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)
765 ext4_discard_preallocations(inode);
766 if (EXT4_JOURNAL(inode))
767 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
768 &EXT4_I(inode)->jinode);
771 static inline void ext4_show_quota_options(struct seq_file *seq,
772 struct super_block *sb)
774 #if defined(CONFIG_QUOTA)
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 if (sbi->s_jquota_fmt) {
780 switch (sbi->s_jquota_fmt) {
791 seq_printf(seq, ",jqfmt=%s", fmtname);
794 if (sbi->s_qf_names[USRQUOTA])
795 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
797 if (sbi->s_qf_names[GRPQUOTA])
798 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
800 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
801 seq_puts(seq, ",usrquota");
803 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
804 seq_puts(seq, ",grpquota");
810 * - it's set to a non-default value OR
811 * - if the per-sb default is different from the global default
813 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
816 unsigned long def_mount_opts;
817 struct super_block *sb = vfs->mnt_sb;
818 struct ext4_sb_info *sbi = EXT4_SB(sb);
819 struct ext4_super_block *es = sbi->s_es;
821 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
822 def_errors = le16_to_cpu(es->s_errors);
824 if (sbi->s_sb_block != 1)
825 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
826 if (test_opt(sb, MINIX_DF))
827 seq_puts(seq, ",minixdf");
828 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
829 seq_puts(seq, ",grpid");
830 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
831 seq_puts(seq, ",nogrpid");
832 if (sbi->s_resuid != EXT4_DEF_RESUID ||
833 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
834 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
836 if (sbi->s_resgid != EXT4_DEF_RESGID ||
837 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
838 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
840 if (test_opt(sb, ERRORS_RO)) {
841 if (def_errors == EXT4_ERRORS_PANIC ||
842 def_errors == EXT4_ERRORS_CONTINUE) {
843 seq_puts(seq, ",errors=remount-ro");
846 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
847 seq_puts(seq, ",errors=continue");
848 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
849 seq_puts(seq, ",errors=panic");
850 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
851 seq_puts(seq, ",nouid32");
852 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
853 seq_puts(seq, ",debug");
854 if (test_opt(sb, OLDALLOC))
855 seq_puts(seq, ",oldalloc");
856 #ifdef CONFIG_EXT4_FS_XATTR
857 if (test_opt(sb, XATTR_USER) &&
858 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
859 seq_puts(seq, ",user_xattr");
860 if (!test_opt(sb, XATTR_USER) &&
861 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
862 seq_puts(seq, ",nouser_xattr");
865 #ifdef CONFIG_EXT4_FS_POSIX_ACL
866 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
867 seq_puts(seq, ",acl");
868 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
869 seq_puts(seq, ",noacl");
871 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
872 seq_printf(seq, ",commit=%u",
873 (unsigned) (sbi->s_commit_interval / HZ));
875 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
876 seq_printf(seq, ",min_batch_time=%u",
877 (unsigned) sbi->s_min_batch_time);
879 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
880 seq_printf(seq, ",max_batch_time=%u",
881 (unsigned) sbi->s_min_batch_time);
885 * We're changing the default of barrier mount option, so
886 * let's always display its mount state so it's clear what its
889 seq_puts(seq, ",barrier=");
890 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
891 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
892 seq_puts(seq, ",journal_async_commit");
893 if (test_opt(sb, NOBH))
894 seq_puts(seq, ",nobh");
895 if (test_opt(sb, I_VERSION))
896 seq_puts(seq, ",i_version");
897 if (!test_opt(sb, DELALLOC))
898 seq_puts(seq, ",nodelalloc");
902 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
904 * journal mode get enabled in different ways
905 * So just print the value even if we didn't specify it
907 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
908 seq_puts(seq, ",data=journal");
909 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
910 seq_puts(seq, ",data=ordered");
911 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
912 seq_puts(seq, ",data=writeback");
914 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
915 seq_printf(seq, ",inode_readahead_blks=%u",
916 sbi->s_inode_readahead_blks);
918 if (test_opt(sb, DATA_ERR_ABORT))
919 seq_puts(seq, ",data_err=abort");
921 if (test_opt(sb, NO_AUTO_DA_ALLOC))
922 seq_puts(seq, ",noauto_da_alloc");
924 if (test_opt(sb, DISCARD))
925 seq_puts(seq, ",discard");
927 if (test_opt(sb, NOLOAD))
928 seq_puts(seq, ",norecovery");
930 ext4_show_quota_options(seq, sb);
935 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
936 u64 ino, u32 generation)
940 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
941 return ERR_PTR(-ESTALE);
942 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
943 return ERR_PTR(-ESTALE);
945 /* iget isn't really right if the inode is currently unallocated!!
947 * ext4_read_inode will return a bad_inode if the inode had been
948 * deleted, so we should be safe.
950 * Currently we don't know the generation for parent directory, so
951 * a generation of 0 means "accept any"
953 inode = ext4_iget(sb, ino);
955 return ERR_CAST(inode);
956 if (generation && inode->i_generation != generation) {
958 return ERR_PTR(-ESTALE);
964 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
965 int fh_len, int fh_type)
967 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
971 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
972 int fh_len, int fh_type)
974 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
979 * Try to release metadata pages (indirect blocks, directories) which are
980 * mapped via the block device. Since these pages could have journal heads
981 * which would prevent try_to_free_buffers() from freeing them, we must use
982 * jbd2 layer's try_to_free_buffers() function to release them.
984 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
987 journal_t *journal = EXT4_SB(sb)->s_journal;
989 WARN_ON(PageChecked(page));
990 if (!page_has_buffers(page))
993 return jbd2_journal_try_to_free_buffers(journal, page,
995 return try_to_free_buffers(page);
999 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1000 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1002 static int ext4_write_dquot(struct dquot *dquot);
1003 static int ext4_acquire_dquot(struct dquot *dquot);
1004 static int ext4_release_dquot(struct dquot *dquot);
1005 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1006 static int ext4_write_info(struct super_block *sb, int type);
1007 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1008 char *path, int remount);
1009 static int ext4_quota_on_mount(struct super_block *sb, int type);
1010 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1011 size_t len, loff_t off);
1012 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1013 const char *data, size_t len, loff_t off);
1015 static const struct dquot_operations ext4_quota_operations = {
1017 .get_reserved_space = ext4_get_reserved_space,
1019 .write_dquot = ext4_write_dquot,
1020 .acquire_dquot = ext4_acquire_dquot,
1021 .release_dquot = ext4_release_dquot,
1022 .mark_dirty = ext4_mark_dquot_dirty,
1023 .write_info = ext4_write_info,
1024 .alloc_dquot = dquot_alloc,
1025 .destroy_dquot = dquot_destroy,
1028 static const struct quotactl_ops ext4_qctl_operations = {
1029 .quota_on = ext4_quota_on,
1030 .quota_off = vfs_quota_off,
1031 .quota_sync = vfs_quota_sync,
1032 .get_info = vfs_get_dqinfo,
1033 .set_info = vfs_set_dqinfo,
1034 .get_dqblk = vfs_get_dqblk,
1035 .set_dqblk = vfs_set_dqblk
1039 static const struct super_operations ext4_sops = {
1040 .alloc_inode = ext4_alloc_inode,
1041 .destroy_inode = ext4_destroy_inode,
1042 .write_inode = ext4_write_inode,
1043 .dirty_inode = ext4_dirty_inode,
1044 .delete_inode = ext4_delete_inode,
1045 .put_super = ext4_put_super,
1046 .sync_fs = ext4_sync_fs,
1047 .freeze_fs = ext4_freeze,
1048 .unfreeze_fs = ext4_unfreeze,
1049 .statfs = ext4_statfs,
1050 .remount_fs = ext4_remount,
1051 .clear_inode = ext4_clear_inode,
1052 .show_options = ext4_show_options,
1054 .quota_read = ext4_quota_read,
1055 .quota_write = ext4_quota_write,
1057 .bdev_try_to_free_page = bdev_try_to_free_page,
1060 static const struct super_operations ext4_nojournal_sops = {
1061 .alloc_inode = ext4_alloc_inode,
1062 .destroy_inode = ext4_destroy_inode,
1063 .write_inode = ext4_write_inode,
1064 .dirty_inode = ext4_dirty_inode,
1065 .delete_inode = ext4_delete_inode,
1066 .write_super = ext4_write_super,
1067 .put_super = ext4_put_super,
1068 .statfs = ext4_statfs,
1069 .remount_fs = ext4_remount,
1070 .clear_inode = ext4_clear_inode,
1071 .show_options = ext4_show_options,
1073 .quota_read = ext4_quota_read,
1074 .quota_write = ext4_quota_write,
1076 .bdev_try_to_free_page = bdev_try_to_free_page,
1079 static const struct export_operations ext4_export_ops = {
1080 .fh_to_dentry = ext4_fh_to_dentry,
1081 .fh_to_parent = ext4_fh_to_parent,
1082 .get_parent = ext4_get_parent,
1086 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1087 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1088 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1089 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1090 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1091 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1092 Opt_journal_update, Opt_journal_dev,
1093 Opt_journal_checksum, Opt_journal_async_commit,
