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);
71 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type = {
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #define IS_EXT3_SB(sb) (0)
87 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
95 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
103 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le32_to_cpu(bg->bg_inode_table_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
111 __u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
119 __u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
127 __u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
135 __u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
143 void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
151 void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
159 void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
167 void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
175 void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
183 void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
191 void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
200 /* Just increment the non-pointer handle value */
201 static handle_t *ext4_get_nojournal(void)
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
209 handle = (handle_t *)ref_cnt;
211 current->journal_info = handle;
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t *handle)
219 unsigned long ref_cnt = (unsigned long)handle;
221 BUG_ON(ref_cnt == 0);
224 handle = (handle_t *)ref_cnt;
226 current->journal_info = handle;
230 * Wrappers for jbd2_journal_start/end.
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
237 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
244 vfs_check_frozen(sb, SB_FREEZE_WRITE);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal;
250 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, __func__, "Detected aborted journal");
252 return ERR_PTR(-EROFS);
254 return jbd2_journal_start(journal, nblocks);
256 return ext4_get_nojournal();
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
265 int __ext4_journal_stop(const char *where, handle_t *handle)
267 struct super_block *sb;
271 if (!ext4_handle_valid(handle)) {
272 ext4_put_nojournal(handle);
275 sb = handle->h_transaction->t_journal->j_private;
277 rc = jbd2_journal_stop(handle);
282 __ext4_std_error(sb, where, err);
286 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
287 struct buffer_head *bh, handle_t *handle, int err)
290 const char *errstr = ext4_decode_error(NULL, err, nbuf);
292 BUG_ON(!ext4_handle_valid(handle));
295 BUFFER_TRACE(bh, "abort");
300 if (is_handle_aborted(handle))
303 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
304 caller, errstr, err_fn);
306 jbd2_journal_abort_handle(handle);
309 /* Deal with the reporting of failure conditions on a filesystem such as
310 * inconsistencies detected or read IO failures.
312 * On ext2, we can store the error state of the filesystem in the
313 * superblock. That is not possible on ext4, because we may have other
314 * write ordering constraints on the superblock which prevent us from
315 * writing it out straight away; and given that the journal is about to
316 * be aborted, we can't rely on the current, or future, transactions to
317 * write out the superblock safely.
319 * We'll just use the jbd2_journal_abort() error code to record an error in
320 * the journal instead. On recovery, the journal will complain about
321 * that error until we've noted it down and cleared it.
324 static void ext4_handle_error(struct super_block *sb)
326 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
328 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
329 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
331 if (sb->s_flags & MS_RDONLY)
334 if (!test_opt(sb, ERRORS_CONT)) {
335 journal_t *journal = EXT4_SB(sb)->s_journal;
337 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
339 jbd2_journal_abort(journal, -EIO);
341 if (test_opt(sb, ERRORS_RO)) {
342 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
343 sb->s_flags |= MS_RDONLY;
345 ext4_commit_super(sb, 1);
346 if (test_opt(sb, ERRORS_PANIC))
347 panic("EXT4-fs (device %s): panic forced after error\n",
351 void __ext4_error(struct super_block *sb, const char *function,
352 const char *fmt, ...)
357 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
362 ext4_handle_error(sb);
365 void ext4_error_inode(const char *function, struct inode *inode,
366 const char *fmt, ...)
371 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
372 inode->i_sb->s_id, function, inode->i_ino, current->comm);
377 ext4_handle_error(inode->i_sb);
380 void ext4_error_file(const char *function, struct file *file,
381 const char *fmt, ...)
384 struct inode *inode = file->f_dentry->d_inode;
385 char pathname[80], *path;
388 path = d_path(&(file->f_path), pathname, sizeof(pathname));
392 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
393 inode->i_sb->s_id, function, inode->i_ino, current->comm, path);
398 ext4_handle_error(inode->i_sb);
401 static const char *ext4_decode_error(struct super_block *sb, int errno,
408 errstr = "IO failure";
411 errstr = "Out of memory";
414 if (!sb || (EXT4_SB(sb)->s_journal &&
415 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
416 errstr = "Journal has aborted";
418 errstr = "Readonly filesystem";
421 /* If the caller passed in an extra buffer for unknown
422 * errors, textualise them now. Else we just return
425 /* Check for truncated error codes... */
426 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
435 /* __ext4_std_error decodes expected errors from journaling functions
436 * automatically and invokes the appropriate error response. */
438 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
443 /* Special case: if the error is EROFS, and we're not already
444 * inside a transaction, then there's really no point in logging
446 if (errno == -EROFS && journal_current_handle() == NULL &&
447 (sb->s_flags & MS_RDONLY))
450 errstr = ext4_decode_error(sb, errno, nbuf);
451 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
452 sb->s_id, function, errstr);
454 ext4_handle_error(sb);
458 * ext4_abort is a much stronger failure handler than ext4_error. The
459 * abort function may be used to deal with unrecoverable failures such
460 * as journal IO errors or ENOMEM at a critical moment in log management.
