vfsv1:
1459 if (sb_any_quota_loaded(sb) &&
1460 sbi->s_jquota_fmt != qfmt) {
1461 ext4_msg(sb, KERN_ERR, "Cannot change "
1462 "journaled quota options when "
1466 sbi->s_jquota_fmt = qfmt;
1470 set_opt(sbi->s_mount_opt, QUOTA);
1471 set_opt(sbi->s_mount_opt, USRQUOTA);
1474 set_opt(sbi->s_mount_opt, QUOTA);
1475 set_opt(sbi->s_mount_opt, GRPQUOTA);
1478 if (sb_any_quota_loaded(sb)) {
1479 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1480 "options when quota turned on");
1483 clear_opt(sbi->s_mount_opt, QUOTA);
1484 clear_opt(sbi->s_mount_opt, USRQUOTA);
1485 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1491 ext4_msg(sb, KERN_ERR,
1492 "quota options not supported");
1496 case Opt_offusrjquota:
1497 case Opt_offgrpjquota:
1498 case Opt_jqfmt_vfsold:
1499 case Opt_jqfmt_vfsv0:
1500 case Opt_jqfmt_vfsv1:
1501 ext4_msg(sb, KERN_ERR,
1502 "journaled quota options not supported");
1508 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1511 clear_opt(sbi->s_mount_opt, BARRIER);
1514 if (match_int(&args[0], &option)) {
1515 set_opt(sbi->s_mount_opt, BARRIER);
1519 set_opt(sbi->s_mount_opt, BARRIER);
1521 clear_opt(sbi->s_mount_opt, BARRIER);
1527 ext4_msg(sb, KERN_ERR,
1528 "resize option only available "
1532 if (match_int(&args[0], &option) != 0)
1534 *n_blocks_count = option;
1537 set_opt(sbi->s_mount_opt, NOBH);
1540 clear_opt(sbi->s_mount_opt, NOBH);
1543 set_opt(sbi->s_mount_opt, I_VERSION);
1544 sb->s_flags |= MS_I_VERSION;
1546 case Opt_nodelalloc:
1547 clear_opt(sbi->s_mount_opt, DELALLOC);
1550 if (match_int(&args[0], &option))
1554 sbi->s_stripe = option;
1557 set_opt(sbi->s_mount_opt, DELALLOC);
1559 case Opt_block_validity:
1560 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1562 case Opt_noblock_validity:
1563 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1565 case Opt_inode_readahead_blks:
1566 if (match_int(&args[0], &option))
1568 if (option < 0 || option > (1 << 30))
1570 if (!is_power_of_2(option)) {
1571 ext4_msg(sb, KERN_ERR,
1572 "EXT4-fs: inode_readahead_blks"
1573 " must be a power of 2");
1576 sbi->s_inode_readahead_blks = option;
1578 case Opt_journal_ioprio:
1579 if (match_int(&args[0], &option))
1581 if (option < 0 || option > 7)
1583 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1586 case Opt_noauto_da_alloc:
1587 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1589 case Opt_auto_da_alloc:
1590 if (match_int(&args[0], &option)) {
1591 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1595 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1597 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1600 set_opt(sbi->s_mount_opt, DISCARD);
1603 clear_opt(sbi->s_mount_opt, DISCARD);
1606 ext4_msg(sb, KERN_ERR,
1607 "Unrecognized mount option \"%s\" "
1608 "or missing value", p);
1613 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1614 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1615 sbi->s_qf_names[USRQUOTA])
1616 clear_opt(sbi->s_mount_opt, USRQUOTA);
1618 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1619 sbi->s_qf_names[GRPQUOTA])
1620 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1622 if ((sbi->s_qf_names[USRQUOTA] &&
1623 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1624 (sbi->s_qf_names[GRPQUOTA] &&
1625 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1626 ext4_msg(sb, KERN_ERR, "old and new quota "
1631 if (!sbi->s_jquota_fmt) {
1632 ext4_msg(sb, KERN_ERR, "journaled quota format "
1637 if (sbi->s_jquota_fmt) {
1638 ext4_msg(sb, KERN_ERR, "journaled quota format "
1639 "specified with no journaling "
1648 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1651 struct ext4_sb_info *sbi = EXT4_SB(sb);
1654 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1655 ext4_msg(sb, KERN_ERR, "revision level too high, "
1656 "forcing read-only mode");
1661 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1662 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1663 "running e2fsck is recommended");
1664 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1665 ext4_msg(sb, KERN_WARNING,
1666 "warning: mounting fs with errors, "
1667 "running e2fsck is recommended");
1668 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1669 le16_to_cpu(es->s_mnt_count) >=
1670 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1671 ext4_msg(sb, KERN_WARNING,
1672 "warning: maximal mount count reached, "
1673 "running e2fsck is recommended");
1674 else if (le32_to_cpu(es->s_checkinterval) &&
1675 (le32_to_cpu(es->s_lastcheck) +
1676 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1677 ext4_msg(sb, KERN_WARNING,
1678 "warning: checktime reached, "
1679 "running e2fsck is recommended");
1680 if (!sbi->s_journal)
1681 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1682 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1683 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1684 le16_add_cpu(&es->s_mnt_count, 1);
1685 es->s_mtime = cpu_to_le32(get_seconds());
1686 ext4_update_dynamic_rev(sb);
1688 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1690 ext4_commit_super(sb, 1);
1691 if (test_opt(sb, DEBUG))
1692 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1693 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1695 sbi->s_groups_count,
1696 EXT4_BLOCKS_PER_GROUP(sb),
1697 EXT4_INODES_PER_GROUP(sb),
1703 static int ext4_fill_flex_info(struct super_block *sb)
1705 struct ext4_sb_info *sbi = EXT4_SB(sb);
1706 struct ext4_group_desc *gdp = NULL;
1707 ext4_group_t flex_group_count;
1708 ext4_group_t flex_group;
1709 int groups_per_flex = 0;
1713 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1714 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1716 if (groups_per_flex < 2) {
1717 sbi->s_log_groups_per_flex = 0;
1721 /* We allocate both existing and potentially added groups */
1722 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1723 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1724 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1725 size = flex_group_count * sizeof(struct flex_groups);
1726 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1727 if (sbi->s_flex_groups == NULL) {
1728 sbi->s_flex_groups = vmalloc(size);
1729 if (sbi->s_flex_groups)
1730 memset(sbi->s_flex_groups, 0, size);
1732 if (sbi->s_flex_groups == NULL) {
1733 ext4_msg(sb, KERN_ERR, "not enough memory for "
1734 "%u flex groups", flex_group_count);
1738 for (i = 0; i < sbi->s_groups_count; i++) {
1739 gdp = ext4_get_group_desc(sb, i, NULL);
1741 flex_group = ext4_flex_group(sbi, i);
1742 atomic_add(ext4_free_inodes_count(sb, gdp),
1743 &sbi->s_flex_groups[flex_group].free_inodes);
1744 atomic_add(ext4_free_blks_count(sb, gdp),
1745 &sbi->s_flex_groups[flex_group].free_blocks);
1746 atomic_add(ext4_used_dirs_count(sb, gdp),
1747 &sbi->s_flex_groups[flex_group].used_dirs);
1755 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1756 struct ext4_group_desc *gdp)
1760 if (sbi->s_es->s_feature_ro_compat &
1761 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1762 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1763 __le32 le_group = cpu_to_le32(block_group);
1765 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1766 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1767 crc = crc16(crc, (__u8 *)gdp, offset);
1768 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1769 /* for checksum of struct ext4_group_desc do the rest...*/
1770 if ((sbi->s_es->s_feature_incompat &
1771 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1772 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1773 crc = crc16(crc, (__u8 *)gdp + offset,
1774 le16_to_cpu(sbi->s_es->s_desc_size) -
1778 return cpu_to_le16(crc);
1781 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1782 struct ext4_group_desc *gdp)
1784 if ((sbi->s_es->s_feature_ro_compat &
1785 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1786 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1792 /* Called at mount-time, super-block is locked */
1793 static int ext4_check_descriptors(struct super_block *sb)
1795 struct ext4_sb_info *sbi = EXT4_SB(sb);
1796 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1797 ext4_fsblk_t last_block;
1798 ext4_fsblk_t block_bitmap;
1799 ext4_fsblk_t inode_bitmap;
