2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
195 handle = (handle_t *)ref_cnt;
197 current->journal_info = handle;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
205 unsigned long ref_cnt = (unsigned long)handle;
207 BUG_ON(ref_cnt == 0);
210 handle = (handle_t *)ref_cnt;
212 current->journal_info = handle;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
239 return jbd2_journal_start(journal, nblocks);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where, handle_t *handle)
252 struct super_block *sb;
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
260 sb = handle->h_transaction->t_journal->j_private;
262 rc = jbd2_journal_stop(handle);
267 __ext4_std_error(sb, where, err);
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
277 BUG_ON(!ext4_handle_valid(handle));
280 BUFFER_TRACE(bh, "abort");
285 if (is_handle_aborted(handle))
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
291 jbd2_journal_abort_handle(handle);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block *sb)
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316 if (sb->s_flags & MS_RDONLY)
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
324 jbd2_journal_abort(journal, -EIO);
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
347 ext4_handle_error(sb);
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
357 errstr = "IO failure";
360 errstr = "Out of memory";
363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 errstr = "Journal has aborted";
367 errstr = "Readonly filesystem";
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
399 errstr = ext4_decode_error(sb, errno, nbuf);
400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
403 ext4_handle_error(sb);
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
416 void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
427 if (test_opt(sb, ERRORS_PANIC))
428 panic("EXT4-fs panic from previous error\n");
430 if (sb->s_flags & MS_RDONLY)
433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 sb->s_flags |= MS_RDONLY;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
453 void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 const char *function, const char *fmt, ...)
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 ext4_commit_super(sb, 0);
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
499 ext4_lock_group(sb, grp);
503 void ext4_update_dynamic_rev(struct super_block *sb)
505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
510 ext4_warning(sb, __func__,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
529 * Open the external journal device
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 __bdevname(dev, b), PTR_ERR(bdev));
548 * Release the journal device
550 static int ext4_blkdev_put(struct block_device *bdev)
553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
558 struct block_device *bdev;
561 bdev = sbi->journal_bdev;
563 ret = ext4_blkdev_put(bdev);
564 sbi->journal_bdev = NULL;
569 static inline struct inode *orphan_list_entry(struct list_head *l)
571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
592 static void ext4_put_super(struct super_block *sb)
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
604 ext4_commit_super(sb, 1);
606 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
614 ext4_release_system_zone(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
619 if (!(sb->s_flags & MS_RDONLY)) {
620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
621 es->s_state = cpu_to_le16(sbi->s_mount_state);
622 ext4_commit_super(sb, 1);
625 remove_proc_entry(sb->s_id, ext4_proc_root);
627 kobject_del(&sbi->s_kobj);
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
635 kfree(sbi->s_flex_groups);
636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
654 invalidate_bdev(sb->s_bdev);
655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
661 sync_blockdev(sbi->journal_bdev);
662 invalidate_bdev(sbi->journal_bdev);
663 ext4_blkdev_remove(sbi);
665 sb->s_fs_info = NULL;
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
674 kfree(sbi->s_blockgroup_lock);
678 static struct kmem_cache *ext4_inode_cachep;
681 * Called inside transaction, so use GFP_NOFS
683 static struct inode *ext4_alloc_inode(struct super_block *sb)
685 struct ext4_inode_info *ei;
687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
691 ei->vfs_inode.i_version = 1;
692 ei->vfs_inode.i_data.writeback_index = 0;
693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
705 ei->i_delalloc_reserved_flag = 0;
706 spin_lock_init(&(ei->i_block_reservation_lock));
708 ei->i_reserved_quota = 0;
710 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
711 ei->cur_aio_dio = NULL;
713 ei->i_datasync_tid = 0;
715 return &ei->vfs_inode;
718 static void ext4_destroy_inode(struct inode *inode)
720 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
721 ext4_msg(inode->i_sb, KERN_ERR,
722 "Inode %lu (%p): orphan list check failed!",
723 inode->i_ino, EXT4_I(inode));
724 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
725 EXT4_I(inode), sizeof(struct ext4_inode_info),
729 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
732 static void init_once(void *foo)
734 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
736 INIT_LIST_HEAD(&ei->i_orphan);
737 #ifdef CONFIG_EXT4_FS_XATTR
738 init_rwsem(&ei->xattr_sem);
740 init_rwsem(&ei->i_data_sem);
741 inode_init_once(&ei->vfs_inode);
744 static int init_inodecache(void)
746 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
747 sizeof(struct ext4_inode_info),
748 0, (SLAB_RECLAIM_ACCOUNT|
751 if (ext4_inode_cachep == NULL)
756 static void destroy_inodecache(void)
758 kmem_cache_destroy(ext4_inode_cachep);
761 static void ext4_clear_inode(struct inode *inode)
763 ext4_discard_preallocations(inode);
764 if (EXT4_JOURNAL(inode))
765 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
766 &EXT4_I(inode)->jinode);
769 static inline void ext4_show_quota_options(struct seq_file *seq,
770 struct super_block *sb)
772 #if defined(CONFIG_QUOTA)
773 struct ext4_sb_info *sbi = EXT4_SB(sb);
775 if (sbi->s_jquota_fmt) {
778 switch (sbi->s_jquota_fmt) {
789 seq_printf(seq, ",jqfmt=%s", fmtname);
792 if (sbi->s_qf_names[USRQUOTA])
793 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
795 if (sbi->s_qf_names[GRPQUOTA])
796 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
798 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
799 seq_puts(seq, ",usrquota");
801 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
802 seq_puts(seq, ",grpquota");
808 * - it's set to a non-default value OR
809 * - if the per-sb default is different from the global default
811 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
814 unsigned long def_mount_opts;
815 struct super_block *sb = vfs->mnt_sb;
816 struct ext4_sb_info *sbi = EXT4_SB(sb);
817 struct ext4_super_block *es = sbi->s_es;
819 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
820 def_errors = le16_to_cpu(es->s_errors);
822 if (sbi->s_sb_block != 1)
823 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
824 if (test_opt(sb, MINIX_DF))
825 seq_puts(seq, ",minixdf");
826 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
827 seq_puts(seq, ",grpid");
828 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
829 seq_puts(seq, ",nogrpid");
830 if (sbi->s_resuid != EXT4_DEF_RESUID ||
831 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
832 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
834 if (sbi->s_resgid != EXT4_DEF_RESGID ||
835 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
836 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
838 if (test_opt(sb, ERRORS_RO)) {
839 if (def_errors == EXT4_ERRORS_PANIC ||
840 def_errors == EXT4_ERRORS_CONTINUE) {
841 seq_puts(seq, ",errors=remount-ro");
844 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
845 seq_puts(seq, ",errors=continue");
846 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
847 seq_puts(seq, ",errors=panic");
848 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
849 seq_puts(seq, ",nouid32");
850 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
851 seq_puts(seq, ",debug");
852 if (test_opt(sb, OLDALLOC))
853 seq_puts(seq, ",oldalloc");
854 #ifdef CONFIG_EXT4_FS_XATTR
855 if (test_opt(sb, XATTR_USER) &&
856 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
857 seq_puts(seq, ",user_xattr");
858 if (!test_opt(sb, XATTR_USER) &&
859 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
860 seq_puts(seq, ",nouser_xattr");
863 #ifdef CONFIG_EXT4_FS_POSIX_ACL
864 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
865 seq_puts(seq, ",acl");
866 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
867 seq_puts(seq, ",noacl");
869 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
870 seq_printf(seq, ",commit=%u",
871 (unsigned) (sbi->s_commit_interval / HZ));
873 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
874 seq_printf(seq, ",min_batch_time=%u",
875 (unsigned) sbi->s_min_batch_time);
877 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
878 seq_printf(seq, ",max_batch_time=%u",
879 (unsigned) sbi->s_min_batch_time);
883 * We're changing the default of barrier mount option, so
884 * let's always display its mount state so it's clear what its
887 seq_puts(seq, ",barrier=");
888 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
889 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
890 seq_puts(seq, ",journal_async_commit");
891 if (test_opt(sb, NOBH))
892 seq_puts(seq, ",nobh");
893 if (test_opt(sb, I_VERSION))
894 seq_puts(seq, ",i_version");
895 if (!test_opt(sb, DELALLOC))
896 seq_puts(seq, ",nodelalloc");
900 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
902 * journal mode get enabled in different ways
903 * So just print the value even if we didn't specify it
905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
906 seq_puts(seq, ",data=journal");
907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
908 seq_puts(seq, ",data=ordered");
909 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
910 seq_puts(seq, ",data=writeback");
912 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
913 seq_printf(seq, ",inode_readahead_blks=%u",
914 sbi->s_inode_readahead_blks);
916 if (test_opt(sb, DATA_ERR_ABORT))
917 seq_puts(seq, ",data_err=abort");
919 if (test_opt(sb, NO_AUTO_DA_ALLOC))
920 seq_puts(seq, ",noauto_da_alloc");
922 if (test_opt(sb, DISCARD))
923 seq_puts(seq, ",discard");
925 if (test_opt(sb, NOLOAD))
926 seq_puts(seq, ",norecovery");
928 ext4_show_quota_options(seq, sb);
933 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
934 u64 ino, u32 generation)
938 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
939 return ERR_PTR(-ESTALE);
940 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
941 return ERR_PTR(-ESTALE);
943 /* iget isn't really right if the inode is currently unallocated!!
945 * ext4_read_inode will return a bad_inode if the inode had been
946 * deleted, so we should be safe.
948 * Currently we don't know the generation for parent directory, so
949 * a generation of 0 means "accept any"
951 inode = ext4_iget(sb, ino);
953 return ERR_CAST(inode);
954 if (generation && inode->i_generation != generation) {
956 return ERR_PTR(-ESTALE);
962 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
963 int fh_len, int fh_type)
965 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
969 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
970 int fh_len, int fh_type)
972 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
977 * Try to release metadata pages (indirect blocks, directories) which are
978 * mapped via the block device. Since these pages could have journal heads
979 * which would prevent try_to_free_buffers() from freeing them, we must use
980 * jbd2 layer's try_to_free_buffers() function to release them.