462 * We unconditionally force the filesystem into an ABORT|READONLY state,
463 * unless the error response on the fs has been set to panic in which
464 * case we take the easy way out and panic immediately.
467 void ext4_abort(struct super_block *sb, const char *function,
468 const char *fmt, ...)
473 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
478 if (test_opt(sb, ERRORS_PANIC))
479 panic("EXT4-fs panic from previous error\n");
481 if (sb->s_flags & MS_RDONLY)
484 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
485 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
486 sb->s_flags |= MS_RDONLY;
487 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
488 if (EXT4_SB(sb)->s_journal)
489 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
492 void ext4_msg (struct super_block * sb, const char *prefix,
493 const char *fmt, ...)
498 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
504 void __ext4_warning(struct super_block *sb, const char *function,
505 const char *fmt, ...)
510 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
517 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
518 const char *function, const char *fmt, ...)
523 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
526 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
531 if (test_opt(sb, ERRORS_CONT)) {
532 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
533 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
534 ext4_commit_super(sb, 0);
537 ext4_unlock_group(sb, grp);
538 ext4_handle_error(sb);
540 * We only get here in the ERRORS_RO case; relocking the group
541 * may be dangerous, but nothing bad will happen since the
542 * filesystem will have already been marked read/only and the
543 * journal has been aborted. We return 1 as a hint to callers
544 * who might what to use the return value from
545 * ext4_grp_locked_error() to distinguish beween the
546 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
547 * aggressively from the ext4 function in question, with a
548 * more appropriate error code.
550 ext4_lock_group(sb, grp);
554 void ext4_update_dynamic_rev(struct super_block *sb)
556 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
558 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
562 "updating to rev %d because of new feature flag, "
563 "running e2fsck is recommended",
566 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
567 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
568 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
569 /* leave es->s_feature_*compat flags alone */
570 /* es->s_uuid will be set by e2fsck if empty */
573 * The rest of the superblock fields should be zero, and if not it
574 * means they are likely already in use, so leave them alone. We
575 * can leave it up to e2fsck to clean up any inconsistencies there.
580 * Open the external journal device
582 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
584 struct block_device *bdev;
585 char b[BDEVNAME_SIZE];
587 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
593 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
594 __bdevname(dev, b), PTR_ERR(bdev));
599 * Release the journal device
601 static int ext4_blkdev_put(struct block_device *bdev)
604 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
607 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
609 struct block_device *bdev;
612 bdev = sbi->journal_bdev;
614 ret = ext4_blkdev_put(bdev);
615 sbi->journal_bdev = NULL;
620 static inline struct inode *orphan_list_entry(struct list_head *l)
622 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
625 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
629 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
630 le32_to_cpu(sbi->s_es->s_last_orphan));
632 printk(KERN_ERR "sb_info orphan list:\n");
633 list_for_each(l, &sbi->s_orphan) {
634 struct inode *inode = orphan_list_entry(l);
636 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
637 inode->i_sb->s_id, inode->i_ino, inode,
638 inode->i_mode, inode->i_nlink,
643 static void ext4_put_super(struct super_block *sb)
645 struct ext4_sb_info *sbi = EXT4_SB(sb);
646 struct ext4_super_block *es = sbi->s_es;
649 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
651 flush_workqueue(sbi->dio_unwritten_wq);
652 destroy_workqueue(sbi->dio_unwritten_wq);
657 ext4_commit_super(sb, 1);
659 if (sbi->s_journal) {
660 err = jbd2_journal_destroy(sbi->s_journal);
661 sbi->s_journal = NULL;
663 ext4_abort(sb, __func__,
664 "Couldn't clean up the journal");
667 ext4_release_system_zone(sb);
