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 ext4_release_system_zone(sb);
608 ext4_ext_release(sb);
609 ext4_xattr_put_super(sb);
610 if (sbi->s_journal) {
611 err = jbd2_journal_destroy(sbi->s_journal);
612 sbi->s_journal = NULL;
614 ext4_abort(sb, __func__,
615 "Couldn't clean up the journal");
617 if (!(sb->s_flags & MS_RDONLY)) {
618 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
619 es->s_state = cpu_to_le16(sbi->s_mount_state);
620 ext4_commit_super(sb, 1);
623 remove_proc_entry(sb->s_id, ext4_proc_root);
625 kobject_del(&sbi->s_kobj);
627 for (i = 0; i < sbi->s_gdb_count; i++)
628 brelse(sbi->s_group_desc[i]);
629 kfree(sbi->s_group_desc);
630 if (is_vmalloc_addr(sbi->s_flex_groups))
631 vfree(sbi->s_flex_groups);
633 kfree(sbi->s_flex_groups);
634 percpu_counter_destroy(&sbi->s_freeblocks_counter);
635 percpu_counter_destroy(&sbi->s_freeinodes_counter);
636 percpu_counter_destroy(&sbi->s_dirs_counter);
637 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
640 for (i = 0; i < MAXQUOTAS; i++)
641 kfree(sbi->s_qf_names[i]);
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi->s_orphan))
649 dump_orphan_list(sb, sbi);
650 J_ASSERT(list_empty(&sbi->s_orphan));
652 invalidate_bdev(sb->s_bdev);
653 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
659 sync_blockdev(sbi->journal_bdev);
660 invalidate_bdev(sbi->journal_bdev);
661 ext4_blkdev_remove(sbi);
663 sb->s_fs_info = NULL;
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
670 kobject_put(&sbi->s_kobj);
671 wait_for_completion(&sbi->s_kobj_unregister);
672 kfree(sbi->s_blockgroup_lock);
676 static struct kmem_cache *ext4_inode_cachep;
679 * Called inside transaction, so use GFP_NOFS
681 static struct inode *ext4_alloc_inode(struct super_block *sb)
683 struct ext4_inode_info *ei;
685 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
689 ei->vfs_inode.i_version = 1;
690 ei->vfs_inode.i_data.writeback_index = 0;
691 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
692 INIT_LIST_HEAD(&ei->i_prealloc_list);
693 spin_lock_init(&ei->i_prealloc_lock);
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
699 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
700 ei->i_reserved_data_blocks = 0;
701 ei->i_reserved_meta_blocks = 0;
702 ei->i_allocated_meta_blocks = 0;
703 ei->i_delalloc_reserved_flag = 0;
704 spin_lock_init(&(ei->i_block_reservation_lock));
705 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
706 ei->cur_aio_dio = NULL;
708 return &ei->vfs_inode;
711 static void ext4_destroy_inode(struct inode *inode)
713 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
714 ext4_msg(inode->i_sb, KERN_ERR,
715 "Inode %lu (%p): orphan list check failed!",
716 inode->i_ino, EXT4_I(inode));
717 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
718 EXT4_I(inode), sizeof(struct ext4_inode_info),
722 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
725 static void init_once(void *foo)
727 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
729 INIT_LIST_HEAD(&ei->i_orphan);
730 #ifdef CONFIG_EXT4_FS_XATTR
731 init_rwsem(&ei->xattr_sem);
733 init_rwsem(&ei->i_data_sem);
734 inode_init_once(&ei->vfs_inode);
737 static int init_inodecache(void)
739 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info),
741 0, (SLAB_RECLAIM_ACCOUNT|
744 if (ext4_inode_cachep == NULL)
749 static void destroy_inodecache(void)
751 kmem_cache_destroy(ext4_inode_cachep);
754 static void ext4_clear_inode(struct inode *inode)
756 ext4_discard_preallocations(inode);
757 if (EXT4_JOURNAL(inode))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
759 &EXT4_I(inode)->jinode);
762 static inline void ext4_show_quota_options(struct seq_file *seq,
763 struct super_block *sb)
765 #if defined(CONFIG_QUOTA)
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
768 if (sbi->s_jquota_fmt)
769 seq_printf(seq, ",jqfmt=%s",
770 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
772 if (sbi->s_qf_names[USRQUOTA])
773 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
775 if (sbi->s_qf_names[GRPQUOTA])
776 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
778 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
779 seq_puts(seq, ",usrquota");
781 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
782 seq_puts(seq, ",grpquota");
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
791 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
794 unsigned long def_mount_opts;
795 struct super_block *sb = vfs->mnt_sb;
796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
799 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
800 def_errors = le16_to_cpu(es->s_errors);
802 if (sbi->s_sb_block != 1)
803 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
804 if (test_opt(sb, MINIX_DF))
805 seq_puts(seq, ",minixdf");
806 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
807 seq_puts(seq, ",grpid");
808 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",nogrpid");
810 if (sbi->s_resuid != EXT4_DEF_RESUID ||
811 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
812 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
814 if (sbi->s_resgid != EXT4_DEF_RESGID ||
815 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
816 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
818 if (test_opt(sb, ERRORS_RO)) {
819 if (def_errors == EXT4_ERRORS_PANIC ||
820 def_errors == EXT4_ERRORS_CONTINUE) {
821 seq_puts(seq, ",errors=remount-ro");
824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
825 seq_puts(seq, ",errors=continue");
826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
827 seq_puts(seq, ",errors=panic");
828 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
829 seq_puts(seq, ",nouid32");
830 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
831 seq_puts(seq, ",debug");
832 if (test_opt(sb, OLDALLOC))
833 seq_puts(seq, ",oldalloc");
834 #ifdef CONFIG_EXT4_FS_XATTR
835 if (test_opt(sb, XATTR_USER) &&
836 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
837 seq_puts(seq, ",user_xattr");
838 if (!test_opt(sb, XATTR_USER) &&
839 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
840 seq_puts(seq, ",nouser_xattr");
843 #ifdef CONFIG_EXT4_FS_POSIX_ACL
844 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
845 seq_puts(seq, ",acl");
846 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",noacl");
849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
850 seq_printf(seq, ",commit=%u",
851 (unsigned) (sbi->s_commit_interval / HZ));
853 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
854 seq_printf(seq, ",min_batch_time=%u",
855 (unsigned) sbi->s_min_batch_time);
857 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
858 seq_printf(seq, ",max_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
867 seq_puts(seq, ",barrier=");
868 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
870 seq_puts(seq, ",journal_async_commit");
871 if (test_opt(sb, NOBH))
872 seq_puts(seq, ",nobh");
873 if (test_opt(sb, I_VERSION))
874 seq_puts(seq, ",i_version");
875 if (!test_opt(sb, DELALLOC))
876 seq_puts(seq, ",nodelalloc");
880 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
886 seq_puts(seq, ",data=journal");
887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
888 seq_puts(seq, ",data=ordered");
889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
890 seq_puts(seq, ",data=writeback");
892 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
893 seq_printf(seq, ",inode_readahead_blks=%u",
894 sbi->s_inode_readahead_blks);
896 if (test_opt(sb, DATA_ERR_ABORT))
897 seq_puts(seq, ",data_err=abort");
899 if (test_opt(sb, NO_AUTO_DA_ALLOC))
900 seq_puts(seq, ",noauto_da_alloc");
902 if (test_opt(sb, DISCARD))
903 seq_puts(seq, ",discard");
905 ext4_show_quota_options(seq, sb);
910 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
911 u64 ino, u32 generation)
915 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
916 return ERR_PTR(-ESTALE);
917 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
918 return ERR_PTR(-ESTALE);
920 /* iget isn't really right if the inode is currently unallocated!!
922 * ext4_read_inode will return a bad_inode if the inode had been
923 * deleted, so we should be safe.
925 * Currently we don't know the generation for parent directory, so
926 * a generation of 0 means "accept any"
928 inode = ext4_iget(sb, ino);
930 return ERR_CAST(inode);
931 if (generation && inode->i_generation != generation) {
933 return ERR_PTR(-ESTALE);
939 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
940 int fh_len, int fh_type)
942 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
946 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
947 int fh_len, int fh_type)
949 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
954 * Try to release metadata pages (indirect blocks, directories) which are
955 * mapped via the block device. Since these pages could have journal heads
956 * which would prevent try_to_free_buffers() from freeing them, we must use
957 * jbd2 layer's try_to_free_buffers() function to release them.