982 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
985 journal_t *journal = EXT4_SB(sb)->s_journal;
987 WARN_ON(PageChecked(page));
988 if (!page_has_buffers(page))
991 return jbd2_journal_try_to_free_buffers(journal, page,
993 return try_to_free_buffers(page);
997 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
998 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1000 static int ext4_write_dquot(struct dquot *dquot);
1001 static int ext4_acquire_dquot(struct dquot *dquot);
1002 static int ext4_release_dquot(struct dquot *dquot);
1003 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1004 static int ext4_write_info(struct super_block *sb, int type);
1005 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1006 char *path, int remount);
1007 static int ext4_quota_on_mount(struct super_block *sb, int type);
1008 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1009 size_t len, loff_t off);
1010 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1011 const char *data, size_t len, loff_t off);
1013 static const struct dquot_operations ext4_quota_operations = {
1014 .initialize = dquot_initialize,
1016 .alloc_space = dquot_alloc_space,
1017 .reserve_space = dquot_reserve_space,
1018 .claim_space = dquot_claim_space,
1019 .release_rsv = dquot_release_reserved_space,
1021 .get_reserved_space = ext4_get_reserved_space,
1023 .alloc_inode = dquot_alloc_inode,
1024 .free_space = dquot_free_space,
1025 .free_inode = dquot_free_inode,
1026 .transfer = dquot_transfer,
1027 .write_dquot = ext4_write_dquot,
1028 .acquire_dquot = ext4_acquire_dquot,
1029 .release_dquot = ext4_release_dquot,
1030 .mark_dirty = ext4_mark_dquot_dirty,
1031 .write_info = ext4_write_info,
1032 .alloc_dquot = dquot_alloc,
1033 .destroy_dquot = dquot_destroy,
1036 static const struct quotactl_ops ext4_qctl_operations = {
1037 .quota_on = ext4_quota_on,
1038 .quota_off = vfs_quota_off,
1039 .quota_sync = vfs_quota_sync,
1040 .get_info = vfs_get_dqinfo,
1041 .set_info = vfs_set_dqinfo,
1042 .get_dqblk = vfs_get_dqblk,
1043 .set_dqblk = vfs_set_dqblk
1047 static const struct super_operations ext4_sops = {
1048 .alloc_inode = ext4_alloc_inode,
1049 .destroy_inode = ext4_destroy_inode,
1050 .write_inode = ext4_write_inode,
1051 .dirty_inode = ext4_dirty_inode,
1052 .delete_inode = ext4_delete_inode,
1053 .put_super = ext4_put_super,
1054 .sync_fs = ext4_sync_fs,
1055 .freeze_fs = ext4_freeze,
1056 .unfreeze_fs = ext4_unfreeze,
1057 .statfs = ext4_statfs,
1058 .remount_fs = ext4_remount,
1059 .clear_inode = ext4_clear_inode,
1060 .show_options = ext4_show_options,
1062 .quota_read = ext4_quota_read,
1063 .quota_write = ext4_quota_write,
1065 .bdev_try_to_free_page = bdev_try_to_free_page,
1068 static const struct super_operations ext4_nojournal_sops = {
1069 .alloc_inode = ext4_alloc_inode,
1070 .destroy_inode = ext4_destroy_inode,
1071 .write_inode = ext4_write_inode,
1072 .dirty_inode = ext4_dirty_inode,
1073 .delete_inode = ext4_delete_inode,
1074 .write_super = ext4_write_super,
1075 .put_super = ext4_put_super,
1076 .statfs = ext4_statfs,
1077 .remount_fs = ext4_remount,
1078 .clear_inode = ext4_clear_inode,
1079 .show_options = ext4_show_options,
1081 .quota_read = ext4_quota_read,
1082 .quota_write = ext4_quota_write,
1084 .bdev_try_to_free_page = bdev_try_to_free_page,
1087 static const struct export_operations ext4_export_ops = {
1088 .fh_to_dentry = ext4_fh_to_dentry,
1089 .fh_to_parent = ext4_fh_to_parent,
1090 .get_parent = ext4_get_parent,
1094 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1095 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1096 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1097 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1098 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1099 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1100 Opt_journal_update, Opt_journal_dev,
1101 Opt_journal_checksum, Opt_journal_async_commit,
1102 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1103 Opt_data_err_abort, Opt_data_err_ignore,
1104 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1105 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1106 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1107 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1108 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1109 Opt_block_validity, Opt_noblock_validity,
1110 Opt_inode_readahead_blks, Opt_journal_ioprio,
1111 Opt_discard, Opt_nodiscard,
1114 static const match_table_t tokens = {
1115 {Opt_bsd_df, "bsddf"},
1116 {Opt_minix_df, "minixdf"},
1117 {Opt_grpid, "grpid"},
1118 {Opt_grpid, "bsdgroups"},
1119 {Opt_nogrpid, "nogrpid"},
1120 {Opt_nogrpid, "sysvgroups"},
1121 {Opt_resgid, "resgid=%u"},
1122 {Opt_resuid, "resuid=%u"},
1124 {Opt_err_cont, "errors=continue"},
1125 {Opt_err_panic, "errors=panic"},
1126 {Opt_err_ro, "errors=remount-ro"},
1127 {Opt_nouid32, "nouid32"},
1128 {Opt_debug, "debug"},
1129 {Opt_oldalloc, "oldalloc"},
1130 {Opt_orlov, "orlov"},
1131 {Opt_user_xattr, "user_xattr"},
1132 {Opt_nouser_xattr, "nouser_xattr"},
1134 {Opt_noacl, "noacl"},
1135 {Opt_noload, "noload"},
1136 {Opt_noload, "norecovery"},
1139 {Opt_commit, "commit=%u"},
1140 {Opt_min_batch_time, "min_batch_time=%u"},
1141 {Opt_max_batch_time, "max_batch_time=%u"},
1142 {Opt_journal_update, "journal=update"},
1143 {Opt_journal_dev, "journal_dev=%u"},
1144 {Opt_journal_checksum, "journal_checksum"},
1145 {Opt_journal_async_commit, "journal_async_commit"},
1146 {Opt_abort, "abort"},
1147 {Opt_data_journal, "data=journal"},
1148 {Opt_data_ordered, "data=ordered"},
1149 {Opt_data_writeback, "data=writeback"},
1150 {Opt_data_err_abort, "data_err=abort"},
1151 {Opt_data_err_ignore, "data_err=ignore"},
1152 {Opt_offusrjquota, "usrjquota="},
1153 {Opt_usrjquota, "usrjquota=%s"},
1154 {Opt_offgrpjquota, "grpjquota="},
1155 {Opt_grpjquota, "grpjquota=%s"},
1156 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1157 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1158 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1159 {Opt_grpquota, "grpquota"},
1160 {Opt_noquota, "noquota"},
1161 {Opt_quota, "quota"},
1162 {Opt_usrquota, "usrquota"},
1163 {Opt_barrier, "barrier=%u"},
1164 {Opt_barrier, "barrier"},
1165 {Opt_nobarrier, "nobarrier"},
1166 {Opt_i_version, "i_version"},
1167 {Opt_stripe, "stripe=%u"},
1168 {Opt_resize, "resize"},
1169 {Opt_delalloc, "delalloc"},
1170 {Opt_nodelalloc, "nodelalloc"},
1171 {Opt_block_validity, "block_validity"},
1172 {Opt_noblock_validity, "noblock_validity"},
1173 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1174 {Opt_journal_ioprio, "journal_ioprio=%u"},
1175 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1176 {Opt_auto_da_alloc, "auto_da_alloc"},
1177 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1178 {Opt_discard, "discard"},
1179 {Opt_nodiscard, "nodiscard"},
1183 static ext4_fsblk_t get_sb_block(void **data)
1185 ext4_fsblk_t sb_block;
1186 char *options = (char *) *data;
1188 if (!options || strncmp(options, "sb=", 3) != 0)
1189 return 1; /* Default location */
1192 /* TODO: use simple_strtoll with >32bit ext4 */
1193 sb_block = simple_strtoul(options, &options, 0);
1194 if (*options && *options != ',') {
1195 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1199 if (*options == ',')
1201 *data = (void *) options;
1206 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1208 static int parse_options(char *options, struct super_block *sb,
1209 unsigned long *journal_devnum,
1210 unsigned int *journal_ioprio,
1211 ext4_fsblk_t *n_blocks_count, int is_remount)
1213 struct ext4_sb_info *sbi = EXT4_SB(sb);
1215 substring_t args[MAX_OPT_ARGS];
1226 while ((p = strsep(&options, ",")) != NULL) {
1231 token = match_token(p, tokens, args);
1234 clear_opt(sbi->s_mount_opt, MINIX_DF);
1237 set_opt(sbi->s_mount_opt, MINIX_DF);
1240 set_opt(sbi->s_mount_opt, GRPID);
1243 clear_opt(sbi->s_mount_opt, GRPID);
1246 if (match_int(&args[0], &option))
1248 sbi->s_resuid = option;
1251 if (match_int(&args[0], &option))
1253 sbi->s_resgid = option;
1256 /* handled by get_sb_block() instead of here */
1257 /* *sb_block = match_int(&args[0]); */
1260 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1261 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1262 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1265 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1266 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1267 set_opt(sbi->s_mount_opt, ERRORS_RO);
1270 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1271 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1272 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1275 set_opt(sbi->s_mount_opt, NO_UID32);
1278 set_opt(sbi->s_mount_opt, DEBUG);
1281 set_opt(sbi->s_mount_opt, OLDALLOC);
1284 clear_opt(sbi->s_mount_opt, OLDALLOC);
1286 #ifdef CONFIG_EXT4_FS_XATTR
1287 case Opt_user_xattr:
1288 set_opt(sbi->s_mount_opt, XATTR_USER);
1290 case Opt_nouser_xattr:
1291 clear_opt(sbi->s_mount_opt, XATTR_USER);
1294 case Opt_user_xattr:
1295 case Opt_nouser_xattr:
1296 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1299 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1301 set_opt(sbi->s_mount_opt, POSIX_ACL);
1304 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1309 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1312 case Opt_journal_update:
1314 /* Eventually we will want to be able to create
1315 a journal file here. For now, only allow the
1316 user to specify an existing inode to be the
1319 ext4_msg(sb, KERN_ERR,
1320 "Cannot specify journal on remount");
1323 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1325 case Opt_journal_dev:
1327 ext4_msg(sb, KERN_ERR,
1328 "Cannot specify journal on remount");
1331 if (match_int(&args[0], &option))
1333 *journal_devnum = option;
1335 case Opt_journal_checksum:
1336 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1338 case Opt_journal_async_commit:
1339 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1340 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1343 set_opt(sbi->s_mount_opt, NOLOAD);
1346 if (match_int(&args[0], &option))
1351 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1352 sbi->s_commit_interval = HZ * option;
1354 case Opt_max_batch_time:
1355 if (match_int(&args[0], &option))
1360 option = EXT4_DEF_MAX_BATCH_TIME;
1361 sbi->s_max_batch_time = option;
1363 case Opt_min_batch_time:
1364 if (match_int(&args[0], &option))
1368 sbi->s_min_batch_time = option;
1370 case Opt_data_journal:
1371 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1373 case Opt_data_ordered:
1374 data_opt = EXT4_MOUNT_ORDERED_DATA;
1376 case Opt_data_writeback:
1377 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1380 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1382 ext4_msg(sb, KERN_ERR,
1383 "Cannot change data mode on remount");
1387 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1388 sbi->s_mount_opt |= data_opt;
1391 case Opt_data_err_abort:
1392 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1394 case Opt_data_err_ignore:
1395 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1404 if (sb_any_quota_loaded(sb) &&
1405 !sbi->s_qf_names[qtype]) {
1406 ext4_msg(sb, KERN_ERR,
1407 "Cannot change journaled "
1408 "quota options when quota turned on");
1411 qname = match_strdup(&args[0]);