1094 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1095 Opt_data_err_abort, Opt_data_err_ignore,
1096 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1097 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1098 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1099 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1100 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1101 Opt_block_validity, Opt_noblock_validity,
1102 Opt_inode_readahead_blks, Opt_journal_ioprio,
1103 Opt_discard, Opt_nodiscard,
1106 static const match_table_t tokens = {
1107 {Opt_bsd_df, "bsddf"},
1108 {Opt_minix_df, "minixdf"},
1109 {Opt_grpid, "grpid"},
1110 {Opt_grpid, "bsdgroups"},
1111 {Opt_nogrpid, "nogrpid"},
1112 {Opt_nogrpid, "sysvgroups"},
1113 {Opt_resgid, "resgid=%u"},
1114 {Opt_resuid, "resuid=%u"},
1116 {Opt_err_cont, "errors=continue"},
1117 {Opt_err_panic, "errors=panic"},
1118 {Opt_err_ro, "errors=remount-ro"},
1119 {Opt_nouid32, "nouid32"},
1120 {Opt_debug, "debug"},
1121 {Opt_oldalloc, "oldalloc"},
1122 {Opt_orlov, "orlov"},
1123 {Opt_user_xattr, "user_xattr"},
1124 {Opt_nouser_xattr, "nouser_xattr"},
1126 {Opt_noacl, "noacl"},
1127 {Opt_noload, "noload"},
1128 {Opt_noload, "norecovery"},
1131 {Opt_commit, "commit=%u"},
1132 {Opt_min_batch_time, "min_batch_time=%u"},
1133 {Opt_max_batch_time, "max_batch_time=%u"},
1134 {Opt_journal_update, "journal=update"},
1135 {Opt_journal_dev, "journal_dev=%u"},
1136 {Opt_journal_checksum, "journal_checksum"},
1137 {Opt_journal_async_commit, "journal_async_commit"},
1138 {Opt_abort, "abort"},
1139 {Opt_data_journal, "data=journal"},
1140 {Opt_data_ordered, "data=ordered"},
1141 {Opt_data_writeback, "data=writeback"},
1142 {Opt_data_err_abort, "data_err=abort"},
1143 {Opt_data_err_ignore, "data_err=ignore"},
1144 {Opt_offusrjquota, "usrjquota="},
1145 {Opt_usrjquota, "usrjquota=%s"},
1146 {Opt_offgrpjquota, "grpjquota="},
1147 {Opt_grpjquota, "grpjquota=%s"},
1148 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1149 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1150 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1151 {Opt_grpquota, "grpquota"},
1152 {Opt_noquota, "noquota"},
1153 {Opt_quota, "quota"},
1154 {Opt_usrquota, "usrquota"},
1155 {Opt_barrier, "barrier=%u"},
1156 {Opt_barrier, "barrier"},
1157 {Opt_nobarrier, "nobarrier"},
1158 {Opt_i_version, "i_version"},
1159 {Opt_stripe, "stripe=%u"},
1160 {Opt_resize, "resize"},
1161 {Opt_delalloc, "delalloc"},
1162 {Opt_nodelalloc, "nodelalloc"},
1163 {Opt_block_validity, "block_validity"},
1164 {Opt_noblock_validity, "noblock_validity"},
1165 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1166 {Opt_journal_ioprio, "journal_ioprio=%u"},
1167 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1168 {Opt_auto_da_alloc, "auto_da_alloc"},
1169 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1170 {Opt_discard, "discard"},
1171 {Opt_nodiscard, "nodiscard"},
1175 static ext4_fsblk_t get_sb_block(void **data)
1177 ext4_fsblk_t sb_block;
1178 char *options = (char *) *data;
1180 if (!options || strncmp(options, "sb=", 3) != 0)
1181 return 1; /* Default location */
1184 /* TODO: use simple_strtoll with >32bit ext4 */
1185 sb_block = simple_strtoul(options, &options, 0);
1186 if (*options && *options != ',') {
1187 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1191 if (*options == ',')
1193 *data = (void *) options;
1198 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1200 static int parse_options(char *options, struct super_block *sb,
1201 unsigned long *journal_devnum,
1202 unsigned int *journal_ioprio,
1203 ext4_fsblk_t *n_blocks_count, int is_remount)
1205 struct ext4_sb_info *sbi = EXT4_SB(sb);
1207 substring_t args[MAX_OPT_ARGS];
1218 while ((p = strsep(&options, ",")) != NULL) {
1223 token = match_token(p, tokens, args);
1226 clear_opt(sbi->s_mount_opt, MINIX_DF);
1229 set_opt(sbi->s_mount_opt, MINIX_DF);
1232 set_opt(sbi->s_mount_opt, GRPID);
1235 clear_opt(sbi->s_mount_opt, GRPID);
1238 if (match_int(&args[0], &option))
1240 sbi->s_resuid = option;
1243 if (match_int(&args[0], &option))
1245 sbi->s_resgid = option;
1248 /* handled by get_sb_block() instead of here */
1249 /* *sb_block = match_int(&args[0]); */
1252 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1253 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1254 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1257 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1258 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1259 set_opt(sbi->s_mount_opt, ERRORS_RO);
1262 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1263 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1264 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1267 set_opt(sbi->s_mount_opt, NO_UID32);
1270 set_opt(sbi->s_mount_opt, DEBUG);
1273 set_opt(sbi->s_mount_opt, OLDALLOC);
1276 clear_opt(sbi->s_mount_opt, OLDALLOC);
1278 #ifdef CONFIG_EXT4_FS_XATTR
1279 case Opt_user_xattr:
1280 set_opt(sbi->s_mount_opt, XATTR_USER);
1282 case Opt_nouser_xattr:
1283 clear_opt(sbi->s_mount_opt, XATTR_USER);
1286 case Opt_user_xattr:
1287 case Opt_nouser_xattr:
1288 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1291 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1293 set_opt(sbi->s_mount_opt, POSIX_ACL);
1296 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1301 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1304 case Opt_journal_update:
1306 /* Eventually we will want to be able to create
1307 a journal file here. For now, only allow the
1308 user to specify an existing inode to be the
1311 ext4_msg(sb, KERN_ERR,
1312 "Cannot specify journal on remount");
1315 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1317 case Opt_journal_dev:
1319 ext4_msg(sb, KERN_ERR,
1320 "Cannot specify journal on remount");
1323 if (match_int(&args[0], &option))
1325 *journal_devnum = option;
1327 case Opt_journal_checksum:
1328 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1330 case Opt_journal_async_commit:
1331 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1332 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1335 set_opt(sbi->s_mount_opt, NOLOAD);
1338 if (match_int(&args[0], &option))
1343 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1344 sbi->s_commit_interval = HZ * option;
1346 case Opt_max_batch_time:
1347 if (match_int(&args[0], &option))
1352 option = EXT4_DEF_MAX_BATCH_TIME;
1353 sbi->s_max_batch_time = option;
1355 case Opt_min_batch_time:
1356 if (match_int(&args[0], &option))
1360 sbi->s_min_batch_time = option;
1362 case Opt_data_journal:
1363 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1365 case Opt_data_ordered:
1366 data_opt = EXT4_MOUNT_ORDERED_DATA;
1368 case Opt_data_writeback:
1369 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1372 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1374 ext4_msg(sb, KERN_ERR,
1375 "Cannot change data mode on remount");
1379 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1380 sbi->s_mount_opt |= data_opt;
1383 case Opt_data_err_abort:
1384 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1386 case Opt_data_err_ignore:
1387 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1396 if (sb_any_quota_loaded(sb) &&
1397 !sbi->s_qf_names[qtype]) {
1398 ext4_msg(sb, KERN_ERR,
1399 "Cannot change journaled "
1400 "quota options when quota turned on");
1403 qname = match_strdup(&args[0]);
1405 ext4_msg(sb, KERN_ERR,
1406 "Not enough memory for "
1407 "storing quotafile name");
1410 if (sbi->s_qf_names[qtype] &&
1411 strcmp(sbi->s_qf_names[qtype], qname)) {
1412 ext4_msg(sb, KERN_ERR,
1413 "%s quota file already "
1414 "specified", QTYPE2NAME(qtype));
1418 sbi->s_qf_names[qtype] = qname;
1419 if (strchr(sbi->s_qf_names[qtype], '/')) {
1420 ext4_msg(sb, KERN_ERR,
1421 "quotafile must be on "
1423 kfree(sbi->s_qf_names[qtype]);
1424 sbi->s_qf_names[qtype] = NULL;
1427 set_opt(sbi->s_mount_opt, QUOTA);
1429 case Opt_offusrjquota:
1432 case Opt_offgrpjquota:
1435 if (sb_any_quota_loaded(sb) &&
1436 sbi->s_qf_names[qtype]) {
1437 ext4_msg(sb, KERN_ERR, "Cannot change "
1438 "journaled quota options when "
1443 * The space will be released later when all options
1444 * are confirmed to be correct
1446 sbi->s_qf_names[qtype] = NULL;
1448 case Opt_jqfmt_vfsold:
1449 qfmt = QFMT_VFS_OLD;
1451 case Opt_jqfmt_vfsv0:
1454 case Opt_jqfmt_vfsv1:
1457 if (sb_any_quota_loaded(sb) &&
1458 sbi->s_jquota_fmt != qfmt) {
1459 ext4_msg(sb, KERN_ERR, "Cannot change "
1460 "journaled quota options when "
1464 sbi->s_jquota_fmt = qfmt;
1468 set_opt(sbi->s_mount_opt, QUOTA);
1469 set_opt(sbi->s_mount_opt, USRQUOTA);
1472 set_opt(sbi->s_mount_opt, QUOTA);
1473 set_opt(sbi->s_mount_opt, GRPQUOTA);
1476 if (sb_any_quota_loaded(sb)) {
1477 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1478 "options when quota turned on");
1481 clear_opt(sbi->s_mount_opt, QUOTA);
1482 clear_opt(sbi->s_mount_opt, USRQUOTA);
1483 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1489 ext4_msg(sb, KERN_ERR,
1490 "quota options not supported");
1494 case Opt_offusrjquota:
1495 case Opt_offgrpjquota:
1496 case Opt_jqfmt_vfsold:
1497 case Opt_jqfmt_vfsv0:
1498 case Opt_jqfmt_vfsv1:
1499 ext4_msg(sb, KERN_ERR,
1500 "journaled quota options not supported");
1506 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1509 clear_opt(sbi->s_mount_opt, BARRIER);
1512 if (match_int(&args[0], &option)) {
1513 set_opt(sbi->s_mount_opt, BARRIER);
1517 set_opt(sbi->s_mount_opt, BARRIER);
1519 clear_opt(sbi->s_mount_opt, BARRIER);
1525 ext4_msg(sb, KERN_ERR,
1526 "resize option only available "
1530 if (match_int(&args[0], &option) != 0)
1532 *n_blocks_count = option;
1535 set_opt(sbi->s_mount_opt, NOBH);
1538 clear_opt(sbi->s_mount_opt, NOBH);
