2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
195 handle = (handle_t *)ref_cnt;
197 current->journal_info = handle;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
205 unsigned long ref_cnt = (unsigned long)handle;
207 BUG_ON(ref_cnt == 0);
210 handle = (handle_t *)ref_cnt;
212 current->journal_info = handle;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
239 return jbd2_journal_start(journal, nblocks);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where, handle_t *handle)
252 struct super_block *sb;
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
260 sb = handle->h_transaction->t_journal->j_private;
262 rc = jbd2_journal_stop(handle);
267 __ext4_std_error(sb, where, err);
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
277 BUG_ON(!ext4_handle_valid(handle));
280 BUFFER_TRACE(bh, "abort");
285 if (is_handle_aborted(handle))
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
291 jbd2_journal_abort_handle(handle);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block *sb)
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316 if (sb->s_flags & MS_RDONLY)
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
324 jbd2_journal_abort(journal, -EIO);
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
347 ext4_handle_error(sb);
350 static const char *ext4_decode_error(struct super_block *sb, int errno,
357 errstr = "IO failure";
360 errstr = "Out of memory";
363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
365 errstr = "Journal has aborted";
367 errstr = "Readonly filesystem";
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
384 /* __ext4_std_error decodes expected errors from journaling functions
385 * automatically and invokes the appropriate error response. */
387 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
399 errstr = ext4_decode_error(sb, errno, nbuf);
400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
403 ext4_handle_error(sb);
407 * ext4_abort is a much stronger failure handler than ext4_error. The
408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
416 void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
427 if (test_opt(sb, ERRORS_PANIC))
428 panic("EXT4-fs panic from previous error\n");
430 if (sb->s_flags & MS_RDONLY)
433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
435 sb->s_flags |= MS_RDONLY;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
441 void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
453 void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
466 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
467 const char *function, const char *fmt, ...)
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
483 ext4_commit_super(sb, 0);
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
499 ext4_lock_group(sb, grp);
503 void ext4_update_dynamic_rev(struct super_block *sb)
505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
510 ext4_warning(sb, __func__,
511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
529 * Open the external journal device
531 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
543 __bdevname(dev, b), PTR_ERR(bdev));
548 * Release the journal device
550 static int ext4_blkdev_put(struct block_device *bdev)
553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
556 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
558 struct block_device *bdev;
561 bdev = sbi->journal_bdev;
563 ret = ext4_blkdev_put(bdev);
564 sbi->journal_bdev = NULL;
569 static inline struct inode *orphan_list_entry(struct list_head *l)
571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
574 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
592 static void ext4_put_super(struct super_block *sb)
594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
604 ext4_commit_super(sb, 1);
606 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
614 ext4_release_system_zone(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
619 if (!(sb->s_flags & MS_RDONLY)) {
620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
621 es->s_state = cpu_to_le16(sbi->s_mount_state);
622 ext4_commit_super(sb, 1);
625 remove_proc_entry(sb->s_id, ext4_proc_root);
627 kobject_del(&sbi->s_kobj);
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
635 kfree(sbi->s_flex_groups);
636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
654 invalidate_bdev(sb->s_bdev);
655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
661 sync_blockdev(sbi->journal_bdev);
662 invalidate_bdev(sbi->journal_bdev);
663 ext4_blkdev_remove(sbi);
665 sb->s_fs_info = NULL;
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
674 kfree(sbi->s_blockgroup_lock);
678 static struct kmem_cache *ext4_inode_cachep;
681 * Called inside transaction, so use GFP_NOFS
683 static struct inode *ext4_alloc_inode(struct super_block *sb)
685 struct ext4_inode_info *ei;
687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
691 ei->vfs_inode.i_version = 1;
692 ei->vfs_inode.i_data.writeback_index = 0;
693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
705 ei->i_delalloc_reserved_flag = 0;
706 spin_lock_init(&(ei->i_block_reservation_lock));
707 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
708 ei->cur_aio_dio = NULL;
710 return &ei->vfs_inode;
713 static void ext4_destroy_inode(struct inode *inode)
715 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
716 ext4_msg(inode->i_sb, KERN_ERR,
717 "Inode %lu (%p): orphan list check failed!",
718 inode->i_ino, EXT4_I(inode));
719 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
720 EXT4_I(inode), sizeof(struct ext4_inode_info),
724 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
727 static void init_once(void *foo)
729 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
731 INIT_LIST_HEAD(&ei->i_orphan);
732 #ifdef CONFIG_EXT4_FS_XATTR
733 init_rwsem(&ei->xattr_sem);
735 init_rwsem(&ei->i_data_sem);
736 inode_init_once(&ei->vfs_inode);
739 static int init_inodecache(void)
741 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
742 sizeof(struct ext4_inode_info),
743 0, (SLAB_RECLAIM_ACCOUNT|
746 if (ext4_inode_cachep == NULL)
751 static void destroy_inodecache(void)
753 kmem_cache_destroy(ext4_inode_cachep);
756 static void ext4_clear_inode(struct inode *inode)
758 ext4_discard_preallocations(inode);
759 if (EXT4_JOURNAL(inode))
760 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
761 &EXT4_I(inode)->jinode);
764 static inline void ext4_show_quota_options(struct seq_file *seq,
765 struct super_block *sb)
767 #if defined(CONFIG_QUOTA)
768 struct ext4_sb_info *sbi = EXT4_SB(sb);
770 if (sbi->s_jquota_fmt)
771 seq_printf(seq, ",jqfmt=%s",
772 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
774 if (sbi->s_qf_names[USRQUOTA])
775 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
777 if (sbi->s_qf_names[GRPQUOTA])
778 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
780 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
781 seq_puts(seq, ",usrquota");
783 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
784 seq_puts(seq, ",grpquota");
790 * - it's set to a non-default value OR
791 * - if the per-sb default is different from the global default
793 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
796 unsigned long def_mount_opts;
797 struct super_block *sb = vfs->mnt_sb;
798 struct ext4_sb_info *sbi = EXT4_SB(sb);
799 struct ext4_super_block *es = sbi->s_es;
801 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
802 def_errors = le16_to_cpu(es->s_errors);
804 if (sbi->s_sb_block != 1)
805 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
806 if (test_opt(sb, MINIX_DF))
807 seq_puts(seq, ",minixdf");
808 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",grpid");
810 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
811 seq_puts(seq, ",nogrpid");
812 if (sbi->s_resuid != EXT4_DEF_RESUID ||
813 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
814 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
816 if (sbi->s_resgid != EXT4_DEF_RESGID ||
817 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
818 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
820 if (test_opt(sb, ERRORS_RO)) {
821 if (def_errors == EXT4_ERRORS_PANIC ||
822 def_errors == EXT4_ERRORS_CONTINUE) {
823 seq_puts(seq, ",errors=remount-ro");
826 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
827 seq_puts(seq, ",errors=continue");
828 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
829 seq_puts(seq, ",errors=panic");
830 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
831 seq_puts(seq, ",nouid32");
832 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
833 seq_puts(seq, ",debug");
834 if (test_opt(sb, OLDALLOC))
835 seq_puts(seq, ",oldalloc");
836 #ifdef CONFIG_EXT4_FS_XATTR
837 if (test_opt(sb, XATTR_USER) &&
838 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
839 seq_puts(seq, ",user_xattr");
840 if (!test_opt(sb, XATTR_USER) &&
841 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
842 seq_puts(seq, ",nouser_xattr");
845 #ifdef CONFIG_EXT4_FS_POSIX_ACL
846 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",acl");
848 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
849 seq_puts(seq, ",noacl");
851 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
852 seq_printf(seq, ",commit=%u",
853 (unsigned) (sbi->s_commit_interval / HZ));
855 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
856 seq_printf(seq, ",min_batch_time=%u",
857 (unsigned) sbi->s_min_batch_time);
859 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
860 seq_printf(seq, ",max_batch_time=%u",
861 (unsigned) sbi->s_min_batch_time);
865 * We're changing the default of barrier mount option, so
866 * let's always display its mount state so it's clear what its
869 seq_puts(seq, ",barrier=");
870 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
871 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
872 seq_puts(seq, ",journal_async_commit");
873 if (test_opt(sb, NOBH))
874 seq_puts(seq, ",nobh");
875 if (test_opt(sb, I_VERSION))
876 seq_puts(seq, ",i_version");
877 if (!test_opt(sb, DELALLOC))
878 seq_puts(seq, ",nodelalloc");
882 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
884 * journal mode get enabled in different ways
885 * So just print the value even if we didn't specify it
887 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
888 seq_puts(seq, ",data=journal");
889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
890 seq_puts(seq, ",data=ordered");
891 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
892 seq_puts(seq, ",data=writeback");
894 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
895 seq_printf(seq, ",inode_readahead_blks=%u",
896 sbi->s_inode_readahead_blks);
898 if (test_opt(sb, DATA_ERR_ABORT))
899 seq_puts(seq, ",data_err=abort");
901 if (test_opt(sb, NO_AUTO_DA_ALLOC))
902 seq_puts(seq, ",noauto_da_alloc");
904 if (test_opt(sb, DISCARD))
905 seq_puts(seq, ",discard");
907 if (test_opt(sb, NOLOAD))
908 seq_puts(seq, ",norecovery");
910 ext4_show_quota_options(seq, sb);
915 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
916 u64 ino, u32 generation)
920 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
921 return ERR_PTR(-ESTALE);
922 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
923 return ERR_PTR(-ESTALE);
925 /* iget isn't really right if the inode is currently unallocated!!
927 * ext4_read_inode will return a bad_inode if the inode had been
928 * deleted, so we should be safe.
930 * Currently we don't know the generation for parent directory, so
931 * a generation of 0 means "accept any"
933 inode = ext4_iget(sb, ino);
935 return ERR_CAST(inode);
936 if (generation && inode->i_generation != generation) {
938 return ERR_PTR(-ESTALE);
944 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
945 int fh_len, int fh_type)
947 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
951 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
952 int fh_len, int fh_type)
954 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
959 * Try to release metadata pages (indirect blocks, directories) which are
960 * mapped via the block device. Since these pages could have journal heads
961 * which would prevent try_to_free_buffers() from freeing them, we must use
962 * jbd2 layer's try_to_free_buffers() function to release them.