1413 ext4_msg(sb, KERN_ERR,
1414 "Not enough memory for "
1415 "storing quotafile name");
1418 if (sbi->s_qf_names[qtype] &&
1419 strcmp(sbi->s_qf_names[qtype], qname)) {
1420 ext4_msg(sb, KERN_ERR,
1421 "%s quota file already "
1422 "specified", QTYPE2NAME(qtype));
1426 sbi->s_qf_names[qtype] = qname;
1427 if (strchr(sbi->s_qf_names[qtype], '/')) {
1428 ext4_msg(sb, KERN_ERR,
1429 "quotafile must be on "
1431 kfree(sbi->s_qf_names[qtype]);
1432 sbi->s_qf_names[qtype] = NULL;
1435 set_opt(sbi->s_mount_opt, QUOTA);
1437 case Opt_offusrjquota:
1440 case Opt_offgrpjquota:
1443 if (sb_any_quota_loaded(sb) &&
1444 sbi->s_qf_names[qtype]) {
1445 ext4_msg(sb, KERN_ERR, "Cannot change "
1446 "journaled quota options when "
1451 * The space will be released later when all options
1452 * are confirmed to be correct
1454 sbi->s_qf_names[qtype] = NULL;
1456 case Opt_jqfmt_vfsold:
1457 qfmt = QFMT_VFS_OLD;
1459 case Opt_jqfmt_vfsv0:
1462 case Opt_jqfmt_vfsv1:
1465 if (sb_any_quota_loaded(sb) &&
1466 sbi->s_jquota_fmt != qfmt) {
1467 ext4_msg(sb, KERN_ERR, "Cannot change "
1468 "journaled quota options when "
1472 sbi->s_jquota_fmt = qfmt;
1476 set_opt(sbi->s_mount_opt, QUOTA);
1477 set_opt(sbi->s_mount_opt, USRQUOTA);
1480 set_opt(sbi->s_mount_opt, QUOTA);
1481 set_opt(sbi->s_mount_opt, GRPQUOTA);
1484 if (sb_any_quota_loaded(sb)) {
1485 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1486 "options when quota turned on");
1489 clear_opt(sbi->s_mount_opt, QUOTA);
1490 clear_opt(sbi->s_mount_opt, USRQUOTA);
1491 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1497 ext4_msg(sb, KERN_ERR,
1498 "quota options not supported");
1502 case Opt_offusrjquota:
1503 case Opt_offgrpjquota:
1504 case Opt_jqfmt_vfsold:
1505 case Opt_jqfmt_vfsv0:
1506 case Opt_jqfmt_vfsv1:
1507 ext4_msg(sb, KERN_ERR,
1508 "journaled quota options not supported");
1514 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1517 clear_opt(sbi->s_mount_opt, BARRIER);
1520 if (match_int(&args[0], &option)) {
1521 set_opt(sbi->s_mount_opt, BARRIER);
1525 set_opt(sbi->s_mount_opt, BARRIER);
1527 clear_opt(sbi->s_mount_opt, BARRIER);
1533 ext4_msg(sb, KERN_ERR,
1534 "resize option only available "
1538 if (match_int(&args[0], &option) != 0)
1540 *n_blocks_count = option;
1543 set_opt(sbi->s_mount_opt, NOBH);
1546 clear_opt(sbi->s_mount_opt, NOBH);
1549 set_opt(sbi->s_mount_opt, I_VERSION);
1550 sb->s_flags |= MS_I_VERSION;
1552 case Opt_nodelalloc:
1553 clear_opt(sbi->s_mount_opt, DELALLOC);
1556 if (match_int(&args[0], &option))
1560 sbi->s_stripe = option;
1563 set_opt(sbi->s_mount_opt, DELALLOC);
1565 case Opt_block_validity:
1566 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1568 case Opt_noblock_validity:
1569 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1571 case Opt_inode_readahead_blks:
1572 if (match_int(&args[0], &option))
1574 if (option < 0 || option > (1 << 30))
1576 if (!is_power_of_2(option)) {
1577 ext4_msg(sb, KERN_ERR,
1578 "EXT4-fs: inode_readahead_blks"
1579 " must be a power of 2");
1582 sbi->s_inode_readahead_blks = option;
1584 case Opt_journal_ioprio:
1585 if (match_int(&args[0], &option))
1587 if (option < 0 || option > 7)
1589 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1592 case Opt_noauto_da_alloc:
1593 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1595 case Opt_auto_da_alloc:
1596 if (match_int(&args[0], &option)) {
1597 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1601 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1603 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1606 set_opt(sbi->s_mount_opt, DISCARD);
1609 clear_opt(sbi->s_mount_opt, DISCARD);
1612 ext4_msg(sb, KERN_ERR,
1613 "Unrecognized mount option \"%s\" "
1614 "or missing value", p);
1619 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1620 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1621 sbi->s_qf_names[USRQUOTA])
1622 clear_opt(sbi->s_mount_opt, USRQUOTA);
1624 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1625 sbi->s_qf_names[GRPQUOTA])
1626 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1628 if ((sbi->s_qf_names[USRQUOTA] &&
1629 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1630 (sbi->s_qf_names[GRPQUOTA] &&
1631 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1632 ext4_msg(sb, KERN_ERR, "old and new quota "
1637 if (!sbi->s_jquota_fmt) {
1638 ext4_msg(sb, KERN_ERR, "journaled quota format "
1643 if (sbi->s_jquota_fmt) {
1644 ext4_msg(sb, KERN_ERR, "journaled quota format "
1645 "specified with no journaling "
1654 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1657 struct ext4_sb_info *sbi = EXT4_SB(sb);
1660 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1661 ext4_msg(sb, KERN_ERR, "revision level too high, "
1662 "forcing read-only mode");
1667 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1668 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1669 "running e2fsck is recommended");
1670 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1671 ext4_msg(sb, KERN_WARNING,
1672 "warning: mounting fs with errors, "
1673 "running e2fsck is recommended");
1674 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1675 le16_to_cpu(es->s_mnt_count) >=
1676 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1677 ext4_msg(sb, KERN_WARNING,
1678 "warning: maximal mount count reached, "
1679 "running e2fsck is recommended");
1680 else if (le32_to_cpu(es->s_checkinterval) &&
1681 (le32_to_cpu(es->s_lastcheck) +
1682 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1683 ext4_msg(sb, KERN_WARNING,
1684 "warning: checktime reached, "
1685 "running e2fsck is recommended");
1686 if (!sbi->s_journal)
1687 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1688 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1689 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1690 le16_add_cpu(&es->s_mnt_count, 1);
1691 es->s_mtime = cpu_to_le32(get_seconds());
1692 ext4_update_dynamic_rev(sb);
1694 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1696 ext4_commit_super(sb, 1);
1697 if (test_opt(sb, DEBUG))
1698 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1699 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1701 sbi->s_groups_count,
1702 EXT4_BLOCKS_PER_GROUP(sb),
1703 EXT4_INODES_PER_GROUP(sb),
1709 static int ext4_fill_flex_info(struct super_block *sb)
1711 struct ext4_sb_info *sbi = EXT4_SB(sb);
1712 struct ext4_group_desc *gdp = NULL;
1713 ext4_group_t flex_group_count;
1714 ext4_group_t flex_group;
1715 int groups_per_flex = 0;
1719 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1720 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1722 if (groups_per_flex < 2) {
1723 sbi->s_log_groups_per_flex = 0;
1727 /* We allocate both existing and potentially added groups */
1728 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1729 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1730 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1731 size = flex_group_count * sizeof(struct flex_groups);
1732 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1733 if (sbi->s_flex_groups == NULL) {
1734 sbi->s_flex_groups = vmalloc(size);
1735 if (sbi->s_flex_groups)
1736 memset(sbi->s_flex_groups, 0, size);
1738 if (sbi->s_flex_groups == NULL) {
1739 ext4_msg(sb, KERN_ERR, "not enough memory for "
1740 "%u flex groups", flex_group_count);
1744 for (i = 0; i < sbi->s_groups_count; i++) {
1745 gdp = ext4_get_group_desc(sb, i, NULL);
1747 flex_group = ext4_flex_group(sbi, i);
1748 atomic_add(ext4_free_inodes_count(sb, gdp),
1749 &sbi->s_flex_groups[flex_group].free_inodes);
1750 atomic_add(ext4_free_blks_count(sb, gdp),
1751 &sbi->s_flex_groups[flex_group].free_blocks);
1752 atomic_add(ext4_used_dirs_count(sb, gdp),
1753 &sbi->s_flex_groups[flex_group].used_dirs);
1761 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1762 struct ext4_group_desc *gdp)
1766 if (sbi->s_es->s_feature_ro_compat &
1767 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1768 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1769 __le32 le_group = cpu_to_le32(block_group);
1771 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1772 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1773 crc = crc16(crc, (__u8 *)gdp, offset);
1774 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1775 /* for checksum of struct ext4_group_desc do the rest...*/
1776 if ((sbi->s_es->s_feature_incompat &
1777 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1778 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1779 crc = crc16(crc, (__u8 *)gdp + offset,
1780 le16_to_cpu(sbi->s_es->s_desc_size) -
1784 return cpu_to_le16(crc);
1787 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1788 struct ext4_group_desc *gdp)
1790 if ((sbi->s_es->s_feature_ro_compat &
1791 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1792 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1798 /* Called at mount-time, super-block is locked */
1799 static int ext4_check_descriptors(struct super_block *sb)
1801 struct ext4_sb_info *sbi = EXT4_SB(sb);
1802 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1803 ext4_fsblk_t last_block;
1804 ext4_fsblk_t block_bitmap;
1805 ext4_fsblk_t inode_bitmap;
1806 ext4_fsblk_t inode_table;
1807 int flexbg_flag = 0;
1810 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1813 ext4_debug("Checking group descriptors");
1815 for (i = 0; i < sbi->s_groups_count; i++) {
1816 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1818 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1819 last_block = ext4_blocks_count(sbi->s_es) - 1;
1821 last_block = first_block +
1822 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1824 block_bitmap = ext4_block_bitmap(sb, gdp);
1825 if (block_bitmap < first_block || block_bitmap > last_block) {
1826 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1827 "Block bitmap for group %u not in group "
1828 "(block %llu)!", i, block_bitmap);
1831 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1832 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1833 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1834 "Inode bitmap for group %u not in group "
1835 "(block %llu)!", i, inode_bitmap);
1838 inode_table = ext4_inode_table(sb, gdp);
1839 if (inode_table < first_block ||
1840 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1841 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1842 "Inode table for group %u not in group "
1843 "(block %llu)!", i, inode_table);
1846 ext4_lock_group(sb, i);
1847 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1848 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1849 "Checksum for group %u failed (%u!=%u)",
1850 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1851 gdp)), le16_to_cpu(gdp->bg_checksum));
1852 if (!(sb->s_flags & MS_RDONLY)) {
1853 ext4_unlock_group(sb, i);
1857 ext4_unlock_group(sb, i);
1859 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1862 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1863 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1867 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1868 * the superblock) which were deleted from all directories, but held open by
1869 * a process at the time of a crash. We walk the list and try to delete these
1870 * inodes at recovery time (only with a read-write filesystem).