1541 set_opt(sbi->s_mount_opt, I_VERSION);
1542 sb->s_flags |= MS_I_VERSION;
1544 case Opt_nodelalloc:
1545 clear_opt(sbi->s_mount_opt, DELALLOC);
1548 if (match_int(&args[0], &option))
1552 sbi->s_stripe = option;
1555 set_opt(sbi->s_mount_opt, DELALLOC);
1557 case Opt_block_validity:
1558 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1560 case Opt_noblock_validity:
1561 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1563 case Opt_inode_readahead_blks:
1564 if (match_int(&args[0], &option))
1566 if (option < 0 || option > (1 << 30))
1568 if (!is_power_of_2(option)) {
1569 ext4_msg(sb, KERN_ERR,
1570 "EXT4-fs: inode_readahead_blks"
1571 " must be a power of 2");
1574 sbi->s_inode_readahead_blks = option;
1576 case Opt_journal_ioprio:
1577 if (match_int(&args[0], &option))
1579 if (option < 0 || option > 7)
1581 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1584 case Opt_noauto_da_alloc:
1585 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1587 case Opt_auto_da_alloc:
1588 if (match_int(&args[0], &option)) {
1589 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1593 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1595 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1598 set_opt(sbi->s_mount_opt, DISCARD);
1601 clear_opt(sbi->s_mount_opt, DISCARD);
1604 ext4_msg(sb, KERN_ERR,
1605 "Unrecognized mount option \"%s\" "
1606 "or missing value", p);
1611 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1612 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1613 sbi->s_qf_names[USRQUOTA])
1614 clear_opt(sbi->s_mount_opt, USRQUOTA);
1616 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1617 sbi->s_qf_names[GRPQUOTA])
1618 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1620 if ((sbi->s_qf_names[USRQUOTA] &&
1621 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1622 (sbi->s_qf_names[GRPQUOTA] &&
1623 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1624 ext4_msg(sb, KERN_ERR, "old and new quota "
1629 if (!sbi->s_jquota_fmt) {
1630 ext4_msg(sb, KERN_ERR, "journaled quota format "
1635 if (sbi->s_jquota_fmt) {
1636 ext4_msg(sb, KERN_ERR, "journaled quota format "
1637 "specified with no journaling "
1646 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1649 struct ext4_sb_info *sbi = EXT4_SB(sb);
1652 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1653 ext4_msg(sb, KERN_ERR, "revision level too high, "
1654 "forcing read-only mode");
1659 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1660 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1661 "running e2fsck is recommended");
1662 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1663 ext4_msg(sb, KERN_WARNING,
1664 "warning: mounting fs with errors, "
1665 "running e2fsck is recommended");
1666 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1667 le16_to_cpu(es->s_mnt_count) >=
1668 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1669 ext4_msg(sb, KERN_WARNING,
1670 "warning: maximal mount count reached, "
1671 "running e2fsck is recommended");
1672 else if (le32_to_cpu(es->s_checkinterval) &&
1673 (le32_to_cpu(es->s_lastcheck) +
1674 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1675 ext4_msg(sb, KERN_WARNING,
1676 "warning: checktime reached, "
1677 "running e2fsck is recommended");
1678 if (!sbi->s_journal)
1679 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1680 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1681 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1682 le16_add_cpu(&es->s_mnt_count, 1);
1683 es->s_mtime = cpu_to_le32(get_seconds());
1684 ext4_update_dynamic_rev(sb);
1686 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1688 ext4_commit_super(sb, 1);
1689 if (test_opt(sb, DEBUG))
1690 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1691 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1693 sbi->s_groups_count,
1694 EXT4_BLOCKS_PER_GROUP(sb),
1695 EXT4_INODES_PER_GROUP(sb),
1701 static int ext4_fill_flex_info(struct super_block *sb)
1703 struct ext4_sb_info *sbi = EXT4_SB(sb);
1704 struct ext4_group_desc *gdp = NULL;
1705 ext4_group_t flex_group_count;
1706 ext4_group_t flex_group;
1707 int groups_per_flex = 0;
1711 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1712 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1714 if (groups_per_flex < 2) {
1715 sbi->s_log_groups_per_flex = 0;
1719 /* We allocate both existing and potentially added groups */
1720 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1721 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1722 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1723 size = flex_group_count * sizeof(struct flex_groups);
1724 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1725 if (sbi->s_flex_groups == NULL) {
1726 sbi->s_flex_groups = vmalloc(size);
1727 if (sbi->s_flex_groups)
1728 memset(sbi->s_flex_groups, 0, size);
1730 if (sbi->s_flex_groups == NULL) {
1731 ext4_msg(sb, KERN_ERR, "not enough memory for "
1732 "%u flex groups", flex_group_count);
1736 for (i = 0; i < sbi->s_groups_count; i++) {
1737 gdp = ext4_get_group_desc(sb, i, NULL);
1739 flex_group = ext4_flex_group(sbi, i);
1740 atomic_add(ext4_free_inodes_count(sb, gdp),
1741 &sbi->s_flex_groups[flex_group].free_inodes);
1742 atomic_add(ext4_free_blks_count(sb, gdp),
1743 &sbi->s_flex_groups[flex_group].free_blocks);
1744 atomic_add(ext4_used_dirs_count(sb, gdp),
1745 &sbi->s_flex_groups[flex_group].used_dirs);
1753 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1754 struct ext4_group_desc *gdp)
1758 if (sbi->s_es->s_feature_ro_compat &
1759 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1760 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1761 __le32 le_group = cpu_to_le32(block_group);
1763 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1764 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1765 crc = crc16(crc, (__u8 *)gdp, offset);
1766 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1767 /* for checksum of struct ext4_group_desc do the rest...*/
1768 if ((sbi->s_es->s_feature_incompat &
1769 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1770 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1771 crc = crc16(crc, (__u8 *)gdp + offset,
1772 le16_to_cpu(sbi->s_es->s_desc_size) -
1776 return cpu_to_le16(crc);
1779 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1780 struct ext4_group_desc *gdp)
1782 if ((sbi->s_es->s_feature_ro_compat &
1783 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1784 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1790 /* Called at mount-time, super-block is locked */
1791 static int ext4_check_descriptors(struct super_block *sb)
1793 struct ext4_sb_info *sbi = EXT4_SB(sb);
1794 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1795 ext4_fsblk_t last_block;
1796 ext4_fsblk_t block_bitmap;
1797 ext4_fsblk_t inode_bitmap;
1798 ext4_fsblk_t inode_table;
1799 int flexbg_flag = 0;
1802 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1805 ext4_debug("Checking group descriptors");
1807 for (i = 0; i < sbi->s_groups_count; i++) {
1808 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1810 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1811 last_block = ext4_blocks_count(sbi->s_es) - 1;
1813 last_block = first_block +
1814 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1816 block_bitmap = ext4_block_bitmap(sb, gdp);
1817 if (block_bitmap < first_block || block_bitmap > last_block) {
1818 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1819 "Block bitmap for group %u not in group "
1820 "(block %llu)!", i, block_bitmap);
1823 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1824 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1825 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1826 "Inode bitmap for group %u not in group "
1827 "(block %llu)!", i, inode_bitmap);
1830 inode_table = ext4_inode_table(sb, gdp);
1831 if (inode_table < first_block ||
1832 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1833 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1834 "Inode table for group %u not in group "
1835 "(block %llu)!", i, inode_table);
1838 ext4_lock_group(sb, i);
1839 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1840 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1841 "Checksum for group %u failed (%u!=%u)",
1842 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1843 gdp)), le16_to_cpu(gdp->bg_checksum));
1844 if (!(sb->s_flags & MS_RDONLY)) {
1845 ext4_unlock_group(sb, i);
1849 ext4_unlock_group(sb, i);
1851 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1854 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1855 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1859 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1860 * the superblock) which were deleted from all directories, but held open by
1861 * a process at the time of a crash. We walk the list and try to delete these
1862 * inodes at recovery time (only with a read-write filesystem).
1864 * In order to keep the orphan inode chain consistent during traversal (in
1865 * case of crash during recovery), we link each inode into the superblock
1866 * orphan list_head and handle it the same way as an inode deletion during
1867 * normal operation (which journals the operations for us).
1869 * We only do an iget() and an iput() on each inode, which is very safe if we
1870 * accidentally point at an in-use or already deleted inode. The worst that
1871 * can happen in this case is that we get a "bit already cleared" message from
1872 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1873 * e2fsck was run on this filesystem, and it must have already done the orphan
1874 * inode cleanup for us, so we can safely abort without any further action.