669 ext4_ext_release(sb);
670 ext4_xattr_put_super(sb);
672 if (!(sb->s_flags & MS_RDONLY)) {
673 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
674 es->s_state = cpu_to_le16(sbi->s_mount_state);
675 ext4_commit_super(sb, 1);
678 remove_proc_entry(sb->s_id, ext4_proc_root);
680 kobject_del(&sbi->s_kobj);
682 for (i = 0; i < sbi->s_gdb_count; i++)
683 brelse(sbi->s_group_desc[i]);
684 kfree(sbi->s_group_desc);
685 if (is_vmalloc_addr(sbi->s_flex_groups))
686 vfree(sbi->s_flex_groups);
688 kfree(sbi->s_flex_groups);
689 percpu_counter_destroy(&sbi->s_freeblocks_counter);
690 percpu_counter_destroy(&sbi->s_freeinodes_counter);
691 percpu_counter_destroy(&sbi->s_dirs_counter);
692 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
695 for (i = 0; i < MAXQUOTAS; i++)
696 kfree(sbi->s_qf_names[i]);
699 /* Debugging code just in case the in-memory inode orphan list
700 * isn't empty. The on-disk one can be non-empty if we've
701 * detected an error and taken the fs readonly, but the
702 * in-memory list had better be clean by this point. */
703 if (!list_empty(&sbi->s_orphan))
704 dump_orphan_list(sb, sbi);
705 J_ASSERT(list_empty(&sbi->s_orphan));
707 invalidate_bdev(sb->s_bdev);
708 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
710 * Invalidate the journal device's buffers. We don't want them
711 * floating about in memory - the physical journal device may
712 * hotswapped, and it breaks the `ro-after' testing code.
714 sync_blockdev(sbi->journal_bdev);
715 invalidate_bdev(sbi->journal_bdev);
716 ext4_blkdev_remove(sbi);
718 sb->s_fs_info = NULL;
720 * Now that we are completely done shutting down the
721 * superblock, we need to actually destroy the kobject.
725 kobject_put(&sbi->s_kobj);
726 wait_for_completion(&sbi->s_kobj_unregister);
727 kfree(sbi->s_blockgroup_lock);
731 static struct kmem_cache *ext4_inode_cachep;
734 * Called inside transaction, so use GFP_NOFS
736 static struct inode *ext4_alloc_inode(struct super_block *sb)
738 struct ext4_inode_info *ei;
740 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
744 ei->vfs_inode.i_version = 1;
745 ei->vfs_inode.i_data.writeback_index = 0;
746 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
747 INIT_LIST_HEAD(&ei->i_prealloc_list);
748 spin_lock_init(&ei->i_prealloc_lock);
750 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
751 * therefore it can be null here. Don't check it, just initialize
754 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
755 ei->i_reserved_data_blocks = 0;
756 ei->i_reserved_meta_blocks = 0;
757 ei->i_allocated_meta_blocks = 0;
758 ei->i_da_metadata_calc_len = 0;
759 ei->i_delalloc_reserved_flag = 0;
760 spin_lock_init(&(ei->i_block_reservation_lock));
762 ei->i_reserved_quota = 0;
764 INIT_LIST_HEAD(&ei->i_completed_io_list);
765 spin_lock_init(&ei->i_completed_io_lock);
766 ei->cur_aio_dio = NULL;
768 ei->i_datasync_tid = 0;
770 return &ei->vfs_inode;
773 static void ext4_destroy_inode(struct inode *inode)
775 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
776 ext4_msg(inode->i_sb, KERN_ERR,
777 "Inode %lu (%p): orphan list check failed!",
778 inode->i_ino, EXT4_I(inode));
779 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
780 EXT4_I(inode), sizeof(struct ext4_inode_info),
784 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
787 static void init_once(void *foo)
789 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
791 INIT_LIST_HEAD(&ei->i_orphan);
792 #ifdef CONFIG_EXT4_FS_XATTR
793 init_rwsem(&ei->xattr_sem);
795 init_rwsem(&ei->i_data_sem);
796 inode_init_once(&ei->vfs_inode);
799 static int init_inodecache(void)
801 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
802 sizeof(struct ext4_inode_info),
803 0, (SLAB_RECLAIM_ACCOUNT|
806 if (ext4_inode_cachep == NULL)
811 static void destroy_inodecache(void)
813 kmem_cache_destroy(ext4_inode_cachep);
816 static void ext4_clear_inode(struct inode *inode)
819 ext4_discard_preallocations(inode);
820 if (EXT4_JOURNAL(inode))
821 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
822 &EXT4_I(inode)->jinode);
825 static inline void ext4_show_quota_options(struct seq_file *seq,
826 struct super_block *sb)
828 #if defined(CONFIG_QUOTA)
829 struct ext4_sb_info *sbi = EXT4_SB(sb);
831 if (sbi->s_jquota_fmt) {
834 switch (sbi->s_jquota_fmt) {
845 seq_printf(seq, ",jqfmt=%s", fmtname);
848 if (sbi->s_qf_names[USRQUOTA])
849 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
851 if (sbi->s_qf_names[GRPQUOTA])
852 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
854 if (test_opt(sb, USRQUOTA))