1872 * In order to keep the orphan inode chain consistent during traversal (in
1873 * case of crash during recovery), we link each inode into the superblock
1874 * orphan list_head and handle it the same way as an inode deletion during
1875 * normal operation (which journals the operations for us).
1877 * We only do an iget() and an iput() on each inode, which is very safe if we
1878 * accidentally point at an in-use or already deleted inode. The worst that
1879 * can happen in this case is that we get a "bit already cleared" message from
1880 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1881 * e2fsck was run on this filesystem, and it must have already done the orphan
1882 * inode cleanup for us, so we can safely abort without any further action.
1884 static void ext4_orphan_cleanup(struct super_block *sb,
1885 struct ext4_super_block *es)
1887 unsigned int s_flags = sb->s_flags;
1888 int nr_orphans = 0, nr_truncates = 0;
1892 if (!es->s_last_orphan) {
1893 jbd_debug(4, "no orphan inodes to clean up\n");
1897 if (bdev_read_only(sb->s_bdev)) {
1898 ext4_msg(sb, KERN_ERR, "write access "
1899 "unavailable, skipping orphan cleanup");
1903 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1904 if (es->s_last_orphan)
1905 jbd_debug(1, "Errors on filesystem, "
1906 "clearing orphan list.\n");
1907 es->s_last_orphan = 0;
1908 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1912 if (s_flags & MS_RDONLY) {
1913 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1914 sb->s_flags &= ~MS_RDONLY;
1917 /* Needed for iput() to work correctly and not trash data */
1918 sb->s_flags |= MS_ACTIVE;
1919 /* Turn on quotas so that they are updated correctly */
1920 for (i = 0; i < MAXQUOTAS; i++) {
1921 if (EXT4_SB(sb)->s_qf_names[i]) {
1922 int ret = ext4_quota_on_mount(sb, i);
1924 ext4_msg(sb, KERN_ERR,
1925 "Cannot turn on journaled "
1926 "quota: error %d", ret);
1931 while (es->s_last_orphan) {
1932 struct inode *inode;
1934 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1935 if (IS_ERR(inode)) {
1936 es->s_last_orphan = 0;
1940 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1942 if (inode->i_nlink) {
1943 ext4_msg(sb, KERN_DEBUG,
1944 "%s: truncating inode %lu to %lld bytes",
1945 __func__, inode->i_ino, inode->i_size);
1946 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1947 inode->i_ino, inode->i_size);
1948 ext4_truncate(inode);
1951 ext4_msg(sb, KERN_DEBUG,
1952 "%s: deleting unreferenced inode %lu",
1953 __func__, inode->i_ino);
1954 jbd_debug(2, "deleting unreferenced inode %lu\n",
1958 iput(inode); /* The delete magic happens here! */
1961 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1964 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1965 PLURAL(nr_orphans));
1967 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1968 PLURAL(nr_truncates));
1970 /* Turn quotas off */
1971 for (i = 0; i < MAXQUOTAS; i++) {
1972 if (sb_dqopt(sb)->files[i])
1973 vfs_quota_off(sb, i, 0);
1976 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1980 * Maximal extent format file size.
1981 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1982 * extent format containers, within a sector_t, and within i_blocks
1983 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1984 * so that won't be a limiting factor.
1986 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1988 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1991 loff_t upper_limit = MAX_LFS_FILESIZE;
1993 /* small i_blocks in vfs inode? */
1994 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1996 * CONFIG_LBDAF is not enabled implies the inode
1997 * i_block represent total blocks in 512 bytes
1998 * 32 == size of vfs inode i_blocks * 8
2000 upper_limit = (1LL << 32) - 1;
2002 /* total blocks in file system block size */
2003 upper_limit >>= (blkbits - 9);
2004 upper_limit <<= blkbits;
2007 /* 32-bit extent-start container, ee_block */
2012 /* Sanity check against vm- & vfs- imposed limits */
2013 if (res > upper_limit)
2020 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2021 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2022 * We need to be 1 filesystem block less than the 2^48 sector limit.
2024 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2026 loff_t res = EXT4_NDIR_BLOCKS;
2029 /* This is calculated to be the largest file size for a dense, block
2030 * mapped file such that the file's total number of 512-byte sectors,
2031 * including data and all indirect blocks, does not exceed (2^48 - 1).
2033 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2034 * number of 512-byte sectors of the file.
2037 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2039 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2040 * the inode i_block field represents total file blocks in
2041 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2043 upper_limit = (1LL << 32) - 1;
2045 /* total blocks in file system block size */
2046 upper_limit >>= (bits - 9);
2050 * We use 48 bit ext4_inode i_blocks
2051 * With EXT4_HUGE_FILE_FL set the i_blocks
2052 * represent total number of blocks in
2053 * file system block size
2055 upper_limit = (1LL << 48) - 1;
2059 /* indirect blocks */
2061 /* double indirect blocks */
2062 meta_blocks += 1 + (1LL << (bits-2));
2063 /* tripple indirect blocks */
2064 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2066 upper_limit -= meta_blocks;
2067 upper_limit <<= bits;
2069 res += 1LL << (bits-2);
2070 res += 1LL << (2*(bits-2));
2071 res += 1LL << (3*(bits-2));
2073 if (res > upper_limit)
2076 if (res > MAX_LFS_FILESIZE)
2077 res = MAX_LFS_FILESIZE;
2082 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2083 ext4_fsblk_t logical_sb_block, int nr)
2085 struct ext4_sb_info *sbi = EXT4_SB(sb);
2086 ext4_group_t bg, first_meta_bg;
2089 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2091 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2093 return logical_sb_block + nr + 1;
2094 bg = sbi->s_desc_per_block * nr;
2095 if (ext4_bg_has_super(sb, bg))
2098 return (has_super + ext4_group_first_block_no(sb, bg));
2102 * ext4_get_stripe_size: Get the stripe size.
2103 * @sbi: In memory super block info
2105 * If we have specified it via mount option, then
2106 * use the mount option value. If the value specified at mount time is
2107 * greater than the blocks per group use the super block value.
2108 * If the super block value is greater than blocks per group return 0.
2109 * Allocator needs it be less than blocks per group.
2112 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2114 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2115 unsigned long stripe_width =
2116 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2118 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2119 return sbi->s_stripe;
2121 if (stripe_width <= sbi->s_blocks_per_group)
2122 return stripe_width;
2124 if (stride <= sbi->s_blocks_per_group)
2133 struct attribute attr;
2134 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2135 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2136 const char *, size_t);
2140 static int parse_strtoul(const char *buf,
2141 unsigned long max, unsigned long *value)
2145 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2146 endp = skip_spaces(endp);
2147 if (*endp || *value > max)
2153 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2154 struct ext4_sb_info *sbi,
2157 return snprintf(buf, PAGE_SIZE, "%llu\n",
2158 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2161 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2162 struct ext4_sb_info *sbi, char *buf)
2164 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2166 return snprintf(buf, PAGE_SIZE, "%lu\n",
2167 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2168 sbi->s_sectors_written_start) >> 1);
2171 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2172 struct ext4_sb_info *sbi, char *buf)
2174 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2176 return snprintf(buf, PAGE_SIZE, "%llu\n",
2177 (unsigned long long)(sbi->s_kbytes_written +
2178 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2179 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2182 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2183 struct ext4_sb_info *sbi,
2184 const char *buf, size_t count)
2188 if (parse_strtoul(buf, 0x40000000, &t))
2191 if (!is_power_of_2(t))
2194 sbi->s_inode_readahead_blks = t;
2198 static ssize_t sbi_ui_show(struct ext4_attr *a,
2199 struct ext4_sb_info *sbi, char *buf)
2201 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2203 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2206 static ssize_t sbi_ui_store(struct ext4_attr *a,
2207 struct ext4_sb_info *sbi,
2208 const char *buf, size_t count)
2210 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2213 if (parse_strtoul(buf, 0xffffffff, &t))
2219 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2220 static struct ext4_attr ext4_attr_##_name = { \
2221 .attr = {.name = __stringify(_name), .mode = _mode }, \
2224 .offset = offsetof(struct ext4_sb_info, _elname), \
2226 #define EXT4_ATTR(name, mode, show, store) \
2227 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2229 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2230 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2231 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2232 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2233 #define ATTR_LIST(name) &ext4_attr_##name.attr
2235 EXT4_RO_ATTR(delayed_allocation_blocks);
2236 EXT4_RO_ATTR(session_write_kbytes);
2237 EXT4_RO_ATTR(lifetime_write_kbytes);
2238 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2239 inode_readahead_blks_store, s_inode_readahead_blks);
2240 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2241 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2242 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2243 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2244 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2245 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2246 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2247 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2249 static struct attribute *ext4_attrs[] = {
2250 ATTR_LIST(delayed_allocation_blocks),
2251 ATTR_LIST(session_write_kbytes),
2252 ATTR_LIST(lifetime_write_kbytes),
2253 ATTR_LIST(inode_readahead_blks),
2254 ATTR_LIST(inode_goal),
2255 ATTR_LIST(mb_stats),
2256 ATTR_LIST(mb_max_to_scan),
2257 ATTR_LIST(mb_min_to_scan),
2258 ATTR_LIST(mb_order2_req),
2259 ATTR_LIST(mb_stream_req),
2260 ATTR_LIST(mb_group_prealloc),
2261 ATTR_LIST(max_writeback_mb_bump),
2265 static ssize_t ext4_attr_show(struct kobject *kobj,
2266 struct attribute *attr, char *buf)
2268 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2270 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2272 return a->show ? a->show(a, sbi, buf) : 0;
2275 static ssize_t ext4_attr_store(struct kobject *kobj,
2276 struct attribute *attr,
2277 const char *buf, size_t len)
2279 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2281 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2283 return a->store ? a->store(a, sbi, buf, len) : 0;
2286 static void ext4_sb_release(struct kobject *kobj)
2288 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2290 complete(&sbi->s_kobj_unregister);
2294 static struct sysfs_ops ext4_attr_ops = {
2295 .show = ext4_attr_show,
2296 .store = ext4_attr_store,
2299 static struct kobj_type ext4_ktype = {
2300 .default_attrs = ext4_attrs,
2301 .sysfs_ops = &ext4_attr_ops,
2302 .release = ext4_sb_release,
2306 * Check whether this filesystem can be mounted based on
2307 * the features present and the RDONLY/RDWR mount requested.