1876 static void ext4_orphan_cleanup(struct super_block *sb,
1877 struct ext4_super_block *es)
1879 unsigned int s_flags = sb->s_flags;
1880 int nr_orphans = 0, nr_truncates = 0;
1884 if (!es->s_last_orphan) {
1885 jbd_debug(4, "no orphan inodes to clean up\n");
1889 if (bdev_read_only(sb->s_bdev)) {
1890 ext4_msg(sb, KERN_ERR, "write access "
1891 "unavailable, skipping orphan cleanup");
1895 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1896 if (es->s_last_orphan)
1897 jbd_debug(1, "Errors on filesystem, "
1898 "clearing orphan list.\n");
1899 es->s_last_orphan = 0;
1900 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1904 if (s_flags & MS_RDONLY) {
1905 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1906 sb->s_flags &= ~MS_RDONLY;
1909 /* Needed for iput() to work correctly and not trash data */
1910 sb->s_flags |= MS_ACTIVE;
1911 /* Turn on quotas so that they are updated correctly */
1912 for (i = 0; i < MAXQUOTAS; i++) {
1913 if (EXT4_SB(sb)->s_qf_names[i]) {
1914 int ret = ext4_quota_on_mount(sb, i);
1916 ext4_msg(sb, KERN_ERR,
1917 "Cannot turn on journaled "
1918 "quota: error %d", ret);
1923 while (es->s_last_orphan) {
1924 struct inode *inode;
1926 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1927 if (IS_ERR(inode)) {
1928 es->s_last_orphan = 0;
1932 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1933 dquot_initialize(inode);
1934 if (inode->i_nlink) {
1935 ext4_msg(sb, KERN_DEBUG,
1936 "%s: truncating inode %lu to %lld bytes",
1937 __func__, inode->i_ino, inode->i_size);
1938 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1939 inode->i_ino, inode->i_size);
1940 ext4_truncate(inode);
1943 ext4_msg(sb, KERN_DEBUG,
1944 "%s: deleting unreferenced inode %lu",
1945 __func__, inode->i_ino);
1946 jbd_debug(2, "deleting unreferenced inode %lu\n",
1950 iput(inode); /* The delete magic happens here! */
1953 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1956 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1957 PLURAL(nr_orphans));
1959 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1960 PLURAL(nr_truncates));
1962 /* Turn quotas off */
1963 for (i = 0; i < MAXQUOTAS; i++) {
1964 if (sb_dqopt(sb)->files[i])
1965 vfs_quota_off(sb, i, 0);
1968 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1972 * Maximal extent format file size.
1973 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1974 * extent format containers, within a sector_t, and within i_blocks
1975 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1976 * so that won't be a limiting factor.
1978 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1980 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1983 loff_t upper_limit = MAX_LFS_FILESIZE;
1985 /* small i_blocks in vfs inode? */
1986 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 * CONFIG_LBDAF is not enabled implies the inode
1989 * i_block represent total blocks in 512 bytes
1990 * 32 == size of vfs inode i_blocks * 8
1992 upper_limit = (1LL << 32) - 1;
1994 /* total blocks in file system block size */
1995 upper_limit >>= (blkbits - 9);
1996 upper_limit <<= blkbits;
1999 /* 32-bit extent-start container, ee_block */
2004 /* Sanity check against vm- & vfs- imposed limits */
2005 if (res > upper_limit)
2012 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2013 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2014 * We need to be 1 filesystem block less than the 2^48 sector limit.
2016 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2018 loff_t res = EXT4_NDIR_BLOCKS;
2021 /* This is calculated to be the largest file size for a dense, block
2022 * mapped file such that the file's total number of 512-byte sectors,
2023 * including data and all indirect blocks, does not exceed (2^48 - 1).
2025 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2026 * number of 512-byte sectors of the file.
2029 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2031 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2032 * the inode i_block field represents total file blocks in
2033 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2035 upper_limit = (1LL << 32) - 1;
2037 /* total blocks in file system block size */
2038 upper_limit >>= (bits - 9);
2042 * We use 48 bit ext4_inode i_blocks
2043 * With EXT4_HUGE_FILE_FL set the i_blocks
2044 * represent total number of blocks in
2045 * file system block size
2047 upper_limit = (1LL << 48) - 1;
2051 /* indirect blocks */
2053 /* double indirect blocks */
2054 meta_blocks += 1 + (1LL << (bits-2));
2055 /* tripple indirect blocks */
2056 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2058 upper_limit -= meta_blocks;
2059 upper_limit <<= bits;
2061 res += 1LL << (bits-2);
2062 res += 1LL << (2*(bits-2));
2063 res += 1LL << (3*(bits-2));
2065 if (res > upper_limit)
2068 if (res > MAX_LFS_FILESIZE)
2069 res = MAX_LFS_FILESIZE;
2074 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2075 ext4_fsblk_t logical_sb_block, int nr)
2077 struct ext4_sb_info *sbi = EXT4_SB(sb);
2078 ext4_group_t bg, first_meta_bg;
2081 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2083 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2085 return logical_sb_block + nr + 1;
2086 bg = sbi->s_desc_per_block * nr;
2087 if (ext4_bg_has_super(sb, bg))
2090 return (has_super + ext4_group_first_block_no(sb, bg));
2094 * ext4_get_stripe_size: Get the stripe size.
2095 * @sbi: In memory super block info
2097 * If we have specified it via mount option, then
2098 * use the mount option value. If the value specified at mount time is
2099 * greater than the blocks per group use the super block value.
2100 * If the super block value is greater than blocks per group return 0.
2101 * Allocator needs it be less than blocks per group.
2104 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2106 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2107 unsigned long stripe_width =
2108 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2110 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2111 return sbi->s_stripe;
2113 if (stripe_width <= sbi->s_blocks_per_group)
2114 return stripe_width;
2116 if (stride <= sbi->s_blocks_per_group)
2125 struct attribute attr;
2126 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2127 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2128 const char *, size_t);
2132 static int parse_strtoul(const char *buf,
2133 unsigned long max, unsigned long *value)
2137 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2138 endp = skip_spaces(endp);
2139 if (*endp || *value > max)
2145 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2146 struct ext4_sb_info *sbi,
2149 return snprintf(buf, PAGE_SIZE, "%llu\n",
2150 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2153 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2154 struct ext4_sb_info *sbi, char *buf)
2156 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2158 return snprintf(buf, PAGE_SIZE, "%lu\n",
2159 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2160 sbi->s_sectors_written_start) >> 1);
2163 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2164 struct ext4_sb_info *sbi, char *buf)
2166 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2168 return snprintf(buf, PAGE_SIZE, "%llu\n",
2169 (unsigned long long)(sbi->s_kbytes_written +
2170 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2171 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2174 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2175 struct ext4_sb_info *sbi,
2176 const char *buf, size_t count)
2180 if (parse_strtoul(buf, 0x40000000, &t))
2183 if (!is_power_of_2(t))
2186 sbi->s_inode_readahead_blks = t;
2190 static ssize_t sbi_ui_show(struct ext4_attr *a,
2191 struct ext4_sb_info *sbi, char *buf)
2193 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2195 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2198 static ssize_t sbi_ui_store(struct ext4_attr *a,
2199 struct ext4_sb_info *sbi,
2200 const char *buf, size_t count)
2202 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2205 if (parse_strtoul(buf, 0xffffffff, &t))
2211 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2212 static struct ext4_attr ext4_attr_##_name = { \
2213 .attr = {.name = __stringify(_name), .mode = _mode }, \
2216 .offset = offsetof(struct ext4_sb_info, _elname), \
2218 #define EXT4_ATTR(name, mode, show, store) \
2219 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2221 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2222 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2223 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2224 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2225 #define ATTR_LIST(name) &ext4_attr_##name.attr
2227 EXT4_RO_ATTR(delayed_allocation_blocks);
2228 EXT4_RO_ATTR(session_write_kbytes);
2229 EXT4_RO_ATTR(lifetime_write_kbytes);
2230 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2231 inode_readahead_blks_store, s_inode_readahead_blks);
2232 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2233 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2234 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2235 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2236 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2237 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2238 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2239 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2241 static struct attribute *ext4_attrs[] = {
2242 ATTR_LIST(delayed_allocation_blocks),
2243 ATTR_LIST(session_write_kbytes),
2244 ATTR_LIST(lifetime_write_kbytes),
2245 ATTR_LIST(inode_readahead_blks),
2246 ATTR_LIST(inode_goal),
2247 ATTR_LIST(mb_stats),
2248 ATTR_LIST(mb_max_to_scan),
2249 ATTR_LIST(mb_min_to_scan),
2250 ATTR_LIST(mb_order2_req),
2251 ATTR_LIST(mb_stream_req),
2252 ATTR_LIST(mb_group_prealloc),
2253 ATTR_LIST(max_writeback_mb_bump),
2257 static ssize_t ext4_attr_show(struct kobject *kobj,
2258 struct attribute *attr, char *buf)
2260 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2262 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2264 return a->show ? a->show(a, sbi, buf) : 0;
2267 static ssize_t ext4_attr_store(struct kobject *kobj,
2268 struct attribute *attr,
2269 const char *buf, size_t len)
2271 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2273 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2275 return a->store ? a->store(a, sbi, buf, len) : 0;
2278 static void ext4_sb_release(struct kobject *kobj)
2280 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2282 complete(&sbi->s_kobj_unregister);
2286 static struct sysfs_ops ext4_attr_ops = {
2287 .show = ext4_attr_show,
2288 .store = ext4_attr_store,
2291 static struct kobj_type ext4_ktype = {
2292 .default_attrs = ext4_attrs,
2293 .sysfs_ops = &ext4_attr_ops,
2294 .release = ext4_sb_release,
2298 * Check whether this filesystem can be mounted based on
2299 * the features present and the RDONLY/RDWR mount requested.
2300 * Returns 1 if this filesystem can be mounted as requested,
2301 * 0 if it cannot be.