959 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
962 journal_t *journal = EXT4_SB(sb)->s_journal;
964 WARN_ON(PageChecked(page));
965 if (!page_has_buffers(page))
968 return jbd2_journal_try_to_free_buffers(journal, page,
970 return try_to_free_buffers(page);
974 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
975 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
977 static int ext4_write_dquot(struct dquot *dquot);
978 static int ext4_acquire_dquot(struct dquot *dquot);
979 static int ext4_release_dquot(struct dquot *dquot);
980 static int ext4_mark_dquot_dirty(struct dquot *dquot);
981 static int ext4_write_info(struct super_block *sb, int type);
982 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
983 char *path, int remount);
984 static int ext4_quota_on_mount(struct super_block *sb, int type);
985 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
986 size_t len, loff_t off);
987 static ssize_t ext4_quota_write(struct super_block *sb, int type,
988 const char *data, size_t len, loff_t off);
990 static const struct dquot_operations ext4_quota_operations = {
991 .initialize = dquot_initialize,
993 .alloc_space = dquot_alloc_space,
994 .reserve_space = dquot_reserve_space,
995 .claim_space = dquot_claim_space,
996 .release_rsv = dquot_release_reserved_space,
997 .get_reserved_space = ext4_get_reserved_space,
998 .alloc_inode = dquot_alloc_inode,
999 .free_space = dquot_free_space,
1000 .free_inode = dquot_free_inode,
1001 .transfer = dquot_transfer,
1002 .write_dquot = ext4_write_dquot,
1003 .acquire_dquot = ext4_acquire_dquot,
1004 .release_dquot = ext4_release_dquot,
1005 .mark_dirty = ext4_mark_dquot_dirty,
1006 .write_info = ext4_write_info,
1007 .alloc_dquot = dquot_alloc,
1008 .destroy_dquot = dquot_destroy,
1011 static const struct quotactl_ops ext4_qctl_operations = {
1012 .quota_on = ext4_quota_on,
1013 .quota_off = vfs_quota_off,
1014 .quota_sync = vfs_quota_sync,
1015 .get_info = vfs_get_dqinfo,
1016 .set_info = vfs_set_dqinfo,
1017 .get_dqblk = vfs_get_dqblk,
1018 .set_dqblk = vfs_set_dqblk
1022 static const struct super_operations ext4_sops = {
1023 .alloc_inode = ext4_alloc_inode,
1024 .destroy_inode = ext4_destroy_inode,
1025 .write_inode = ext4_write_inode,
1026 .dirty_inode = ext4_dirty_inode,
1027 .delete_inode = ext4_delete_inode,
1028 .put_super = ext4_put_super,
1029 .sync_fs = ext4_sync_fs,
1030 .freeze_fs = ext4_freeze,
1031 .unfreeze_fs = ext4_unfreeze,
1032 .statfs = ext4_statfs,
1033 .remount_fs = ext4_remount,
1034 .clear_inode = ext4_clear_inode,
1035 .show_options = ext4_show_options,
1037 .quota_read = ext4_quota_read,
1038 .quota_write = ext4_quota_write,
1040 .bdev_try_to_free_page = bdev_try_to_free_page,
1043 static const struct super_operations ext4_nojournal_sops = {
1044 .alloc_inode = ext4_alloc_inode,
1045 .destroy_inode = ext4_destroy_inode,
1046 .write_inode = ext4_write_inode,
1047 .dirty_inode = ext4_dirty_inode,
1048 .delete_inode = ext4_delete_inode,
1049 .write_super = ext4_write_super,
1050 .put_super = ext4_put_super,
1051 .statfs = ext4_statfs,
1052 .remount_fs = ext4_remount,
1053 .clear_inode = ext4_clear_inode,
1054 .show_options = ext4_show_options,
1056 .quota_read = ext4_quota_read,
1057 .quota_write = ext4_quota_write,
1059 .bdev_try_to_free_page = bdev_try_to_free_page,
1062 static const struct export_operations ext4_export_ops = {
1063 .fh_to_dentry = ext4_fh_to_dentry,
1064 .fh_to_parent = ext4_fh_to_parent,
1065 .get_parent = ext4_get_parent,
1069 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1070 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1071 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1072 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1073 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1074 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1075 Opt_journal_update, Opt_journal_dev,
1076 Opt_journal_checksum, Opt_journal_async_commit,
1077 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1078 Opt_data_err_abort, Opt_data_err_ignore,
1079 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1080 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1081 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1082 Opt_usrquota, Opt_grpquota, Opt_i_version,
1083 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1084 Opt_block_validity, Opt_noblock_validity,
1085 Opt_inode_readahead_blks, Opt_journal_ioprio,
1086 Opt_discard, Opt_nodiscard,
1089 static const match_table_t tokens = {
1090 {Opt_bsd_df, "bsddf"},
1091 {Opt_minix_df, "minixdf"},
1092 {Opt_grpid, "grpid"},
1093 {Opt_grpid, "bsdgroups"},
1094 {Opt_nogrpid, "nogrpid"},
1095 {Opt_nogrpid, "sysvgroups"},
1096 {Opt_resgid, "resgid=%u"},
1097 {Opt_resuid, "resuid=%u"},
1099 {Opt_err_cont, "errors=continue"},
1100 {Opt_err_panic, "errors=panic"},
1101 {Opt_err_ro, "errors=remount-ro"},
1102 {Opt_nouid32, "nouid32"},
1103 {Opt_debug, "debug"},
1104 {Opt_oldalloc, "oldalloc"},
1105 {Opt_orlov, "orlov"},
1106 {Opt_user_xattr, "user_xattr"},
1107 {Opt_nouser_xattr, "nouser_xattr"},
1109 {Opt_noacl, "noacl"},
1110 {Opt_noload, "noload"},
1113 {Opt_commit, "commit=%u"},
1114 {Opt_min_batch_time, "min_batch_time=%u"},
1115 {Opt_max_batch_time, "max_batch_time=%u"},
1116 {Opt_journal_update, "journal=update"},
1117 {Opt_journal_dev, "journal_dev=%u"},
1118 {Opt_journal_checksum, "journal_checksum"},
1119 {Opt_journal_async_commit, "journal_async_commit"},
1120 {Opt_abort, "abort"},
1121 {Opt_data_journal, "data=journal"},
1122 {Opt_data_ordered, "data=ordered"},
1123 {Opt_data_writeback, "data=writeback"},
1124 {Opt_data_err_abort, "data_err=abort"},
1125 {Opt_data_err_ignore, "data_err=ignore"},
1126 {Opt_offusrjquota, "usrjquota="},
1127 {Opt_usrjquota, "usrjquota=%s"},
1128 {Opt_offgrpjquota, "grpjquota="},
1129 {Opt_grpjquota, "grpjquota=%s"},
1130 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1131 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1132 {Opt_grpquota, "grpquota"},
1133 {Opt_noquota, "noquota"},
1134 {Opt_quota, "quota"},
1135 {Opt_usrquota, "usrquota"},
1136 {Opt_barrier, "barrier=%u"},
1137 {Opt_barrier, "barrier"},
1138 {Opt_nobarrier, "nobarrier"},
1139 {Opt_i_version, "i_version"},
1140 {Opt_stripe, "stripe=%u"},
1141 {Opt_resize, "resize"},
1142 {Opt_delalloc, "delalloc"},
1143 {Opt_nodelalloc, "nodelalloc"},
1144 {Opt_block_validity, "block_validity"},
1145 {Opt_noblock_validity, "noblock_validity"},
1146 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1147 {Opt_journal_ioprio, "journal_ioprio=%u"},
1148 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1149 {Opt_auto_da_alloc, "auto_da_alloc"},
1150 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1151 {Opt_discard, "discard"},
1152 {Opt_nodiscard, "nodiscard"},
1156 static ext4_fsblk_t get_sb_block(void **data)
1158 ext4_fsblk_t sb_block;
1159 char *options = (char *) *data;
1161 if (!options || strncmp(options, "sb=", 3) != 0)
1162 return 1; /* Default location */
1165 /* TODO: use simple_strtoll with >32bit ext4 */
1166 sb_block = simple_strtoul(options, &options, 0);
1167 if (*options && *options != ',') {
1168 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1172 if (*options == ',')
1174 *data = (void *) options;
1179 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1181 static int parse_options(char *options, struct super_block *sb,
1182 unsigned long *journal_devnum,
1183 unsigned int *journal_ioprio,
1184 ext4_fsblk_t *n_blocks_count, int is_remount)
1186 struct ext4_sb_info *sbi = EXT4_SB(sb);
1188 substring_t args[MAX_OPT_ARGS];
1199 while ((p = strsep(&options, ",")) != NULL) {
1204 token = match_token(p, tokens, args);
1207 clear_opt(sbi->s_mount_opt, MINIX_DF);
1210 set_opt(sbi->s_mount_opt, MINIX_DF);
1213 set_opt(sbi->s_mount_opt, GRPID);
1216 clear_opt(sbi->s_mount_opt, GRPID);
1219 if (match_int(&args[0], &option))
1221 sbi->s_resuid = option;
1224 if (match_int(&args[0], &option))
1226 sbi->s_resgid = option;
1229 /* handled by get_sb_block() instead of here */
1230 /* *sb_block = match_int(&args[0]); */
1233 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1234 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1235 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1238 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1239 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1240 set_opt(sbi->s_mount_opt, ERRORS_RO);
1243 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1244 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1245 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1248 set_opt(sbi->s_mount_opt, NO_UID32);
1251 set_opt(sbi->s_mount_opt, DEBUG);
1254 set_opt(sbi->s_mount_opt, OLDALLOC);
1257 clear_opt(sbi->s_mount_opt, OLDALLOC);
1259 #ifdef CONFIG_EXT4_FS_XATTR
1260 case Opt_user_xattr:
1261 set_opt(sbi->s_mount_opt, XATTR_USER);
1263 case Opt_nouser_xattr:
1264 clear_opt(sbi->s_mount_opt, XATTR_USER);
1267 case Opt_user_xattr:
1268 case Opt_nouser_xattr:
1269 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1272 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1274 set_opt(sbi->s_mount_opt, POSIX_ACL);
1277 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1282 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1285 case Opt_journal_update:
1287 /* Eventually we will want to be able to create
1288 a journal file here. For now, only allow the
1289 user to specify an existing inode to be the
1292 ext4_msg(sb, KERN_ERR,
1293 "Cannot specify journal on remount");
1296 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1298 case Opt_journal_dev:
1300 ext4_msg(sb, KERN_ERR,
1301 "Cannot specify journal on remount");
1304 if (match_int(&args[0], &option))
1306 *journal_devnum = option;
1308 case Opt_journal_checksum:
1309 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1311 case Opt_journal_async_commit:
1312 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1313 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1316 set_opt(sbi->s_mount_opt, NOLOAD);
1319 if (match_int(&args[0], &option))
1324 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1325 sbi->s_commit_interval = HZ * option;
1327 case Opt_max_batch_time:
1328 if (match_int(&args[0], &option))
1333 option = EXT4_DEF_MAX_BATCH_TIME;
1334 sbi->s_max_batch_time = option;
1336 case Opt_min_batch_time:
1337 if (match_int(&args[0], &option))
1341 sbi->s_min_batch_time = option;
1343 case Opt_data_journal:
1344 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1346 case Opt_data_ordered:
1347 data_opt = EXT4_MOUNT_ORDERED_DATA;
1349 case Opt_data_writeback:
1350 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1353 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1355 ext4_msg(sb, KERN_ERR,
1356 "Cannot change data mode on remount");
1360 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1361 sbi->s_mount_opt |= data_opt;
1364 case Opt_data_err_abort:
1365 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1367 case Opt_data_err_ignore:
1368 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1377 if (sb_any_quota_loaded(sb) &&
1378 !sbi->s_qf_names[qtype]) {
1379 ext4_msg(sb, KERN_ERR,
1380 "Cannot change journaled "
1381 "quota options when quota turned on");
1384 qname = match_strdup(&args[0]);
1386 ext4_msg(sb, KERN_ERR,