964 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
967 journal_t *journal = EXT4_SB(sb)->s_journal;
969 WARN_ON(PageChecked(page));
970 if (!page_has_buffers(page))
973 return jbd2_journal_try_to_free_buffers(journal, page,
975 return try_to_free_buffers(page);
979 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
980 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
982 static int ext4_write_dquot(struct dquot *dquot);
983 static int ext4_acquire_dquot(struct dquot *dquot);
984 static int ext4_release_dquot(struct dquot *dquot);
985 static int ext4_mark_dquot_dirty(struct dquot *dquot);
986 static int ext4_write_info(struct super_block *sb, int type);
987 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
988 char *path, int remount);
989 static int ext4_quota_on_mount(struct super_block *sb, int type);
990 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
991 size_t len, loff_t off);
992 static ssize_t ext4_quota_write(struct super_block *sb, int type,
993 const char *data, size_t len, loff_t off);
995 static const struct dquot_operations ext4_quota_operations = {
996 .initialize = dquot_initialize,
998 .alloc_space = dquot_alloc_space,
999 .reserve_space = dquot_reserve_space,
1000 .claim_space = dquot_claim_space,
1001 .release_rsv = dquot_release_reserved_space,
1002 .get_reserved_space = ext4_get_reserved_space,
1003 .alloc_inode = dquot_alloc_inode,
1004 .free_space = dquot_free_space,
1005 .free_inode = dquot_free_inode,
1006 .transfer = dquot_transfer,
1007 .write_dquot = ext4_write_dquot,
1008 .acquire_dquot = ext4_acquire_dquot,
1009 .release_dquot = ext4_release_dquot,
1010 .mark_dirty = ext4_mark_dquot_dirty,
1011 .write_info = ext4_write_info,
1012 .alloc_dquot = dquot_alloc,
1013 .destroy_dquot = dquot_destroy,
1016 static const struct quotactl_ops ext4_qctl_operations = {
1017 .quota_on = ext4_quota_on,
1018 .quota_off = vfs_quota_off,
1019 .quota_sync = vfs_quota_sync,
1020 .get_info = vfs_get_dqinfo,
1021 .set_info = vfs_set_dqinfo,
1022 .get_dqblk = vfs_get_dqblk,
1023 .set_dqblk = vfs_set_dqblk
1027 static const struct super_operations ext4_sops = {
1028 .alloc_inode = ext4_alloc_inode,
1029 .destroy_inode = ext4_destroy_inode,
1030 .write_inode = ext4_write_inode,
1031 .dirty_inode = ext4_dirty_inode,
1032 .delete_inode = ext4_delete_inode,
1033 .put_super = ext4_put_super,
1034 .sync_fs = ext4_sync_fs,
1035 .freeze_fs = ext4_freeze,
1036 .unfreeze_fs = ext4_unfreeze,
1037 .statfs = ext4_statfs,
1038 .remount_fs = ext4_remount,
1039 .clear_inode = ext4_clear_inode,
1040 .show_options = ext4_show_options,
1042 .quota_read = ext4_quota_read,
1043 .quota_write = ext4_quota_write,
1045 .bdev_try_to_free_page = bdev_try_to_free_page,
1048 static const struct super_operations ext4_nojournal_sops = {
1049 .alloc_inode = ext4_alloc_inode,
1050 .destroy_inode = ext4_destroy_inode,
1051 .write_inode = ext4_write_inode,
1052 .dirty_inode = ext4_dirty_inode,
1053 .delete_inode = ext4_delete_inode,
1054 .write_super = ext4_write_super,
1055 .put_super = ext4_put_super,
1056 .statfs = ext4_statfs,
1057 .remount_fs = ext4_remount,
1058 .clear_inode = ext4_clear_inode,
1059 .show_options = ext4_show_options,
1061 .quota_read = ext4_quota_read,
1062 .quota_write = ext4_quota_write,
1064 .bdev_try_to_free_page = bdev_try_to_free_page,
1067 static const struct export_operations ext4_export_ops = {
1068 .fh_to_dentry = ext4_fh_to_dentry,
1069 .fh_to_parent = ext4_fh_to_parent,
1070 .get_parent = ext4_get_parent,
1074 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1075 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1076 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1077 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1078 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1079 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1080 Opt_journal_update, Opt_journal_dev,
1081 Opt_journal_checksum, Opt_journal_async_commit,
1082 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1083 Opt_data_err_abort, Opt_data_err_ignore,
1084 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1085 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1086 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1087 Opt_usrquota, Opt_grpquota, Opt_i_version,
1088 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1089 Opt_block_validity, Opt_noblock_validity,
1090 Opt_inode_readahead_blks, Opt_journal_ioprio,
1091 Opt_discard, Opt_nodiscard,
1094 static const match_table_t tokens = {
1095 {Opt_bsd_df, "bsddf"},
1096 {Opt_minix_df, "minixdf"},
1097 {Opt_grpid, "grpid"},
1098 {Opt_grpid, "bsdgroups"},
1099 {Opt_nogrpid, "nogrpid"},
1100 {Opt_nogrpid, "sysvgroups"},
1101 {Opt_resgid, "resgid=%u"},
1102 {Opt_resuid, "resuid=%u"},
1104 {Opt_err_cont, "errors=continue"},
1105 {Opt_err_panic, "errors=panic"},
1106 {Opt_err_ro, "errors=remount-ro"},
1107 {Opt_nouid32, "nouid32"},
1108 {Opt_debug, "debug"},
1109 {Opt_oldalloc, "oldalloc"},
1110 {Opt_orlov, "orlov"},
1111 {Opt_user_xattr, "user_xattr"},
1112 {Opt_nouser_xattr, "nouser_xattr"},
1114 {Opt_noacl, "noacl"},
1115 {Opt_noload, "noload"},
1116 {Opt_noload, "norecovery"},
1119 {Opt_commit, "commit=%u"},
1120 {Opt_min_batch_time, "min_batch_time=%u"},
1121 {Opt_max_batch_time, "max_batch_time=%u"},
1122 {Opt_journal_update, "journal=update"},
1123 {Opt_journal_dev, "journal_dev=%u"},
1124 {Opt_journal_checksum, "journal_checksum"},
1125 {Opt_journal_async_commit, "journal_async_commit"},
1126 {Opt_abort, "abort"},
1127 {Opt_data_journal, "data=journal"},
1128 {Opt_data_ordered, "data=ordered"},
1129 {Opt_data_writeback, "data=writeback"},
1130 {Opt_data_err_abort, "data_err=abort"},
1131 {Opt_data_err_ignore, "data_err=ignore"},
1132 {Opt_offusrjquota, "usrjquota="},
1133 {Opt_usrjquota, "usrjquota=%s"},
1134 {Opt_offgrpjquota, "grpjquota="},
1135 {Opt_grpjquota, "grpjquota=%s"},
1136 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1137 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1138 {Opt_grpquota, "grpquota"},
1139 {Opt_noquota, "noquota"},
1140 {Opt_quota, "quota"},
1141 {Opt_usrquota, "usrquota"},
1142 {Opt_barrier, "barrier=%u"},
1143 {Opt_barrier, "barrier"},
1144 {Opt_nobarrier, "nobarrier"},
1145 {Opt_i_version, "i_version"},
1146 {Opt_stripe, "stripe=%u"},
1147 {Opt_resize, "resize"},
1148 {Opt_delalloc, "delalloc"},
1149 {Opt_nodelalloc, "nodelalloc"},
1150 {Opt_block_validity, "block_validity"},
1151 {Opt_noblock_validity, "noblock_validity"},
1152 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1153 {Opt_journal_ioprio, "journal_ioprio=%u"},
1154 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1155 {Opt_auto_da_alloc, "auto_da_alloc"},
1156 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1157 {Opt_discard, "discard"},
1158 {Opt_nodiscard, "nodiscard"},
1162 static ext4_fsblk_t get_sb_block(void **data)
1164 ext4_fsblk_t sb_block;
1165 char *options = (char *) *data;
1167 if (!options || strncmp(options, "sb=", 3) != 0)
1168 return 1; /* Default location */
1171 /* TODO: use simple_strtoll with >32bit ext4 */
1172 sb_block = simple_strtoul(options, &options, 0);
1173 if (*options && *options != ',') {
1174 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1178 if (*options == ',')
1180 *data = (void *) options;
1185 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1187 static int parse_options(char *options, struct super_block *sb,
1188 unsigned long *journal_devnum,
1189 unsigned int *journal_ioprio,
1190 ext4_fsblk_t *n_blocks_count, int is_remount)
1192 struct ext4_sb_info *sbi = EXT4_SB(sb);
1194 substring_t args[MAX_OPT_ARGS];
1205 while ((p = strsep(&options, ",")) != NULL) {
1210 token = match_token(p, tokens, args);
1213 clear_opt(sbi->s_mount_opt, MINIX_DF);
1216 set_opt(sbi->s_mount_opt, MINIX_DF);
1219 set_opt(sbi->s_mount_opt, GRPID);
1222 clear_opt(sbi->s_mount_opt, GRPID);
1225 if (match_int(&args[0], &option))
1227 sbi->s_resuid = option;
1230 if (match_int(&args[0], &option))
1232 sbi->s_resgid = option;
1235 /* handled by get_sb_block() instead of here */
1236 /* *sb_block = match_int(&args[0]); */
1239 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1240 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1241 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1244 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1245 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1246 set_opt(sbi->s_mount_opt, ERRORS_RO);
1249 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1250 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1251 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1254 set_opt(sbi->s_mount_opt, NO_UID32);
1257 set_opt(sbi->s_mount_opt, DEBUG);
1260 set_opt(sbi->s_mount_opt, OLDALLOC);
1263 clear_opt(sbi->s_mount_opt, OLDALLOC);
1265 #ifdef CONFIG_EXT4_FS_XATTR
1266 case Opt_user_xattr:
1267 set_opt(sbi->s_mount_opt, XATTR_USER);
1269 case Opt_nouser_xattr:
1270 clear_opt(sbi->s_mount_opt, XATTR_USER);
1273 case Opt_user_xattr:
1274 case Opt_nouser_xattr:
1275 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1278 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1280 set_opt(sbi->s_mount_opt, POSIX_ACL);
1283 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1288 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1291 case Opt_journal_update:
1293 /* Eventually we will want to be able to create
1294 a journal file here. For now, only allow the
1295 user to specify an existing inode to be the
1298 ext4_msg(sb, KERN_ERR,
1299 "Cannot specify journal on remount");
1302 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1304 case Opt_journal_dev:
1306 ext4_msg(sb, KERN_ERR,
1307 "Cannot specify journal on remount");
1310 if (match_int(&args[0], &option))
1312 *journal_devnum = option;
1314 case Opt_journal_checksum:
1315 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1317 case Opt_journal_async_commit:
1318 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1319 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1322 set_opt(sbi->s_mount_opt, NOLOAD);
1325 if (match_int(&args[0], &option))
1330 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1331 sbi->s_commit_interval = HZ * option;
1333 case Opt_max_batch_time:
1334 if (match_int(&args[0], &option))
1339 option = EXT4_DEF_MAX_BATCH_TIME;
1340 sbi->s_max_batch_time = option;
1342 case Opt_min_batch_time:
1343 if (match_int(&args[0], &option))
1347 sbi->s_min_batch_time = option;
1349 case Opt_data_journal:
1350 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1352 case Opt_data_ordered:
1353 data_opt = EXT4_MOUNT_ORDERED_DATA;
1355 case Opt_data_writeback:
1356 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1359 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1361 ext4_msg(sb, KERN_ERR,
1362 "Cannot change data mode on remount");
1366 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1367 sbi->s_mount_opt |= data_opt;
1370 case Opt_data_err_abort:
1371 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1373 case Opt_data_err_ignore:
1374 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1383 if (sb_any_quota_loaded(sb) &&
1384 !sbi->s_qf_names[qtype]) {
1385 ext4_msg(sb, KERN_ERR,
1386 "Cannot change journaled "
1387 "quota options when quota turned on");
1390 qname = match_strdup(&args[0]);