2303 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2305 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2306 ext4_msg(sb, KERN_ERR,
2307 "Couldn't mount because of "
2308 "unsupported optional features (%x)",
2309 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2310 ~EXT4_FEATURE_INCOMPAT_SUPP));
2317 /* Check that feature set is OK for a read-write mount */
2318 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2319 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2320 "unsupported optional features (%x)",
2321 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2322 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2326 * Large file size enabled file system can only be mounted
2327 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2329 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2330 if (sizeof(blkcnt_t) < sizeof(u64)) {
2331 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2332 "cannot be mounted RDWR without "
2340 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2341 __releases(kernel_lock)
2342 __acquires(kernel_lock)
2344 struct buffer_head *bh;
2345 struct ext4_super_block *es = NULL;
2346 struct ext4_sb_info *sbi;
2348 ext4_fsblk_t sb_block = get_sb_block(&data);
2349 ext4_fsblk_t logical_sb_block;
2350 unsigned long offset = 0;
2351 unsigned long journal_devnum = 0;
2352 unsigned long def_mount_opts;
2358 unsigned int db_count;
2360 int needs_recovery, has_huge_files;
2363 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2365 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2369 sbi->s_blockgroup_lock =
2370 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2371 if (!sbi->s_blockgroup_lock) {
2375 sb->s_fs_info = sbi;
2376 sbi->s_mount_opt = 0;
2377 sbi->s_resuid = EXT4_DEF_RESUID;
2378 sbi->s_resgid = EXT4_DEF_RESGID;
2379 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2380 sbi->s_sb_block = sb_block;
2381 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2386 /* Cleanup superblock name */
2387 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2390 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2392 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2397 * The ext4 superblock will not be buffer aligned for other than 1kB
2398 * block sizes. We need to calculate the offset from buffer start.
2400 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2401 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2402 offset = do_div(logical_sb_block, blocksize);
2404 logical_sb_block = sb_block;
2407 if (!(bh = sb_bread(sb, logical_sb_block))) {
2408 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2412 * Note: s_es must be initialized as soon as possible because
2413 * some ext4 macro-instructions depend on its value
2415 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2417 sb->s_magic = le16_to_cpu(es->s_magic);
2418 if (sb->s_magic != EXT4_SUPER_MAGIC)
2420 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2422 /* Set defaults before we parse the mount options */
2423 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2424 if (def_mount_opts & EXT4_DEFM_DEBUG)
2425 set_opt(sbi->s_mount_opt, DEBUG);
2426 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2427 set_opt(sbi->s_mount_opt, GRPID);
2428 if (def_mount_opts & EXT4_DEFM_UID16)
2429 set_opt(sbi->s_mount_opt, NO_UID32);
2430 #ifdef CONFIG_EXT4_FS_XATTR
2431 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2432 set_opt(sbi->s_mount_opt, XATTR_USER);
2434 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2435 if (def_mount_opts & EXT4_DEFM_ACL)
2436 set_opt(sbi->s_mount_opt, POSIX_ACL);
2438 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2439 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2440 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2441 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2442 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2443 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2445 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2446 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2447 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2448 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2450 set_opt(sbi->s_mount_opt, ERRORS_RO);
2452 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2453 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2454 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2455 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2456 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2458 set_opt(sbi->s_mount_opt, BARRIER);
2461 * enable delayed allocation by default
2462 * Use -o nodelalloc to turn it off
2464 set_opt(sbi->s_mount_opt, DELALLOC);
2466 if (!parse_options((char *) data, sb, &journal_devnum,
2467 &journal_ioprio, NULL, 0))
2470 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2471 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2473 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2474 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2475 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2476 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2477 ext4_msg(sb, KERN_WARNING,
2478 "feature flags set on rev 0 fs, "
2479 "running e2fsck is recommended");
2482 * Check feature flags regardless of the revision level, since we
2483 * previously didn't change the revision level when setting the flags,
2484 * so there is a chance incompat flags are set on a rev 0 filesystem.
2486 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2489 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2491 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2492 blocksize > EXT4_MAX_BLOCK_SIZE) {
2493 ext4_msg(sb, KERN_ERR,
2494 "Unsupported filesystem blocksize %d", blocksize);
2498 if (sb->s_blocksize != blocksize) {
2499 /* Validate the filesystem blocksize */
2500 if (!sb_set_blocksize(sb, blocksize)) {
2501 ext4_msg(sb, KERN_ERR, "bad block size %d",
2507 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2508 offset = do_div(logical_sb_block, blocksize);
2509 bh = sb_bread(sb, logical_sb_block);
2511 ext4_msg(sb, KERN_ERR,
2512 "Can't read superblock on 2nd try");
2515 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2517 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2518 ext4_msg(sb, KERN_ERR,
2519 "Magic mismatch, very weird!");
2524 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2525 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2526 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2528 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2530 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2531 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2532 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2534 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2535 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2536 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2537 (!is_power_of_2(sbi->s_inode_size)) ||
2538 (sbi->s_inode_size > blocksize)) {
2539 ext4_msg(sb, KERN_ERR,
2540 "unsupported inode size: %d",
2544 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2545 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2548 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2549 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2550 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2551 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2552 !is_power_of_2(sbi->s_desc_size)) {
2553 ext4_msg(sb, KERN_ERR,
2554 "unsupported descriptor size %lu",
2559 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2561 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2562 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2563 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2566 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2567 if (sbi->s_inodes_per_block == 0)
2569 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2570 sbi->s_inodes_per_block;
2571 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2573 sbi->s_mount_state = le16_to_cpu(es->s_state);
2574 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2575 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2577 for (i = 0; i < 4; i++)
2578 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2579 sbi->s_def_hash_version = es->s_def_hash_version;
2580 i = le32_to_cpu(es->s_flags);
2581 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2582 sbi->s_hash_unsigned = 3;
2583 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2584 #ifdef __CHAR_UNSIGNED__
2585 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2586 sbi->s_hash_unsigned = 3;
2588 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2593 if (sbi->s_blocks_per_group > blocksize * 8) {
2594 ext4_msg(sb, KERN_ERR,
2595 "#blocks per group too big: %lu",
2596 sbi->s_blocks_per_group);
2599 if (sbi->s_inodes_per_group > blocksize * 8) {
2600 ext4_msg(sb, KERN_ERR,
2601 "#inodes per group too big: %lu",
2602 sbi->s_inodes_per_group);
2607 * Test whether we have more sectors than will fit in sector_t,
2608 * and whether the max offset is addressable by the page cache.
2610 if ((ext4_blocks_count(es) >
2611 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2612 (ext4_blocks_count(es) >
2613 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2614 ext4_msg(sb, KERN_ERR, "filesystem"
2615 " too large to mount safely on this system");
2616 if (sizeof(sector_t) < 8)
2617 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2622 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2625 /* check blocks count against device size */
2626 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2627 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2628 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2629 "exceeds size of device (%llu blocks)",
2630 ext4_blocks_count(es), blocks_count);
2635 * It makes no sense for the first data block to be beyond the end
2636 * of the filesystem.
2638 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2639 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2640 "block %u is beyond end of filesystem (%llu)",
2641 le32_to_cpu(es->s_first_data_block),
2642 ext4_blocks_count(es));
2645 blocks_count = (ext4_blocks_count(es) -
2646 le32_to_cpu(es->s_first_data_block) +
2647 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2648 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2649 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2650 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2651 "(block count %llu, first data block %u, "
2652 "blocks per group %lu)", sbi->s_groups_count,
2653 ext4_blocks_count(es),
2654 le32_to_cpu(es->s_first_data_block),
2655 EXT4_BLOCKS_PER_GROUP(sb));
2658 sbi->s_groups_count = blocks_count;
2659 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2660 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2661 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2662 EXT4_DESC_PER_BLOCK(sb);
2663 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2665 if (sbi->s_group_desc == NULL) {
2666 ext4_msg(sb, KERN_ERR, "not enough memory");
2670 #ifdef CONFIG_PROC_FS
2672 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2675 bgl_lock_init(sbi->s_blockgroup_lock);
2677 for (i = 0; i < db_count; i++) {
2678 block = descriptor_loc(sb, logical_sb_block, i);
2679 sbi->s_group_desc[i] = sb_bread(sb, block);
2680 if (!sbi->s_group_desc[i]) {
2681 ext4_msg(sb, KERN_ERR,
2682 "can't read group descriptor %d", i);
2687 if (!ext4_check_descriptors(sb)) {
2688 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2691 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2692 if (!ext4_fill_flex_info(sb)) {
2693 ext4_msg(sb, KERN_ERR,
2694 "unable to initialize "
2695 "flex_bg meta info!");
2699 sbi->s_gdb_count = db_count;
2700 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2701 spin_lock_init(&sbi->s_next_gen_lock);
2703 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2704 ext4_count_free_blocks(sb));
2706 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2707 ext4_count_free_inodes(sb));
2710 err = percpu_counter_init(&sbi->s_dirs_counter,
2711 ext4_count_dirs(sb));
2714 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2717 ext4_msg(sb, KERN_ERR, "insufficient memory");
2721 sbi->s_stripe = ext4_get_stripe_size(sbi);
2722 sbi->s_max_writeback_mb_bump = 128;
2725 * set up enough so that it can read an inode
2727 if (!test_opt(sb, NOLOAD) &&
2728 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2729 sb->s_op = &ext4_sops;
2731 sb->s_op = &ext4_nojournal_sops;
2732 sb->s_export_op = &ext4_export_ops;
2733 sb->s_xattr = ext4_xattr_handlers;
2735 sb->s_qcop = &ext4_qctl_operations;
2736 sb->dq_op = &ext4_quota_operations;
2738 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2739 mutex_init(&sbi->s_orphan_lock);
2740 mutex_init(&sbi->s_resize_lock);
2744 needs_recovery = (es->s_last_orphan != 0 ||
2745 EXT4_HAS_INCOMPAT_FEATURE(sb,
2746 EXT4_FEATURE_INCOMPAT_RECOVER));
2749 * The first inode we look at is the journal inode. Don't try
2750 * root first: it may be modified in the journal!