1387 "Not enough memory for "
1388 "storing quotafile name");
1391 if (sbi->s_qf_names[qtype] &&
1392 strcmp(sbi->s_qf_names[qtype], qname)) {
1393 ext4_msg(sb, KERN_ERR,
1394 "%s quota file already "
1395 "specified", QTYPE2NAME(qtype));
1399 sbi->s_qf_names[qtype] = qname;
1400 if (strchr(sbi->s_qf_names[qtype], '/')) {
1401 ext4_msg(sb, KERN_ERR,
1402 "quotafile must be on "
1404 kfree(sbi->s_qf_names[qtype]);
1405 sbi->s_qf_names[qtype] = NULL;
1408 set_opt(sbi->s_mount_opt, QUOTA);
1410 case Opt_offusrjquota:
1413 case Opt_offgrpjquota:
1416 if (sb_any_quota_loaded(sb) &&
1417 sbi->s_qf_names[qtype]) {
1418 ext4_msg(sb, KERN_ERR, "Cannot change "
1419 "journaled quota options when "
1424 * The space will be released later when all options
1425 * are confirmed to be correct
1427 sbi->s_qf_names[qtype] = NULL;
1429 case Opt_jqfmt_vfsold:
1430 qfmt = QFMT_VFS_OLD;
1432 case Opt_jqfmt_vfsv0:
1435 if (sb_any_quota_loaded(sb) &&
1436 sbi->s_jquota_fmt != qfmt) {
1437 ext4_msg(sb, KERN_ERR, "Cannot change "
1438 "journaled quota options when "
1442 sbi->s_jquota_fmt = qfmt;
1446 set_opt(sbi->s_mount_opt, QUOTA);
1447 set_opt(sbi->s_mount_opt, USRQUOTA);
1450 set_opt(sbi->s_mount_opt, QUOTA);
1451 set_opt(sbi->s_mount_opt, GRPQUOTA);
1454 if (sb_any_quota_loaded(sb)) {
1455 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1456 "options when quota turned on");
1459 clear_opt(sbi->s_mount_opt, QUOTA);
1460 clear_opt(sbi->s_mount_opt, USRQUOTA);
1461 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1467 ext4_msg(sb, KERN_ERR,
1468 "quota options not supported");
1472 case Opt_offusrjquota:
1473 case Opt_offgrpjquota:
1474 case Opt_jqfmt_vfsold:
1475 case Opt_jqfmt_vfsv0:
1476 ext4_msg(sb, KERN_ERR,
1477 "journaled quota options not supported");
1483 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1486 clear_opt(sbi->s_mount_opt, BARRIER);
1489 if (match_int(&args[0], &option)) {
1490 set_opt(sbi->s_mount_opt, BARRIER);
1494 set_opt(sbi->s_mount_opt, BARRIER);
1496 clear_opt(sbi->s_mount_opt, BARRIER);
1502 ext4_msg(sb, KERN_ERR,
1503 "resize option only available "
1507 if (match_int(&args[0], &option) != 0)
1509 *n_blocks_count = option;
1512 set_opt(sbi->s_mount_opt, NOBH);
1515 clear_opt(sbi->s_mount_opt, NOBH);
1518 set_opt(sbi->s_mount_opt, I_VERSION);
1519 sb->s_flags |= MS_I_VERSION;
1521 case Opt_nodelalloc:
1522 clear_opt(sbi->s_mount_opt, DELALLOC);
1525 if (match_int(&args[0], &option))
1529 sbi->s_stripe = option;
1532 set_opt(sbi->s_mount_opt, DELALLOC);
1534 case Opt_block_validity:
1535 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1537 case Opt_noblock_validity:
1538 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1540 case Opt_inode_readahead_blks:
1541 if (match_int(&args[0], &option))
1543 if (option < 0 || option > (1 << 30))
1545 if (!is_power_of_2(option)) {
1546 ext4_msg(sb, KERN_ERR,
1547 "EXT4-fs: inode_readahead_blks"
1548 " must be a power of 2");
1551 sbi->s_inode_readahead_blks = option;
1553 case Opt_journal_ioprio:
1554 if (match_int(&args[0], &option))
1556 if (option < 0 || option > 7)
1558 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1561 case Opt_noauto_da_alloc:
1562 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1564 case Opt_auto_da_alloc:
1565 if (match_int(&args[0], &option)) {
1566 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1570 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1572 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1575 set_opt(sbi->s_mount_opt, DISCARD);
1578 clear_opt(sbi->s_mount_opt, DISCARD);
1581 ext4_msg(sb, KERN_ERR,
1582 "Unrecognized mount option \"%s\" "
1583 "or missing value", p);
1588 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1589 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1590 sbi->s_qf_names[USRQUOTA])
1591 clear_opt(sbi->s_mount_opt, USRQUOTA);
1593 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1594 sbi->s_qf_names[GRPQUOTA])
1595 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1597 if ((sbi->s_qf_names[USRQUOTA] &&
1598 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1599 (sbi->s_qf_names[GRPQUOTA] &&
1600 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1601 ext4_msg(sb, KERN_ERR, "old and new quota "
1606 if (!sbi->s_jquota_fmt) {
1607 ext4_msg(sb, KERN_ERR, "journaled quota format "
1612 if (sbi->s_jquota_fmt) {
1613 ext4_msg(sb, KERN_ERR, "journaled quota format "
1614 "specified with no journaling "
1623 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1626 struct ext4_sb_info *sbi = EXT4_SB(sb);
1629 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1630 ext4_msg(sb, KERN_ERR, "revision level too high, "
1631 "forcing read-only mode");
1636 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1637 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1638 "running e2fsck is recommended");
1639 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1640 ext4_msg(sb, KERN_WARNING,
1641 "warning: mounting fs with errors, "
1642 "running e2fsck is recommended");
1643 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1644 le16_to_cpu(es->s_mnt_count) >=
1645 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1646 ext4_msg(sb, KERN_WARNING,
1647 "warning: maximal mount count reached, "
1648 "running e2fsck is recommended");
1649 else if (le32_to_cpu(es->s_checkinterval) &&
1650 (le32_to_cpu(es->s_lastcheck) +
1651 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1652 ext4_msg(sb, KERN_WARNING,
1653 "warning: checktime reached, "
1654 "running e2fsck is recommended");
1655 if (!sbi->s_journal)
1656 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1657 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1658 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1659 le16_add_cpu(&es->s_mnt_count, 1);
1660 es->s_mtime = cpu_to_le32(get_seconds());
1661 ext4_update_dynamic_rev(sb);
1663 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1665 ext4_commit_super(sb, 1);
1666 if (test_opt(sb, DEBUG))
1667 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1668 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1670 sbi->s_groups_count,
1671 EXT4_BLOCKS_PER_GROUP(sb),
1672 EXT4_INODES_PER_GROUP(sb),
1678 static int ext4_fill_flex_info(struct super_block *sb)
1680 struct ext4_sb_info *sbi = EXT4_SB(sb);
1681 struct ext4_group_desc *gdp = NULL;
1682 ext4_group_t flex_group_count;
1683 ext4_group_t flex_group;
1684 int groups_per_flex = 0;
1688 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1689 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1691 if (groups_per_flex < 2) {
1692 sbi->s_log_groups_per_flex = 0;
1696 /* We allocate both existing and potentially added groups */
1697 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1698 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1699 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1700 size = flex_group_count * sizeof(struct flex_groups);
1701 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1702 if (sbi->s_flex_groups == NULL) {
1703 sbi->s_flex_groups = vmalloc(size);
1704 if (sbi->s_flex_groups)
1705 memset(sbi->s_flex_groups, 0, size);
1707 if (sbi->s_flex_groups == NULL) {
1708 ext4_msg(sb, KERN_ERR, "not enough memory for "
1709 "%u flex groups", flex_group_count);
1713 for (i = 0; i < sbi->s_groups_count; i++) {
1714 gdp = ext4_get_group_desc(sb, i, NULL);
1716 flex_group = ext4_flex_group(sbi, i);
1717 atomic_add(ext4_free_inodes_count(sb, gdp),
1718 &sbi->s_flex_groups[flex_group].free_inodes);
1719 atomic_add(ext4_free_blks_count(sb, gdp),
1720 &sbi->s_flex_groups[flex_group].free_blocks);
1721 atomic_add(ext4_used_dirs_count(sb, gdp),
1722 &sbi->s_flex_groups[flex_group].used_dirs);
1730 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1731 struct ext4_group_desc *gdp)
1735 if (sbi->s_es->s_feature_ro_compat &
1736 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1737 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1738 __le32 le_group = cpu_to_le32(block_group);
1740 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1741 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1742 crc = crc16(crc, (__u8 *)gdp, offset);
1743 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1744 /* for checksum of struct ext4_group_desc do the rest...*/
1745 if ((sbi->s_es->s_feature_incompat &
1746 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1747 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1748 crc = crc16(crc, (__u8 *)gdp + offset,
1749 le16_to_cpu(sbi->s_es->s_desc_size) -
1753 return cpu_to_le16(crc);
1756 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1757 struct ext4_group_desc *gdp)
1759 if ((sbi->s_es->s_feature_ro_compat &
1760 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1761 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1767 /* Called at mount-time, super-block is locked */
1768 static int ext4_check_descriptors(struct super_block *sb)
1770 struct ext4_sb_info *sbi = EXT4_SB(sb);
1771 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1772 ext4_fsblk_t last_block;
1773 ext4_fsblk_t block_bitmap;
1774 ext4_fsblk_t inode_bitmap;
1775 ext4_fsblk_t inode_table;
1776 int flexbg_flag = 0;
1779 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1782 ext4_debug("Checking group descriptors");
1784 for (i = 0; i < sbi->s_groups_count; i++) {
1785 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1787 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1788 last_block = ext4_blocks_count(sbi->s_es) - 1;
1790 last_block = first_block +
1791 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1793 block_bitmap = ext4_block_bitmap(sb, gdp);
1794 if (block_bitmap < first_block || block_bitmap > last_block) {
1795 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1796 "Block bitmap for group %u not in group "
1797 "(block %llu)!", i, block_bitmap);
1800 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1801 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1802 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1803 "Inode bitmap for group %u not in group "
1804 "(block %llu)!", i, inode_bitmap);
1807 inode_table = ext4_inode_table(sb, gdp);
1808 if (inode_table < first_block ||
1809 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1810 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1811 "Inode table for group %u not in group "
1812 "(block %llu)!", i, inode_table);
1815 ext4_lock_group(sb, i);
1816 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1817 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1818 "Checksum for group %u failed (%u!=%u)",
1819 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1820 gdp)), le16_to_cpu(gdp->bg_checksum));
1821 if (!(sb->s_flags & MS_RDONLY)) {
1822 ext4_unlock_group(sb, i);
1826 ext4_unlock_group(sb, i);
1828 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1831 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1832 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1836 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1837 * the superblock) which were deleted from all directories, but held open by
1838 * a process at the time of a crash. We walk the list and try to delete these
1839 * inodes at recovery time (only with a read-write filesystem).