1392 ext4_msg(sb, KERN_ERR,
1393 "Not enough memory for "
1394 "storing quotafile name");
1397 if (sbi->s_qf_names[qtype] &&
1398 strcmp(sbi->s_qf_names[qtype], qname)) {
1399 ext4_msg(sb, KERN_ERR,
1400 "%s quota file already "
1401 "specified", QTYPE2NAME(qtype));
1405 sbi->s_qf_names[qtype] = qname;
1406 if (strchr(sbi->s_qf_names[qtype], '/')) {
1407 ext4_msg(sb, KERN_ERR,
1408 "quotafile must be on "
1410 kfree(sbi->s_qf_names[qtype]);
1411 sbi->s_qf_names[qtype] = NULL;
1414 set_opt(sbi->s_mount_opt, QUOTA);
1416 case Opt_offusrjquota:
1419 case Opt_offgrpjquota:
1422 if (sb_any_quota_loaded(sb) &&
1423 sbi->s_qf_names[qtype]) {
1424 ext4_msg(sb, KERN_ERR, "Cannot change "
1425 "journaled quota options when "
1430 * The space will be released later when all options
1431 * are confirmed to be correct
1433 sbi->s_qf_names[qtype] = NULL;
1435 case Opt_jqfmt_vfsold:
1436 qfmt = QFMT_VFS_OLD;
1438 case Opt_jqfmt_vfsv0:
1441 if (sb_any_quota_loaded(sb) &&
1442 sbi->s_jquota_fmt != qfmt) {
1443 ext4_msg(sb, KERN_ERR, "Cannot change "
1444 "journaled quota options when "
1448 sbi->s_jquota_fmt = qfmt;
1452 set_opt(sbi->s_mount_opt, QUOTA);
1453 set_opt(sbi->s_mount_opt, USRQUOTA);
1456 set_opt(sbi->s_mount_opt, QUOTA);
1457 set_opt(sbi->s_mount_opt, GRPQUOTA);
1460 if (sb_any_quota_loaded(sb)) {
1461 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1462 "options when quota turned on");
1465 clear_opt(sbi->s_mount_opt, QUOTA);
1466 clear_opt(sbi->s_mount_opt, USRQUOTA);
1467 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1473 ext4_msg(sb, KERN_ERR,
1474 "quota options not supported");
1478 case Opt_offusrjquota:
1479 case Opt_offgrpjquota:
1480 case Opt_jqfmt_vfsold:
1481 case Opt_jqfmt_vfsv0:
1482 ext4_msg(sb, KERN_ERR,
1483 "journaled quota options not supported");
1489 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1492 clear_opt(sbi->s_mount_opt, BARRIER);
1495 if (match_int(&args[0], &option)) {
1496 set_opt(sbi->s_mount_opt, BARRIER);
1500 set_opt(sbi->s_mount_opt, BARRIER);
1502 clear_opt(sbi->s_mount_opt, BARRIER);
1508 ext4_msg(sb, KERN_ERR,
1509 "resize option only available "
1513 if (match_int(&args[0], &option) != 0)
1515 *n_blocks_count = option;
1518 set_opt(sbi->s_mount_opt, NOBH);
1521 clear_opt(sbi->s_mount_opt, NOBH);
1524 set_opt(sbi->s_mount_opt, I_VERSION);
1525 sb->s_flags |= MS_I_VERSION;
1527 case Opt_nodelalloc:
1528 clear_opt(sbi->s_mount_opt, DELALLOC);
1531 if (match_int(&args[0], &option))
1535 sbi->s_stripe = option;
1538 set_opt(sbi->s_mount_opt, DELALLOC);
1540 case Opt_block_validity:
1541 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1543 case Opt_noblock_validity:
1544 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1546 case Opt_inode_readahead_blks:
1547 if (match_int(&args[0], &option))
1549 if (option < 0 || option > (1 << 30))
1551 if (!is_power_of_2(option)) {
1552 ext4_msg(sb, KERN_ERR,
1553 "EXT4-fs: inode_readahead_blks"
1554 " must be a power of 2");
1557 sbi->s_inode_readahead_blks = option;
1559 case Opt_journal_ioprio:
1560 if (match_int(&args[0], &option))
1562 if (option < 0 || option > 7)
1564 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1567 case Opt_noauto_da_alloc:
1568 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1570 case Opt_auto_da_alloc:
1571 if (match_int(&args[0], &option)) {
1572 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1576 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1578 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1581 set_opt(sbi->s_mount_opt, DISCARD);
1584 clear_opt(sbi->s_mount_opt, DISCARD);
1587 ext4_msg(sb, KERN_ERR,
1588 "Unrecognized mount option \"%s\" "
1589 "or missing value", p);
1594 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1595 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1596 sbi->s_qf_names[USRQUOTA])
1597 clear_opt(sbi->s_mount_opt, USRQUOTA);
1599 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1600 sbi->s_qf_names[GRPQUOTA])
1601 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1603 if ((sbi->s_qf_names[USRQUOTA] &&
1604 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1605 (sbi->s_qf_names[GRPQUOTA] &&
1606 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1607 ext4_msg(sb, KERN_ERR, "old and new quota "
1612 if (!sbi->s_jquota_fmt) {
1613 ext4_msg(sb, KERN_ERR, "journaled quota format "
1618 if (sbi->s_jquota_fmt) {
1619 ext4_msg(sb, KERN_ERR, "journaled quota format "
1620 "specified with no journaling "
1629 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1632 struct ext4_sb_info *sbi = EXT4_SB(sb);
1635 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1636 ext4_msg(sb, KERN_ERR, "revision level too high, "
1637 "forcing read-only mode");
1642 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1643 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1644 "running e2fsck is recommended");
1645 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1646 ext4_msg(sb, KERN_WARNING,
1647 "warning: mounting fs with errors, "
1648 "running e2fsck is recommended");
1649 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1650 le16_to_cpu(es->s_mnt_count) >=
1651 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1652 ext4_msg(sb, KERN_WARNING,
1653 "warning: maximal mount count reached, "
1654 "running e2fsck is recommended");
1655 else if (le32_to_cpu(es->s_checkinterval) &&
1656 (le32_to_cpu(es->s_lastcheck) +
1657 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1658 ext4_msg(sb, KERN_WARNING,
1659 "warning: checktime reached, "
1660 "running e2fsck is recommended");
1661 if (!sbi->s_journal)
1662 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1663 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1664 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1665 le16_add_cpu(&es->s_mnt_count, 1);
1666 es->s_mtime = cpu_to_le32(get_seconds());
1667 ext4_update_dynamic_rev(sb);
1669 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1671 ext4_commit_super(sb, 1);
1672 if (test_opt(sb, DEBUG))
1673 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1674 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1676 sbi->s_groups_count,
1677 EXT4_BLOCKS_PER_GROUP(sb),
1678 EXT4_INODES_PER_GROUP(sb),
1684 static int ext4_fill_flex_info(struct super_block *sb)
1686 struct ext4_sb_info *sbi = EXT4_SB(sb);
1687 struct ext4_group_desc *gdp = NULL;
1688 ext4_group_t flex_group_count;
1689 ext4_group_t flex_group;
1690 int groups_per_flex = 0;
1694 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1695 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1697 if (groups_per_flex < 2) {
1698 sbi->s_log_groups_per_flex = 0;
1702 /* We allocate both existing and potentially added groups */
1703 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1704 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1705 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1706 size = flex_group_count * sizeof(struct flex_groups);
1707 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1708 if (sbi->s_flex_groups == NULL) {
1709 sbi->s_flex_groups = vmalloc(size);
1710 if (sbi->s_flex_groups)
1711 memset(sbi->s_flex_groups, 0, size);
1713 if (sbi->s_flex_groups == NULL) {
1714 ext4_msg(sb, KERN_ERR, "not enough memory for "
1715 "%u flex groups", flex_group_count);
1719 for (i = 0; i < sbi->s_groups_count; i++) {
1720 gdp = ext4_get_group_desc(sb, i, NULL);
1722 flex_group = ext4_flex_group(sbi, i);
1723 atomic_add(ext4_free_inodes_count(sb, gdp),
1724 &sbi->s_flex_groups[flex_group].free_inodes);
1725 atomic_add(ext4_free_blks_count(sb, gdp),
1726 &sbi->s_flex_groups[flex_group].free_blocks);
1727 atomic_add(ext4_used_dirs_count(sb, gdp),
1728 &sbi->s_flex_groups[flex_group].used_dirs);
1736 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1737 struct ext4_group_desc *gdp)
1741 if (sbi->s_es->s_feature_ro_compat &
1742 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1743 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1744 __le32 le_group = cpu_to_le32(block_group);
1746 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1747 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1748 crc = crc16(crc, (__u8 *)gdp, offset);
1749 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1750 /* for checksum of struct ext4_group_desc do the rest...*/
1751 if ((sbi->s_es->s_feature_incompat &
1752 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1753 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1754 crc = crc16(crc, (__u8 *)gdp + offset,
1755 le16_to_cpu(sbi->s_es->s_desc_size) -
1759 return cpu_to_le16(crc);
1762 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1763 struct ext4_group_desc *gdp)
1765 if ((sbi->s_es->s_feature_ro_compat &
1766 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1767 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1773 /* Called at mount-time, super-block is locked */
1774 static int ext4_check_descriptors(struct super_block *sb)
1776 struct ext4_sb_info *sbi = EXT4_SB(sb);
1777 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1778 ext4_fsblk_t last_block;
1779 ext4_fsblk_t block_bitmap;
1780 ext4_fsblk_t inode_bitmap;
1781 ext4_fsblk_t inode_table;
1782 int flexbg_flag = 0;
1785 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1788 ext4_debug("Checking group descriptors");
1790 for (i = 0; i < sbi->s_groups_count; i++) {
1791 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1793 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1794 last_block = ext4_blocks_count(sbi->s_es) - 1;
1796 last_block = first_block +
1797 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1799 block_bitmap = ext4_block_bitmap(sb, gdp);
1800 if (block_bitmap < first_block || block_bitmap > last_block) {
1801 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1802 "Block bitmap for group %u not in group "
1803 "(block %llu)!", i, block_bitmap);
1806 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1807 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1808 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1809 "Inode bitmap for group %u not in group "
1810 "(block %llu)!", i, inode_bitmap);
1813 inode_table = ext4_inode_table(sb, gdp);
1814 if (inode_table < first_block ||
1815 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1816 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1817 "Inode table for group %u not in group "
1818 "(block %llu)!", i, inode_table);
1821 ext4_lock_group(sb, i);
1822 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1823 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1824 "Checksum for group %u failed (%u!=%u)",
1825 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1826 gdp)), le16_to_cpu(gdp->bg_checksum));
1827 if (!(sb->s_flags & MS_RDONLY)) {
1828 ext4_unlock_group(sb, i);
1832 ext4_unlock_group(sb, i);
1834 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1837 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1838 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1842 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1843 * the superblock) which were deleted from all directories, but held open by
1844 * a process at the time of a crash. We walk the list and try to delete these
1845 * inodes at recovery time (only with a read-write filesystem).