1800 ext4_fsblk_t inode_table;
1801 int flexbg_flag = 0;
1804 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1807 ext4_debug("Checking group descriptors");
1809 for (i = 0; i < sbi->s_groups_count; i++) {
1810 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1812 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1813 last_block = ext4_blocks_count(sbi->s_es) - 1;
1815 last_block = first_block +
1816 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1818 block_bitmap = ext4_block_bitmap(sb, gdp);
1819 if (block_bitmap < first_block || block_bitmap > last_block) {
1820 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1821 "Block bitmap for group %u not in group "
1822 "(block %llu)!", i, block_bitmap);
1825 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1826 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1827 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1828 "Inode bitmap for group %u not in group "
1829 "(block %llu)!", i, inode_bitmap);
1832 inode_table = ext4_inode_table(sb, gdp);
1833 if (inode_table < first_block ||
1834 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1835 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1836 "Inode table for group %u not in group "
1837 "(block %llu)!", i, inode_table);
1840 ext4_lock_group(sb, i);
1841 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1842 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1843 "Checksum for group %u failed (%u!=%u)",
1844 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1845 gdp)), le16_to_cpu(gdp->bg_checksum));
1846 if (!(sb->s_flags & MS_RDONLY)) {
1847 ext4_unlock_group(sb, i);
1851 ext4_unlock_group(sb, i);
1853 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1856 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1857 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1861 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1862 * the superblock) which were deleted from all directories, but held open by
1863 * a process at the time of a crash. We walk the list and try to delete these
1864 * inodes at recovery time (only with a read-write filesystem).
1866 * In order to keep the orphan inode chain consistent during traversal (in
1867 * case of crash during recovery), we link each inode into the superblock
1868 * orphan list_head and handle it the same way as an inode deletion during
1869 * normal operation (which journals the operations for us).
1871 * We only do an iget() and an iput() on each inode, which is very safe if we
1872 * accidentally point at an in-use or already deleted inode. The worst that
1873 * can happen in this case is that we get a "bit already cleared" message from
1874 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1875 * e2fsck was run on this filesystem, and it must have already done the orphan
1876 * inode cleanup for us, so we can safely abort without any further action.
1878 static void ext4_orphan_cleanup(struct super_block *sb,
1879 struct ext4_super_block *es)
1881 unsigned int s_flags = sb->s_flags;
1882 int nr_orphans = 0, nr_truncates = 0;
1886 if (!es->s_last_orphan) {
1887 jbd_debug(4, "no orphan inodes to clean up\n");
1891 if (bdev_read_only(sb->s_bdev)) {
1892 ext4_msg(sb, KERN_ERR, "write access "
1893 "unavailable, skipping orphan cleanup");
1897 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1898 if (es->s_last_orphan)
1899 jbd_debug(1, "Errors on filesystem, "
1900 "clearing orphan list.\n");
1901 es->s_last_orphan = 0;
1902 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1906 if (s_flags & MS_RDONLY) {
1907 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1908 sb->s_flags &= ~MS_RDONLY;
1911 /* Needed for iput() to work correctly and not trash data */
1912 sb->s_flags |= MS_ACTIVE;
1913 /* Turn on quotas so that they are updated correctly */
1914 for (i = 0; i < MAXQUOTAS; i++) {
1915 if (EXT4_SB(sb)->s_qf_names[i]) {
1916 int ret = ext4_quota_on_mount(sb, i);
1918 ext4_msg(sb, KERN_ERR,
1919 "Cannot turn on journaled "
1920 "quota: error %d", ret);
1925 while (es->s_last_orphan) {
1926 struct inode *inode;
1928 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1929 if (IS_ERR(inode)) {
1930 es->s_last_orphan = 0;
1934 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1936 if (inode->i_nlink) {
1937 ext4_msg(sb, KERN_DEBUG,
1938 "%s: truncating inode %lu to %lld bytes",
1939 __func__, inode->i_ino, inode->i_size);
1940 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1941 inode->i_ino, inode->i_size);
1942 ext4_truncate(inode);
1945 ext4_msg(sb, KERN_DEBUG,
1946 "%s: deleting unreferenced inode %lu",
1947 __func__, inode->i_ino);
1948 jbd_debug(2, "deleting unreferenced inode %lu\n",
1952 iput(inode); /* The delete magic happens here! */
1955 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1958 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1959 PLURAL(nr_orphans));
1961 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1962 PLURAL(nr_truncates));
1964 /* Turn quotas off */
1965 for (i = 0; i < MAXQUOTAS; i++) {
1966 if (sb_dqopt(sb)->files[i])
1967 vfs_quota_off(sb, i, 0);
1970 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1974 * Maximal extent format file size.
1975 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1976 * extent format containers, within a sector_t, and within i_blocks
1977 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1978 * so that won't be a limiting factor.
1980 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1982 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1985 loff_t upper_limit = MAX_LFS_FILESIZE;
1987 /* small i_blocks in vfs inode? */
1988 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1990 * CONFIG_LBDAF is not enabled implies the inode
1991 * i_block represent total blocks in 512 bytes
1992 * 32 == size of vfs inode i_blocks * 8
1994 upper_limit = (1LL << 32) - 1;
1996 /* total blocks in file system block size */
1997 upper_limit >>= (blkbits - 9);
1998 upper_limit <<= blkbits;
2001 /* 32-bit extent-start container, ee_block */
2006 /* Sanity check against vm- & vfs- imposed limits */
2007 if (res > upper_limit)
2014 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2015 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2016 * We need to be 1 filesystem block less than the 2^48 sector limit.
2018 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2020 loff_t res = EXT4_NDIR_BLOCKS;
2023 /* This is calculated to be the largest file size for a dense, block
2024 * mapped file such that the file's total number of 512-byte sectors,
2025 * including data and all indirect blocks, does not exceed (2^48 - 1).
2027 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2028 * number of 512-byte sectors of the file.
2031 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2033 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2034 * the inode i_block field represents total file blocks in
2035 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2037 upper_limit = (1LL << 32) - 1;
2039 /* total blocks in file system block size */
2040 upper_limit >>= (bits - 9);
2044 * We use 48 bit ext4_inode i_blocks
2045 * With EXT4_HUGE_FILE_FL set the i_blocks
2046 * represent total number of blocks in
2047 * file system block size
2049 upper_limit = (1LL << 48) - 1;
2053 /* indirect blocks */
2055 /* double indirect blocks */
2056 meta_blocks += 1 + (1LL << (bits-2));
2057 /* tripple indirect blocks */
2058 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2060 upper_limit -= meta_blocks;
2061 upper_limit <<= bits;
2063 res += 1LL << (bits-2);
2064 res += 1LL << (2*(bits-2));
2065 res += 1LL << (3*(bits-2));
2067 if (res > upper_limit)
2070 if (res > MAX_LFS_FILESIZE)
2071 res = MAX_LFS_FILESIZE;
2076 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2077 ext4_fsblk_t logical_sb_block, int nr)
2079 struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 ext4_group_t bg, first_meta_bg;
2083 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2085 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2087 return logical_sb_block + nr + 1;
2088 bg = sbi->s_desc_per_block * nr;
2089 if (ext4_bg_has_super(sb, bg))
2092 return (has_super + ext4_group_first_block_no(sb, bg));
2096 * ext4_get_stripe_size: Get the stripe size.