1841 * In order to keep the orphan inode chain consistent during traversal (in
1842 * case of crash during recovery), we link each inode into the superblock
1843 * orphan list_head and handle it the same way as an inode deletion during
1844 * normal operation (which journals the operations for us).
1846 * We only do an iget() and an iput() on each inode, which is very safe if we
1847 * accidentally point at an in-use or already deleted inode. The worst that
1848 * can happen in this case is that we get a "bit already cleared" message from
1849 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1850 * e2fsck was run on this filesystem, and it must have already done the orphan
1851 * inode cleanup for us, so we can safely abort without any further action.
1853 static void ext4_orphan_cleanup(struct super_block *sb,
1854 struct ext4_super_block *es)
1856 unsigned int s_flags = sb->s_flags;
1857 int nr_orphans = 0, nr_truncates = 0;
1861 if (!es->s_last_orphan) {
1862 jbd_debug(4, "no orphan inodes to clean up\n");
1866 if (bdev_read_only(sb->s_bdev)) {
1867 ext4_msg(sb, KERN_ERR, "write access "
1868 "unavailable, skipping orphan cleanup");
1872 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1873 if (es->s_last_orphan)
1874 jbd_debug(1, "Errors on filesystem, "
1875 "clearing orphan list.\n");
1876 es->s_last_orphan = 0;
1877 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1881 if (s_flags & MS_RDONLY) {
1882 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1883 sb->s_flags &= ~MS_RDONLY;
1886 /* Needed for iput() to work correctly and not trash data */
1887 sb->s_flags |= MS_ACTIVE;
1888 /* Turn on quotas so that they are updated correctly */
1889 for (i = 0; i < MAXQUOTAS; i++) {
1890 if (EXT4_SB(sb)->s_qf_names[i]) {
1891 int ret = ext4_quota_on_mount(sb, i);
1893 ext4_msg(sb, KERN_ERR,
1894 "Cannot turn on journaled "
1895 "quota: error %d", ret);
1900 while (es->s_last_orphan) {
1901 struct inode *inode;
1903 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1904 if (IS_ERR(inode)) {
1905 es->s_last_orphan = 0;
1909 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1911 if (inode->i_nlink) {
1912 ext4_msg(sb, KERN_DEBUG,
1913 "%s: truncating inode %lu to %lld bytes",
1914 __func__, inode->i_ino, inode->i_size);
1915 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1916 inode->i_ino, inode->i_size);
1917 ext4_truncate(inode);
1920 ext4_msg(sb, KERN_DEBUG,
1921 "%s: deleting unreferenced inode %lu",
1922 __func__, inode->i_ino);
1923 jbd_debug(2, "deleting unreferenced inode %lu\n",
1927 iput(inode); /* The delete magic happens here! */
1930 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1933 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1934 PLURAL(nr_orphans));
1936 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1937 PLURAL(nr_truncates));
1939 /* Turn quotas off */
1940 for (i = 0; i < MAXQUOTAS; i++) {
1941 if (sb_dqopt(sb)->files[i])
1942 vfs_quota_off(sb, i, 0);
1945 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1949 * Maximal extent format file size.
1950 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1951 * extent format containers, within a sector_t, and within i_blocks
1952 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1953 * so that won't be a limiting factor.
1955 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1957 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1960 loff_t upper_limit = MAX_LFS_FILESIZE;
1962 /* small i_blocks in vfs inode? */
1963 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1965 * CONFIG_LBDAF is not enabled implies the inode
1966 * i_block represent total blocks in 512 bytes
1967 * 32 == size of vfs inode i_blocks * 8
1969 upper_limit = (1LL << 32) - 1;
1971 /* total blocks in file system block size */
1972 upper_limit >>= (blkbits - 9);
1973 upper_limit <<= blkbits;
1976 /* 32-bit extent-start container, ee_block */
1981 /* Sanity check against vm- & vfs- imposed limits */
1982 if (res > upper_limit)
1989 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1990 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1991 * We need to be 1 filesystem block less than the 2^48 sector limit.
1993 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1995 loff_t res = EXT4_NDIR_BLOCKS;
1998 /* This is calculated to be the largest file size for a dense, block
1999 * mapped file such that the file's total number of 512-byte sectors,
2000 * including data and all indirect blocks, does not exceed (2^48 - 1).
2002 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2003 * number of 512-byte sectors of the file.
2006 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2008 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2009 * the inode i_block field represents total file blocks in
2010 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2012 upper_limit = (1LL << 32) - 1;
2014 /* total blocks in file system block size */
2015 upper_limit >>= (bits - 9);
2019 * We use 48 bit ext4_inode i_blocks
2020 * With EXT4_HUGE_FILE_FL set the i_blocks
2021 * represent total number of blocks in
2022 * file system block size
2024 upper_limit = (1LL << 48) - 1;
2028 /* indirect blocks */
2030 /* double indirect blocks */
2031 meta_blocks += 1 + (1LL << (bits-2));
2032 /* tripple indirect blocks */
2033 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2035 upper_limit -= meta_blocks;
2036 upper_limit <<= bits;
2038 res += 1LL << (bits-2);
2039 res += 1LL << (2*(bits-2));
2040 res += 1LL << (3*(bits-2));
2042 if (res > upper_limit)
2045 if (res > MAX_LFS_FILESIZE)
2046 res = MAX_LFS_FILESIZE;
2051 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2052 ext4_fsblk_t logical_sb_block, int nr)
2054 struct ext4_sb_info *sbi = EXT4_SB(sb);
2055 ext4_group_t bg, first_meta_bg;
2058 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2060 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2062 return logical_sb_block + nr + 1;
2063 bg = sbi->s_desc_per_block * nr;
2064 if (ext4_bg_has_super(sb, bg))
2067 return (has_super + ext4_group_first_block_no(sb, bg));
2071 * ext4_get_stripe_size: Get the stripe size.
2072 * @sbi: In memory super block info
2074 * If we have specified it via mount option, then
2075 * use the mount option value. If the value specified at mount time is
2076 * greater than the blocks per group use the super block value.
2077 * If the super block value is greater than blocks per group return 0.
2078 * Allocator needs it be less than blocks per group.
2081 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2083 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2084 unsigned long stripe_width =
2085 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2087 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2088 return sbi->s_stripe;
2090 if (stripe_width <= sbi->s_blocks_per_group)
2091 return stripe_width;
2093 if (stride <= sbi->s_blocks_per_group)
2102 struct attribute attr;
2103 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2104 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2105 const char *, size_t);
2109 static int parse_strtoul(const char *buf,
2110 unsigned long max, unsigned long *value)
2114 while (*buf && isspace(*buf))
2116 *value = simple_strtoul(buf, &endp, 0);
2117 while (*endp && isspace(*endp))
2119 if (*endp || *value > max)
2125 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2126 struct ext4_sb_info *sbi,
2129 return snprintf(buf, PAGE_SIZE, "%llu\n",
2130 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2133 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2134 struct ext4_sb_info *sbi, char *buf)
2136 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2138 return snprintf(buf, PAGE_SIZE, "%lu\n",
2139 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2140 sbi->s_sectors_written_start) >> 1);
2143 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2144 struct ext4_sb_info *sbi, char *buf)
2146 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2148 return snprintf(buf, PAGE_SIZE, "%llu\n",
2149 sbi->s_kbytes_written +
2150 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2151 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2154 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2155 struct ext4_sb_info *sbi,
2156 const char *buf, size_t count)
2160 if (parse_strtoul(buf, 0x40000000, &t))
2163 if (!is_power_of_2(t))
2166 sbi->s_inode_readahead_blks = t;
2170 static ssize_t sbi_ui_show(struct ext4_attr *a,
2171 struct ext4_sb_info *sbi, char *buf)
2173 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2175 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2178 static ssize_t sbi_ui_store(struct ext4_attr *a,
2179 struct ext4_sb_info *sbi,
2180 const char *buf, size_t count)
2182 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2185 if (parse_strtoul(buf, 0xffffffff, &t))
2191 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2192 static struct ext4_attr ext4_attr_##_name = { \
2193 .attr = {.name = __stringify(_name), .mode = _mode }, \
2196 .offset = offsetof(struct ext4_sb_info, _elname), \
2198 #define EXT4_ATTR(name, mode, show, store) \
2199 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2201 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2202 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2203 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2204 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2205 #define ATTR_LIST(name) &ext4_attr_##name.attr
2207 EXT4_RO_ATTR(delayed_allocation_blocks);
2208 EXT4_RO_ATTR(session_write_kbytes);
2209 EXT4_RO_ATTR(lifetime_write_kbytes);
2210 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2211 inode_readahead_blks_store, s_inode_readahead_blks);
2212 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2213 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2214 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2215 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2216 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2217 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2218 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2219 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2221 static struct attribute *ext4_attrs[] = {
2222 ATTR_LIST(delayed_allocation_blocks),
2223 ATTR_LIST(session_write_kbytes),
2224 ATTR_LIST(lifetime_write_kbytes),
2225 ATTR_LIST(inode_readahead_blks),
2226 ATTR_LIST(inode_goal),
2227 ATTR_LIST(mb_stats),
2228 ATTR_LIST(mb_max_to_scan),
2229 ATTR_LIST(mb_min_to_scan),
2230 ATTR_LIST(mb_order2_req),
2231 ATTR_LIST(mb_stream_req),
2232 ATTR_LIST(mb_group_prealloc),
2233 ATTR_LIST(max_writeback_mb_bump),
2237 static ssize_t ext4_attr_show(struct kobject *kobj,
2238 struct attribute *attr, char *buf)
2240 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2242 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2244 return a->show ? a->show(a, sbi, buf) : 0;
2247 static ssize_t ext4_attr_store(struct kobject *kobj,
2248 struct attribute *attr,
2249 const char *buf, size_t len)
2251 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2253 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2255 return a->store ? a->store(a, sbi, buf, len) : 0;
2258 static void ext4_sb_release(struct kobject *kobj)
2260 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2262 complete(&sbi->s_kobj_unregister);
2266 static struct sysfs_ops ext4_attr_ops = {
2267 .show = ext4_attr_show,
2268 .store = ext4_attr_store,
2271 static struct kobj_type ext4_ktype = {
2272 .default_attrs = ext4_attrs,
2273 .sysfs_ops = &ext4_attr_ops,
2274 .release = ext4_sb_release,
2278 * Check whether this filesystem can be mounted based on
2279 * the features present and the RDONLY/RDWR mount requested.