1847 * In order to keep the orphan inode chain consistent during traversal (in
1848 * case of crash during recovery), we link each inode into the superblock
1849 * orphan list_head and handle it the same way as an inode deletion during
1850 * normal operation (which journals the operations for us).
1852 * We only do an iget() and an iput() on each inode, which is very safe if we
1853 * accidentally point at an in-use or already deleted inode. The worst that
1854 * can happen in this case is that we get a "bit already cleared" message from
1855 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1856 * e2fsck was run on this filesystem, and it must have already done the orphan
1857 * inode cleanup for us, so we can safely abort without any further action.
1859 static void ext4_orphan_cleanup(struct super_block *sb,
1860 struct ext4_super_block *es)
1862 unsigned int s_flags = sb->s_flags;
1863 int nr_orphans = 0, nr_truncates = 0;
1867 if (!es->s_last_orphan) {
1868 jbd_debug(4, "no orphan inodes to clean up\n");
1872 if (bdev_read_only(sb->s_bdev)) {
1873 ext4_msg(sb, KERN_ERR, "write access "
1874 "unavailable, skipping orphan cleanup");
1878 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1879 if (es->s_last_orphan)
1880 jbd_debug(1, "Errors on filesystem, "
1881 "clearing orphan list.\n");
1882 es->s_last_orphan = 0;
1883 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1887 if (s_flags & MS_RDONLY) {
1888 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1889 sb->s_flags &= ~MS_RDONLY;
1892 /* Needed for iput() to work correctly and not trash data */
1893 sb->s_flags |= MS_ACTIVE;
1894 /* Turn on quotas so that they are updated correctly */
1895 for (i = 0; i < MAXQUOTAS; i++) {
1896 if (EXT4_SB(sb)->s_qf_names[i]) {
1897 int ret = ext4_quota_on_mount(sb, i);
1899 ext4_msg(sb, KERN_ERR,
1900 "Cannot turn on journaled "
1901 "quota: error %d", ret);
1906 while (es->s_last_orphan) {
1907 struct inode *inode;
1909 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1910 if (IS_ERR(inode)) {
1911 es->s_last_orphan = 0;
1915 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1917 if (inode->i_nlink) {
1918 ext4_msg(sb, KERN_DEBUG,
1919 "%s: truncating inode %lu to %lld bytes",
1920 __func__, inode->i_ino, inode->i_size);
1921 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1922 inode->i_ino, inode->i_size);
1923 ext4_truncate(inode);
1926 ext4_msg(sb, KERN_DEBUG,
1927 "%s: deleting unreferenced inode %lu",
1928 __func__, inode->i_ino);
1929 jbd_debug(2, "deleting unreferenced inode %lu\n",
1933 iput(inode); /* The delete magic happens here! */
1936 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1939 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1940 PLURAL(nr_orphans));
1942 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1943 PLURAL(nr_truncates));
1945 /* Turn quotas off */
1946 for (i = 0; i < MAXQUOTAS; i++) {
1947 if (sb_dqopt(sb)->files[i])
1948 vfs_quota_off(sb, i, 0);
1951 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1955 * Maximal extent format file size.
1956 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1957 * extent format containers, within a sector_t, and within i_blocks
1958 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1959 * so that won't be a limiting factor.
1961 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1963 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1966 loff_t upper_limit = MAX_LFS_FILESIZE;
1968 /* small i_blocks in vfs inode? */
1969 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1971 * CONFIG_LBDAF is not enabled implies the inode
1972 * i_block represent total blocks in 512 bytes
1973 * 32 == size of vfs inode i_blocks * 8
1975 upper_limit = (1LL << 32) - 1;
1977 /* total blocks in file system block size */
1978 upper_limit >>= (blkbits - 9);
1979 upper_limit <<= blkbits;
1982 /* 32-bit extent-start container, ee_block */
1987 /* Sanity check against vm- & vfs- imposed limits */
1988 if (res > upper_limit)
1995 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1996 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1997 * We need to be 1 filesystem block less than the 2^48 sector limit.
1999 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2001 loff_t res = EXT4_NDIR_BLOCKS;
2004 /* This is calculated to be the largest file size for a dense, block
2005 * mapped file such that the file's total number of 512-byte sectors,
2006 * including data and all indirect blocks, does not exceed (2^48 - 1).
2008 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2009 * number of 512-byte sectors of the file.
2012 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2014 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2015 * the inode i_block field represents total file blocks in
2016 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2018 upper_limit = (1LL << 32) - 1;
2020 /* total blocks in file system block size */
2021 upper_limit >>= (bits - 9);
2025 * We use 48 bit ext4_inode i_blocks
2026 * With EXT4_HUGE_FILE_FL set the i_blocks
2027 * represent total number of blocks in
2028 * file system block size
2030 upper_limit = (1LL << 48) - 1;
2034 /* indirect blocks */
2036 /* double indirect blocks */
2037 meta_blocks += 1 + (1LL << (bits-2));
2038 /* tripple indirect blocks */
2039 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2041 upper_limit -= meta_blocks;
2042 upper_limit <<= bits;
2044 res += 1LL << (bits-2);
2045 res += 1LL << (2*(bits-2));
2046 res += 1LL << (3*(bits-2));
2048 if (res > upper_limit)
2051 if (res > MAX_LFS_FILESIZE)
2052 res = MAX_LFS_FILESIZE;
2057 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2058 ext4_fsblk_t logical_sb_block, int nr)
2060 struct ext4_sb_info *sbi = EXT4_SB(sb);
2061 ext4_group_t bg, first_meta_bg;
2064 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2066 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2068 return logical_sb_block + nr + 1;
2069 bg = sbi->s_desc_per_block * nr;
2070 if (ext4_bg_has_super(sb, bg))
2073 return (has_super + ext4_group_first_block_no(sb, bg));
2077 * ext4_get_stripe_size: Get the stripe size.
2078 * @sbi: In memory super block info
2080 * If we have specified it via mount option, then
2081 * use the mount option value. If the value specified at mount time is
2082 * greater than the blocks per group use the super block value.
2083 * If the super block value is greater than blocks per group return 0.
2084 * Allocator needs it be less than blocks per group.