2097 * @sbi: In memory super block info
2099 * If we have specified it via mount option, then
2100 * use the mount option value. If the value specified at mount time is
2101 * greater than the blocks per group use the super block value.
2102 * If the super block value is greater than blocks per group return 0.
2103 * Allocator needs it be less than blocks per group.
2106 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2108 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2109 unsigned long stripe_width =
2110 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2112 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2113 return sbi->s_stripe;
2115 if (stripe_width <= sbi->s_blocks_per_group)
2116 return stripe_width;
2118 if (stride <= sbi->s_blocks_per_group)
2127 struct attribute attr;
2128 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2129 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2130 const char *, size_t);
2134 static int parse_strtoul(const char *buf,
2135 unsigned long max, unsigned long *value)
2139 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2140 endp = skip_spaces(endp);
2141 if (*endp || *value > max)
2147 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2148 struct ext4_sb_info *sbi,
2151 return snprintf(buf, PAGE_SIZE, "%llu\n",
2152 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2155 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2156 struct ext4_sb_info *sbi, char *buf)
2158 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2160 return snprintf(buf, PAGE_SIZE, "%lu\n",
2161 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2162 sbi->s_sectors_written_start) >> 1);
2165 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2166 struct ext4_sb_info *sbi, char *buf)
2168 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2170 return snprintf(buf, PAGE_SIZE, "%llu\n",
2171 (unsigned long long)(sbi->s_kbytes_written +
2172 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2173 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2176 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2177 struct ext4_sb_info *sbi,
2178 const char *buf, size_t count)
2182 if (parse_strtoul(buf, 0x40000000, &t))
2185 if (!is_power_of_2(t))
2188 sbi->s_inode_readahead_blks = t;
2192 static ssize_t sbi_ui_show(struct ext4_attr *a,
2193 struct ext4_sb_info *sbi, char *buf)
2195 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2197 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2200 static ssize_t sbi_ui_store(struct ext4_attr *a,
2201 struct ext4_sb_info *sbi,
2202 const char *buf, size_t count)
2204 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2207 if (parse_strtoul(buf, 0xffffffff, &t))
2213 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2214 static struct ext4_attr ext4_attr_##_name = { \
2215 .attr = {.name = __stringify(_name), .mode = _mode }, \
2218 .offset = offsetof(struct ext4_sb_info, _elname), \
2220 #define EXT4_ATTR(name, mode, show, store) \
2221 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2223 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2224 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2225 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2226 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2227 #define ATTR_LIST(name) &ext4_attr_##name.attr
2229 EXT4_RO_ATTR(delayed_allocation_blocks);
2230 EXT4_RO_ATTR(session_write_kbytes);
2231 EXT4_RO_ATTR(lifetime_write_kbytes);
2232 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2233 inode_readahead_blks_store, s_inode_readahead_blks);
2234 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2235 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2236 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2237 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2238 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2239 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2240 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2241 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2243 static struct attribute *ext4_attrs[] = {
2244 ATTR_LIST(delayed_allocation_blocks),
2245 ATTR_LIST(session_write_kbytes),
2246 ATTR_LIST(lifetime_write_kbytes),
2247 ATTR_LIST(inode_readahead_blks),
2248 ATTR_LIST(inode_goal),
2249 ATTR_LIST(mb_stats),
2250 ATTR_LIST(mb_max_to_scan),
2251 ATTR_LIST(mb_min_to_scan),
2252 ATTR_LIST(mb_order2_req),
2253 ATTR_LIST(mb_stream_req),
2254 ATTR_LIST(mb_group_prealloc),
2255 ATTR_LIST(max_writeback_mb_bump),
2259 static ssize_t ext4_attr_show(struct kobject *kobj,
2260 struct attribute *attr, char *buf)
2262 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2264 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2266 return a->show ? a->show(a, sbi, buf) : 0;
2269 static ssize_t ext4_attr_store(struct kobject *kobj,
2270 struct attribute *attr,
2271 const char *buf, size_t len)
2273 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2275 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2277 return a->store ? a->store(a, sbi, buf, len) : 0;
2280 static void ext4_sb_release(struct kobject *kobj)
2282 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2284 complete(&sbi->s_kobj_unregister);
2288 static struct sysfs_ops ext4_attr_ops = {
2289 .show = ext4_attr_show,
2290 .store = ext4_attr_store,
2293 static struct kobj_type ext4_ktype = {
2294 .default_attrs = ext4_attrs,
2295 .sysfs_ops = &ext4_attr_ops,
2296 .release = ext4_sb_release,
2300 * Check whether this filesystem can be mounted based on
2301 * the features present and the RDONLY/RDWR mount requested.
2302 * Returns 1 if this filesystem can be mounted as requested,
2303 * 0 if it cannot be.
2305 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2307 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2308 ext4_msg(sb, KERN_ERR,
2309 "Couldn't mount because of "
2310 "unsupported optional features (%x)",
2311 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2312 ~EXT4_FEATURE_INCOMPAT_SUPP));
2319 /* Check that feature set is OK for a read-write mount */
2320 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2321 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2322 "unsupported optional features (%x)",
2323 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2324 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2328 * Large file size enabled file system can only be mounted
2329 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2331 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2332 if (sizeof(blkcnt_t) < sizeof(u64)) {
2333 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2334 "cannot be mounted RDWR without "
2342 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2343 __releases(kernel_lock)
2344 __acquires(kernel_lock)
2346 struct buffer_head *bh;
2347 struct ext4_super_block *es = NULL;
2348 struct ext4_sb_info *sbi;
2350 ext4_fsblk_t sb_block = get_sb_block(&data);
2351 ext4_fsblk_t logical_sb_block;
2352 unsigned long offset = 0;
2353 unsigned long journal_devnum = 0;
2354 unsigned long def_mount_opts;
2360 unsigned int db_count;
2362 int needs_recovery, has_huge_files;
2365 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2367 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2371 sbi->s_blockgroup_lock =
2372 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2373 if (!sbi->s_blockgroup_lock) {
2377 sb->s_fs_info = sbi;
2378 sbi->s_mount_opt = 0;
2379 sbi->s_resuid = EXT4_DEF_RESUID;
2380 sbi->s_resgid = EXT4_DEF_RESGID;
2381 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2382 sbi->s_sb_block = sb_block;
2383 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2388 /* Cleanup superblock name */
2389 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2392 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2394 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2399 * The ext4 superblock will not be buffer aligned for other than 1kB
2400 * block sizes. We need to calculate the offset from buffer start.