2280 * Returns 1 if this filesystem can be mounted as requested,
2281 * 0 if it cannot be.
2283 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2285 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2286 ext4_msg(sb, KERN_ERR,
2287 "Couldn't mount because of "
2288 "unsupported optional features (%x)",
2289 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2290 ~EXT4_FEATURE_INCOMPAT_SUPP));
2297 /* Check that feature set is OK for a read-write mount */
2298 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2299 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2300 "unsupported optional features (%x)",
2301 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2302 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2306 * Large file size enabled file system can only be mounted
2307 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2309 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2310 if (sizeof(blkcnt_t) < sizeof(u64)) {
2311 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2312 "cannot be mounted RDWR without "
2320 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2321 __releases(kernel_lock)
2322 __acquires(kernel_lock)
2324 struct buffer_head *bh;
2325 struct ext4_super_block *es = NULL;
2326 struct ext4_sb_info *sbi;
2328 ext4_fsblk_t sb_block = get_sb_block(&data);
2329 ext4_fsblk_t logical_sb_block;
2330 unsigned long offset = 0;
2331 unsigned long journal_devnum = 0;
2332 unsigned long def_mount_opts;
2338 unsigned int db_count;
2340 int needs_recovery, has_huge_files;
2343 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2345 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2349 sbi->s_blockgroup_lock =
2350 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2351 if (!sbi->s_blockgroup_lock) {
2355 sb->s_fs_info = sbi;
2356 sbi->s_mount_opt = 0;
2357 sbi->s_resuid = EXT4_DEF_RESUID;
2358 sbi->s_resgid = EXT4_DEF_RESGID;
2359 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2360 sbi->s_sb_block = sb_block;
2361 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2366 /* Cleanup superblock name */
2367 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2370 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2372 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2377 * The ext4 superblock will not be buffer aligned for other than 1kB
2378 * block sizes. We need to calculate the offset from buffer start.
2380 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2381 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2382 offset = do_div(logical_sb_block, blocksize);
2384 logical_sb_block = sb_block;
2387 if (!(bh = sb_bread(sb, logical_sb_block))) {
2388 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2392 * Note: s_es must be initialized as soon as possible because
2393 * some ext4 macro-instructions depend on its value
2395 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2397 sb->s_magic = le16_to_cpu(es->s_magic);
2398 if (sb->s_magic != EXT4_SUPER_MAGIC)
2400 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2402 /* Set defaults before we parse the mount options */
2403 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2404 if (def_mount_opts & EXT4_DEFM_DEBUG)
2405 set_opt(sbi->s_mount_opt, DEBUG);
2406 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2407 set_opt(sbi->s_mount_opt, GRPID);
2408 if (def_mount_opts & EXT4_DEFM_UID16)
2409 set_opt(sbi->s_mount_opt, NO_UID32);
2410 #ifdef CONFIG_EXT4_FS_XATTR
2411 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2412 set_opt(sbi->s_mount_opt, XATTR_USER);
2414 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2415 if (def_mount_opts & EXT4_DEFM_ACL)
2416 set_opt(sbi->s_mount_opt, POSIX_ACL);
2418 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2419 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2420 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2421 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2422 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2423 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2425 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2426 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2427 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2428 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2430 set_opt(sbi->s_mount_opt, ERRORS_RO);
2432 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2433 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2434 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2435 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2436 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2438 set_opt(sbi->s_mount_opt, BARRIER);
2441 * enable delayed allocation by default
2442 * Use -o nodelalloc to turn it off
2444 set_opt(sbi->s_mount_opt, DELALLOC);
2446 if (!parse_options((char *) data, sb, &journal_devnum,
2447 &journal_ioprio, NULL, 0))
2450 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2451 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2453 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2454 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2455 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2456 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2457 ext4_msg(sb, KERN_WARNING,
2458 "feature flags set on rev 0 fs, "
2459 "running e2fsck is recommended");
2462 * Check feature flags regardless of the revision level, since we
2463 * previously didn't change the revision level when setting the flags,
2464 * so there is a chance incompat flags are set on a rev 0 filesystem.
2466 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2469 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2471 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2472 blocksize > EXT4_MAX_BLOCK_SIZE) {
2473 ext4_msg(sb, KERN_ERR,
2474 "Unsupported filesystem blocksize %d", blocksize);
2478 if (sb->s_blocksize != blocksize) {
2479 /* Validate the filesystem blocksize */
2480 if (!sb_set_blocksize(sb, blocksize)) {
2481 ext4_msg(sb, KERN_ERR, "bad block size %d",
2487 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2488 offset = do_div(logical_sb_block, blocksize);
2489 bh = sb_bread(sb, logical_sb_block);
2491 ext4_msg(sb, KERN_ERR,
2492 "Can't read superblock on 2nd try");
2495 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2497 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2498 ext4_msg(sb, KERN_ERR,
2499 "Magic mismatch, very weird!");
2504 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2505 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2506 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2508 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2510 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2511 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2512 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2514 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2515 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2516 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2517 (!is_power_of_2(sbi->s_inode_size)) ||
2518 (sbi->s_inode_size > blocksize)) {
2519 ext4_msg(sb, KERN_ERR,
2520 "unsupported inode size: %d",
2524 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2525 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2528 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2529 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2530 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2531 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2532 !is_power_of_2(sbi->s_desc_size)) {
2533 ext4_msg(sb, KERN_ERR,
2534 "unsupported descriptor size %lu",
2539 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2541 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2542 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2543 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2546 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2547 if (sbi->s_inodes_per_block == 0)
2549 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2550 sbi->s_inodes_per_block;
2551 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2553 sbi->s_mount_state = le16_to_cpu(es->s_state);
2554 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2555 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2557 for (i = 0; i < 4; i++)
2558 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2559 sbi->s_def_hash_version = es->s_def_hash_version;
2560 i = le32_to_cpu(es->s_flags);
2561 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2562 sbi->s_hash_unsigned = 3;
2563 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2564 #ifdef __CHAR_UNSIGNED__
2565 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2566 sbi->s_hash_unsigned = 3;
2568 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2573 if (sbi->s_blocks_per_group > blocksize * 8) {
2574 ext4_msg(sb, KERN_ERR,
2575 "#blocks per group too big: %lu",
2576 sbi->s_blocks_per_group);
2579 if (sbi->s_inodes_per_group > blocksize * 8) {
2580 ext4_msg(sb, KERN_ERR,
2581 "#inodes per group too big: %lu",
2582 sbi->s_inodes_per_group);
2587 * Test whether we have more sectors than will fit in sector_t,
2588 * and whether the max offset is addressable by the page cache.
2590 if ((ext4_blocks_count(es) >
2591 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2592 (ext4_blocks_count(es) >
2593 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2594 ext4_msg(sb, KERN_ERR, "filesystem"
2595 " too large to mount safely on this system");
2596 if (sizeof(sector_t) < 8)
2597 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2602 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2605 /* check blocks count against device size */
2606 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2607 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2608 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2609 "exceeds size of device (%llu blocks)",
2610 ext4_blocks_count(es), blocks_count);
2615 * It makes no sense for the first data block to be beyond the end
2616 * of the filesystem.