2087 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2089 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2090 unsigned long stripe_width =
2091 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2093 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2094 return sbi->s_stripe;
2096 if (stripe_width <= sbi->s_blocks_per_group)
2097 return stripe_width;
2099 if (stride <= sbi->s_blocks_per_group)
2108 struct attribute attr;
2109 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2110 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2111 const char *, size_t);
2115 static int parse_strtoul(const char *buf,
2116 unsigned long max, unsigned long *value)
2120 while (*buf && isspace(*buf))
2122 *value = simple_strtoul(buf, &endp, 0);
2123 while (*endp && isspace(*endp))
2125 if (*endp || *value > max)
2131 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2132 struct ext4_sb_info *sbi,
2135 return snprintf(buf, PAGE_SIZE, "%llu\n",
2136 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2139 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2140 struct ext4_sb_info *sbi, char *buf)
2142 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2144 return snprintf(buf, PAGE_SIZE, "%lu\n",
2145 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2146 sbi->s_sectors_written_start) >> 1);
2149 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2150 struct ext4_sb_info *sbi, char *buf)
2152 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2154 return snprintf(buf, PAGE_SIZE, "%llu\n",
2155 sbi->s_kbytes_written +
2156 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2157 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2160 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2161 struct ext4_sb_info *sbi,
2162 const char *buf, size_t count)
2166 if (parse_strtoul(buf, 0x40000000, &t))
2169 if (!is_power_of_2(t))
2172 sbi->s_inode_readahead_blks = t;
2176 static ssize_t sbi_ui_show(struct ext4_attr *a,
2177 struct ext4_sb_info *sbi, char *buf)
2179 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2181 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2184 static ssize_t sbi_ui_store(struct ext4_attr *a,
2185 struct ext4_sb_info *sbi,
2186 const char *buf, size_t count)
2188 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2191 if (parse_strtoul(buf, 0xffffffff, &t))
2197 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2198 static struct ext4_attr ext4_attr_##_name = { \
2199 .attr = {.name = __stringify(_name), .mode = _mode }, \
2202 .offset = offsetof(struct ext4_sb_info, _elname), \
2204 #define EXT4_ATTR(name, mode, show, store) \
2205 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2207 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2208 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2209 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2210 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2211 #define ATTR_LIST(name) &ext4_attr_##name.attr
2213 EXT4_RO_ATTR(delayed_allocation_blocks);
2214 EXT4_RO_ATTR(session_write_kbytes);
2215 EXT4_RO_ATTR(lifetime_write_kbytes);
2216 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2217 inode_readahead_blks_store, s_inode_readahead_blks);
2218 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2219 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2220 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2221 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2222 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2223 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2224 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2225 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2227 static struct attribute *ext4_attrs[] = {
2228 ATTR_LIST(delayed_allocation_blocks),
2229 ATTR_LIST(session_write_kbytes),
2230 ATTR_LIST(lifetime_write_kbytes),
2231 ATTR_LIST(inode_readahead_blks),
2232 ATTR_LIST(inode_goal),
2233 ATTR_LIST(mb_stats),
2234 ATTR_LIST(mb_max_to_scan),
2235 ATTR_LIST(mb_min_to_scan),
2236 ATTR_LIST(mb_order2_req),
2237 ATTR_LIST(mb_stream_req),
2238 ATTR_LIST(mb_group_prealloc),
2239 ATTR_LIST(max_writeback_mb_bump),
2243 static ssize_t ext4_attr_show(struct kobject *kobj,
2244 struct attribute *attr, char *buf)
2246 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2248 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2250 return a->show ? a->show(a, sbi, buf) : 0;
2253 static ssize_t ext4_attr_store(struct kobject *kobj,
2254 struct attribute *attr,
2255 const char *buf, size_t len)
2257 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2259 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2261 return a->store ? a->store(a, sbi, buf, len) : 0;
2264 static void ext4_sb_release(struct kobject *kobj)
2266 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2268 complete(&sbi->s_kobj_unregister);
2272 static struct sysfs_ops ext4_attr_ops = {
2273 .show = ext4_attr_show,
2274 .store = ext4_attr_store,
2277 static struct kobj_type ext4_ktype = {
2278 .default_attrs = ext4_attrs,
2279 .sysfs_ops = &ext4_attr_ops,
2280 .release = ext4_sb_release,
2284 * Check whether this filesystem can be mounted based on
2285 * the features present and the RDONLY/RDWR mount requested.
2286 * Returns 1 if this filesystem can be mounted as requested,
2287 * 0 if it cannot be.
2289 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2291 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2292 ext4_msg(sb, KERN_ERR,
2293 "Couldn't mount because of "
2294 "unsupported optional features (%x)",
2295 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2296 ~EXT4_FEATURE_INCOMPAT_SUPP));
2303 /* Check that feature set is OK for a read-write mount */
2304 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2305 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2306 "unsupported optional features (%x)",
2307 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2308 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2312 * Large file size enabled file system can only be mounted
2313 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2315 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2316 if (sizeof(blkcnt_t) < sizeof(u64)) {
2317 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2318 "cannot be mounted RDWR without "
2326 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2327 __releases(kernel_lock)
2328 __acquires(kernel_lock)
2330 struct buffer_head *bh;
2331 struct ext4_super_block *es = NULL;
2332 struct ext4_sb_info *sbi;
2334 ext4_fsblk_t sb_block = get_sb_block(&data);
2335 ext4_fsblk_t logical_sb_block;
2336 unsigned long offset = 0;
2337 unsigned long journal_devnum = 0;
2338 unsigned long def_mount_opts;
2344 unsigned int db_count;
2346 int needs_recovery, has_huge_files;
2349 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2351 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2355 sbi->s_blockgroup_lock =
2356 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2357 if (!sbi->s_blockgroup_lock) {
2361 sb->s_fs_info = sbi;
2362 sbi->s_mount_opt = 0;
2363 sbi->s_resuid = EXT4_DEF_RESUID;
2364 sbi->s_resgid = EXT4_DEF_RESGID;
2365 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2366 sbi->s_sb_block = sb_block;
2367 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2372 /* Cleanup superblock name */
2373 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2376 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2378 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2383 * The ext4 superblock will not be buffer aligned for other than 1kB
2384 * block sizes. We need to calculate the offset from buffer start.
2386 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2387 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2388 offset = do_div(logical_sb_block, blocksize);
2390 logical_sb_block = sb_block;
2393 if (!(bh = sb_bread(sb, logical_sb_block))) {
2394 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2398 * Note: s_es must be initialized as soon as possible because
2399 * some ext4 macro-instructions depend on its value
2401 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2403 sb->s_magic = le16_to_cpu(es->s_magic);
2404 if (sb->s_magic != EXT4_SUPER_MAGIC)
2406 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2408 /* Set defaults before we parse the mount options */
2409 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2410 if (def_mount_opts & EXT4_DEFM_DEBUG)
2411 set_opt(sbi->s_mount_opt, DEBUG);
2412 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2413 set_opt(sbi->s_mount_opt, GRPID);
2414 if (def_mount_opts & EXT4_DEFM_UID16)
2415 set_opt(sbi->s_mount_opt, NO_UID32);
2416 #ifdef CONFIG_EXT4_FS_XATTR
2417 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2418 set_opt(sbi->s_mount_opt, XATTR_USER);
2420 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2421 if (def_mount_opts & EXT4_DEFM_ACL)
2422 set_opt(sbi->s_mount_opt, POSIX_ACL);
2424 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2425 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2426 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2427 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2428 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2429 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2431 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2432 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2433 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2434 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2436 set_opt(sbi->s_mount_opt, ERRORS_RO);
2438 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2439 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2440 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2441 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2442 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2444 set_opt(sbi->s_mount_opt, BARRIER);
2447 * enable delayed allocation by default
2448 * Use -o nodelalloc to turn it off
2450 set_opt(sbi->s_mount_opt, DELALLOC);
2452 if (!parse_options((char *) data, sb, &journal_devnum,
2453 &journal_ioprio, NULL, 0))
2456 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2457 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2459 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2460 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2461 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2462 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2463 ext4_msg(sb, KERN_WARNING,
2464 "feature flags set on rev 0 fs, "
2465 "running e2fsck is recommended");
2468 * Check feature flags regardless of the revision level, since we
2469 * previously didn't change the revision level when setting the flags,
2470 * so there is a chance incompat flags are set on a rev 0 filesystem.
2472 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2475 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2477 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2478 blocksize > EXT4_MAX_BLOCK_SIZE) {
2479 ext4_msg(sb, KERN_ERR,
2480 "Unsupported filesystem blocksize %d", blocksize);
2484 if (sb->s_blocksize != blocksize) {
2485 /* Validate the filesystem blocksize */
2486 if (!sb_set_blocksize(sb, blocksize)) {
2487 ext4_msg(sb, KERN_ERR, "bad block size %d",
2493 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2494 offset = do_div(logical_sb_block, blocksize);
2495 bh = sb_bread(sb, logical_sb_block);
2497 ext4_msg(sb, KERN_ERR,
2498 "Can't read superblock on 2nd try");
2501 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2503 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2504 ext4_msg(sb, KERN_ERR,
2505 "Magic mismatch, very weird!");
2510 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2511 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2512 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2514 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2516 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2517 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2518 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2520 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2521 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2522 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2523 (!is_power_of_2(sbi->s_inode_size)) ||
2524 (sbi->s_inode_size > blocksize)) {
2525 ext4_msg(sb, KERN_ERR,
2526 "unsupported inode size: %d",
2530 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2531 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2534 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2535 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2536 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2537 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2538 !is_power_of_2(sbi->s_desc_size)) {
2539 ext4_msg(sb, KERN_ERR,
2540 "unsupported descriptor size %lu",
2545 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2547 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2548 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2549 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2552 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2553 if (sbi->s_inodes_per_block == 0)
2555 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2556 sbi->s_inodes_per_block;
2557 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2559 sbi->s_mount_state = le16_to_cpu(es->s_state);
2560 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2561 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2563 for (i = 0; i < 4; i++)
2564 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2565 sbi->s_def_hash_version = es->s_def_hash_version;
2566 i = le32_to_cpu(es->s_flags);
2567 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2568 sbi->s_hash_unsigned = 3;
2569 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2570 #ifdef __CHAR_UNSIGNED__
2571 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2572 sbi->s_hash_unsigned = 3;
2574 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2579 if (sbi->s_blocks_per_group > blocksize * 8) {
2580 ext4_msg(sb, KERN_ERR,
2581 "#blocks per group too big: %lu",
2582 sbi->s_blocks_per_group);
2585 if (sbi->s_inodes_per_group > blocksize * 8) {
2586 ext4_msg(sb, KERN_ERR,
2587 "#inodes per group too big: %lu",
2588 sbi->s_inodes_per_group);
2593 * Test whether we have more sectors than will fit in sector_t,
2594 * and whether the max offset is addressable by the page cache.
2596 if ((ext4_blocks_count(es) >
2597 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2598 (ext4_blocks_count(es) >
2599 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2600 ext4_msg(sb, KERN_ERR, "filesystem"
2601 " too large to mount safely on this system");
2602 if (sizeof(sector_t) < 8)
2603 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2608 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2611 /* check blocks count against device size */
2612 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2613 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2614 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2615 "exceeds size of device (%llu blocks)",
2616 ext4_blocks_count(es), blocks_count);
2621 * It makes no sense for the first data block to be beyond the end
2622 * of the filesystem.