2402 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2403 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2404 offset = do_div(logical_sb_block, blocksize);
2406 logical_sb_block = sb_block;
2409 if (!(bh = sb_bread(sb, logical_sb_block))) {
2410 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2414 * Note: s_es must be initialized as soon as possible because
2415 * some ext4 macro-instructions depend on its value
2417 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2419 sb->s_magic = le16_to_cpu(es->s_magic);
2420 if (sb->s_magic != EXT4_SUPER_MAGIC)
2422 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2424 /* Set defaults before we parse the mount options */
2425 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2426 if (def_mount_opts & EXT4_DEFM_DEBUG)
2427 set_opt(sbi->s_mount_opt, DEBUG);
2428 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2429 set_opt(sbi->s_mount_opt, GRPID);
2430 if (def_mount_opts & EXT4_DEFM_UID16)
2431 set_opt(sbi->s_mount_opt, NO_UID32);
2432 #ifdef CONFIG_EXT4_FS_XATTR
2433 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2434 set_opt(sbi->s_mount_opt, XATTR_USER);
2436 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2437 if (def_mount_opts & EXT4_DEFM_ACL)
2438 set_opt(sbi->s_mount_opt, POSIX_ACL);
2440 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2441 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2442 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2443 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2444 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2445 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2447 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2448 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2449 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2450 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2452 set_opt(sbi->s_mount_opt, ERRORS_RO);
2454 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2455 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2456 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2457 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2458 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2460 set_opt(sbi->s_mount_opt, BARRIER);
2463 * enable delayed allocation by default
2464 * Use -o nodelalloc to turn it off
2466 set_opt(sbi->s_mount_opt, DELALLOC);
2468 if (!parse_options((char *) data, sb, &journal_devnum,
2469 &journal_ioprio, NULL, 0))
2472 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2473 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2475 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2476 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2477 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2478 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2479 ext4_msg(sb, KERN_WARNING,
2480 "feature flags set on rev 0 fs, "
2481 "running e2fsck is recommended");
2484 * Check feature flags regardless of the revision level, since we
2485 * previously didn't change the revision level when setting the flags,
2486 * so there is a chance incompat flags are set on a rev 0 filesystem.
2488 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2491 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2493 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2494 blocksize > EXT4_MAX_BLOCK_SIZE) {
2495 ext4_msg(sb, KERN_ERR,
2496 "Unsupported filesystem blocksize %d", blocksize);
2500 if (sb->s_blocksize != blocksize) {
2501 /* Validate the filesystem blocksize */
2502 if (!sb_set_blocksize(sb, blocksize)) {
2503 ext4_msg(sb, KERN_ERR, "bad block size %d",
2509 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2510 offset = do_div(logical_sb_block, blocksize);
2511 bh = sb_bread(sb, logical_sb_block);
2513 ext4_msg(sb, KERN_ERR,
2514 "Can't read superblock on 2nd try");
2517 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2519 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2520 ext4_msg(sb, KERN_ERR,
2521 "Magic mismatch, very weird!");
2526 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2527 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2528 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2530 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2532 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2533 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2534 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2536 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2537 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2538 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2539 (!is_power_of_2(sbi->s_inode_size)) ||
2540 (sbi->s_inode_size > blocksize)) {
2541 ext4_msg(sb, KERN_ERR,
2542 "unsupported inode size: %d",
2546 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2547 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2550 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2552 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2553 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2554 !is_power_of_2(sbi->s_desc_size)) {
2555 ext4_msg(sb, KERN_ERR,
2556 "unsupported descriptor size %lu",
2561 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2563 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2564 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2565 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2568 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2569 if (sbi->s_inodes_per_block == 0)
2571 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2572 sbi->s_inodes_per_block;
2573 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2575 sbi->s_mount_state = le16_to_cpu(es->s_state);
2576 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2577 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2579 for (i = 0; i < 4; i++)
2580 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2581 sbi->s_def_hash_version = es->s_def_hash_version;
2582 i = le32_to_cpu(es->s_flags);
2583 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2584 sbi->s_hash_unsigned = 3;
2585 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2586 #ifdef __CHAR_UNSIGNED__
2587 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2588 sbi->s_hash_unsigned = 3;
2590 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2595 if (sbi->s_blocks_per_group > blocksize * 8) {
2596 ext4_msg(sb, KERN_ERR,
2597 "#blocks per group too big: %lu",
2598 sbi->s_blocks_per_group);
2601 if (sbi->s_inodes_per_group > blocksize * 8) {
2602 ext4_msg(sb, KERN_ERR,
2603 "#inodes per group too big: %lu",
2604 sbi->s_inodes_per_group);
2609 * Test whether we have more sectors than will fit in sector_t,
2610 * and whether the max offset is addressable by the page cache.
2612 if ((ext4_blocks_count(es) >
2613 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2614 (ext4_blocks_count(es) >
2615 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2616 ext4_msg(sb, KERN_ERR, "filesystem"
2617 " too large to mount safely on this system");
2618 if (sizeof(sector_t) < 8)
2619 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2624 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2627 /* check blocks count against device size */
2628 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2629 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2630 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2631 "exceeds size of device (%llu blocks)",
2632 ext4_blocks_count(es), blocks_count);
2637 * It makes no sense for the first data block to be beyond the end
2638 * of the filesystem.
2640 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2641 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2642 "block %u is beyond end of filesystem (%llu)",
2643 le32_to_cpu(es->s_first_data_block),
2644 ext4_blocks_count(es));
2647 blocks_count = (ext4_blocks_count(es) -
2648 le32_to_cpu(es->s_first_data_block) +
2649 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2650 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2651 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2652 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2653 "(block count %llu, first data block %u, "
2654 "blocks per group %lu)", sbi->s_groups_count,
2655 ext4_blocks_count(es),
2656 le32_to_cpu(es->s_first_data_block),
2657 EXT4_BLOCKS_PER_GROUP(sb));
2660 sbi->s_groups_count = blocks_count;
2661 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2662 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2663 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2664 EXT4_DESC_PER_BLOCK(sb);
2665 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2667 if (sbi->s_group_desc == NULL) {
2668 ext4_msg(sb, KERN_ERR, "not enough memory");
2672 #ifdef CONFIG_PROC_FS
2674 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2677 bgl_lock_init(sbi->s_blockgroup_lock);
2679 for (i = 0; i < db_count; i++) {
2680 block = descriptor_loc(sb, logical_sb_block, i);
2681 sbi->s_group_desc[i] = sb_bread(sb, block);
2682 if (!sbi->s_group_desc[i]) {
2683 ext4_msg(sb, KERN_ERR,
2684 "can't read group descriptor %d", i);
2689 if (!ext4_check_descriptors(sb)) {
2690 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2693 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2694 if (!ext4_fill_flex_info(sb)) {
2695 ext4_msg(sb, KERN_ERR,
2696 "unable to initialize "
2697 "flex_bg meta info!");
2701 sbi->s_gdb_count = db_count;
2702 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2703 spin_lock_init(&sbi->s_next_gen_lock);
2705 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2706 ext4_count_free_blocks(sb));
2708 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2709 ext4_count_free_inodes(sb));
2712 err = percpu_counter_init(&sbi->s_dirs_counter,
2713 ext4_count_dirs(sb));
2716 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2719 ext4_msg(sb, KERN_ERR, "insufficient memory");
2723 sbi->s_stripe = ext4_get_stripe_size(sbi);
2724 sbi->s_max_writeback_mb_bump = 128;
2727 * set up enough so that it can read an inode
2729 if (!test_opt(sb, NOLOAD) &&
2730 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2731 sb->s_op = &ext4_sops;
2733 sb->s_op = &ext4_nojournal_sops;
2734 sb->s_export_op = &ext4_export_ops;
2735 sb->s_xattr = ext4_xattr_handlers;
2737 sb->s_qcop = &ext4_qctl_operations;
2738 sb->dq_op = &ext4_quota_operations;
2740 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2741 mutex_init(&sbi->s_orphan_lock);
2742 mutex_init(&sbi->s_resize_lock);
2746 needs_recovery = (es->s_last_orphan != 0 ||
2747 EXT4_HAS_INCOMPAT_FEATURE(sb,
2748 EXT4_FEATURE_INCOMPAT_RECOVER));
2751 * The first inode we look at is the journal inode. Don't try
2752 * root first: it may be modified in the journal!