855 seq_puts(seq, ",usrquota");
857 if (test_opt(sb, GRPQUOTA))
858 seq_puts(seq, ",grpquota");
864 * - it's set to a non-default value OR
865 * - if the per-sb default is different from the global default
867 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
870 unsigned long def_mount_opts;
871 struct super_block *sb = vfs->mnt_sb;
872 struct ext4_sb_info *sbi = EXT4_SB(sb);
873 struct ext4_super_block *es = sbi->s_es;
875 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
876 def_errors = le16_to_cpu(es->s_errors);
878 if (sbi->s_sb_block != 1)
879 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
880 if (test_opt(sb, MINIX_DF))
881 seq_puts(seq, ",minixdf");
882 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
883 seq_puts(seq, ",grpid");
884 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
885 seq_puts(seq, ",nogrpid");
886 if (sbi->s_resuid != EXT4_DEF_RESUID ||
887 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
888 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
890 if (sbi->s_resgid != EXT4_DEF_RESGID ||
891 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
892 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
894 if (test_opt(sb, ERRORS_RO)) {
895 if (def_errors == EXT4_ERRORS_PANIC ||
896 def_errors == EXT4_ERRORS_CONTINUE) {
897 seq_puts(seq, ",errors=remount-ro");
900 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
901 seq_puts(seq, ",errors=continue");
902 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
903 seq_puts(seq, ",errors=panic");
904 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
905 seq_puts(seq, ",nouid32");
906 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
907 seq_puts(seq, ",debug");
908 if (test_opt(sb, OLDALLOC))
909 seq_puts(seq, ",oldalloc");
910 #ifdef CONFIG_EXT4_FS_XATTR
911 if (test_opt(sb, XATTR_USER) &&
912 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
913 seq_puts(seq, ",user_xattr");
914 if (!test_opt(sb, XATTR_USER) &&
915 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
916 seq_puts(seq, ",nouser_xattr");
919 #ifdef CONFIG_EXT4_FS_POSIX_ACL
920 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
921 seq_puts(seq, ",acl");
922 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
923 seq_puts(seq, ",noacl");
925 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
926 seq_printf(seq, ",commit=%u",
927 (unsigned) (sbi->s_commit_interval / HZ));
929 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
930 seq_printf(seq, ",min_batch_time=%u",
931 (unsigned) sbi->s_min_batch_time);
933 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
934 seq_printf(seq, ",max_batch_time=%u",
935 (unsigned) sbi->s_min_batch_time);
939 * We're changing the default of barrier mount option, so
940 * let's always display its mount state so it's clear what its
943 seq_puts(seq, ",barrier=");
944 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
945 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
946 seq_puts(seq, ",journal_async_commit");
947 else if (test_opt(sb, JOURNAL_CHECKSUM))
948 seq_puts(seq, ",journal_checksum");
949 if (test_opt(sb, I_VERSION))
950 seq_puts(seq, ",i_version");
951 if (!test_opt(sb, DELALLOC))
952 seq_puts(seq, ",nodelalloc");
956 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
958 * journal mode get enabled in different ways
959 * So just print the value even if we didn't specify it
961 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
962 seq_puts(seq, ",data=journal");
963 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
964 seq_puts(seq, ",data=ordered");
965 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
966 seq_puts(seq, ",data=writeback");
968 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
969 seq_printf(seq, ",inode_readahead_blks=%u",
970 sbi->s_inode_readahead_blks);
972 if (test_opt(sb, DATA_ERR_ABORT))
973 seq_puts(seq, ",data_err=abort");
975 if (test_opt(sb, NO_AUTO_DA_ALLOC))
976 seq_puts(seq, ",noauto_da_alloc");
978 if (test_opt(sb, DISCARD))
979 seq_puts(seq, ",discard");
981 if (test_opt(sb, NOLOAD))
982 seq_puts(seq, ",norecovery");
984 if (test_opt(sb, DIOREAD_NOLOCK))
985 seq_puts(seq, ",dioread_nolock");
987 ext4_show_quota_options(seq, sb);
992 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
993 u64 ino, u32 generation)
997 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
998 return ERR_PTR(-ESTALE);
999 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1000 return ERR_PTR(-ESTALE);
1002 /* iget isn't really right if the inode is currently unallocated!!