2308 * Returns 1 if this filesystem can be mounted as requested,
2309 * 0 if it cannot be.
2311 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2313 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2314 ext4_msg(sb, KERN_ERR,
2315 "Couldn't mount because of "
2316 "unsupported optional features (%x)",
2317 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2318 ~EXT4_FEATURE_INCOMPAT_SUPP));
2325 /* Check that feature set is OK for a read-write mount */
2326 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2327 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2328 "unsupported optional features (%x)",
2329 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2330 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2334 * Large file size enabled file system can only be mounted
2335 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2338 if (sizeof(blkcnt_t) < sizeof(u64)) {
2339 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2340 "cannot be mounted RDWR without "
2348 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2349 __releases(kernel_lock)
2350 __acquires(kernel_lock)
2352 struct buffer_head *bh;
2353 struct ext4_super_block *es = NULL;
2354 struct ext4_sb_info *sbi;
2356 ext4_fsblk_t sb_block = get_sb_block(&data);
2357 ext4_fsblk_t logical_sb_block;
2358 unsigned long offset = 0;
2359 unsigned long journal_devnum = 0;
2360 unsigned long def_mount_opts;
2366 unsigned int db_count;
2368 int needs_recovery, has_huge_files;
2371 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2373 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2377 sbi->s_blockgroup_lock =
2378 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2379 if (!sbi->s_blockgroup_lock) {
2383 sb->s_fs_info = sbi;
2384 sbi->s_mount_opt = 0;
2385 sbi->s_resuid = EXT4_DEF_RESUID;
2386 sbi->s_resgid = EXT4_DEF_RESGID;
2387 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2388 sbi->s_sb_block = sb_block;
2389 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2394 /* Cleanup superblock name */
2395 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2398 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2400 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2405 * The ext4 superblock will not be buffer aligned for other than 1kB
2406 * block sizes. We need to calculate the offset from buffer start.
2408 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2409 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2410 offset = do_div(logical_sb_block, blocksize);
2412 logical_sb_block = sb_block;
2415 if (!(bh = sb_bread(sb, logical_sb_block))) {
2416 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2420 * Note: s_es must be initialized as soon as possible because
2421 * some ext4 macro-instructions depend on its value
2423 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2425 sb->s_magic = le16_to_cpu(es->s_magic);
2426 if (sb->s_magic != EXT4_SUPER_MAGIC)
2428 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2430 /* Set defaults before we parse the mount options */
2431 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2432 if (def_mount_opts & EXT4_DEFM_DEBUG)
2433 set_opt(sbi->s_mount_opt, DEBUG);
2434 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2435 set_opt(sbi->s_mount_opt, GRPID);
2436 if (def_mount_opts & EXT4_DEFM_UID16)
2437 set_opt(sbi->s_mount_opt, NO_UID32);
2438 #ifdef CONFIG_EXT4_FS_XATTR
2439 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2440 set_opt(sbi->s_mount_opt, XATTR_USER);
2442 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2443 if (def_mount_opts & EXT4_DEFM_ACL)
2444 set_opt(sbi->s_mount_opt, POSIX_ACL);
2446 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2447 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2448 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2449 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2450 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2451 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2453 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2454 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2455 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2456 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2458 set_opt(sbi->s_mount_opt, ERRORS_RO);
2460 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2461 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2462 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2463 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2464 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2466 set_opt(sbi->s_mount_opt, BARRIER);
2469 * enable delayed allocation by default
2470 * Use -o nodelalloc to turn it off
2472 set_opt(sbi->s_mount_opt, DELALLOC);
2474 if (!parse_options((char *) data, sb, &journal_devnum,
2475 &journal_ioprio, NULL, 0))
2478 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2479 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2481 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2482 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2483 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2484 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2485 ext4_msg(sb, KERN_WARNING,
2486 "feature flags set on rev 0 fs, "
2487 "running e2fsck is recommended");
2490 * Check feature flags regardless of the revision level, since we
2491 * previously didn't change the revision level when setting the flags,
2492 * so there is a chance incompat flags are set on a rev 0 filesystem.
2494 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2497 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2499 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2500 blocksize > EXT4_MAX_BLOCK_SIZE) {
2501 ext4_msg(sb, KERN_ERR,
2502 "Unsupported filesystem blocksize %d", blocksize);
2506 if (sb->s_blocksize != blocksize) {
2507 /* Validate the filesystem blocksize */
2508 if (!sb_set_blocksize(sb, blocksize)) {
2509 ext4_msg(sb, KERN_ERR, "bad block size %d",
2515 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2516 offset = do_div(logical_sb_block, blocksize);
2517 bh = sb_bread(sb, logical_sb_block);
2519 ext4_msg(sb, KERN_ERR,
2520 "Can't read superblock on 2nd try");
2523 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2525 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2526 ext4_msg(sb, KERN_ERR,
2527 "Magic mismatch, very weird!");
2532 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2533 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2534 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2536 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2538 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2539 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2540 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2542 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2543 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2544 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2545 (!is_power_of_2(sbi->s_inode_size)) ||
2546 (sbi->s_inode_size > blocksize)) {
2547 ext4_msg(sb, KERN_ERR,
2548 "unsupported inode size: %d",
2552 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2553 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2556 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2557 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2558 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2559 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2560 !is_power_of_2(sbi->s_desc_size)) {
2561 ext4_msg(sb, KERN_ERR,
2562 "unsupported descriptor size %lu",
2567 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2569 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2570 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2571 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2574 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2575 if (sbi->s_inodes_per_block == 0)
2577 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2578 sbi->s_inodes_per_block;
2579 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2581 sbi->s_mount_state = le16_to_cpu(es->s_state);
2582 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2583 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2585 for (i = 0; i < 4; i++)
2586 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2587 sbi->s_def_hash_version = es->s_def_hash_version;
2588 i = le32_to_cpu(es->s_flags);
2589 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2590 sbi->s_hash_unsigned = 3;
2591 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2592 #ifdef __CHAR_UNSIGNED__
2593 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2594 sbi->s_hash_unsigned = 3;
2596 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2601 if (sbi->s_blocks_per_group > blocksize * 8) {
2602 ext4_msg(sb, KERN_ERR,
2603 "#blocks per group too big: %lu",
2604 sbi->s_blocks_per_group);
2607 if (sbi->s_inodes_per_group > blocksize * 8) {
2608 ext4_msg(sb, KERN_ERR,
2609 "#inodes per group too big: %lu",
2610 sbi->s_inodes_per_group);
2615 * Test whether we have more sectors than will fit in sector_t,
2616 * and whether the max offset is addressable by the page cache.
2618 if ((ext4_blocks_count(es) >
2619 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2620 (ext4_blocks_count(es) >
2621 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2622 ext4_msg(sb, KERN_ERR, "filesystem"
2623 " too large to mount safely on this system");
2624 if (sizeof(sector_t) < 8)
2625 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2630 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2633 /* check blocks count against device size */
2634 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2635 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2636 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2637 "exceeds size of device (%llu blocks)",
2638 ext4_blocks_count(es), blocks_count);
2643 * It makes no sense for the first data block to be beyond the end
2644 * of the filesystem.
2646 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2647 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2648 "block %u is beyond end of filesystem (%llu)",
2649 le32_to_cpu(es->s_first_data_block),
2650 ext4_blocks_count(es));
2653 blocks_count = (ext4_blocks_count(es) -
2654 le32_to_cpu(es->s_first_data_block) +
2655 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2656 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2657 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2658 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2659 "(block count %llu, first data block %u, "
2660 "blocks per group %lu)", sbi->s_groups_count,
2661 ext4_blocks_count(es),
2662 le32_to_cpu(es->s_first_data_block),
2663 EXT4_BLOCKS_PER_GROUP(sb));
2666 sbi->s_groups_count = blocks_count;
2667 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2668 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2669 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2670 EXT4_DESC_PER_BLOCK(sb);
2671 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2673 if (sbi->s_group_desc == NULL) {
2674 ext4_msg(sb, KERN_ERR, "not enough memory");
2678 #ifdef CONFIG_PROC_FS
2680 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2683 bgl_lock_init(sbi->s_blockgroup_lock);
2685 for (i = 0; i < db_count; i++) {
2686 block = descriptor_loc(sb, logical_sb_block, i);
2687 sbi->s_group_desc[i] = sb_bread(sb, block);
2688 if (!sbi->s_group_desc[i]) {
2689 ext4_msg(sb, KERN_ERR,
2690 "can't read group descriptor %d", i);
2695 if (!ext4_check_descriptors(sb)) {
2696 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2700 if (!ext4_fill_flex_info(sb)) {
2701 ext4_msg(sb, KERN_ERR,
2702 "unable to initialize "
2703 "flex_bg meta info!");
2707 sbi->s_gdb_count = db_count;
2708 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2709 spin_lock_init(&sbi->s_next_gen_lock);
2711 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2712 ext4_count_free_blocks(sb));
2714 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2715 ext4_count_free_inodes(sb));
2718 err = percpu_counter_init(&sbi->s_dirs_counter,
2719 ext4_count_dirs(sb));
2722 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2725 ext4_msg(sb, KERN_ERR, "insufficient memory");
2729 sbi->s_stripe = ext4_get_stripe_size(sbi);
2730 sbi->s_max_writeback_mb_bump = 128;
2733 * set up enough so that it can read an inode
2735 if (!test_opt(sb, NOLOAD) &&
2736 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2737 sb->s_op = &ext4_sops;
2739 sb->s_op = &ext4_nojournal_sops;
2740 sb->s_export_op = &ext4_export_ops;
2741 sb->s_xattr = ext4_xattr_handlers;
2743 sb->s_qcop = &ext4_qctl_operations;
2744 sb->dq_op = &ext4_quota_operations;
2746 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2747 mutex_init(&sbi->s_orphan_lock);
2748 mutex_init(&sbi->s_resize_lock);
2752 needs_recovery = (es->s_last_orphan != 0 ||
2753 EXT4_HAS_INCOMPAT_FEATURE(sb,
2754 EXT4_FEATURE_INCOMPAT_RECOVER));
2757 * The first inode we look at is the journal inode. Don't try
2758 * root first: it may be modified in the journal!