2754 if (!test_opt(sb, NOLOAD) &&
2755 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2756 if (ext4_load_journal(sb, es, journal_devnum))
2758 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2759 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2760 ext4_msg(sb, KERN_ERR, "required journal recovery "
2761 "suppressed and not mounted read-only");
2764 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2765 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2766 sbi->s_journal = NULL;
2771 if (ext4_blocks_count(es) > 0xffffffffULL &&
2772 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2773 JBD2_FEATURE_INCOMPAT_64BIT)) {
2774 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2778 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2779 jbd2_journal_set_features(sbi->s_journal,
2780 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2781 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2782 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2783 jbd2_journal_set_features(sbi->s_journal,
2784 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2785 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2786 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2788 jbd2_journal_clear_features(sbi->s_journal,
2789 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2790 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2793 /* We have now updated the journal if required, so we can
2794 * validate the data journaling mode. */
2795 switch (test_opt(sb, DATA_FLAGS)) {
2797 /* No mode set, assume a default based on the journal
2798 * capabilities: ORDERED_DATA if the journal can
2799 * cope, else JOURNAL_DATA
2801 if (jbd2_journal_check_available_features
2802 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2803 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2805 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2808 case EXT4_MOUNT_ORDERED_DATA:
2809 case EXT4_MOUNT_WRITEBACK_DATA:
2810 if (!jbd2_journal_check_available_features
2811 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2812 ext4_msg(sb, KERN_ERR, "Journal does not support "
2813 "requested data journaling mode");
2819 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2823 if (test_opt(sb, NOBH)) {
2824 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2825 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2826 "its supported only with writeback mode");
2827 clear_opt(sbi->s_mount_opt, NOBH);
2830 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2831 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2832 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2833 goto failed_mount_wq;
2837 * The jbd2_journal_load will have done any necessary log recovery,
2838 * so we can safely mount the rest of the filesystem now.
2841 root = ext4_iget(sb, EXT4_ROOT_INO);
2843 ext4_msg(sb, KERN_ERR, "get root inode failed");
2844 ret = PTR_ERR(root);
2847 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2849 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2852 sb->s_root = d_alloc_root(root);
2854 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2860 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2862 /* determine the minimum size of new large inodes, if present */
2863 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2864 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2865 EXT4_GOOD_OLD_INODE_SIZE;
2866 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2867 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2868 if (sbi->s_want_extra_isize <
2869 le16_to_cpu(es->s_want_extra_isize))
2870 sbi->s_want_extra_isize =
2871 le16_to_cpu(es->s_want_extra_isize);
2872 if (sbi->s_want_extra_isize <
2873 le16_to_cpu(es->s_min_extra_isize))
2874 sbi->s_want_extra_isize =
2875 le16_to_cpu(es->s_min_extra_isize);
2878 /* Check if enough inode space is available */
2879 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2880 sbi->s_inode_size) {
2881 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2882 EXT4_GOOD_OLD_INODE_SIZE;
2883 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2887 if (test_opt(sb, DELALLOC) &&
2888 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2889 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2890 "requested data journaling mode");
2891 clear_opt(sbi->s_mount_opt, DELALLOC);
2894 err = ext4_setup_system_zone(sb);
2896 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2897 "zone (%d)\n", err);
2902 err = ext4_mb_init(sb, needs_recovery);
2904 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2909 sbi->s_kobj.kset = ext4_kset;
2910 init_completion(&sbi->s_kobj_unregister);
2911 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2914 ext4_mb_release(sb);
2915 ext4_ext_release(sb);
2919 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2920 ext4_orphan_cleanup(sb, es);
2921 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2922 if (needs_recovery) {
2923 ext4_msg(sb, KERN_INFO, "recovery complete");
2924 ext4_mark_recovery_complete(sb, es);
2926 if (EXT4_SB(sb)->s_journal) {
2927 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2928 descr = " journalled data mode";
2929 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2930 descr = " ordered data mode";
2932 descr = " writeback data mode";
2934 descr = "out journal";
2936 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2943 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2947 ext4_msg(sb, KERN_ERR, "mount failed");
2948 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2950 ext4_release_system_zone(sb);
2951 if (sbi->s_journal) {
2952 jbd2_journal_destroy(sbi->s_journal);
2953 sbi->s_journal = NULL;
2956 if (sbi->s_flex_groups) {
2957 if (is_vmalloc_addr(sbi->s_flex_groups))
2958 vfree(sbi->s_flex_groups);
2960 kfree(sbi->s_flex_groups);
2962 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2963 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2964 percpu_counter_destroy(&sbi->s_dirs_counter);
2965 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2967 for (i = 0; i < db_count; i++)
2968 brelse(sbi->s_group_desc[i]);
2969 kfree(sbi->s_group_desc);
2972 remove_proc_entry(sb->s_id, ext4_proc_root);
2975 for (i = 0; i < MAXQUOTAS; i++)
2976 kfree(sbi->s_qf_names[i]);
2978 ext4_blkdev_remove(sbi);
2981 sb->s_fs_info = NULL;
2982 kfree(sbi->s_blockgroup_lock);
2989 * Setup any per-fs journal parameters now. We'll do this both on
2990 * initial mount, once the journal has been initialised but before we've
2991 * done any recovery; and again on any subsequent remount.