2624 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2625 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2626 "block %u is beyond end of filesystem (%llu)",
2627 le32_to_cpu(es->s_first_data_block),
2628 ext4_blocks_count(es));
2631 blocks_count = (ext4_blocks_count(es) -
2632 le32_to_cpu(es->s_first_data_block) +
2633 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2634 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2635 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2636 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2637 "(block count %llu, first data block %u, "
2638 "blocks per group %lu)", sbi->s_groups_count,
2639 ext4_blocks_count(es),
2640 le32_to_cpu(es->s_first_data_block),
2641 EXT4_BLOCKS_PER_GROUP(sb));
2644 sbi->s_groups_count = blocks_count;
2645 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2646 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2647 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2648 EXT4_DESC_PER_BLOCK(sb);
2649 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2651 if (sbi->s_group_desc == NULL) {
2652 ext4_msg(sb, KERN_ERR, "not enough memory");
2656 #ifdef CONFIG_PROC_FS
2658 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2661 bgl_lock_init(sbi->s_blockgroup_lock);
2663 for (i = 0; i < db_count; i++) {
2664 block = descriptor_loc(sb, logical_sb_block, i);
2665 sbi->s_group_desc[i] = sb_bread(sb, block);
2666 if (!sbi->s_group_desc[i]) {
2667 ext4_msg(sb, KERN_ERR,
2668 "can't read group descriptor %d", i);
2673 if (!ext4_check_descriptors(sb)) {
2674 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2677 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2678 if (!ext4_fill_flex_info(sb)) {
2679 ext4_msg(sb, KERN_ERR,
2680 "unable to initialize "
2681 "flex_bg meta info!");
2685 sbi->s_gdb_count = db_count;
2686 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2687 spin_lock_init(&sbi->s_next_gen_lock);
2689 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2690 ext4_count_free_blocks(sb));
2692 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2693 ext4_count_free_inodes(sb));
2696 err = percpu_counter_init(&sbi->s_dirs_counter,
2697 ext4_count_dirs(sb));
2700 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2703 ext4_msg(sb, KERN_ERR, "insufficient memory");
2707 sbi->s_stripe = ext4_get_stripe_size(sbi);
2708 sbi->s_max_writeback_mb_bump = 128;
2711 * set up enough so that it can read an inode
2713 if (!test_opt(sb, NOLOAD) &&
2714 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2715 sb->s_op = &ext4_sops;
2717 sb->s_op = &ext4_nojournal_sops;
2718 sb->s_export_op = &ext4_export_ops;
2719 sb->s_xattr = ext4_xattr_handlers;
2721 sb->s_qcop = &ext4_qctl_operations;
2722 sb->dq_op = &ext4_quota_operations;
2724 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2725 mutex_init(&sbi->s_orphan_lock);
2726 mutex_init(&sbi->s_resize_lock);
2730 needs_recovery = (es->s_last_orphan != 0 ||
2731 EXT4_HAS_INCOMPAT_FEATURE(sb,
2732 EXT4_FEATURE_INCOMPAT_RECOVER));
2735 * The first inode we look at is the journal inode. Don't try
2736 * root first: it may be modified in the journal!
2738 if (!test_opt(sb, NOLOAD) &&
2739 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2740 if (ext4_load_journal(sb, es, journal_devnum))
2742 if (!(sb->s_flags & MS_RDONLY) &&
2743 EXT4_SB(sb)->s_journal->j_failed_commit) {
2744 ext4_msg(sb, KERN_CRIT, "error: "
2745 "ext4_fill_super: Journal transaction "
2747 EXT4_SB(sb)->s_journal->j_failed_commit);
2748 if (test_opt(sb, ERRORS_RO)) {
2749 ext4_msg(sb, KERN_CRIT,
2750 "Mounting filesystem read-only");
2751 sb->s_flags |= MS_RDONLY;
2752 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2753 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2755 if (test_opt(sb, ERRORS_PANIC)) {
2756 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2757 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2758 ext4_commit_super(sb, 1);
2762 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2763 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2764 ext4_msg(sb, KERN_ERR, "required journal recovery "
2765 "suppressed and not mounted read-only");
2768 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2769 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2770 sbi->s_journal = NULL;
2775 if (ext4_blocks_count(es) > 0xffffffffULL &&
2776 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2777 JBD2_FEATURE_INCOMPAT_64BIT)) {
2778 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2782 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2783 jbd2_journal_set_features(sbi->s_journal,
2784 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2785 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2786 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2787 jbd2_journal_set_features(sbi->s_journal,
2788 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2789 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2790 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2792 jbd2_journal_clear_features(sbi->s_journal,
2793 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2794 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2797 /* We have now updated the journal if required, so we can
2798 * validate the data journaling mode. */
2799 switch (test_opt(sb, DATA_FLAGS)) {
2801 /* No mode set, assume a default based on the journal
2802 * capabilities: ORDERED_DATA if the journal can
2803 * cope, else JOURNAL_DATA
2805 if (jbd2_journal_check_available_features
2806 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2807 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2809 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2812 case EXT4_MOUNT_ORDERED_DATA:
2813 case EXT4_MOUNT_WRITEBACK_DATA:
2814 if (!jbd2_journal_check_available_features
2815 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2816 ext4_msg(sb, KERN_ERR, "Journal does not support "
2817 "requested data journaling mode");
2823 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2827 if (test_opt(sb, NOBH)) {
2828 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2829 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2830 "its supported only with writeback mode");
2831 clear_opt(sbi->s_mount_opt, NOBH);
2834 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2835 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2836 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2837 goto failed_mount_wq;
2841 * The jbd2_journal_load will have done any necessary log recovery,
2842 * so we can safely mount the rest of the filesystem now.
2845 root = ext4_iget(sb, EXT4_ROOT_INO);
2847 ext4_msg(sb, KERN_ERR, "get root inode failed");
2848 ret = PTR_ERR(root);
2851 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2853 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2856 sb->s_root = d_alloc_root(root);
2858 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2864 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2866 /* determine the minimum size of new large inodes, if present */
2867 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2868 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2869 EXT4_GOOD_OLD_INODE_SIZE;
2870 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2871 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2872 if (sbi->s_want_extra_isize <
2873 le16_to_cpu(es->s_want_extra_isize))
2874 sbi->s_want_extra_isize =
2875 le16_to_cpu(es->s_want_extra_isize);
2876 if (sbi->s_want_extra_isize <
2877 le16_to_cpu(es->s_min_extra_isize))
2878 sbi->s_want_extra_isize =
2879 le16_to_cpu(es->s_min_extra_isize);
2882 /* Check if enough inode space is available */
2883 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2884 sbi->s_inode_size) {
2885 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2886 EXT4_GOOD_OLD_INODE_SIZE;
2887 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2891 if (test_opt(sb, DELALLOC) &&
2892 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2893 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2894 "requested data journaling mode");
2895 clear_opt(sbi->s_mount_opt, DELALLOC);
2898 err = ext4_setup_system_zone(sb);
2900 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2901 "zone (%d)\n", err);
2906 err = ext4_mb_init(sb, needs_recovery);
2908 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2913 sbi->s_kobj.kset = ext4_kset;
2914 init_completion(&sbi->s_kobj_unregister);
2915 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2918 ext4_mb_release(sb);
2919 ext4_ext_release(sb);
2923 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2924 ext4_orphan_cleanup(sb, es);
2925 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2926 if (needs_recovery) {
2927 ext4_msg(sb, KERN_INFO, "recovery complete");
2928 ext4_mark_recovery_complete(sb, es);
2930 if (EXT4_SB(sb)->s_journal) {
2931 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2932 descr = " journalled data mode";
2933 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2934 descr = " ordered data mode";
2936 descr = " writeback data mode";
2938 descr = "out journal";
2940 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2947 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2951 ext4_msg(sb, KERN_ERR, "mount failed");
2952 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2954 ext4_release_system_zone(sb);
2955 if (sbi->s_journal) {
2956 jbd2_journal_destroy(sbi->s_journal);
2957 sbi->s_journal = NULL;
2960 if (sbi->s_flex_groups) {
2961 if (is_vmalloc_addr(sbi->s_flex_groups))
2962 vfree(sbi->s_flex_groups);
2964 kfree(sbi->s_flex_groups);
2966 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2967 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2968 percpu_counter_destroy(&sbi->s_dirs_counter);
2969 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2971 for (i = 0; i < db_count; i++)
2972 brelse(sbi->s_group_desc[i]);
2973 kfree(sbi->s_group_desc);
2976 remove_proc_entry(sb->s_id, ext4_proc_root);
2979 for (i = 0; i < MAXQUOTAS; i++)
2980 kfree(sbi->s_qf_names[i]);
2982 ext4_blkdev_remove(sbi);
2985 sb->s_fs_info = NULL;
2986 kfree(sbi->s_blockgroup_lock);
2993 * Setup any per-fs journal parameters now. We'll do this both on
2994 * initial mount, once the journal has been initialised but before we've
2995 * done any recovery; and again on any subsequent remount.
2997 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2999 struct ext4_sb_info *sbi = EXT4_SB(sb);
3001 journal->j_commit_interval = sbi->s_commit_interval;
3002 journal->j_min_batch_time = sbi->s_min_batch_time;
3003 journal->j_max_batch_time = sbi->s_max_batch_time;
3005 spin_lock(&journal->j_state_lock);
3006 if (test_opt(sb, BARRIER))
3007 journal->j_flags |= JBD2_BARRIER;
3009 journal->j_flags &= ~JBD2_BARRIER;
3010 if (test_opt(sb, DATA_ERR_ABORT))
3011 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3013 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3014 spin_unlock(&journal->j_state_lock);
3017 static journal_t *ext4_get_journal(struct super_block *sb,
3018 unsigned int journal_inum)
3020 struct inode *journal_inode;
3023 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3025 /* First, test for the existence of a valid inode on disk. Bad
3026 * things happen if we iget() an unused inode, as the subsequent
3027 * iput() will try to delete it. */
3029 journal_inode = ext4_iget(sb, journal_inum);
3030 if (IS_ERR(journal_inode)) {
3031 ext4_msg(sb, KERN_ERR, "no journal found");
3034 if (!journal_inode->i_nlink) {
3035 make_bad_inode(journal_inode);
3036 iput(journal_inode);
3037 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3041 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3042 journal_inode, journal_inode->i_size);
3043 if (!S_ISREG(journal_inode->i_mode)) {
3044 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3045 iput(journal_inode);
3049 journal = jbd2_journal_init_inode(journal_inode);
3051 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3052 iput(journal_inode);
3055 journal->j_private = sb;
3056 ext4_init_journal_params(sb, journal);
3060 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3063 struct buffer_head *bh;
3067 int hblock, blocksize;
3068 ext4_fsblk_t sb_block;
3069 unsigned long offset;
3070 struct ext4_super_block *es;
3071 struct block_device *bdev;
3073 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3075 bdev = ext4_blkdev_get(j_dev, sb);
3079 if (bd_claim(bdev, sb)) {
3080 ext4_msg(sb, KERN_ERR,
3081 "failed to claim external journal device");
3082 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3086 blocksize = sb->s_blocksize;
3087 hblock = bdev_logical_block_size(bdev);
3088 if (blocksize < hblock) {
3089 ext4_msg(sb, KERN_ERR,
3090 "blocksize too small for journal device");
3094 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3095 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3096 set_blocksize(bdev, blocksize);
3097 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3098 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3099 "external journal");
3103 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3104 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3105 !(le32_to_cpu(es->s_feature_incompat) &
3106 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3107 ext4_msg(sb, KERN_ERR, "external journal has "
3113 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3114 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3119 len = ext4_blocks_count(es);
3120 start = sb_block + 1;
3121 brelse(bh); /* we're done with the superblock */
3123 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3124 start, len, blocksize);
3126 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3129 journal->j_private = sb;
3130 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3131 wait_on_buffer(journal->j_sb_buffer);
3132 if (!buffer_uptodate(journal->j_sb_buffer)) {
3133 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3136 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3137 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3138 "user (unsupported) - %d",
3139 be32_to_cpu(journal->j_superblock->s_nr_users));
3142 EXT4_SB(sb)->journal_bdev = bdev;
3143 ext4_init_journal_params(sb, journal);
3147 jbd2_journal_destroy(journal);
3149 ext4_blkdev_put(bdev);
3153 static int ext4_load_journal(struct super_block *sb,
3154 struct ext4_super_block *es,
3155 unsigned long journal_devnum)
3158 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3161 int really_read_only;
3163 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3165 if (journal_devnum &&
3166 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3167 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3168 "numbers have changed");
3169 journal_dev = new_decode_dev(journal_devnum);
3171 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3173 really_read_only = bdev_read_only(sb->s_bdev);
3176 * Are we loading a blank journal or performing recovery after a
3177 * crash? For recovery, we need to check in advance whether we
3178 * can get read-write access to the device.