2618 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2619 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2620 "block %u is beyond end of filesystem (%llu)",
2621 le32_to_cpu(es->s_first_data_block),
2622 ext4_blocks_count(es));
2625 blocks_count = (ext4_blocks_count(es) -
2626 le32_to_cpu(es->s_first_data_block) +
2627 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2628 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2629 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2630 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2631 "(block count %llu, first data block %u, "
2632 "blocks per group %lu)", sbi->s_groups_count,
2633 ext4_blocks_count(es),
2634 le32_to_cpu(es->s_first_data_block),
2635 EXT4_BLOCKS_PER_GROUP(sb));
2638 sbi->s_groups_count = blocks_count;
2639 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2640 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2641 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2642 EXT4_DESC_PER_BLOCK(sb);
2643 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2645 if (sbi->s_group_desc == NULL) {
2646 ext4_msg(sb, KERN_ERR, "not enough memory");
2650 #ifdef CONFIG_PROC_FS
2652 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2655 bgl_lock_init(sbi->s_blockgroup_lock);
2657 for (i = 0; i < db_count; i++) {
2658 block = descriptor_loc(sb, logical_sb_block, i);
2659 sbi->s_group_desc[i] = sb_bread(sb, block);
2660 if (!sbi->s_group_desc[i]) {
2661 ext4_msg(sb, KERN_ERR,
2662 "can't read group descriptor %d", i);
2667 if (!ext4_check_descriptors(sb)) {
2668 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2671 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2672 if (!ext4_fill_flex_info(sb)) {
2673 ext4_msg(sb, KERN_ERR,
2674 "unable to initialize "
2675 "flex_bg meta info!");
2679 sbi->s_gdb_count = db_count;
2680 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2681 spin_lock_init(&sbi->s_next_gen_lock);
2683 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2684 ext4_count_free_blocks(sb));
2686 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2687 ext4_count_free_inodes(sb));
2690 err = percpu_counter_init(&sbi->s_dirs_counter,
2691 ext4_count_dirs(sb));
2694 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2697 ext4_msg(sb, KERN_ERR, "insufficient memory");
2701 sbi->s_stripe = ext4_get_stripe_size(sbi);
2702 sbi->s_max_writeback_mb_bump = 128;
2705 * set up enough so that it can read an inode
2707 if (!test_opt(sb, NOLOAD) &&
2708 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2709 sb->s_op = &ext4_sops;
2711 sb->s_op = &ext4_nojournal_sops;
2712 sb->s_export_op = &ext4_export_ops;
2713 sb->s_xattr = ext4_xattr_handlers;
2715 sb->s_qcop = &ext4_qctl_operations;
2716 sb->dq_op = &ext4_quota_operations;
2718 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2719 mutex_init(&sbi->s_orphan_lock);
2720 mutex_init(&sbi->s_resize_lock);
2724 needs_recovery = (es->s_last_orphan != 0 ||
2725 EXT4_HAS_INCOMPAT_FEATURE(sb,
2726 EXT4_FEATURE_INCOMPAT_RECOVER));
2729 * The first inode we look at is the journal inode. Don't try
2730 * root first: it may be modified in the journal!
2732 if (!test_opt(sb, NOLOAD) &&
2733 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2734 if (ext4_load_journal(sb, es, journal_devnum))
2736 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2737 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2738 ext4_msg(sb, KERN_ERR, "required journal recovery "
2739 "suppressed and not mounted read-only");
2742 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2743 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2744 sbi->s_journal = NULL;
2749 if (ext4_blocks_count(es) > 0xffffffffULL &&
2750 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2751 JBD2_FEATURE_INCOMPAT_64BIT)) {
2752 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2756 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2757 jbd2_journal_set_features(sbi->s_journal,
2758 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2759 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2760 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2761 jbd2_journal_set_features(sbi->s_journal,
2762 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2763 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2764 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2766 jbd2_journal_clear_features(sbi->s_journal,
2767 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2768 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2771 /* We have now updated the journal if required, so we can
2772 * validate the data journaling mode. */
2773 switch (test_opt(sb, DATA_FLAGS)) {
2775 /* No mode set, assume a default based on the journal
2776 * capabilities: ORDERED_DATA if the journal can
2777 * cope, else JOURNAL_DATA
2779 if (jbd2_journal_check_available_features
2780 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2781 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2783 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2786 case EXT4_MOUNT_ORDERED_DATA:
2787 case EXT4_MOUNT_WRITEBACK_DATA:
2788 if (!jbd2_journal_check_available_features
2789 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2790 ext4_msg(sb, KERN_ERR, "Journal does not support "
2791 "requested data journaling mode");
2797 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2801 if (test_opt(sb, NOBH)) {
2802 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2803 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2804 "its supported only with writeback mode");
2805 clear_opt(sbi->s_mount_opt, NOBH);
2808 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2809 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2810 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2811 goto failed_mount_wq;
2815 * The jbd2_journal_load will have done any necessary log recovery,
2816 * so we can safely mount the rest of the filesystem now.
2819 root = ext4_iget(sb, EXT4_ROOT_INO);
2821 ext4_msg(sb, KERN_ERR, "get root inode failed");
2822 ret = PTR_ERR(root);
2825 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2827 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2830 sb->s_root = d_alloc_root(root);
2832 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2838 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2840 /* determine the minimum size of new large inodes, if present */
2841 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2842 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2843 EXT4_GOOD_OLD_INODE_SIZE;
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2845 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2846 if (sbi->s_want_extra_isize <
2847 le16_to_cpu(es->s_want_extra_isize))
2848 sbi->s_want_extra_isize =
2849 le16_to_cpu(es->s_want_extra_isize);
2850 if (sbi->s_want_extra_isize <
2851 le16_to_cpu(es->s_min_extra_isize))
2852 sbi->s_want_extra_isize =
2853 le16_to_cpu(es->s_min_extra_isize);
2856 /* Check if enough inode space is available */
2857 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2858 sbi->s_inode_size) {
2859 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2860 EXT4_GOOD_OLD_INODE_SIZE;
2861 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2865 if (test_opt(sb, DELALLOC) &&
2866 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2867 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2868 "requested data journaling mode");
2869 clear_opt(sbi->s_mount_opt, DELALLOC);
2872 err = ext4_setup_system_zone(sb);
2874 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2875 "zone (%d)\n", err);
2880 err = ext4_mb_init(sb, needs_recovery);
2882 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2887 sbi->s_kobj.kset = ext4_kset;
2888 init_completion(&sbi->s_kobj_unregister);
2889 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2892 ext4_mb_release(sb);
2893 ext4_ext_release(sb);
2897 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2898 ext4_orphan_cleanup(sb, es);
2899 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2900 if (needs_recovery) {
2901 ext4_msg(sb, KERN_INFO, "recovery complete");
2902 ext4_mark_recovery_complete(sb, es);
2904 if (EXT4_SB(sb)->s_journal) {
2905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2906 descr = " journalled data mode";
2907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2908 descr = " ordered data mode";
2910 descr = " writeback data mode";
2912 descr = "out journal";
2914 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2921 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2925 ext4_msg(sb, KERN_ERR, "mount failed");
2926 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2928 ext4_release_system_zone(sb);
2929 if (sbi->s_journal) {
2930 jbd2_journal_destroy(sbi->s_journal);
2931 sbi->s_journal = NULL;
2934 if (sbi->s_flex_groups) {
2935 if (is_vmalloc_addr(sbi->s_flex_groups))
2936 vfree(sbi->s_flex_groups);
2938 kfree(sbi->s_flex_groups);
2940 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2941 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2942 percpu_counter_destroy(&sbi->s_dirs_counter);
2943 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2945 for (i = 0; i < db_count; i++)
2946 brelse(sbi->s_group_desc[i]);
2947 kfree(sbi->s_group_desc);
2950 remove_proc_entry(sb->s_id, ext4_proc_root);
2953 for (i = 0; i < MAXQUOTAS; i++)
2954 kfree(sbi->s_qf_names[i]);
2956 ext4_blkdev_remove(sbi);
2959 sb->s_fs_info = NULL;
2960 kfree(sbi->s_blockgroup_lock);
2967 * Setup any per-fs journal parameters now. We'll do this both on
2968 * initial mount, once the journal has been initialised but before we've
2969 * done any recovery; and again on any subsequent remount.
2971 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2973 struct ext4_sb_info *sbi = EXT4_SB(sb);
2975 journal->j_commit_interval = sbi->s_commit_interval;
2976 journal->j_min_batch_time = sbi->s_min_batch_time;
2977 journal->j_max_batch_time = sbi->s_max_batch_time;
2979 spin_lock(&journal->j_state_lock);
2980 if (test_opt(sb, BARRIER))
2981 journal->j_flags |= JBD2_BARRIER;
2983 journal->j_flags &= ~JBD2_BARRIER;
2984 if (test_opt(sb, DATA_ERR_ABORT))
2985 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2987 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2988 spin_unlock(&journal->j_state_lock);
2991 static journal_t *ext4_get_journal(struct super_block *sb,
2992 unsigned int journal_inum)
2994 struct inode *journal_inode;
2997 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2999 /* First, test for the existence of a valid inode on disk. Bad
3000 * things happen if we iget() an unused inode, as the subsequent
3001 * iput() will try to delete it. */
3003 journal_inode = ext4_iget(sb, journal_inum);
3004 if (IS_ERR(journal_inode)) {
3005 ext4_msg(sb, KERN_ERR, "no journal found");
3008 if (!journal_inode->i_nlink) {
3009 make_bad_inode(journal_inode);
3010 iput(journal_inode);
3011 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3015 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3016 journal_inode, journal_inode->i_size);
3017 if (!S_ISREG(journal_inode->i_mode)) {
3018 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3019 iput(journal_inode);
3023 journal = jbd2_journal_init_inode(journal_inode);
3025 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3026 iput(journal_inode);
3029 journal->j_private = sb;
3030 ext4_init_journal_params(sb, journal);
3034 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3037 struct buffer_head *bh;
3041 int hblock, blocksize;
3042 ext4_fsblk_t sb_block;
3043 unsigned long offset;
3044 struct ext4_super_block *es;
3045 struct block_device *bdev;
3047 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3049 bdev = ext4_blkdev_get(j_dev, sb);
3053 if (bd_claim(bdev, sb)) {
3054 ext4_msg(sb, KERN_ERR,
3055 "failed to claim external journal device");
3056 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3060 blocksize = sb->s_blocksize;
3061 hblock = bdev_logical_block_size(bdev);
3062 if (blocksize < hblock) {
3063 ext4_msg(sb, KERN_ERR,
3064 "blocksize too small for journal device");
3068 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3069 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3070 set_blocksize(bdev, blocksize);
3071 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3072 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3073 "external journal");
3077 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3078 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3079 !(le32_to_cpu(es->s_feature_incompat) &
3080 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3081 ext4_msg(sb, KERN_ERR, "external journal has "
3087 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3088 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3093 len = ext4_blocks_count(es);
3094 start = sb_block + 1;
3095 brelse(bh); /* we're done with the superblock */
3097 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3098 start, len, blocksize);
3100 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3103 journal->j_private = sb;
3104 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3105 wait_on_buffer(journal->j_sb_buffer);
3106 if (!buffer_uptodate(journal->j_sb_buffer)) {
3107 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3110 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3111 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3112 "user (unsupported) - %d",
3113 be32_to_cpu(journal->j_superblock->s_nr_users));
3116 EXT4_SB(sb)->journal_bdev = bdev;
3117 ext4_init_journal_params(sb, journal);
3121 jbd2_journal_destroy(journal);
3123 ext4_blkdev_put(bdev);
3127 static int ext4_load_journal(struct super_block *sb,
3128 struct ext4_super_block *es,
3129 unsigned long journal_devnum)
3132 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3135 int really_read_only;
3137 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3139 if (journal_devnum &&
3140 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3141 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3142 "numbers have changed");
3143 journal_dev = new_decode_dev(journal_devnum);
3145 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3147 really_read_only = bdev_read_only(sb->s_bdev);
3150 * Are we loading a blank journal or performing recovery after a
3151 * crash? For recovery, we need to check in advance whether we
3152 * can get read-write access to the device.