2760 if (!test_opt(sb, NOLOAD) &&
2761 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2762 if (ext4_load_journal(sb, es, journal_devnum))
2764 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2765 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2766 ext4_msg(sb, KERN_ERR, "required journal recovery "
2767 "suppressed and not mounted read-only");
2770 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2771 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2772 sbi->s_journal = NULL;
2777 if (ext4_blocks_count(es) > 0xffffffffULL &&
2778 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2779 JBD2_FEATURE_INCOMPAT_64BIT)) {
2780 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2784 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2785 jbd2_journal_set_features(sbi->s_journal,
2786 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2787 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2788 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2789 jbd2_journal_set_features(sbi->s_journal,
2790 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2791 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2792 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2794 jbd2_journal_clear_features(sbi->s_journal,
2795 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2796 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2799 /* We have now updated the journal if required, so we can
2800 * validate the data journaling mode. */
2801 switch (test_opt(sb, DATA_FLAGS)) {
2803 /* No mode set, assume a default based on the journal
2804 * capabilities: ORDERED_DATA if the journal can
2805 * cope, else JOURNAL_DATA
2807 if (jbd2_journal_check_available_features
2808 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2809 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2811 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2814 case EXT4_MOUNT_ORDERED_DATA:
2815 case EXT4_MOUNT_WRITEBACK_DATA:
2816 if (!jbd2_journal_check_available_features
2817 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2818 ext4_msg(sb, KERN_ERR, "Journal does not support "
2819 "requested data journaling mode");
2825 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2829 if (test_opt(sb, NOBH)) {
2830 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2831 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2832 "its supported only with writeback mode");
2833 clear_opt(sbi->s_mount_opt, NOBH);
2836 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2837 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2838 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2839 goto failed_mount_wq;
2843 * The jbd2_journal_load will have done any necessary log recovery,
2844 * so we can safely mount the rest of the filesystem now.
2847 root = ext4_iget(sb, EXT4_ROOT_INO);
2849 ext4_msg(sb, KERN_ERR, "get root inode failed");
2850 ret = PTR_ERR(root);
2853 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2855 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2858 sb->s_root = d_alloc_root(root);
2860 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2866 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2868 /* determine the minimum size of new large inodes, if present */
2869 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2870 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2871 EXT4_GOOD_OLD_INODE_SIZE;
2872 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2873 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2874 if (sbi->s_want_extra_isize <
2875 le16_to_cpu(es->s_want_extra_isize))
2876 sbi->s_want_extra_isize =
2877 le16_to_cpu(es->s_want_extra_isize);
2878 if (sbi->s_want_extra_isize <
2879 le16_to_cpu(es->s_min_extra_isize))
2880 sbi->s_want_extra_isize =
2881 le16_to_cpu(es->s_min_extra_isize);
2884 /* Check if enough inode space is available */
2885 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2886 sbi->s_inode_size) {
2887 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2888 EXT4_GOOD_OLD_INODE_SIZE;
2889 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2893 if (test_opt(sb, DELALLOC) &&
2894 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2895 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2896 "requested data journaling mode");
2897 clear_opt(sbi->s_mount_opt, DELALLOC);
2900 err = ext4_setup_system_zone(sb);
2902 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2903 "zone (%d)\n", err);
2908 err = ext4_mb_init(sb, needs_recovery);
2910 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2915 sbi->s_kobj.kset = ext4_kset;
2916 init_completion(&sbi->s_kobj_unregister);
2917 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2920 ext4_mb_release(sb);
2921 ext4_ext_release(sb);
2925 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2926 ext4_orphan_cleanup(sb, es);
2927 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2928 if (needs_recovery) {
2929 ext4_msg(sb, KERN_INFO, "recovery complete");
2930 ext4_mark_recovery_complete(sb, es);
2932 if (EXT4_SB(sb)->s_journal) {
2933 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2934 descr = " journalled data mode";
2935 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2936 descr = " ordered data mode";
2938 descr = " writeback data mode";
2940 descr = "out journal";
2942 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2949 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2953 ext4_msg(sb, KERN_ERR, "mount failed");
2954 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2956 ext4_release_system_zone(sb);
2957 if (sbi->s_journal) {
2958 jbd2_journal_destroy(sbi->s_journal);
2959 sbi->s_journal = NULL;
2962 if (sbi->s_flex_groups) {
2963 if (is_vmalloc_addr(sbi->s_flex_groups))
2964 vfree(sbi->s_flex_groups);
2966 kfree(sbi->s_flex_groups);
2968 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2969 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2970 percpu_counter_destroy(&sbi->s_dirs_counter);
2971 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2973 for (i = 0; i < db_count; i++)
2974 brelse(sbi->s_group_desc[i]);
2975 kfree(sbi->s_group_desc);
2978 remove_proc_entry(sb->s_id, ext4_proc_root);
2981 for (i = 0; i < MAXQUOTAS; i++)
2982 kfree(sbi->s_qf_names[i]);
2984 ext4_blkdev_remove(sbi);
2987 sb->s_fs_info = NULL;
2988 kfree(sbi->s_blockgroup_lock);
2995 * Setup any per-fs journal parameters now. We'll do this both on
2996 * initial mount, once the journal has been initialised but before we've
2997 * done any recovery; and again on any subsequent remount.
2999 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3001 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 journal->j_commit_interval = sbi->s_commit_interval;
3004 journal->j_min_batch_time = sbi->s_min_batch_time;
3005 journal->j_max_batch_time = sbi->s_max_batch_time;
3007 spin_lock(&journal->j_state_lock);
3008 if (test_opt(sb, BARRIER))
3009 journal->j_flags |= JBD2_BARRIER;
3011 journal->j_flags &= ~JBD2_BARRIER;
3012 if (test_opt(sb, DATA_ERR_ABORT))
3013 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3015 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3016 spin_unlock(&journal->j_state_lock);
3019 static journal_t *ext4_get_journal(struct super_block *sb,
3020 unsigned int journal_inum)
3022 struct inode *journal_inode;
3025 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3027 /* First, test for the existence of a valid inode on disk. Bad
3028 * things happen if we iget() an unused inode, as the subsequent
3029 * iput() will try to delete it. */
3031 journal_inode = ext4_iget(sb, journal_inum);
3032 if (IS_ERR(journal_inode)) {
3033 ext4_msg(sb, KERN_ERR, "no journal found");
3036 if (!journal_inode->i_nlink) {
3037 make_bad_inode(journal_inode);
3038 iput(journal_inode);
3039 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3043 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3044 journal_inode, journal_inode->i_size);
3045 if (!S_ISREG(journal_inode->i_mode)) {
3046 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3047 iput(journal_inode);
3051 journal = jbd2_journal_init_inode(journal_inode);
3053 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3054 iput(journal_inode);
3057 journal->j_private = sb;
3058 ext4_init_journal_params(sb, journal);
3062 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3065 struct buffer_head *bh;
3069 int hblock, blocksize;
3070 ext4_fsblk_t sb_block;
3071 unsigned long offset;
3072 struct ext4_super_block *es;
3073 struct block_device *bdev;
3075 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3077 bdev = ext4_blkdev_get(j_dev, sb);
3081 if (bd_claim(bdev, sb)) {
3082 ext4_msg(sb, KERN_ERR,
3083 "failed to claim external journal device");
3084 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3088 blocksize = sb->s_blocksize;
3089 hblock = bdev_logical_block_size(bdev);
3090 if (blocksize < hblock) {
3091 ext4_msg(sb, KERN_ERR,
3092 "blocksize too small for journal device");
3096 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3097 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3098 set_blocksize(bdev, blocksize);
3099 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3100 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3101 "external journal");
3105 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3106 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3107 !(le32_to_cpu(es->s_feature_incompat) &
3108 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3109 ext4_msg(sb, KERN_ERR, "external journal has "
3115 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3116 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3121 len = ext4_blocks_count(es);
3122 start = sb_block + 1;
3123 brelse(bh); /* we're done with the superblock */
3125 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3126 start, len, blocksize);
3128 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3131 journal->j_private = sb;
3132 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3133 wait_on_buffer(journal->j_sb_buffer);
3134 if (!buffer_uptodate(journal->j_sb_buffer)) {
3135 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3138 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3139 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3140 "user (unsupported) - %d",
3141 be32_to_cpu(journal->j_superblock->s_nr_users));
3144 EXT4_SB(sb)->journal_bdev = bdev;
3145 ext4_init_journal_params(sb, journal);
3149 jbd2_journal_destroy(journal);
3151 ext4_blkdev_put(bdev);
3155 static int ext4_load_journal(struct super_block *sb,
3156 struct ext4_super_block *es,
3157 unsigned long journal_devnum)
3160 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3163 int really_read_only;
3165 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3167 if (journal_devnum &&
3168 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3169 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3170 "numbers have changed");
3171 journal_dev = new_decode_dev(journal_devnum);
3173 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3175 really_read_only = bdev_read_only(sb->s_bdev);
3178 * Are we loading a blank journal or performing recovery after a
3179 * crash? For recovery, we need to check in advance whether we
3180 * can get read-write access to the device.