3180 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3181 if (sb->s_flags & MS_RDONLY) {
3182 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3183 "required on readonly filesystem");
3184 if (really_read_only) {
3185 ext4_msg(sb, KERN_ERR, "write access "
3186 "unavailable, cannot proceed");
3189 ext4_msg(sb, KERN_INFO, "write access will "
3190 "be enabled during recovery");
3194 if (journal_inum && journal_dev) {
3195 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3196 "and inode journals!");
3201 if (!(journal = ext4_get_journal(sb, journal_inum)))
3204 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3208 if (!(journal->j_flags & JBD2_BARRIER))
3209 ext4_msg(sb, KERN_INFO, "barriers disabled");
3211 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3212 err = jbd2_journal_update_format(journal);
3214 ext4_msg(sb, KERN_ERR, "error updating journal");
3215 jbd2_journal_destroy(journal);
3220 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3221 err = jbd2_journal_wipe(journal, !really_read_only);
3223 err = jbd2_journal_load(journal);
3226 ext4_msg(sb, KERN_ERR, "error loading journal");
3227 jbd2_journal_destroy(journal);
3231 EXT4_SB(sb)->s_journal = journal;
3232 ext4_clear_journal_err(sb, es);
3234 if (journal_devnum &&
3235 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3236 es->s_journal_dev = cpu_to_le32(journal_devnum);
3238 /* Make sure we flush the recovery flag to disk. */
3239 ext4_commit_super(sb, 1);
3245 static int ext4_commit_super(struct super_block *sb, int sync)
3247 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3248 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3253 if (buffer_write_io_error(sbh)) {
3255 * Oh, dear. A previous attempt to write the
3256 * superblock failed. This could happen because the
3257 * USB device was yanked out. Or it could happen to
3258 * be a transient write error and maybe the block will
3259 * be remapped. Nothing we can do but to retry the
3260 * write and hope for the best.
3262 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3263 "superblock detected");
3264 clear_buffer_write_io_error(sbh);
3265 set_buffer_uptodate(sbh);
3268 * If the file system is mounted read-only, don't update the
3269 * superblock write time. This avoids updating the superblock
3270 * write time when we are mounting the root file system
3271 * read/only but we need to replay the journal; at that point,
3272 * for people who are east of GMT and who make their clock
3273 * tick in localtime for Windows bug-for-bug compatibility,
3274 * the clock is set in the future, and this will cause e2fsck
3275 * to complain and force a full file system check.
3277 if (!(sb->s_flags & MS_RDONLY))
3278 es->s_wtime = cpu_to_le32(get_seconds());
3279 es->s_kbytes_written =
3280 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3281 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3282 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3283 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3284 &EXT4_SB(sb)->s_freeblocks_counter));
3285 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3286 &EXT4_SB(sb)->s_freeinodes_counter));
3288 BUFFER_TRACE(sbh, "marking dirty");
3289 mark_buffer_dirty(sbh);
3291 error = sync_dirty_buffer(sbh);
3295 error = buffer_write_io_error(sbh);
3297 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3299 clear_buffer_write_io_error(sbh);
3300 set_buffer_uptodate(sbh);
3307 * Have we just finished recovery? If so, and if we are mounting (or
3308 * remounting) the filesystem readonly, then we will end up with a
3309 * consistent fs on disk. Record that fact.
3311 static void ext4_mark_recovery_complete(struct super_block *sb,
3312 struct ext4_super_block *es)
3314 journal_t *journal = EXT4_SB(sb)->s_journal;
3316 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3317 BUG_ON(journal != NULL);
3320 jbd2_journal_lock_updates(journal);
3321 if (jbd2_journal_flush(journal) < 0)
3324 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3325 sb->s_flags & MS_RDONLY) {
3326 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3327 ext4_commit_super(sb, 1);
3331 jbd2_journal_unlock_updates(journal);
3335 * If we are mounting (or read-write remounting) a filesystem whose journal
3336 * has recorded an error from a previous lifetime, move that error to the
3337 * main filesystem now.
3339 static void ext4_clear_journal_err(struct super_block *sb,
3340 struct ext4_super_block *es)
3346 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3348 journal = EXT4_SB(sb)->s_journal;
3351 * Now check for any error status which may have been recorded in the
3352 * journal by a prior ext4_error() or ext4_abort()
3355 j_errno = jbd2_journal_errno(journal);
3359 errstr = ext4_decode_error(sb, j_errno, nbuf);
3360 ext4_warning(sb, __func__, "Filesystem error recorded "
3361 "from previous mount: %s", errstr);
3362 ext4_warning(sb, __func__, "Marking fs in need of "
3363 "filesystem check.");
3365 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3367 ext4_commit_super(sb, 1);
3369 jbd2_journal_clear_err(journal);
3374 * Force the running and committing transactions to commit,
3375 * and wait on the commit.
3377 int ext4_force_commit(struct super_block *sb)
3382 if (sb->s_flags & MS_RDONLY)
3385 journal = EXT4_SB(sb)->s_journal;
3387 ret = ext4_journal_force_commit(journal);
3392 static void ext4_write_super(struct super_block *sb)
3395 ext4_commit_super(sb, 1);
3399 static int ext4_sync_fs(struct super_block *sb, int wait)
3403 struct ext4_sb_info *sbi = EXT4_SB(sb);
3405 trace_ext4_sync_fs(sb, wait);
3406 flush_workqueue(sbi->dio_unwritten_wq);
3407 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3409 jbd2_log_wait_commit(sbi->s_journal, target);
3415 * LVM calls this function before a (read-only) snapshot is created. This
3416 * gives us a chance to flush the journal completely and mark the fs clean.
3418 static int ext4_freeze(struct super_block *sb)
3423 if (sb->s_flags & MS_RDONLY)
3426 journal = EXT4_SB(sb)->s_journal;
3428 /* Now we set up the journal barrier. */
3429 jbd2_journal_lock_updates(journal);
3432 * Don't clear the needs_recovery flag if we failed to flush
3435 error = jbd2_journal_flush(journal);
3438 jbd2_journal_unlock_updates(journal);
3442 /* Journal blocked and flushed, clear needs_recovery flag. */
3443 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3444 error = ext4_commit_super(sb, 1);
3451 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3452 * flag here, even though the filesystem is not technically dirty yet.
3454 static int ext4_unfreeze(struct super_block *sb)
3456 if (sb->s_flags & MS_RDONLY)
3460 /* Reset the needs_recovery flag before the fs is unlocked. */
3461 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3462 ext4_commit_super(sb, 1);
3464 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3468 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3470 struct ext4_super_block *es;
3471 struct ext4_sb_info *sbi = EXT4_SB(sb);
3472 ext4_fsblk_t n_blocks_count = 0;
3473 unsigned long old_sb_flags;
3474 struct ext4_mount_options old_opts;
3476 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3484 /* Store the original options */
3486 old_sb_flags = sb->s_flags;
3487 old_opts.s_mount_opt = sbi->s_mount_opt;
3488 old_opts.s_resuid = sbi->s_resuid;
3489 old_opts.s_resgid = sbi->s_resgid;
3490 old_opts.s_commit_interval = sbi->s_commit_interval;
3491 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3492 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3494 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3495 for (i = 0; i < MAXQUOTAS; i++)
3496 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3498 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3499 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3502 * Allow the "check" option to be passed as a remount option.
3504 if (!parse_options(data, sb, NULL, &journal_ioprio,
3505 &n_blocks_count, 1)) {
3510 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3511 ext4_abort(sb, __func__, "Abort forced by user");
3513 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3514 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3518 if (sbi->s_journal) {
3519 ext4_init_journal_params(sb, sbi->s_journal);
3520 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3523 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3524 n_blocks_count > ext4_blocks_count(es)) {
3525 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3530 if (*flags & MS_RDONLY) {
3532 * First of all, the unconditional stuff we have to do
3533 * to disable replay of the journal when we next remount
3535 sb->s_flags |= MS_RDONLY;
3538 * OK, test if we are remounting a valid rw partition
3539 * readonly, and if so set the rdonly flag and then
3540 * mark the partition as valid again.
3542 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3543 (sbi->s_mount_state & EXT4_VALID_FS))
3544 es->s_state = cpu_to_le16(sbi->s_mount_state);
3547 ext4_mark_recovery_complete(sb, es);
3549 /* Make sure we can mount this feature set readwrite */
3550 if (!ext4_feature_set_ok(sb, 0)) {
3555 * Make sure the group descriptor checksums
3556 * are sane. If they aren't, refuse to remount r/w.
3558 for (g = 0; g < sbi->s_groups_count; g++) {
3559 struct ext4_group_desc *gdp =
3560 ext4_get_group_desc(sb, g, NULL);
3562 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3563 ext4_msg(sb, KERN_ERR,
3564 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3565 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3566 le16_to_cpu(gdp->bg_checksum));
3573 * If we have an unprocessed orphan list hanging
3574 * around from a previously readonly bdev mount,
3575 * require a full umount/remount for now.