2993 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2995 struct ext4_sb_info *sbi = EXT4_SB(sb);
2997 journal->j_commit_interval = sbi->s_commit_interval;
2998 journal->j_min_batch_time = sbi->s_min_batch_time;
2999 journal->j_max_batch_time = sbi->s_max_batch_time;
3001 spin_lock(&journal->j_state_lock);
3002 if (test_opt(sb, BARRIER))
3003 journal->j_flags |= JBD2_BARRIER;
3005 journal->j_flags &= ~JBD2_BARRIER;
3006 if (test_opt(sb, DATA_ERR_ABORT))
3007 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3009 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3010 spin_unlock(&journal->j_state_lock);
3013 static journal_t *ext4_get_journal(struct super_block *sb,
3014 unsigned int journal_inum)
3016 struct inode *journal_inode;
3019 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3021 /* First, test for the existence of a valid inode on disk. Bad
3022 * things happen if we iget() an unused inode, as the subsequent
3023 * iput() will try to delete it. */
3025 journal_inode = ext4_iget(sb, journal_inum);
3026 if (IS_ERR(journal_inode)) {
3027 ext4_msg(sb, KERN_ERR, "no journal found");
3030 if (!journal_inode->i_nlink) {
3031 make_bad_inode(journal_inode);
3032 iput(journal_inode);
3033 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3037 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3038 journal_inode, journal_inode->i_size);
3039 if (!S_ISREG(journal_inode->i_mode)) {
3040 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3041 iput(journal_inode);
3045 journal = jbd2_journal_init_inode(journal_inode);
3047 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3048 iput(journal_inode);
3051 journal->j_private = sb;
3052 ext4_init_journal_params(sb, journal);
3056 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3059 struct buffer_head *bh;
3063 int hblock, blocksize;
3064 ext4_fsblk_t sb_block;
3065 unsigned long offset;
3066 struct ext4_super_block *es;
3067 struct block_device *bdev;
3069 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3071 bdev = ext4_blkdev_get(j_dev, sb);
3075 if (bd_claim(bdev, sb)) {
3076 ext4_msg(sb, KERN_ERR,
3077 "failed to claim external journal device");
3078 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3082 blocksize = sb->s_blocksize;
3083 hblock = bdev_logical_block_size(bdev);
3084 if (blocksize < hblock) {
3085 ext4_msg(sb, KERN_ERR,
3086 "blocksize too small for journal device");
3090 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3091 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3092 set_blocksize(bdev, blocksize);
3093 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3094 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3095 "external journal");
3099 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3100 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3101 !(le32_to_cpu(es->s_feature_incompat) &
3102 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3103 ext4_msg(sb, KERN_ERR, "external journal has "
3109 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3110 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3115 len = ext4_blocks_count(es);
3116 start = sb_block + 1;
3117 brelse(bh); /* we're done with the superblock */
3119 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3120 start, len, blocksize);
3122 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3125 journal->j_private = sb;
3126 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3127 wait_on_buffer(journal->j_sb_buffer);
3128 if (!buffer_uptodate(journal->j_sb_buffer)) {
3129 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3132 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3133 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3134 "user (unsupported) - %d",
3135 be32_to_cpu(journal->j_superblock->s_nr_users));
3138 EXT4_SB(sb)->journal_bdev = bdev;
3139 ext4_init_journal_params(sb, journal);
3143 jbd2_journal_destroy(journal);
3145 ext4_blkdev_put(bdev);
3149 static int ext4_load_journal(struct super_block *sb,
3150 struct ext4_super_block *es,
3151 unsigned long journal_devnum)
3154 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3157 int really_read_only;
3159 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3161 if (journal_devnum &&
3162 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3163 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3164 "numbers have changed");
3165 journal_dev = new_decode_dev(journal_devnum);
3167 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3169 really_read_only = bdev_read_only(sb->s_bdev);
3172 * Are we loading a blank journal or performing recovery after a
3173 * crash? For recovery, we need to check in advance whether we
3174 * can get read-write access to the device.
3176 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3177 if (sb->s_flags & MS_RDONLY) {
3178 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3179 "required on readonly filesystem");
3180 if (really_read_only) {
3181 ext4_msg(sb, KERN_ERR, "write access "
3182 "unavailable, cannot proceed");
3185 ext4_msg(sb, KERN_INFO, "write access will "
3186 "be enabled during recovery");
3190 if (journal_inum && journal_dev) {
3191 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3192 "and inode journals!");
3197 if (!(journal = ext4_get_journal(sb, journal_inum)))
3200 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3204 if (!(journal->j_flags & JBD2_BARRIER))
3205 ext4_msg(sb, KERN_INFO, "barriers disabled");
3207 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3208 err = jbd2_journal_update_format(journal);
3210 ext4_msg(sb, KERN_ERR, "error updating journal");
3211 jbd2_journal_destroy(journal);
3216 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3217 err = jbd2_journal_wipe(journal, !really_read_only);
3219 err = jbd2_journal_load(journal);
3222 ext4_msg(sb, KERN_ERR, "error loading journal");
3223 jbd2_journal_destroy(journal);
3227 EXT4_SB(sb)->s_journal = journal;
3228 ext4_clear_journal_err(sb, es);
3230 if (journal_devnum &&
3231 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3232 es->s_journal_dev = cpu_to_le32(journal_devnum);
3234 /* Make sure we flush the recovery flag to disk. */
3235 ext4_commit_super(sb, 1);
3241 static int ext4_commit_super(struct super_block *sb, int sync)
3243 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3244 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3249 if (buffer_write_io_error(sbh)) {
3251 * Oh, dear. A previous attempt to write the
3252 * superblock failed. This could happen because the
3253 * USB device was yanked out. Or it could happen to
3254 * be a transient write error and maybe the block will
3255 * be remapped. Nothing we can do but to retry the
3256 * write and hope for the best.
3258 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3259 "superblock detected");
3260 clear_buffer_write_io_error(sbh);
3261 set_buffer_uptodate(sbh);
3264 * If the file system is mounted read-only, don't update the
3265 * superblock write time. This avoids updating the superblock
3266 * write time when we are mounting the root file system
3267 * read/only but we need to replay the journal; at that point,
3268 * for people who are east of GMT and who make their clock
3269 * tick in localtime for Windows bug-for-bug compatibility,
3270 * the clock is set in the future, and this will cause e2fsck
3271 * to complain and force a full file system check.
3273 if (!(sb->s_flags & MS_RDONLY))
3274 es->s_wtime = cpu_to_le32(get_seconds());
3275 es->s_kbytes_written =
3276 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3277 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3278 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3279 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3280 &EXT4_SB(sb)->s_freeblocks_counter));
3281 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3282 &EXT4_SB(sb)->s_freeinodes_counter));
3284 BUFFER_TRACE(sbh, "marking dirty");
3285 mark_buffer_dirty(sbh);
3287 error = sync_dirty_buffer(sbh);
3291 error = buffer_write_io_error(sbh);
3293 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3295 clear_buffer_write_io_error(sbh);
3296 set_buffer_uptodate(sbh);
3303 * Have we just finished recovery? If so, and if we are mounting (or
3304 * remounting) the filesystem readonly, then we will end up with a
3305 * consistent fs on disk. Record that fact.
3307 static void ext4_mark_recovery_complete(struct super_block *sb,
3308 struct ext4_super_block *es)
3310 journal_t *journal = EXT4_SB(sb)->s_journal;
3312 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3313 BUG_ON(journal != NULL);
3316 jbd2_journal_lock_updates(journal);
3317 if (jbd2_journal_flush(journal) < 0)
3320 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3321 sb->s_flags & MS_RDONLY) {
3322 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3323 ext4_commit_super(sb, 1);
3327 jbd2_journal_unlock_updates(journal);
3331 * If we are mounting (or read-write remounting) a filesystem whose journal
3332 * has recorded an error from a previous lifetime, move that error to the
3333 * main filesystem now.
3335 static void ext4_clear_journal_err(struct super_block *sb,
3336 struct ext4_super_block *es)
3342 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3344 journal = EXT4_SB(sb)->s_journal;
3347 * Now check for any error status which may have been recorded in the
3348 * journal by a prior ext4_error() or ext4_abort()
3351 j_errno = jbd2_journal_errno(journal);
3355 errstr = ext4_decode_error(sb, j_errno, nbuf);
3356 ext4_warning(sb, __func__, "Filesystem error recorded "
3357 "from previous mount: %s", errstr);
3358 ext4_warning(sb, __func__, "Marking fs in need of "
3359 "filesystem check.");
3361 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3362 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3363 ext4_commit_super(sb, 1);
3365 jbd2_journal_clear_err(journal);
3370 * Force the running and committing transactions to commit,
3371 * and wait on the commit.
3373 int ext4_force_commit(struct super_block *sb)
3378 if (sb->s_flags & MS_RDONLY)
3381 journal = EXT4_SB(sb)->s_journal;
3383 ret = ext4_journal_force_commit(journal);
3388 static void ext4_write_super(struct super_block *sb)
3391 ext4_commit_super(sb, 1);
3395 static int ext4_sync_fs(struct super_block *sb, int wait)
3399 struct ext4_sb_info *sbi = EXT4_SB(sb);
3401 trace_ext4_sync_fs(sb, wait);
3402 flush_workqueue(sbi->dio_unwritten_wq);
3403 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3405 jbd2_log_wait_commit(sbi->s_journal, target);
3411 * LVM calls this function before a (read-only) snapshot is created. This
3412 * gives us a chance to flush the journal completely and mark the fs clean.