2752 if (!test_opt(sb, NOLOAD) &&
2753 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2754 if (ext4_load_journal(sb, es, journal_devnum))
2756 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2757 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2758 ext4_msg(sb, KERN_ERR, "required journal recovery "
2759 "suppressed and not mounted read-only");
2762 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2763 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2764 sbi->s_journal = NULL;
2769 if (ext4_blocks_count(es) > 0xffffffffULL &&
2770 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2771 JBD2_FEATURE_INCOMPAT_64BIT)) {
2772 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2776 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2777 jbd2_journal_set_features(sbi->s_journal,
2778 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2779 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2780 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2781 jbd2_journal_set_features(sbi->s_journal,
2782 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2783 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2784 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2786 jbd2_journal_clear_features(sbi->s_journal,
2787 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2788 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2791 /* We have now updated the journal if required, so we can
2792 * validate the data journaling mode. */
2793 switch (test_opt(sb, DATA_FLAGS)) {
2795 /* No mode set, assume a default based on the journal
2796 * capabilities: ORDERED_DATA if the journal can
2797 * cope, else JOURNAL_DATA
2799 if (jbd2_journal_check_available_features
2800 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2801 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2803 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2806 case EXT4_MOUNT_ORDERED_DATA:
2807 case EXT4_MOUNT_WRITEBACK_DATA:
2808 if (!jbd2_journal_check_available_features
2809 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2810 ext4_msg(sb, KERN_ERR, "Journal does not support "
2811 "requested data journaling mode");
2817 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2821 if (test_opt(sb, NOBH)) {
2822 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2823 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2824 "its supported only with writeback mode");
2825 clear_opt(sbi->s_mount_opt, NOBH);
2828 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2829 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2830 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2831 goto failed_mount_wq;
2835 * The jbd2_journal_load will have done any necessary log recovery,
2836 * so we can safely mount the rest of the filesystem now.
2839 root = ext4_iget(sb, EXT4_ROOT_INO);
2841 ext4_msg(sb, KERN_ERR, "get root inode failed");
2842 ret = PTR_ERR(root);
2845 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2847 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2850 sb->s_root = d_alloc_root(root);
2852 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2858 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2860 /* determine the minimum size of new large inodes, if present */
2861 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2862 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2863 EXT4_GOOD_OLD_INODE_SIZE;
2864 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2865 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2866 if (sbi->s_want_extra_isize <
2867 le16_to_cpu(es->s_want_extra_isize))
2868 sbi->s_want_extra_isize =
2869 le16_to_cpu(es->s_want_extra_isize);
2870 if (sbi->s_want_extra_isize <
2871 le16_to_cpu(es->s_min_extra_isize))
2872 sbi->s_want_extra_isize =
2873 le16_to_cpu(es->s_min_extra_isize);
2876 /* Check if enough inode space is available */
2877 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2878 sbi->s_inode_size) {
2879 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2880 EXT4_GOOD_OLD_INODE_SIZE;
2881 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2885 if (test_opt(sb, DELALLOC) &&
2886 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2887 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2888 "requested data journaling mode");
2889 clear_opt(sbi->s_mount_opt, DELALLOC);
2892 err = ext4_setup_system_zone(sb);
2894 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2895 "zone (%d)\n", err);
2900 err = ext4_mb_init(sb, needs_recovery);
2902 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2907 sbi->s_kobj.kset = ext4_kset;
2908 init_completion(&sbi->s_kobj_unregister);
2909 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2912 ext4_mb_release(sb);
2913 ext4_ext_release(sb);
2917 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2918 ext4_orphan_cleanup(sb, es);
2919 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2920 if (needs_recovery) {
2921 ext4_msg(sb, KERN_INFO, "recovery complete");
2922 ext4_mark_recovery_complete(sb, es);
2924 if (EXT4_SB(sb)->s_journal) {
2925 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2926 descr = " journalled data mode";
2927 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2928 descr = " ordered data mode";
2930 descr = " writeback data mode";
2932 descr = "out journal";
2934 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2941 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2945 ext4_msg(sb, KERN_ERR, "mount failed");
2946 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2948 ext4_release_system_zone(sb);
2949 if (sbi->s_journal) {
2950 jbd2_journal_destroy(sbi->s_journal);
2951 sbi->s_journal = NULL;
2954 if (sbi->s_flex_groups) {
2955 if (is_vmalloc_addr(sbi->s_flex_groups))
2956 vfree(sbi->s_flex_groups);
2958 kfree(sbi->s_flex_groups);
2960 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2961 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2962 percpu_counter_destroy(&sbi->s_dirs_counter);
2963 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2965 for (i = 0; i < db_count; i++)
2966 brelse(sbi->s_group_desc[i]);
2967 kfree(sbi->s_group_desc);
2970 remove_proc_entry(sb->s_id, ext4_proc_root);
2973 for (i = 0; i < MAXQUOTAS; i++)
2974 kfree(sbi->s_qf_names[i]);
2976 ext4_blkdev_remove(sbi);
2979 sb->s_fs_info = NULL;
2980 kfree(sbi->s_blockgroup_lock);
2987 * Setup any per-fs journal parameters now. We'll do this both on
2988 * initial mount, once the journal has been initialised but before we've
2989 * done any recovery; and again on any subsequent remount.
2991 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2993 struct ext4_sb_info *sbi = EXT4_SB(sb);
2995 journal->j_commit_interval = sbi->s_commit_interval;
2996 journal->j_min_batch_time = sbi->s_min_batch_time;
2997 journal->j_max_batch_time = sbi->s_max_batch_time;
2999 spin_lock(&journal->j_state_lock);
3000 if (test_opt(sb, BARRIER))
3001 journal->j_flags |= JBD2_BARRIER;
3003 journal->j_flags &= ~JBD2_BARRIER;
3004 if (test_opt(sb, DATA_ERR_ABORT))
3005 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3007 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3008 spin_unlock(&journal->j_state_lock);
3011 static journal_t *ext4_get_journal(struct super_block *sb,
3012 unsigned int journal_inum)
3014 struct inode *journal_inode;
3017 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3019 /* First, test for the existence of a valid inode on disk. Bad
3020 * things happen if we iget() an unused inode, as the subsequent
3021 * iput() will try to delete it. */
3023 journal_inode = ext4_iget(sb, journal_inum);
3024 if (IS_ERR(journal_inode)) {
3025 ext4_msg(sb, KERN_ERR, "no journal found");
3028 if (!journal_inode->i_nlink) {
3029 make_bad_inode(journal_inode);
3030 iput(journal_inode);
3031 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3035 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3036 journal_inode, journal_inode->i_size);
3037 if (!S_ISREG(journal_inode->i_mode)) {
3038 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3039 iput(journal_inode);
3043 journal = jbd2_journal_init_inode(journal_inode);
3045 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3046 iput(journal_inode);
3049 journal->j_private = sb;
3050 ext4_init_journal_params(sb, journal);
3054 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3057 struct buffer_head *bh;
3061 int hblock, blocksize;
3062 ext4_fsblk_t sb_block;
3063 unsigned long offset;
3064 struct ext4_super_block *es;
3065 struct block_device *bdev;
3067 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3069 bdev = ext4_blkdev_get(j_dev, sb);
3073 if (bd_claim(bdev, sb)) {
3074 ext4_msg(sb, KERN_ERR,
3075 "failed to claim external journal device");
3076 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3080 blocksize = sb->s_blocksize;
3081 hblock = bdev_logical_block_size(bdev);
3082 if (blocksize < hblock) {
3083 ext4_msg(sb, KERN_ERR,
3084 "blocksize too small for journal device");
3088 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3089 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3090 set_blocksize(bdev, blocksize);
3091 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3092 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3093 "external journal");
3097 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3098 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3099 !(le32_to_cpu(es->s_feature_incompat) &
3100 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3101 ext4_msg(sb, KERN_ERR, "external journal has "
3107 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3108 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3113 len = ext4_blocks_count(es);
3114 start = sb_block + 1;
3115 brelse(bh); /* we're done with the superblock */
3117 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3118 start, len, blocksize);
3120 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3123 journal->j_private = sb;
3124 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3125 wait_on_buffer(journal->j_sb_buffer);
3126 if (!buffer_uptodate(journal->j_sb_buffer)) {
3127 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3130 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3131 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3132 "user (unsupported) - %d",
3133 be32_to_cpu(journal->j_superblock->s_nr_users));
3136 EXT4_SB(sb)->journal_bdev = bdev;
3137 ext4_init_journal_params(sb, journal);
3141 jbd2_journal_destroy(journal);
3143 ext4_blkdev_put(bdev);
3147 static int ext4_load_journal(struct super_block *sb,
3148 struct ext4_super_block *es,
3149 unsigned long journal_devnum)
3152 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3155 int really_read_only;
3157 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3159 if (journal_devnum &&
3160 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3161 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3162 "numbers have changed");
3163 journal_dev = new_decode_dev(journal_devnum);
3165 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3167 really_read_only = bdev_read_only(sb->s_bdev);
3170 * Are we loading a blank journal or performing recovery after a
3171 * crash? For recovery, we need to check in advance whether we
3172 * can get read-write access to the device.
3174 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3175 if (sb->s_flags & MS_RDONLY) {
3176 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3177 "required on readonly filesystem");
3178 if (really_read_only) {
3179 ext4_msg(sb, KERN_ERR, "write access "
3180 "unavailable, cannot proceed");
3183 ext4_msg(sb, KERN_INFO, "write access will "
3184 "be enabled during recovery");
3188 if (journal_inum && journal_dev) {
3189 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3190 "and inode journals!");
3195 if (!(journal = ext4_get_journal(sb, journal_inum)))
3198 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3202 if (!(journal->j_flags & JBD2_BARRIER))
3203 ext4_msg(sb, KERN_INFO, "barriers disabled");
3205 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3206 err = jbd2_journal_update_format(journal);
3208 ext4_msg(sb, KERN_ERR, "error updating journal");
3209 jbd2_journal_destroy(journal);
3214 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3215 err = jbd2_journal_wipe(journal, !really_read_only);
3217 err = jbd2_journal_load(journal);
3220 ext4_msg(sb, KERN_ERR, "error loading journal");
3221 jbd2_journal_destroy(journal);
3225 EXT4_SB(sb)->s_journal = journal;
3226 ext4_clear_journal_err(sb, es);
3228 if (journal_devnum &&
3229 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3230 es->s_journal_dev = cpu_to_le32(journal_devnum);
3232 /* Make sure we flush the recovery flag to disk. */
3233 ext4_commit_super(sb, 1);
3239 static int ext4_commit_super(struct super_block *sb, int sync)
3241 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3242 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3247 if (buffer_write_io_error(sbh)) {
3249 * Oh, dear. A previous attempt to write the
3250 * superblock failed. This could happen because the
3251 * USB device was yanked out. Or it could happen to
3252 * be a transient write error and maybe the block will
3253 * be remapped. Nothing we can do but to retry the
3254 * write and hope for the best.