3577 if (es->s_last_orphan) {
3578 ext4_msg(sb, KERN_WARNING, "Couldn't "
3579 "remount RDWR because of unprocessed "
3580 "orphan inode list. Please "
3581 "umount/remount instead");
3587 * Mounting a RDONLY partition read-write, so reread
3588 * and store the current valid flag. (It may have
3589 * been changed by e2fsck since we originally mounted
3593 ext4_clear_journal_err(sb, es);
3594 sbi->s_mount_state = le16_to_cpu(es->s_state);
3595 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3597 if (!ext4_setup_super(sb, es, 0))
3598 sb->s_flags &= ~MS_RDONLY;
3601 ext4_setup_system_zone(sb);
3602 if (sbi->s_journal == NULL)
3603 ext4_commit_super(sb, 1);
3606 /* Release old quota file names */
3607 for (i = 0; i < MAXQUOTAS; i++)
3608 if (old_opts.s_qf_names[i] &&
3609 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3610 kfree(old_opts.s_qf_names[i]);
3617 sb->s_flags = old_sb_flags;
3618 sbi->s_mount_opt = old_opts.s_mount_opt;
3619 sbi->s_resuid = old_opts.s_resuid;
3620 sbi->s_resgid = old_opts.s_resgid;
3621 sbi->s_commit_interval = old_opts.s_commit_interval;
3622 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3623 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3625 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3626 for (i = 0; i < MAXQUOTAS; i++) {
3627 if (sbi->s_qf_names[i] &&
3628 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3629 kfree(sbi->s_qf_names[i]);
3630 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3638 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3640 struct super_block *sb = dentry->d_sb;
3641 struct ext4_sb_info *sbi = EXT4_SB(sb);
3642 struct ext4_super_block *es = sbi->s_es;
3645 if (test_opt(sb, MINIX_DF)) {
3646 sbi->s_overhead_last = 0;
3647 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3648 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3649 ext4_fsblk_t overhead = 0;
3652 * Compute the overhead (FS structures). This is constant
3653 * for a given filesystem unless the number of block groups
3654 * changes so we cache the previous value until it does.
3658 * All of the blocks before first_data_block are
3661 overhead = le32_to_cpu(es->s_first_data_block);
3664 * Add the overhead attributed to the superblock and
3665 * block group descriptors. If the sparse superblocks
3666 * feature is turned on, then not all groups have this.
3668 for (i = 0; i < ngroups; i++) {
3669 overhead += ext4_bg_has_super(sb, i) +
3670 ext4_bg_num_gdb(sb, i);
3675 * Every block group has an inode bitmap, a block
3676 * bitmap, and an inode table.
3678 overhead += ngroups * (2 + sbi->s_itb_per_group);
3679 sbi->s_overhead_last = overhead;
3681 sbi->s_blocks_last = ext4_blocks_count(es);
3684 buf->f_type = EXT4_SUPER_MAGIC;
3685 buf->f_bsize = sb->s_blocksize;
3686 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3687 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3688 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3689 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3690 if (buf->f_bfree < ext4_r_blocks_count(es))
3692 buf->f_files = le32_to_cpu(es->s_inodes_count);
3693 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3694 buf->f_namelen = EXT4_NAME_LEN;
3695 fsid = le64_to_cpup((void *)es->s_uuid) ^
3696 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3697 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3698 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3703 /* Helper function for writing quotas on sync - we need to start transaction
3704 * before quota file is locked for write. Otherwise the are possible deadlocks:
3705 * Process 1 Process 2
3706 * ext4_create() quota_sync()
3707 * jbd2_journal_start() write_dquot()
3708 * vfs_dq_init() down(dqio_mutex)
3709 * down(dqio_mutex) jbd2_journal_start()
3715 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3717 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3720 static int ext4_write_dquot(struct dquot *dquot)
3724 struct inode *inode;
3726 inode = dquot_to_inode(dquot);
3727 handle = ext4_journal_start(inode,
3728 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3730 return PTR_ERR(handle);
3731 ret = dquot_commit(dquot);
3732 err = ext4_journal_stop(handle);
3738 static int ext4_acquire_dquot(struct dquot *dquot)
3743 handle = ext4_journal_start(dquot_to_inode(dquot),
3744 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3746 return PTR_ERR(handle);
3747 ret = dquot_acquire(dquot);
3748 err = ext4_journal_stop(handle);
3754 static int ext4_release_dquot(struct dquot *dquot)
3759 handle = ext4_journal_start(dquot_to_inode(dquot),
3760 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3761 if (IS_ERR(handle)) {
3762 /* Release dquot anyway to avoid endless cycle in dqput() */
3763 dquot_release(dquot);
3764 return PTR_ERR(handle);
3766 ret = dquot_release(dquot);
3767 err = ext4_journal_stop(handle);
3773 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3775 /* Are we journaling quotas? */
3776 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3777 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3778 dquot_mark_dquot_dirty(dquot);
3779 return ext4_write_dquot(dquot);
3781 return dquot_mark_dquot_dirty(dquot);
3785 static int ext4_write_info(struct super_block *sb, int type)
3790 /* Data block + inode block */
3791 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3793 return PTR_ERR(handle);
3794 ret = dquot_commit_info(sb, type);
3795 err = ext4_journal_stop(handle);
3802 * Turn on quotas during mount time - we need to find
3803 * the quota file and such...
3805 static int ext4_quota_on_mount(struct super_block *sb, int type)
3807 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3808 EXT4_SB(sb)->s_jquota_fmt, type);
3812 * Standard function to be called on quota_on
3814 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3815 char *name, int remount)
3820 if (!test_opt(sb, QUOTA))
3822 /* When remounting, no checks are needed and in fact, name is NULL */
3824 return vfs_quota_on(sb, type, format_id, name, remount);
3826 err = kern_path(name, LOOKUP_FOLLOW, &path);
3830 /* Quotafile not on the same filesystem? */
3831 if (path.mnt->mnt_sb != sb) {
3835 /* Journaling quota? */
3836 if (EXT4_SB(sb)->s_qf_names[type]) {
3837 /* Quotafile not in fs root? */
3838 if (path.dentry->d_parent != sb->s_root)
3839 ext4_msg(sb, KERN_WARNING,
3840 "Quota file not on filesystem root. "
3841 "Journaled quota will not work");
3845 * When we journal data on quota file, we have to flush journal to see
3846 * all updates to the file when we bypass pagecache...
3848 if (EXT4_SB(sb)->s_journal &&
3849 ext4_should_journal_data(path.dentry->d_inode)) {
3851 * We don't need to lock updates but journal_flush() could
3852 * otherwise be livelocked...
3854 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3855 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3856 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3863 err = vfs_quota_on_path(sb, type, format_id, &path);
3868 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3869 * acquiring the locks... As quota files are never truncated and quota code
3870 * itself serializes the operations (and noone else should touch the files)
3871 * we don't have to be afraid of races */
3872 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3873 size_t len, loff_t off)
3875 struct inode *inode = sb_dqopt(sb)->files[type];
3876 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3878 int offset = off & (sb->s_blocksize - 1);
3881 struct buffer_head *bh;
3882 loff_t i_size = i_size_read(inode);
3886 if (off+len > i_size)
3889 while (toread > 0) {
3890 tocopy = sb->s_blocksize - offset < toread ?
3891 sb->s_blocksize - offset : toread;
3892 bh = ext4_bread(NULL, inode, blk, 0, &err);
3895 if (!bh) /* A hole? */
3896 memset(data, 0, tocopy);
3898 memcpy(data, bh->b_data+offset, tocopy);
3908 /* Write to quotafile (we know the transaction is already started and has
3909 * enough credits) */
3910 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3911 const char *data, size_t len, loff_t off)
3913 struct inode *inode = sb_dqopt(sb)->files[type];
3914 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3916 int offset = off & (sb->s_blocksize - 1);
3918 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3919 size_t towrite = len;
3920 struct buffer_head *bh;
3921 handle_t *handle = journal_current_handle();
3923 if (EXT4_SB(sb)->s_journal && !handle) {
3924 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3925 " cancelled because transaction is not started",
3926 (unsigned long long)off, (unsigned long long)len);
3929 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3930 while (towrite > 0) {
3931 tocopy = sb->s_blocksize - offset < towrite ?
3932 sb->s_blocksize - offset : towrite;
3933 bh = ext4_bread(handle, inode, blk, 1, &err);
3936 if (journal_quota) {
3937 err = ext4_journal_get_write_access(handle, bh);
3944 memcpy(bh->b_data+offset, data, tocopy);
3945 flush_dcache_page(bh->b_page);
3948 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3950 /* Always do at least ordered writes for quotas */
3951 err = ext4_jbd2_file_inode(handle, inode);
3952 mark_buffer_dirty(bh);
3963 if (len == towrite) {
3964 mutex_unlock(&inode->i_mutex);
3967 if (inode->i_size < off+len-towrite) {
3968 i_size_write(inode, off+len-towrite);
3969 EXT4_I(inode)->i_disksize = inode->i_size;
3971 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3972 ext4_mark_inode_dirty(handle, inode);
3973 mutex_unlock(&inode->i_mutex);
3974 return len - towrite;
3979 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3980 const char *dev_name, void *data, struct vfsmount *mnt)
3982 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3985 static struct file_system_type ext4_fs_type = {
3986 .owner = THIS_MODULE,
3988 .get_sb = ext4_get_sb,
3989 .kill_sb = kill_block_super,
3990 .fs_flags = FS_REQUIRES_DEV,
3993 static int __init init_ext4_fs(void)
3997 err = init_ext4_system_zone();
4000 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4003 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4004 err = init_ext4_mballoc();
4008 err = init_ext4_xattr();
4011 err = init_inodecache();
4014 err = register_filesystem(&ext4_fs_type);
4019 destroy_inodecache();
4023 exit_ext4_mballoc();
4025 remove_proc_entry("fs/ext4", NULL);
4026 kset_unregister(ext4_kset);
4028 exit_ext4_system_zone();
4032 static void __exit exit_ext4_fs(void)
4034 unregister_filesystem(&ext4_fs_type);
4035 destroy_inodecache();
4037 exit_ext4_mballoc();
4038 remove_proc_entry("fs/ext4", NULL);
4039 kset_unregister(ext4_kset);
4040 exit_ext4_system_zone();
4043 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4044 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4045 MODULE_LICENSE("GPL");
4046 module_init(init_ext4_fs)
4047 module_exit(exit_ext4_fs)