3182 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3183 if (sb->s_flags & MS_RDONLY) {
3184 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3185 "required on readonly filesystem");
3186 if (really_read_only) {
3187 ext4_msg(sb, KERN_ERR, "write access "
3188 "unavailable, cannot proceed");
3191 ext4_msg(sb, KERN_INFO, "write access will "
3192 "be enabled during recovery");
3196 if (journal_inum && journal_dev) {
3197 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3198 "and inode journals!");
3203 if (!(journal = ext4_get_journal(sb, journal_inum)))
3206 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3210 if (!(journal->j_flags & JBD2_BARRIER))
3211 ext4_msg(sb, KERN_INFO, "barriers disabled");
3213 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3214 err = jbd2_journal_update_format(journal);
3216 ext4_msg(sb, KERN_ERR, "error updating journal");
3217 jbd2_journal_destroy(journal);
3222 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3223 err = jbd2_journal_wipe(journal, !really_read_only);
3225 err = jbd2_journal_load(journal);
3228 ext4_msg(sb, KERN_ERR, "error loading journal");
3229 jbd2_journal_destroy(journal);
3233 EXT4_SB(sb)->s_journal = journal;
3234 ext4_clear_journal_err(sb, es);
3236 if (journal_devnum &&
3237 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3238 es->s_journal_dev = cpu_to_le32(journal_devnum);
3240 /* Make sure we flush the recovery flag to disk. */
3241 ext4_commit_super(sb, 1);
3247 static int ext4_commit_super(struct super_block *sb, int sync)
3249 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3250 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3255 if (buffer_write_io_error(sbh)) {
3257 * Oh, dear. A previous attempt to write the
3258 * superblock failed. This could happen because the
3259 * USB device was yanked out. Or it could happen to
3260 * be a transient write error and maybe the block will
3261 * be remapped. Nothing we can do but to retry the
3262 * write and hope for the best.
3264 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3265 "superblock detected");
3266 clear_buffer_write_io_error(sbh);
3267 set_buffer_uptodate(sbh);
3270 * If the file system is mounted read-only, don't update the
3271 * superblock write time. This avoids updating the superblock
3272 * write time when we are mounting the root file system
3273 * read/only but we need to replay the journal; at that point,
3274 * for people who are east of GMT and who make their clock
3275 * tick in localtime for Windows bug-for-bug compatibility,
3276 * the clock is set in the future, and this will cause e2fsck
3277 * to complain and force a full file system check.
3279 if (!(sb->s_flags & MS_RDONLY))
3280 es->s_wtime = cpu_to_le32(get_seconds());
3281 es->s_kbytes_written =
3282 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3283 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3284 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3285 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3286 &EXT4_SB(sb)->s_freeblocks_counter));
3287 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3288 &EXT4_SB(sb)->s_freeinodes_counter));
3290 BUFFER_TRACE(sbh, "marking dirty");
3291 mark_buffer_dirty(sbh);
3293 error = sync_dirty_buffer(sbh);
3297 error = buffer_write_io_error(sbh);
3299 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3301 clear_buffer_write_io_error(sbh);
3302 set_buffer_uptodate(sbh);
3309 * Have we just finished recovery? If so, and if we are mounting (or
3310 * remounting) the filesystem readonly, then we will end up with a
3311 * consistent fs on disk. Record that fact.
3313 static void ext4_mark_recovery_complete(struct super_block *sb,
3314 struct ext4_super_block *es)
3316 journal_t *journal = EXT4_SB(sb)->s_journal;
3318 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3319 BUG_ON(journal != NULL);
3322 jbd2_journal_lock_updates(journal);
3323 if (jbd2_journal_flush(journal) < 0)
3326 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3327 sb->s_flags & MS_RDONLY) {
3328 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3329 ext4_commit_super(sb, 1);
3333 jbd2_journal_unlock_updates(journal);
3337 * If we are mounting (or read-write remounting) a filesystem whose journal
3338 * has recorded an error from a previous lifetime, move that error to the
3339 * main filesystem now.
3341 static void ext4_clear_journal_err(struct super_block *sb,
3342 struct ext4_super_block *es)
3348 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3350 journal = EXT4_SB(sb)->s_journal;
3353 * Now check for any error status which may have been recorded in the
3354 * journal by a prior ext4_error() or ext4_abort()
3357 j_errno = jbd2_journal_errno(journal);
3361 errstr = ext4_decode_error(sb, j_errno, nbuf);
3362 ext4_warning(sb, __func__, "Filesystem error recorded "
3363 "from previous mount: %s", errstr);
3364 ext4_warning(sb, __func__, "Marking fs in need of "
3365 "filesystem check.");
3367 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3368 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3369 ext4_commit_super(sb, 1);
3371 jbd2_journal_clear_err(journal);
3376 * Force the running and committing transactions to commit,
3377 * and wait on the commit.
3379 int ext4_force_commit(struct super_block *sb)
3384 if (sb->s_flags & MS_RDONLY)
3387 journal = EXT4_SB(sb)->s_journal;
3389 ret = ext4_journal_force_commit(journal);
3394 static void ext4_write_super(struct super_block *sb)
3397 ext4_commit_super(sb, 1);
3401 static int ext4_sync_fs(struct super_block *sb, int wait)
3405 struct ext4_sb_info *sbi = EXT4_SB(sb);
3407 trace_ext4_sync_fs(sb, wait);
3408 flush_workqueue(sbi->dio_unwritten_wq);
3409 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3411 jbd2_log_wait_commit(sbi->s_journal, target);
3417 * LVM calls this function before a (read-only) snapshot is created. This
3418 * gives us a chance to flush the journal completely and mark the fs clean.
3420 static int ext4_freeze(struct super_block *sb)
3425 if (sb->s_flags & MS_RDONLY)
3428 journal = EXT4_SB(sb)->s_journal;
3430 /* Now we set up the journal barrier. */
3431 jbd2_journal_lock_updates(journal);
3434 * Don't clear the needs_recovery flag if we failed to flush
3437 error = jbd2_journal_flush(journal);
3440 jbd2_journal_unlock_updates(journal);
3444 /* Journal blocked and flushed, clear needs_recovery flag. */
3445 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3446 error = ext4_commit_super(sb, 1);
3453 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3454 * flag here, even though the filesystem is not technically dirty yet.
3456 static int ext4_unfreeze(struct super_block *sb)
3458 if (sb->s_flags & MS_RDONLY)
3462 /* Reset the needs_recovery flag before the fs is unlocked. */
3463 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3464 ext4_commit_super(sb, 1);
3466 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3470 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3472 struct ext4_super_block *es;
3473 struct ext4_sb_info *sbi = EXT4_SB(sb);
3474 ext4_fsblk_t n_blocks_count = 0;
3475 unsigned long old_sb_flags;
3476 struct ext4_mount_options old_opts;
3478 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3486 /* Store the original options */
3488 old_sb_flags = sb->s_flags;
3489 old_opts.s_mount_opt = sbi->s_mount_opt;
3490 old_opts.s_resuid = sbi->s_resuid;
3491 old_opts.s_resgid = sbi->s_resgid;
3492 old_opts.s_commit_interval = sbi->s_commit_interval;
3493 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3494 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3496 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3497 for (i = 0; i < MAXQUOTAS; i++)
3498 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3500 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3501 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3504 * Allow the "check" option to be passed as a remount option.
3506 if (!parse_options(data, sb, NULL, &journal_ioprio,
3507 &n_blocks_count, 1)) {
3512 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3513 ext4_abort(sb, __func__, "Abort forced by user");
3515 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3516 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3520 if (sbi->s_journal) {
3521 ext4_init_journal_params(sb, sbi->s_journal);
3522 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3525 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3526 n_blocks_count > ext4_blocks_count(es)) {
3527 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3532 if (*flags & MS_RDONLY) {
3534 * First of all, the unconditional stuff we have to do
3535 * to disable replay of the journal when we next remount
3537 sb->s_flags |= MS_RDONLY;
3540 * OK, test if we are remounting a valid rw partition
3541 * readonly, and if so set the rdonly flag and then
3542 * mark the partition as valid again.
3544 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3545 (sbi->s_mount_state & EXT4_VALID_FS))
3546 es->s_state = cpu_to_le16(sbi->s_mount_state);
3549 ext4_mark_recovery_complete(sb, es);
3551 /* Make sure we can mount this feature set readwrite */
3552 if (!ext4_feature_set_ok(sb, 0)) {
3557 * Make sure the group descriptor checksums
3558 * are sane. If they aren't, refuse to remount r/w.
3560 for (g = 0; g < sbi->s_groups_count; g++) {
3561 struct ext4_group_desc *gdp =
3562 ext4_get_group_desc(sb, g, NULL);
3564 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3565 ext4_msg(sb, KERN_ERR,
3566 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3567 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3568 le16_to_cpu(gdp->bg_checksum));
3575 * If we have an unprocessed orphan list hanging
3576 * around from a previously readonly bdev mount,
3577 * require a full umount/remount for now.
3579 if (es->s_last_orphan) {
3580 ext4_msg(sb, KERN_WARNING, "Couldn't "
3581 "remount RDWR because of unprocessed "
3582 "orphan inode list. Please "
3583 "umount/remount instead");
3589 * Mounting a RDONLY partition read-write, so reread
3590 * and store the current valid flag. (It may have
3591 * been changed by e2fsck since we originally mounted
3595 ext4_clear_journal_err(sb, es);
3596 sbi->s_mount_state = le16_to_cpu(es->s_state);
3597 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3599 if (!ext4_setup_super(sb, es, 0))
3600 sb->s_flags &= ~MS_RDONLY;
3603 ext4_setup_system_zone(sb);
3604 if (sbi->s_journal == NULL)
3605 ext4_commit_super(sb, 1);
3608 /* Release old quota file names */
3609 for (i = 0; i < MAXQUOTAS; i++)
3610 if (old_opts.s_qf_names[i] &&
3611 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3612 kfree(old_opts.s_qf_names[i]);
3619 sb->s_flags = old_sb_flags;
3620 sbi->s_mount_opt = old_opts.s_mount_opt;
3621 sbi->s_resuid = old_opts.s_resuid;
3622 sbi->s_resgid = old_opts.s_resgid;
3623 sbi->s_commit_interval = old_opts.s_commit_interval;
3624 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3625 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3627 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3628 for (i = 0; i < MAXQUOTAS; i++) {
3629 if (sbi->s_qf_names[i] &&
3630 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3631 kfree(sbi->s_qf_names[i]);
3632 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3640 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3642 struct super_block *sb = dentry->d_sb;
3643 struct ext4_sb_info *sbi = EXT4_SB(sb);
3644 struct ext4_super_block *es = sbi->s_es;
3647 if (test_opt(sb, MINIX_DF)) {
3648 sbi->s_overhead_last = 0;
3649 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3650 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3651 ext4_fsblk_t overhead = 0;
3654 * Compute the overhead (FS structures). This is constant
3655 * for a given filesystem unless the number of block groups
3656 * changes so we cache the previous value until it does.