3154 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3155 if (sb->s_flags & MS_RDONLY) {
3156 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3157 "required on readonly filesystem");
3158 if (really_read_only) {
3159 ext4_msg(sb, KERN_ERR, "write access "
3160 "unavailable, cannot proceed");
3163 ext4_msg(sb, KERN_INFO, "write access will "
3164 "be enabled during recovery");
3168 if (journal_inum && journal_dev) {
3169 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3170 "and inode journals!");
3175 if (!(journal = ext4_get_journal(sb, journal_inum)))
3178 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3182 if (!(journal->j_flags & JBD2_BARRIER))
3183 ext4_msg(sb, KERN_INFO, "barriers disabled");
3185 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3186 err = jbd2_journal_update_format(journal);
3188 ext4_msg(sb, KERN_ERR, "error updating journal");
3189 jbd2_journal_destroy(journal);
3194 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3195 err = jbd2_journal_wipe(journal, !really_read_only);
3197 err = jbd2_journal_load(journal);
3200 ext4_msg(sb, KERN_ERR, "error loading journal");
3201 jbd2_journal_destroy(journal);
3205 EXT4_SB(sb)->s_journal = journal;
3206 ext4_clear_journal_err(sb, es);
3208 if (journal_devnum &&
3209 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3210 es->s_journal_dev = cpu_to_le32(journal_devnum);
3212 /* Make sure we flush the recovery flag to disk. */
3213 ext4_commit_super(sb, 1);
3219 static int ext4_commit_super(struct super_block *sb, int sync)
3221 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3222 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3227 if (buffer_write_io_error(sbh)) {
3229 * Oh, dear. A previous attempt to write the
3230 * superblock failed. This could happen because the
3231 * USB device was yanked out. Or it could happen to
3232 * be a transient write error and maybe the block will
3233 * be remapped. Nothing we can do but to retry the
3234 * write and hope for the best.
3236 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3237 "superblock detected");
3238 clear_buffer_write_io_error(sbh);
3239 set_buffer_uptodate(sbh);
3242 * If the file system is mounted read-only, don't update the
3243 * superblock write time. This avoids updating the superblock
3244 * write time when we are mounting the root file system
3245 * read/only but we need to replay the journal; at that point,
3246 * for people who are east of GMT and who make their clock
3247 * tick in localtime for Windows bug-for-bug compatibility,
3248 * the clock is set in the future, and this will cause e2fsck
3249 * to complain and force a full file system check.
3251 if (!(sb->s_flags & MS_RDONLY))
3252 es->s_wtime = cpu_to_le32(get_seconds());
3253 es->s_kbytes_written =
3254 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3255 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3256 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3257 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3258 &EXT4_SB(sb)->s_freeblocks_counter));
3259 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3260 &EXT4_SB(sb)->s_freeinodes_counter));
3262 BUFFER_TRACE(sbh, "marking dirty");
3263 mark_buffer_dirty(sbh);
3265 error = sync_dirty_buffer(sbh);
3269 error = buffer_write_io_error(sbh);
3271 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3273 clear_buffer_write_io_error(sbh);
3274 set_buffer_uptodate(sbh);
3281 * Have we just finished recovery? If so, and if we are mounting (or
3282 * remounting) the filesystem readonly, then we will end up with a
3283 * consistent fs on disk. Record that fact.
3285 static void ext4_mark_recovery_complete(struct super_block *sb,
3286 struct ext4_super_block *es)
3288 journal_t *journal = EXT4_SB(sb)->s_journal;
3290 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3291 BUG_ON(journal != NULL);
3294 jbd2_journal_lock_updates(journal);
3295 if (jbd2_journal_flush(journal) < 0)
3298 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3299 sb->s_flags & MS_RDONLY) {
3300 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3301 ext4_commit_super(sb, 1);
3305 jbd2_journal_unlock_updates(journal);
3309 * If we are mounting (or read-write remounting) a filesystem whose journal
3310 * has recorded an error from a previous lifetime, move that error to the
3311 * main filesystem now.
3313 static void ext4_clear_journal_err(struct super_block *sb,
3314 struct ext4_super_block *es)
3320 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3322 journal = EXT4_SB(sb)->s_journal;
3325 * Now check for any error status which may have been recorded in the
3326 * journal by a prior ext4_error() or ext4_abort()
3329 j_errno = jbd2_journal_errno(journal);
3333 errstr = ext4_decode_error(sb, j_errno, nbuf);
3334 ext4_warning(sb, __func__, "Filesystem error recorded "
3335 "from previous mount: %s", errstr);
3336 ext4_warning(sb, __func__, "Marking fs in need of "
3337 "filesystem check.");
3339 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3340 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3341 ext4_commit_super(sb, 1);
3343 jbd2_journal_clear_err(journal);
3348 * Force the running and committing transactions to commit,
3349 * and wait on the commit.
3351 int ext4_force_commit(struct super_block *sb)
3356 if (sb->s_flags & MS_RDONLY)
3359 journal = EXT4_SB(sb)->s_journal;
3361 ret = ext4_journal_force_commit(journal);
3366 static void ext4_write_super(struct super_block *sb)
3369 ext4_commit_super(sb, 1);
3373 static int ext4_sync_fs(struct super_block *sb, int wait)
3377 struct ext4_sb_info *sbi = EXT4_SB(sb);
3379 trace_ext4_sync_fs(sb, wait);
3380 flush_workqueue(sbi->dio_unwritten_wq);
3381 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3383 jbd2_log_wait_commit(sbi->s_journal, target);
3389 * LVM calls this function before a (read-only) snapshot is created. This
3390 * gives us a chance to flush the journal completely and mark the fs clean.
3392 static int ext4_freeze(struct super_block *sb)
3397 if (sb->s_flags & MS_RDONLY)
3400 journal = EXT4_SB(sb)->s_journal;
3402 /* Now we set up the journal barrier. */
3403 jbd2_journal_lock_updates(journal);
3406 * Don't clear the needs_recovery flag if we failed to flush
3409 error = jbd2_journal_flush(journal);
3412 jbd2_journal_unlock_updates(journal);
3416 /* Journal blocked and flushed, clear needs_recovery flag. */
3417 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3418 error = ext4_commit_super(sb, 1);
3425 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3426 * flag here, even though the filesystem is not technically dirty yet.
3428 static int ext4_unfreeze(struct super_block *sb)
3430 if (sb->s_flags & MS_RDONLY)
3434 /* Reset the needs_recovery flag before the fs is unlocked. */
3435 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3436 ext4_commit_super(sb, 1);
3438 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3442 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3444 struct ext4_super_block *es;
3445 struct ext4_sb_info *sbi = EXT4_SB(sb);
3446 ext4_fsblk_t n_blocks_count = 0;
3447 unsigned long old_sb_flags;
3448 struct ext4_mount_options old_opts;
3450 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3458 /* Store the original options */
3460 old_sb_flags = sb->s_flags;
3461 old_opts.s_mount_opt = sbi->s_mount_opt;
3462 old_opts.s_resuid = sbi->s_resuid;
3463 old_opts.s_resgid = sbi->s_resgid;
3464 old_opts.s_commit_interval = sbi->s_commit_interval;
3465 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3466 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3468 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3469 for (i = 0; i < MAXQUOTAS; i++)
3470 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3472 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3473 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3476 * Allow the "check" option to be passed as a remount option.
3478 if (!parse_options(data, sb, NULL, &journal_ioprio,
3479 &n_blocks_count, 1)) {
3484 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3485 ext4_abort(sb, __func__, "Abort forced by user");
3487 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3488 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3492 if (sbi->s_journal) {
3493 ext4_init_journal_params(sb, sbi->s_journal);
3494 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3497 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3498 n_blocks_count > ext4_blocks_count(es)) {
3499 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3504 if (*flags & MS_RDONLY) {
3506 * First of all, the unconditional stuff we have to do
3507 * to disable replay of the journal when we next remount
3509 sb->s_flags |= MS_RDONLY;
3512 * OK, test if we are remounting a valid rw partition
3513 * readonly, and if so set the rdonly flag and then
3514 * mark the partition as valid again.
3516 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3517 (sbi->s_mount_state & EXT4_VALID_FS))
3518 es->s_state = cpu_to_le16(sbi->s_mount_state);
3521 ext4_mark_recovery_complete(sb, es);
3523 /* Make sure we can mount this feature set readwrite */
3524 if (!ext4_feature_set_ok(sb, 0)) {
3529 * Make sure the group descriptor checksums
3530 * are sane. If they aren't, refuse to remount r/w.
3532 for (g = 0; g < sbi->s_groups_count; g++) {
3533 struct ext4_group_desc *gdp =
3534 ext4_get_group_desc(sb, g, NULL);
3536 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3537 ext4_msg(sb, KERN_ERR,
3538 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3539 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3540 le16_to_cpu(gdp->bg_checksum));
3547 * If we have an unprocessed orphan list hanging
3548 * around from a previously readonly bdev mount,
3549 * require a full umount/remount for now.