3256 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3257 "superblock detected");
3258 clear_buffer_write_io_error(sbh);
3259 set_buffer_uptodate(sbh);
3262 * If the file system is mounted read-only, don't update the
3263 * superblock write time. This avoids updating the superblock
3264 * write time when we are mounting the root file system
3265 * read/only but we need to replay the journal; at that point,
3266 * for people who are east of GMT and who make their clock
3267 * tick in localtime for Windows bug-for-bug compatibility,
3268 * the clock is set in the future, and this will cause e2fsck
3269 * to complain and force a full file system check.
3271 if (!(sb->s_flags & MS_RDONLY))
3272 es->s_wtime = cpu_to_le32(get_seconds());
3273 es->s_kbytes_written =
3274 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3275 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3276 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3277 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3278 &EXT4_SB(sb)->s_freeblocks_counter));
3279 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3280 &EXT4_SB(sb)->s_freeinodes_counter));
3282 BUFFER_TRACE(sbh, "marking dirty");
3283 mark_buffer_dirty(sbh);
3285 error = sync_dirty_buffer(sbh);
3289 error = buffer_write_io_error(sbh);
3291 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3293 clear_buffer_write_io_error(sbh);
3294 set_buffer_uptodate(sbh);
3301 * Have we just finished recovery? If so, and if we are mounting (or
3302 * remounting) the filesystem readonly, then we will end up with a
3303 * consistent fs on disk. Record that fact.
3305 static void ext4_mark_recovery_complete(struct super_block *sb,
3306 struct ext4_super_block *es)
3308 journal_t *journal = EXT4_SB(sb)->s_journal;
3310 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3311 BUG_ON(journal != NULL);
3314 jbd2_journal_lock_updates(journal);
3315 if (jbd2_journal_flush(journal) < 0)
3318 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3319 sb->s_flags & MS_RDONLY) {
3320 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3321 ext4_commit_super(sb, 1);
3325 jbd2_journal_unlock_updates(journal);
3329 * If we are mounting (or read-write remounting) a filesystem whose journal
3330 * has recorded an error from a previous lifetime, move that error to the
3331 * main filesystem now.
3333 static void ext4_clear_journal_err(struct super_block *sb,
3334 struct ext4_super_block *es)
3340 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3342 journal = EXT4_SB(sb)->s_journal;
3345 * Now check for any error status which may have been recorded in the
3346 * journal by a prior ext4_error() or ext4_abort()
3349 j_errno = jbd2_journal_errno(journal);
3353 errstr = ext4_decode_error(sb, j_errno, nbuf);
3354 ext4_warning(sb, __func__, "Filesystem error recorded "
3355 "from previous mount: %s", errstr);
3356 ext4_warning(sb, __func__, "Marking fs in need of "
3357 "filesystem check.");
3359 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3360 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3361 ext4_commit_super(sb, 1);
3363 jbd2_journal_clear_err(journal);
3368 * Force the running and committing transactions to commit,
3369 * and wait on the commit.
3371 int ext4_force_commit(struct super_block *sb)
3376 if (sb->s_flags & MS_RDONLY)
3379 journal = EXT4_SB(sb)->s_journal;
3381 ret = ext4_journal_force_commit(journal);
3386 static void ext4_write_super(struct super_block *sb)
3389 ext4_commit_super(sb, 1);
3393 static int ext4_sync_fs(struct super_block *sb, int wait)
3397 struct ext4_sb_info *sbi = EXT4_SB(sb);
3399 trace_ext4_sync_fs(sb, wait);
3400 flush_workqueue(sbi->dio_unwritten_wq);
3401 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3403 jbd2_log_wait_commit(sbi->s_journal, target);
3409 * LVM calls this function before a (read-only) snapshot is created. This
3410 * gives us a chance to flush the journal completely and mark the fs clean.
3412 static int ext4_freeze(struct super_block *sb)
3417 if (sb->s_flags & MS_RDONLY)
3420 journal = EXT4_SB(sb)->s_journal;
3422 /* Now we set up the journal barrier. */
3423 jbd2_journal_lock_updates(journal);
3426 * Don't clear the needs_recovery flag if we failed to flush
3429 error = jbd2_journal_flush(journal);
3432 jbd2_journal_unlock_updates(journal);
3436 /* Journal blocked and flushed, clear needs_recovery flag. */
3437 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3438 error = ext4_commit_super(sb, 1);
3445 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3446 * flag here, even though the filesystem is not technically dirty yet.
3448 static int ext4_unfreeze(struct super_block *sb)
3450 if (sb->s_flags & MS_RDONLY)
3454 /* Reset the needs_recovery flag before the fs is unlocked. */
3455 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3456 ext4_commit_super(sb, 1);
3458 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3462 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3464 struct ext4_super_block *es;
3465 struct ext4_sb_info *sbi = EXT4_SB(sb);
3466 ext4_fsblk_t n_blocks_count = 0;
3467 unsigned long old_sb_flags;
3468 struct ext4_mount_options old_opts;
3470 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3478 /* Store the original options */
3480 old_sb_flags = sb->s_flags;
3481 old_opts.s_mount_opt = sbi->s_mount_opt;
3482 old_opts.s_resuid = sbi->s_resuid;
3483 old_opts.s_resgid = sbi->s_resgid;
3484 old_opts.s_commit_interval = sbi->s_commit_interval;
3485 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3486 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3488 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3489 for (i = 0; i < MAXQUOTAS; i++)
3490 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3492 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3493 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3496 * Allow the "check" option to be passed as a remount option.
3498 if (!parse_options(data, sb, NULL, &journal_ioprio,
3499 &n_blocks_count, 1)) {
3504 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3505 ext4_abort(sb, __func__, "Abort forced by user");
3507 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3508 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3512 if (sbi->s_journal) {
3513 ext4_init_journal_params(sb, sbi->s_journal);
3514 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3517 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3518 n_blocks_count > ext4_blocks_count(es)) {
3519 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3524 if (*flags & MS_RDONLY) {
3526 * First of all, the unconditional stuff we have to do
3527 * to disable replay of the journal when we next remount
3529 sb->s_flags |= MS_RDONLY;
3532 * OK, test if we are remounting a valid rw partition
3533 * readonly, and if so set the rdonly flag and then
3534 * mark the partition as valid again.
3536 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3537 (sbi->s_mount_state & EXT4_VALID_FS))
3538 es->s_state = cpu_to_le16(sbi->s_mount_state);
3541 ext4_mark_recovery_complete(sb, es);
3543 /* Make sure we can mount this feature set readwrite */
3544 if (!ext4_feature_set_ok(sb, 0)) {
3549 * Make sure the group descriptor checksums
3550 * are sane. If they aren't, refuse to remount r/w.
3552 for (g = 0; g < sbi->s_groups_count; g++) {
3553 struct ext4_group_desc *gdp =
3554 ext4_get_group_desc(sb, g, NULL);
3556 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3557 ext4_msg(sb, KERN_ERR,
3558 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3559 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3560 le16_to_cpu(gdp->bg_checksum));
3567 * If we have an unprocessed orphan list hanging
3568 * around from a previously readonly bdev mount,
3569 * require a full umount/remount for now.
3571 if (es->s_last_orphan) {
3572 ext4_msg(sb, KERN_WARNING, "Couldn't "
3573 "remount RDWR because of unprocessed "
3574 "orphan inode list. Please "
3575 "umount/remount instead");
3581 * Mounting a RDONLY partition read-write, so reread
3582 * and store the current valid flag. (It may have
3583 * been changed by e2fsck since we originally mounted
3587 ext4_clear_journal_err(sb, es);
3588 sbi->s_mount_state = le16_to_cpu(es->s_state);
3589 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3591 if (!ext4_setup_super(sb, es, 0))
3592 sb->s_flags &= ~MS_RDONLY;
3595 ext4_setup_system_zone(sb);
3596 if (sbi->s_journal == NULL)
3597 ext4_commit_super(sb, 1);
3600 /* Release old quota file names */
3601 for (i = 0; i < MAXQUOTAS; i++)
3602 if (old_opts.s_qf_names[i] &&
3603 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3604 kfree(old_opts.s_qf_names[i]);
3611 sb->s_flags = old_sb_flags;
3612 sbi->s_mount_opt = old_opts.s_mount_opt;
3613 sbi->s_resuid = old_opts.s_resuid;
3614 sbi->s_resgid = old_opts.s_resgid;
3615 sbi->s_commit_interval = old_opts.s_commit_interval;
3616 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3617 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3619 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3620 for (i = 0; i < MAXQUOTAS; i++) {
3621 if (sbi->s_qf_names[i] &&
3622 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3623 kfree(sbi->s_qf_names[i]);
3624 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3632 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3634 struct super_block *sb = dentry->d_sb;
3635 struct ext4_sb_info *sbi = EXT4_SB(sb);
3636 struct ext4_super_block *es = sbi->s_es;
3639 if (test_opt(sb, MINIX_DF)) {
3640 sbi->s_overhead_last = 0;
3641 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3642 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3643 ext4_fsblk_t overhead = 0;
3646 * Compute the overhead (FS structures). This is constant
3647 * for a given filesystem unless the number of block groups
3648 * changes so we cache the previous value until it does.