3660 * All of the blocks before first_data_block are
3663 overhead = le32_to_cpu(es->s_first_data_block);
3666 * Add the overhead attributed to the superblock and
3667 * block group descriptors. If the sparse superblocks
3668 * feature is turned on, then not all groups have this.
3670 for (i = 0; i < ngroups; i++) {
3671 overhead += ext4_bg_has_super(sb, i) +
3672 ext4_bg_num_gdb(sb, i);
3677 * Every block group has an inode bitmap, a block
3678 * bitmap, and an inode table.
3680 overhead += ngroups * (2 + sbi->s_itb_per_group);
3681 sbi->s_overhead_last = overhead;
3683 sbi->s_blocks_last = ext4_blocks_count(es);
3686 buf->f_type = EXT4_SUPER_MAGIC;
3687 buf->f_bsize = sb->s_blocksize;
3688 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3689 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3690 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3691 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3692 if (buf->f_bfree < ext4_r_blocks_count(es))
3694 buf->f_files = le32_to_cpu(es->s_inodes_count);
3695 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3696 buf->f_namelen = EXT4_NAME_LEN;
3697 fsid = le64_to_cpup((void *)es->s_uuid) ^
3698 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3699 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3700 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3705 /* Helper function for writing quotas on sync - we need to start transaction
3706 * before quota file is locked for write. Otherwise the are possible deadlocks:
3707 * Process 1 Process 2
3708 * ext4_create() quota_sync()
3709 * jbd2_journal_start() write_dquot()
3710 * vfs_dq_init() down(dqio_mutex)
3711 * down(dqio_mutex) jbd2_journal_start()
3717 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3719 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3722 static int ext4_write_dquot(struct dquot *dquot)
3726 struct inode *inode;
3728 inode = dquot_to_inode(dquot);
3729 handle = ext4_journal_start(inode,
3730 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3732 return PTR_ERR(handle);
3733 ret = dquot_commit(dquot);
3734 err = ext4_journal_stop(handle);
3740 static int ext4_acquire_dquot(struct dquot *dquot)
3745 handle = ext4_journal_start(dquot_to_inode(dquot),
3746 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3748 return PTR_ERR(handle);
3749 ret = dquot_acquire(dquot);
3750 err = ext4_journal_stop(handle);
3756 static int ext4_release_dquot(struct dquot *dquot)
3761 handle = ext4_journal_start(dquot_to_inode(dquot),
3762 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3763 if (IS_ERR(handle)) {
3764 /* Release dquot anyway to avoid endless cycle in dqput() */
3765 dquot_release(dquot);
3766 return PTR_ERR(handle);
3768 ret = dquot_release(dquot);
3769 err = ext4_journal_stop(handle);
3775 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3777 /* Are we journaling quotas? */
3778 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3779 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3780 dquot_mark_dquot_dirty(dquot);
3781 return ext4_write_dquot(dquot);
3783 return dquot_mark_dquot_dirty(dquot);
3787 static int ext4_write_info(struct super_block *sb, int type)
3792 /* Data block + inode block */
3793 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3795 return PTR_ERR(handle);
3796 ret = dquot_commit_info(sb, type);
3797 err = ext4_journal_stop(handle);
3804 * Turn on quotas during mount time - we need to find
3805 * the quota file and such...
3807 static int ext4_quota_on_mount(struct super_block *sb, int type)
3809 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3810 EXT4_SB(sb)->s_jquota_fmt, type);
3814 * Standard function to be called on quota_on
3816 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3817 char *name, int remount)
3822 if (!test_opt(sb, QUOTA))
3824 /* When remounting, no checks are needed and in fact, name is NULL */
3826 return vfs_quota_on(sb, type, format_id, name, remount);
3828 err = kern_path(name, LOOKUP_FOLLOW, &path);
3832 /* Quotafile not on the same filesystem? */
3833 if (path.mnt->mnt_sb != sb) {
3837 /* Journaling quota? */
3838 if (EXT4_SB(sb)->s_qf_names[type]) {
3839 /* Quotafile not in fs root? */
3840 if (path.dentry->d_parent != sb->s_root)
3841 ext4_msg(sb, KERN_WARNING,
3842 "Quota file not on filesystem root. "
3843 "Journaled quota will not work");
3847 * When we journal data on quota file, we have to flush journal to see
3848 * all updates to the file when we bypass pagecache...
3850 if (EXT4_SB(sb)->s_journal &&
3851 ext4_should_journal_data(path.dentry->d_inode)) {
3853 * We don't need to lock updates but journal_flush() could
3854 * otherwise be livelocked...
3856 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3857 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3858 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3865 err = vfs_quota_on_path(sb, type, format_id, &path);
3870 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3871 * acquiring the locks... As quota files are never truncated and quota code
3872 * itself serializes the operations (and noone else should touch the files)
3873 * we don't have to be afraid of races */
3874 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3875 size_t len, loff_t off)
3877 struct inode *inode = sb_dqopt(sb)->files[type];
3878 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3880 int offset = off & (sb->s_blocksize - 1);
3883 struct buffer_head *bh;
3884 loff_t i_size = i_size_read(inode);
3888 if (off+len > i_size)
3891 while (toread > 0) {
3892 tocopy = sb->s_blocksize - offset < toread ?
3893 sb->s_blocksize - offset : toread;
3894 bh = ext4_bread(NULL, inode, blk, 0, &err);
3897 if (!bh) /* A hole? */
3898 memset(data, 0, tocopy);
3900 memcpy(data, bh->b_data+offset, tocopy);
3910 /* Write to quotafile (we know the transaction is already started and has
3911 * enough credits) */
3912 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3913 const char *data, size_t len, loff_t off)
3915 struct inode *inode = sb_dqopt(sb)->files[type];
3916 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3918 int offset = off & (sb->s_blocksize - 1);
3920 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3921 size_t towrite = len;
3922 struct buffer_head *bh;
3923 handle_t *handle = journal_current_handle();
3925 if (EXT4_SB(sb)->s_journal && !handle) {
3926 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3927 " cancelled because transaction is not started",
3928 (unsigned long long)off, (unsigned long long)len);
3931 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3932 while (towrite > 0) {
3933 tocopy = sb->s_blocksize - offset < towrite ?
3934 sb->s_blocksize - offset : towrite;
3935 bh = ext4_bread(handle, inode, blk, 1, &err);
3938 if (journal_quota) {
3939 err = ext4_journal_get_write_access(handle, bh);
3946 memcpy(bh->b_data+offset, data, tocopy);
3947 flush_dcache_page(bh->b_page);
3950 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3952 /* Always do at least ordered writes for quotas */
3953 err = ext4_jbd2_file_inode(handle, inode);
3954 mark_buffer_dirty(bh);
3965 if (len == towrite) {
3966 mutex_unlock(&inode->i_mutex);
3969 if (inode->i_size < off+len-towrite) {
3970 i_size_write(inode, off+len-towrite);
3971 EXT4_I(inode)->i_disksize = inode->i_size;
3973 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3974 ext4_mark_inode_dirty(handle, inode);
3975 mutex_unlock(&inode->i_mutex);
3976 return len - towrite;
3981 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3982 const char *dev_name, void *data, struct vfsmount *mnt)
3984 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3987 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
3988 static struct file_system_type ext2_fs_type = {
3989 .owner = THIS_MODULE,
3991 .get_sb = ext4_get_sb,
3992 .kill_sb = kill_block_super,
3993 .fs_flags = FS_REQUIRES_DEV,
3996 static inline void register_as_ext2(void)
3998 int err = register_filesystem(&ext2_fs_type);
4001 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4004 static inline void unregister_as_ext2(void)
4006 unregister_filesystem(&ext2_fs_type);
4008 MODULE_ALIAS("ext2");
4010 static inline void register_as_ext2(void) { }
4011 static inline void unregister_as_ext2(void) { }
4014 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4015 static struct file_system_type ext3_fs_type = {
4016 .owner = THIS_MODULE,
4018 .get_sb = ext4_get_sb,
4019 .kill_sb = kill_block_super,
4020 .fs_flags = FS_REQUIRES_DEV,
4023 static inline void register_as_ext3(void)
4025 int err = register_filesystem(&ext3_fs_type);
4028 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4031 static inline void unregister_as_ext3(void)
4033 unregister_filesystem(&ext3_fs_type);
4035 MODULE_ALIAS("ext3");
4037 static inline void register_as_ext3(void) { }
4038 static inline void unregister_as_ext3(void) { }
4041 static struct file_system_type ext4_fs_type = {
4042 .owner = THIS_MODULE,
4044 .get_sb = ext4_get_sb,
4045 .kill_sb = kill_block_super,
4046 .fs_flags = FS_REQUIRES_DEV,
4049 static int __init init_ext4_fs(void)
4053 err = init_ext4_system_zone();
4056 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4059 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4060 err = init_ext4_mballoc();
4064 err = init_ext4_xattr();
4067 err = init_inodecache();
4072 err = register_filesystem(&ext4_fs_type);
4077 unregister_as_ext2();
4078 unregister_as_ext3();
4079 destroy_inodecache();
4083 exit_ext4_mballoc();
4085 remove_proc_entry("fs/ext4", NULL);
4086 kset_unregister(ext4_kset);
4088 exit_ext4_system_zone();
4092 static void __exit exit_ext4_fs(void)
4094 unregister_as_ext2();
4095 unregister_as_ext3();
4096 unregister_filesystem(&ext4_fs_type);
4097 destroy_inodecache();
4099 exit_ext4_mballoc();
4100 remove_proc_entry("fs/ext4", NULL);
4101 kset_unregister(ext4_kset);
4102 exit_ext4_system_zone();
4105 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4106 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4107 MODULE_LICENSE("GPL");
4108 module_init(init_ext4_fs)
4109 module_exit(exit_ext4_fs)