3551 if (es->s_last_orphan) {
3552 ext4_msg(sb, KERN_WARNING, "Couldn't "
3553 "remount RDWR because of unprocessed "
3554 "orphan inode list. Please "
3555 "umount/remount instead");
3561 * Mounting a RDONLY partition read-write, so reread
3562 * and store the current valid flag. (It may have
3563 * been changed by e2fsck since we originally mounted
3567 ext4_clear_journal_err(sb, es);
3568 sbi->s_mount_state = le16_to_cpu(es->s_state);
3569 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3571 if (!ext4_setup_super(sb, es, 0))
3572 sb->s_flags &= ~MS_RDONLY;
3575 ext4_setup_system_zone(sb);
3576 if (sbi->s_journal == NULL)
3577 ext4_commit_super(sb, 1);
3580 /* Release old quota file names */
3581 for (i = 0; i < MAXQUOTAS; i++)
3582 if (old_opts.s_qf_names[i] &&
3583 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3584 kfree(old_opts.s_qf_names[i]);
3591 sb->s_flags = old_sb_flags;
3592 sbi->s_mount_opt = old_opts.s_mount_opt;
3593 sbi->s_resuid = old_opts.s_resuid;
3594 sbi->s_resgid = old_opts.s_resgid;
3595 sbi->s_commit_interval = old_opts.s_commit_interval;
3596 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3597 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3599 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3600 for (i = 0; i < MAXQUOTAS; i++) {
3601 if (sbi->s_qf_names[i] &&
3602 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3603 kfree(sbi->s_qf_names[i]);
3604 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3612 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3614 struct super_block *sb = dentry->d_sb;
3615 struct ext4_sb_info *sbi = EXT4_SB(sb);
3616 struct ext4_super_block *es = sbi->s_es;
3619 if (test_opt(sb, MINIX_DF)) {
3620 sbi->s_overhead_last = 0;
3621 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3622 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3623 ext4_fsblk_t overhead = 0;
3626 * Compute the overhead (FS structures). This is constant
3627 * for a given filesystem unless the number of block groups
3628 * changes so we cache the previous value until it does.
3632 * All of the blocks before first_data_block are
3635 overhead = le32_to_cpu(es->s_first_data_block);
3638 * Add the overhead attributed to the superblock and
3639 * block group descriptors. If the sparse superblocks
3640 * feature is turned on, then not all groups have this.
3642 for (i = 0; i < ngroups; i++) {
3643 overhead += ext4_bg_has_super(sb, i) +
3644 ext4_bg_num_gdb(sb, i);
3649 * Every block group has an inode bitmap, a block
3650 * bitmap, and an inode table.
3652 overhead += ngroups * (2 + sbi->s_itb_per_group);
3653 sbi->s_overhead_last = overhead;
3655 sbi->s_blocks_last = ext4_blocks_count(es);
3658 buf->f_type = EXT4_SUPER_MAGIC;
3659 buf->f_bsize = sb->s_blocksize;
3660 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3661 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3662 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3663 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3664 if (buf->f_bfree < ext4_r_blocks_count(es))
3666 buf->f_files = le32_to_cpu(es->s_inodes_count);
3667 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3668 buf->f_namelen = EXT4_NAME_LEN;
3669 fsid = le64_to_cpup((void *)es->s_uuid) ^
3670 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3671 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3672 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3677 /* Helper function for writing quotas on sync - we need to start transaction
3678 * before quota file is locked for write. Otherwise the are possible deadlocks:
3679 * Process 1 Process 2
3680 * ext4_create() quota_sync()
3681 * jbd2_journal_start() write_dquot()
3682 * vfs_dq_init() down(dqio_mutex)
3683 * down(dqio_mutex) jbd2_journal_start()
3689 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3691 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3694 static int ext4_write_dquot(struct dquot *dquot)
3698 struct inode *inode;
3700 inode = dquot_to_inode(dquot);
3701 handle = ext4_journal_start(inode,
3702 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3704 return PTR_ERR(handle);
3705 ret = dquot_commit(dquot);
3706 err = ext4_journal_stop(handle);
3712 static int ext4_acquire_dquot(struct dquot *dquot)
3717 handle = ext4_journal_start(dquot_to_inode(dquot),
3718 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3720 return PTR_ERR(handle);
3721 ret = dquot_acquire(dquot);
3722 err = ext4_journal_stop(handle);
3728 static int ext4_release_dquot(struct dquot *dquot)
3733 handle = ext4_journal_start(dquot_to_inode(dquot),
3734 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3735 if (IS_ERR(handle)) {
3736 /* Release dquot anyway to avoid endless cycle in dqput() */
3737 dquot_release(dquot);
3738 return PTR_ERR(handle);
3740 ret = dquot_release(dquot);
3741 err = ext4_journal_stop(handle);
3747 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3749 /* Are we journaling quotas? */
3750 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3751 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3752 dquot_mark_dquot_dirty(dquot);
3753 return ext4_write_dquot(dquot);
3755 return dquot_mark_dquot_dirty(dquot);
3759 static int ext4_write_info(struct super_block *sb, int type)
3764 /* Data block + inode block */
3765 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3767 return PTR_ERR(handle);
3768 ret = dquot_commit_info(sb, type);
3769 err = ext4_journal_stop(handle);
3776 * Turn on quotas during mount time - we need to find
3777 * the quota file and such...
3779 static int ext4_quota_on_mount(struct super_block *sb, int type)
3781 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3782 EXT4_SB(sb)->s_jquota_fmt, type);
3786 * Standard function to be called on quota_on
3788 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3789 char *name, int remount)
3794 if (!test_opt(sb, QUOTA))
3796 /* When remounting, no checks are needed and in fact, name is NULL */
3798 return vfs_quota_on(sb, type, format_id, name, remount);
3800 err = kern_path(name, LOOKUP_FOLLOW, &path);
3804 /* Quotafile not on the same filesystem? */
3805 if (path.mnt->mnt_sb != sb) {
3809 /* Journaling quota? */
3810 if (EXT4_SB(sb)->s_qf_names[type]) {
3811 /* Quotafile not in fs root? */
3812 if (path.dentry->d_parent != sb->s_root)
3813 ext4_msg(sb, KERN_WARNING,
3814 "Quota file not on filesystem root. "
3815 "Journaled quota will not work");
3819 * When we journal data on quota file, we have to flush journal to see
3820 * all updates to the file when we bypass pagecache...
3822 if (EXT4_SB(sb)->s_journal &&
3823 ext4_should_journal_data(path.dentry->d_inode)) {
3825 * We don't need to lock updates but journal_flush() could
3826 * otherwise be livelocked...
3828 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3829 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3830 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3837 err = vfs_quota_on_path(sb, type, format_id, &path);
3842 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3843 * acquiring the locks... As quota files are never truncated and quota code
3844 * itself serializes the operations (and noone else should touch the files)
3845 * we don't have to be afraid of races */
3846 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3847 size_t len, loff_t off)
3849 struct inode *inode = sb_dqopt(sb)->files[type];
3850 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3852 int offset = off & (sb->s_blocksize - 1);
3855 struct buffer_head *bh;
3856 loff_t i_size = i_size_read(inode);
3860 if (off+len > i_size)
3863 while (toread > 0) {
3864 tocopy = sb->s_blocksize - offset < toread ?
3865 sb->s_blocksize - offset : toread;
3866 bh = ext4_bread(NULL, inode, blk, 0, &err);
3869 if (!bh) /* A hole? */
3870 memset(data, 0, tocopy);
3872 memcpy(data, bh->b_data+offset, tocopy);
3882 /* Write to quotafile (we know the transaction is already started and has
3883 * enough credits) */
3884 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3885 const char *data, size_t len, loff_t off)
3887 struct inode *inode = sb_dqopt(sb)->files[type];
3888 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3890 int offset = off & (sb->s_blocksize - 1);
3892 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3893 size_t towrite = len;
3894 struct buffer_head *bh;
3895 handle_t *handle = journal_current_handle();
3897 if (EXT4_SB(sb)->s_journal && !handle) {
3898 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3899 " cancelled because transaction is not started",
3900 (unsigned long long)off, (unsigned long long)len);
3903 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3904 while (towrite > 0) {
3905 tocopy = sb->s_blocksize - offset < towrite ?
3906 sb->s_blocksize - offset : towrite;
3907 bh = ext4_bread(handle, inode, blk, 1, &err);
3910 if (journal_quota) {
3911 err = ext4_journal_get_write_access(handle, bh);
3918 memcpy(bh->b_data+offset, data, tocopy);
3919 flush_dcache_page(bh->b_page);
3922 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3924 /* Always do at least ordered writes for quotas */
3925 err = ext4_jbd2_file_inode(handle, inode);
3926 mark_buffer_dirty(bh);
3937 if (len == towrite) {
3938 mutex_unlock(&inode->i_mutex);
3941 if (inode->i_size < off+len-towrite) {
3942 i_size_write(inode, off+len-towrite);
3943 EXT4_I(inode)->i_disksize = inode->i_size;
3945 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3946 ext4_mark_inode_dirty(handle, inode);
3947 mutex_unlock(&inode->i_mutex);
3948 return len - towrite;
3953 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3954 const char *dev_name, void *data, struct vfsmount *mnt)
3956 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3959 static struct file_system_type ext4_fs_type = {
3960 .owner = THIS_MODULE,
3962 .get_sb = ext4_get_sb,
3963 .kill_sb = kill_block_super,
3964 .fs_flags = FS_REQUIRES_DEV,
3967 static int __init init_ext4_fs(void)
3971 err = init_ext4_system_zone();
3974 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3977 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3978 err = init_ext4_mballoc();
3982 err = init_ext4_xattr();
3985 err = init_inodecache();
3988 err = register_filesystem(&ext4_fs_type);
3993 destroy_inodecache();
3997 exit_ext4_mballoc();
3999 remove_proc_entry("fs/ext4", NULL);
4000 kset_unregister(ext4_kset);
4002 exit_ext4_system_zone();
4006 static void __exit exit_ext4_fs(void)
4008 unregister_filesystem(&ext4_fs_type);
4009 destroy_inodecache();
4011 exit_ext4_mballoc();
4012 remove_proc_entry("fs/ext4", NULL);
4013 kset_unregister(ext4_kset);
4014 exit_ext4_system_zone();
4017 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4018 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4019 MODULE_LICENSE("GPL");
4020 module_init(init_ext4_fs)
4021 module_exit(exit_ext4_fs)