3652 * All of the blocks before first_data_block are
3655 overhead = le32_to_cpu(es->s_first_data_block);
3658 * Add the overhead attributed to the superblock and
3659 * block group descriptors. If the sparse superblocks
3660 * feature is turned on, then not all groups have this.
3662 for (i = 0; i < ngroups; i++) {
3663 overhead += ext4_bg_has_super(sb, i) +
3664 ext4_bg_num_gdb(sb, i);
3669 * Every block group has an inode bitmap, a block
3670 * bitmap, and an inode table.
3672 overhead += ngroups * (2 + sbi->s_itb_per_group);
3673 sbi->s_overhead_last = overhead;
3675 sbi->s_blocks_last = ext4_blocks_count(es);
3678 buf->f_type = EXT4_SUPER_MAGIC;
3679 buf->f_bsize = sb->s_blocksize;
3680 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3681 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3682 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3683 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3684 if (buf->f_bfree < ext4_r_blocks_count(es))
3686 buf->f_files = le32_to_cpu(es->s_inodes_count);
3687 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3688 buf->f_namelen = EXT4_NAME_LEN;
3689 fsid = le64_to_cpup((void *)es->s_uuid) ^
3690 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3691 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3692 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3697 /* Helper function for writing quotas on sync - we need to start transaction
3698 * before quota file is locked for write. Otherwise the are possible deadlocks:
3699 * Process 1 Process 2
3700 * ext4_create() quota_sync()
3701 * jbd2_journal_start() write_dquot()
3702 * dquot_initialize() down(dqio_mutex)
3703 * down(dqio_mutex) jbd2_journal_start()
3709 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3711 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3714 static int ext4_write_dquot(struct dquot *dquot)
3718 struct inode *inode;
3720 inode = dquot_to_inode(dquot);
3721 handle = ext4_journal_start(inode,
3722 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3724 return PTR_ERR(handle);
3725 ret = dquot_commit(dquot);
3726 err = ext4_journal_stop(handle);
3732 static int ext4_acquire_dquot(struct dquot *dquot)
3737 handle = ext4_journal_start(dquot_to_inode(dquot),
3738 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3740 return PTR_ERR(handle);
3741 ret = dquot_acquire(dquot);
3742 err = ext4_journal_stop(handle);
3748 static int ext4_release_dquot(struct dquot *dquot)
3753 handle = ext4_journal_start(dquot_to_inode(dquot),
3754 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3755 if (IS_ERR(handle)) {
3756 /* Release dquot anyway to avoid endless cycle in dqput() */
3757 dquot_release(dquot);
3758 return PTR_ERR(handle);
3760 ret = dquot_release(dquot);
3761 err = ext4_journal_stop(handle);
3767 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3769 /* Are we journaling quotas? */
3770 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3771 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3772 dquot_mark_dquot_dirty(dquot);
3773 return ext4_write_dquot(dquot);
3775 return dquot_mark_dquot_dirty(dquot);
3779 static int ext4_write_info(struct super_block *sb, int type)
3784 /* Data block + inode block */
3785 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3787 return PTR_ERR(handle);
3788 ret = dquot_commit_info(sb, type);
3789 err = ext4_journal_stop(handle);
3796 * Turn on quotas during mount time - we need to find
3797 * the quota file and such...
3799 static int ext4_quota_on_mount(struct super_block *sb, int type)
3801 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3802 EXT4_SB(sb)->s_jquota_fmt, type);
3806 * Standard function to be called on quota_on
3808 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3809 char *name, int remount)
3814 if (!test_opt(sb, QUOTA))
3816 /* When remounting, no checks are needed and in fact, name is NULL */
3818 return vfs_quota_on(sb, type, format_id, name, remount);
3820 err = kern_path(name, LOOKUP_FOLLOW, &path);
3824 /* Quotafile not on the same filesystem? */
3825 if (path.mnt->mnt_sb != sb) {
3829 /* Journaling quota? */
3830 if (EXT4_SB(sb)->s_qf_names[type]) {
3831 /* Quotafile not in fs root? */
3832 if (path.dentry->d_parent != sb->s_root)
3833 ext4_msg(sb, KERN_WARNING,
3834 "Quota file not on filesystem root. "
3835 "Journaled quota will not work");
3839 * When we journal data on quota file, we have to flush journal to see
3840 * all updates to the file when we bypass pagecache...
3842 if (EXT4_SB(sb)->s_journal &&
3843 ext4_should_journal_data(path.dentry->d_inode)) {
3845 * We don't need to lock updates but journal_flush() could
3846 * otherwise be livelocked...
3848 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3849 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3850 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3857 err = vfs_quota_on_path(sb, type, format_id, &path);
3862 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3863 * acquiring the locks... As quota files are never truncated and quota code
3864 * itself serializes the operations (and noone else should touch the files)
3865 * we don't have to be afraid of races */
3866 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3867 size_t len, loff_t off)
3869 struct inode *inode = sb_dqopt(sb)->files[type];
3870 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3872 int offset = off & (sb->s_blocksize - 1);
3875 struct buffer_head *bh;
3876 loff_t i_size = i_size_read(inode);
3880 if (off+len > i_size)
3883 while (toread > 0) {
3884 tocopy = sb->s_blocksize - offset < toread ?
3885 sb->s_blocksize - offset : toread;
3886 bh = ext4_bread(NULL, inode, blk, 0, &err);
3889 if (!bh) /* A hole? */
3890 memset(data, 0, tocopy);
3892 memcpy(data, bh->b_data+offset, tocopy);
3902 /* Write to quotafile (we know the transaction is already started and has
3903 * enough credits) */
3904 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3905 const char *data, size_t len, loff_t off)
3907 struct inode *inode = sb_dqopt(sb)->files[type];
3908 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3910 int offset = off & (sb->s_blocksize - 1);
3912 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3913 size_t towrite = len;
3914 struct buffer_head *bh;
3915 handle_t *handle = journal_current_handle();
3917 if (EXT4_SB(sb)->s_journal && !handle) {
3918 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3919 " cancelled because transaction is not started",
3920 (unsigned long long)off, (unsigned long long)len);
3923 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3924 while (towrite > 0) {
3925 tocopy = sb->s_blocksize - offset < towrite ?
3926 sb->s_blocksize - offset : towrite;
3927 bh = ext4_bread(handle, inode, blk, 1, &err);
3930 if (journal_quota) {
3931 err = ext4_journal_get_write_access(handle, bh);
3938 memcpy(bh->b_data+offset, data, tocopy);
3939 flush_dcache_page(bh->b_page);
3942 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3944 /* Always do at least ordered writes for quotas */
3945 err = ext4_jbd2_file_inode(handle, inode);
3946 mark_buffer_dirty(bh);
3957 if (len == towrite) {
3958 mutex_unlock(&inode->i_mutex);
3961 if (inode->i_size < off+len-towrite) {
3962 i_size_write(inode, off+len-towrite);
3963 EXT4_I(inode)->i_disksize = inode->i_size;
3965 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3966 ext4_mark_inode_dirty(handle, inode);
3967 mutex_unlock(&inode->i_mutex);
3968 return len - towrite;
3973 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3974 const char *dev_name, void *data, struct vfsmount *mnt)
3976 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3979 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3980 static struct file_system_type ext2_fs_type = {
3981 .owner = THIS_MODULE,
3983 .get_sb = ext4_get_sb,
3984 .kill_sb = kill_block_super,
3985 .fs_flags = FS_REQUIRES_DEV,
3988 static inline void register_as_ext2(void)
3990 int err = register_filesystem(&ext2_fs_type);
3993 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
3996 static inline void unregister_as_ext2(void)
3998 unregister_filesystem(&ext2_fs_type);
4000 MODULE_ALIAS("ext2");
4002 static inline void register_as_ext2(void) { }
4003 static inline void unregister_as_ext2(void) { }
4006 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4007 static struct file_system_type ext3_fs_type = {
4008 .owner = THIS_MODULE,
4010 .get_sb = ext4_get_sb,
4011 .kill_sb = kill_block_super,
4012 .fs_flags = FS_REQUIRES_DEV,
4015 static inline void register_as_ext3(void)
4017 int err = register_filesystem(&ext3_fs_type);
4020 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4023 static inline void unregister_as_ext3(void)
4025 unregister_filesystem(&ext3_fs_type);
4027 MODULE_ALIAS("ext3");
4029 static inline void register_as_ext3(void) { }
4030 static inline void unregister_as_ext3(void) { }
4033 static struct file_system_type ext4_fs_type = {
4034 .owner = THIS_MODULE,
4036 .get_sb = ext4_get_sb,
4037 .kill_sb = kill_block_super,
4038 .fs_flags = FS_REQUIRES_DEV,
4041 static int __init init_ext4_fs(void)
4045 err = init_ext4_system_zone();
4048 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4051 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4052 err = init_ext4_mballoc();
4056 err = init_ext4_xattr();
4059 err = init_inodecache();
4064 err = register_filesystem(&ext4_fs_type);
4069 unregister_as_ext2();
4070 unregister_as_ext3();
4071 destroy_inodecache();
4075 exit_ext4_mballoc();
4077 remove_proc_entry("fs/ext4", NULL);
4078 kset_unregister(ext4_kset);
4080 exit_ext4_system_zone();
4084 static void __exit exit_ext4_fs(void)
4086 unregister_as_ext2();
4087 unregister_as_ext3();
4088 unregister_filesystem(&ext4_fs_type);
4089 destroy_inodecache();
4091 exit_ext4_mballoc();
4092 remove_proc_entry("fs/ext4", NULL);
4093 kset_unregister(ext4_kset);
4094 exit_ext4_system_zone();
4097 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4098 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4099 MODULE_LICENSE("GPL");
4100 module_init(init_ext4_fs)
4101 module_exit(exit_ext4_fs)