3414 static int ext4_freeze(struct super_block *sb)
3419 if (sb->s_flags & MS_RDONLY)
3422 journal = EXT4_SB(sb)->s_journal;
3424 /* Now we set up the journal barrier. */
3425 jbd2_journal_lock_updates(journal);
3428 * Don't clear the needs_recovery flag if we failed to flush
3431 error = jbd2_journal_flush(journal);
3434 jbd2_journal_unlock_updates(journal);
3438 /* Journal blocked and flushed, clear needs_recovery flag. */
3439 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3440 error = ext4_commit_super(sb, 1);
3447 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3448 * flag here, even though the filesystem is not technically dirty yet.
3450 static int ext4_unfreeze(struct super_block *sb)
3452 if (sb->s_flags & MS_RDONLY)
3456 /* Reset the needs_recovery flag before the fs is unlocked. */
3457 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3458 ext4_commit_super(sb, 1);
3460 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3464 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3466 struct ext4_super_block *es;
3467 struct ext4_sb_info *sbi = EXT4_SB(sb);
3468 ext4_fsblk_t n_blocks_count = 0;
3469 unsigned long old_sb_flags;
3470 struct ext4_mount_options old_opts;
3472 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3480 /* Store the original options */
3482 old_sb_flags = sb->s_flags;
3483 old_opts.s_mount_opt = sbi->s_mount_opt;
3484 old_opts.s_resuid = sbi->s_resuid;
3485 old_opts.s_resgid = sbi->s_resgid;
3486 old_opts.s_commit_interval = sbi->s_commit_interval;
3487 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3488 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3490 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3491 for (i = 0; i < MAXQUOTAS; i++)
3492 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3494 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3495 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3498 * Allow the "check" option to be passed as a remount option.
3500 if (!parse_options(data, sb, NULL, &journal_ioprio,
3501 &n_blocks_count, 1)) {
3506 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3507 ext4_abort(sb, __func__, "Abort forced by user");
3509 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3510 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3514 if (sbi->s_journal) {
3515 ext4_init_journal_params(sb, sbi->s_journal);
3516 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3519 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3520 n_blocks_count > ext4_blocks_count(es)) {
3521 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3526 if (*flags & MS_RDONLY) {
3528 * First of all, the unconditional stuff we have to do
3529 * to disable replay of the journal when we next remount
3531 sb->s_flags |= MS_RDONLY;
3534 * OK, test if we are remounting a valid rw partition
3535 * readonly, and if so set the rdonly flag and then
3536 * mark the partition as valid again.
3538 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3539 (sbi->s_mount_state & EXT4_VALID_FS))
3540 es->s_state = cpu_to_le16(sbi->s_mount_state);
3543 ext4_mark_recovery_complete(sb, es);
3545 /* Make sure we can mount this feature set readwrite */
3546 if (!ext4_feature_set_ok(sb, 0)) {
3551 * Make sure the group descriptor checksums
3552 * are sane. If they aren't, refuse to remount r/w.
3554 for (g = 0; g < sbi->s_groups_count; g++) {
3555 struct ext4_group_desc *gdp =
3556 ext4_get_group_desc(sb, g, NULL);
3558 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3559 ext4_msg(sb, KERN_ERR,
3560 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3561 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3562 le16_to_cpu(gdp->bg_checksum));
3569 * If we have an unprocessed orphan list hanging
3570 * around from a previously readonly bdev mount,
3571 * require a full umount/remount for now.
3573 if (es->s_last_orphan) {
3574 ext4_msg(sb, KERN_WARNING, "Couldn't "
3575 "remount RDWR because of unprocessed "
3576 "orphan inode list. Please "
3577 "umount/remount instead");
3583 * Mounting a RDONLY partition read-write, so reread
3584 * and store the current valid flag. (It may have
3585 * been changed by e2fsck since we originally mounted
3589 ext4_clear_journal_err(sb, es);
3590 sbi->s_mount_state = le16_to_cpu(es->s_state);
3591 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3593 if (!ext4_setup_super(sb, es, 0))
3594 sb->s_flags &= ~MS_RDONLY;
3597 ext4_setup_system_zone(sb);
3598 if (sbi->s_journal == NULL)
3599 ext4_commit_super(sb, 1);
3602 /* Release old quota file names */
3603 for (i = 0; i < MAXQUOTAS; i++)
3604 if (old_opts.s_qf_names[i] &&
3605 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3606 kfree(old_opts.s_qf_names[i]);
3613 sb->s_flags = old_sb_flags;
3614 sbi->s_mount_opt = old_opts.s_mount_opt;
3615 sbi->s_resuid = old_opts.s_resuid;
3616 sbi->s_resgid = old_opts.s_resgid;
3617 sbi->s_commit_interval = old_opts.s_commit_interval;
3618 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3619 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3621 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3622 for (i = 0; i < MAXQUOTAS; i++) {
3623 if (sbi->s_qf_names[i] &&
3624 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3625 kfree(sbi->s_qf_names[i]);
3626 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3634 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3636 struct super_block *sb = dentry->d_sb;
3637 struct ext4_sb_info *sbi = EXT4_SB(sb);
3638 struct ext4_super_block *es = sbi->s_es;
3641 if (test_opt(sb, MINIX_DF)) {
3642 sbi->s_overhead_last = 0;
3643 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3644 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3645 ext4_fsblk_t overhead = 0;
3648 * Compute the overhead (FS structures). This is constant
3649 * for a given filesystem unless the number of block groups
3650 * changes so we cache the previous value until it does.
3654 * All of the blocks before first_data_block are
3657 overhead = le32_to_cpu(es->s_first_data_block);
3660 * Add the overhead attributed to the superblock and
3661 * block group descriptors. If the sparse superblocks
3662 * feature is turned on, then not all groups have this.
3664 for (i = 0; i < ngroups; i++) {
3665 overhead += ext4_bg_has_super(sb, i) +
3666 ext4_bg_num_gdb(sb, i);
3671 * Every block group has an inode bitmap, a block
3672 * bitmap, and an inode table.
3674 overhead += ngroups * (2 + sbi->s_itb_per_group);
3675 sbi->s_overhead_last = overhead;
3677 sbi->s_blocks_last = ext4_blocks_count(es);
3680 buf->f_type = EXT4_SUPER_MAGIC;
3681 buf->f_bsize = sb->s_blocksize;
3682 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3683 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3684 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3685 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3686 if (buf->f_bfree < ext4_r_blocks_count(es))
3688 buf->f_files = le32_to_cpu(es->s_inodes_count);
3689 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3690 buf->f_namelen = EXT4_NAME_LEN;
3691 fsid = le64_to_cpup((void *)es->s_uuid) ^
3692 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3693 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3694 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3699 /* Helper function for writing quotas on sync - we need to start transaction
3700 * before quota file is locked for write. Otherwise the are possible deadlocks:
3701 * Process 1 Process 2
3702 * ext4_create() quota_sync()
3703 * jbd2_journal_start() write_dquot()
3704 * vfs_dq_init() down(dqio_mutex)
3705 * down(dqio_mutex) jbd2_journal_start()
3711 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3713 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3716 static int ext4_write_dquot(struct dquot *dquot)
3720 struct inode *inode;
3722 inode = dquot_to_inode(dquot);
3723 handle = ext4_journal_start(inode,
3724 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3726 return PTR_ERR(handle);
3727 ret = dquot_commit(dquot);
3728 err = ext4_journal_stop(handle);
3734 static int ext4_acquire_dquot(struct dquot *dquot)
3739 handle = ext4_journal_start(dquot_to_inode(dquot),
3740 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3742 return PTR_ERR(handle);
3743 ret = dquot_acquire(dquot);
3744 err = ext4_journal_stop(handle);
3750 static int ext4_release_dquot(struct dquot *dquot)
3755 handle = ext4_journal_start(dquot_to_inode(dquot),
3756 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3757 if (IS_ERR(handle)) {
3758 /* Release dquot anyway to avoid endless cycle in dqput() */
3759 dquot_release(dquot);
3760 return PTR_ERR(handle);
3762 ret = dquot_release(dquot);
3763 err = ext4_journal_stop(handle);
3769 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3771 /* Are we journaling quotas? */
3772 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3773 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3774 dquot_mark_dquot_dirty(dquot);
3775 return ext4_write_dquot(dquot);
3777 return dquot_mark_dquot_dirty(dquot);
3781 static int ext4_write_info(struct super_block *sb, int type)
3786 /* Data block + inode block */
3787 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3789 return PTR_ERR(handle);
3790 ret = dquot_commit_info(sb, type);
3791 err = ext4_journal_stop(handle);
3798 * Turn on quotas during mount time - we need to find
3799 * the quota file and such...
3801 static int ext4_quota_on_mount(struct super_block *sb, int type)
3803 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3804 EXT4_SB(sb)->s_jquota_fmt, type);
3808 * Standard function to be called on quota_on
3810 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3811 char *name, int remount)
3816 if (!test_opt(sb, QUOTA))
3818 /* When remounting, no checks are needed and in fact, name is NULL */
3820 return vfs_quota_on(sb, type, format_id, name, remount);
3822 err = kern_path(name, LOOKUP_FOLLOW, &path);
3826 /* Quotafile not on the same filesystem? */
3827 if (path.mnt->mnt_sb != sb) {
3831 /* Journaling quota? */
3832 if (EXT4_SB(sb)->s_qf_names[type]) {
3833 /* Quotafile not in fs root? */
3834 if (path.dentry->d_parent != sb->s_root)
3835 ext4_msg(sb, KERN_WARNING,
3836 "Quota file not on filesystem root. "
3837 "Journaled quota will not work");
3841 * When we journal data on quota file, we have to flush journal to see
3842 * all updates to the file when we bypass pagecache...
3844 if (EXT4_SB(sb)->s_journal &&
3845 ext4_should_journal_data(path.dentry->d_inode)) {
3847 * We don't need to lock updates but journal_flush() could
3848 * otherwise be livelocked...
3850 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3851 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3852 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3859 err = vfs_quota_on_path(sb, type, format_id, &path);
3864 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3865 * acquiring the locks... As quota files are never truncated and quota code
3866 * itself serializes the operations (and noone else should touch the files)
3867 * we don't have to be afraid of races */
3868 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3869 size_t len, loff_t off)
3871 struct inode *inode = sb_dqopt(sb)->files[type];
3872 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3874 int offset = off & (sb->s_blocksize - 1);
3877 struct buffer_head *bh;
3878 loff_t i_size = i_size_read(inode);
3882 if (off+len > i_size)
3885 while (toread > 0) {
3886 tocopy = sb->s_blocksize - offset < toread ?
3887 sb->s_blocksize - offset : toread;
3888 bh = ext4_bread(NULL, inode, blk, 0, &err);
3891 if (!bh) /* A hole? */
3892 memset(data, 0, tocopy);
3894 memcpy(data, bh->b_data+offset, tocopy);
3904 /* Write to quotafile (we know the transaction is already started and has
3905 * enough credits) */
3906 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3907 const char *data, size_t len, loff_t off)
3909 struct inode *inode = sb_dqopt(sb)->files[type];
3910 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3912 int offset = off & (sb->s_blocksize - 1);
3914 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3915 size_t towrite = len;
3916 struct buffer_head *bh;
3917 handle_t *handle = journal_current_handle();
3919 if (EXT4_SB(sb)->s_journal && !handle) {
3920 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3921 " cancelled because transaction is not started",
3922 (unsigned long long)off, (unsigned long long)len);
3925 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3926 while (towrite > 0) {
3927 tocopy = sb->s_blocksize - offset < towrite ?
3928 sb->s_blocksize - offset : towrite;
3929 bh = ext4_bread(handle, inode, blk, 1, &err);
3932 if (journal_quota) {
3933 err = ext4_journal_get_write_access(handle, bh);
3940 memcpy(bh->b_data+offset, data, tocopy);
3941 flush_dcache_page(bh->b_page);
3944 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3946 /* Always do at least ordered writes for quotas */
3947 err = ext4_jbd2_file_inode(handle, inode);
3948 mark_buffer_dirty(bh);
3959 if (len == towrite) {
3960 mutex_unlock(&inode->i_mutex);
3963 if (inode->i_size < off+len-towrite) {
3964 i_size_write(inode, off+len-towrite);
3965 EXT4_I(inode)->i_disksize = inode->i_size;
3967 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3968 ext4_mark_inode_dirty(handle, inode);
3969 mutex_unlock(&inode->i_mutex);
3970 return len - towrite;
3975 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3976 const char *dev_name, void *data, struct vfsmount *mnt)
3978 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3981 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3982 static struct file_system_type ext2_fs_type = {
3983 .owner = THIS_MODULE,
3985 .get_sb = ext4_get_sb,
3986 .kill_sb = kill_block_super,
3987 .fs_flags = FS_REQUIRES_DEV,
3990 static inline void register_as_ext2(void)
3992 int err = register_filesystem(&ext2_fs_type);
3995 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
3998 static inline void unregister_as_ext2(void)
4000 unregister_filesystem(&ext2_fs_type);
4002 MODULE_ALIAS("ext2");
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);
4029 MODULE_ALIAS("ext3");
4031 static inline void register_as_ext3(void) { }
4032 static inline void unregister_as_ext3(void) { }
4035 static struct file_system_type ext4_fs_type = {
4036 .owner = THIS_MODULE,
4038 .get_sb = ext4_get_sb,
4039 .kill_sb = kill_block_super,
4040 .fs_flags = FS_REQUIRES_DEV,
4043 static int __init init_ext4_fs(void)
4047 err = init_ext4_system_zone();
4050 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4053 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4054 err = init_ext4_mballoc();
4058 err = init_ext4_xattr();
4061 err = init_inodecache();
4066 err = register_filesystem(&ext4_fs_type);
4071 unregister_as_ext2();
4072 unregister_as_ext3();
4073 destroy_inodecache();
4077 exit_ext4_mballoc();
4079 remove_proc_entry("fs/ext4", NULL);
4080 kset_unregister(ext4_kset);
4082 exit_ext4_system_zone();
4086 static void __exit exit_ext4_fs(void)
4088 unregister_as_ext2();
4089 unregister_as_ext3();
4090 unregister_filesystem(&ext4_fs_type);
4091 destroy_inodecache();
4093 exit_ext4_mballoc();
4094 remove_proc_entry("fs/ext4", NULL);
4095 kset_unregister(ext4_kset);
4096 exit_ext4_system_zone();
4099 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4100 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4101 MODULE_LICENSE("GPL");
4102 module_init(init_ext4_fs)
4103 module_exit(exit_ext4_fs)