1004 * ext4_read_inode will return a bad_inode if the inode had been
1005 * deleted, so we should be safe.
1007 * Currently we don't know the generation for parent directory, so
1008 * a generation of 0 means "accept any"
1010 inode = ext4_iget(sb, ino);
1012 return ERR_CAST(inode);
1013 if (generation && inode->i_generation != generation) {
1015 return ERR_PTR(-ESTALE);
1021 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1022 int fh_len, int fh_type)
1024 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1025 ext4_nfs_get_inode);
1028 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1029 int fh_len, int fh_type)
1031 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1032 ext4_nfs_get_inode);
1036 * Try to release metadata pages (indirect blocks, directories) which are
1037 * mapped via the block device. Since these pages could have journal heads
1038 * which would prevent try_to_free_buffers() from freeing them, we must use
1039 * jbd2 layer's try_to_free_buffers() function to release them.
1041 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1044 journal_t *journal = EXT4_SB(sb)->s_journal;
1046 WARN_ON(PageChecked(page));
1047 if (!page_has_buffers(page))
1050 return jbd2_journal_try_to_free_buffers(journal, page,
1051 wait & ~__GFP_WAIT);
1052 return try_to_free_buffers(page);
1056 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1057 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1059 static int ext4_write_dquot(struct dquot *dquot);
1060 static int ext4_acquire_dquot(struct dquot *dquot);
1061 static int ext4_release_dquot(struct dquot *dquot);
1062 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1063 static int ext4_write_info(struct super_block *sb, int type);
1064 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1066 static int ext4_quota_on_mount(struct super_block *sb, int type);
1067 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1068 size_t len, loff_t off);
1069 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1070 const char *data, size_t len, loff_t off);
1072 static const struct dquot_operations ext4_quota_operations = {
1074 .get_reserved_space = ext4_get_reserved_space,
1076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
1080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
1085 static const struct quotactl_ops ext4_qctl_operations = {
1086 .quota_on = ext4_quota_on,
1087 .quota_off = dquot_quota_off,
1088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
1096 static const struct super_operations ext4_sops = {
1097 .alloc_inode = ext4_alloc_inode,
1098 .destroy_inode = ext4_destroy_inode,
1099 .write_inode = ext4_write_inode,
1100 .dirty_inode = ext4_dirty_inode,
1101 .delete_inode = ext4_delete_inode,
1102 .put_super = ext4_put_super,
1103 .sync_fs = ext4_sync_fs,
1104 .freeze_fs = ext4_freeze,
1105 .unfreeze_fs = ext4_unfreeze,
1106 .statfs = ext4_statfs,
1107 .remount_fs = ext4_remount,
1108 .clear_inode = ext4_clear_inode,
1109 .show_options = ext4_show_options,
1111 .quota_read = ext4_quota_read,
1112 .quota_write = ext4_quota_write,
1114 .bdev_try_to_free_page = bdev_try_to_free_page,
1117 static const struct super_operations ext4_nojournal_sops = {
1118 .alloc_inode = ext4_alloc_inode,
1119 .destroy_inode = ext4_destroy_inode,
1120 .write_inode = ext4_write_inode,
1121 .dirty_inode = ext4_dirty_inode,
1122 .delete_inode = ext4_delete_inode,
1123 .write_super = ext4_write_super,
1124 .put_super = ext4_put_super,
1125 .statfs = ext4_statfs,
1126 .remount_fs = ext4_remount,
1127 .clear_inode = ext4_clear_inode,
1128 .show_options = ext4_show_options,
1130 .quota_read = ext4_quota_read,
1131 .quota_write = ext4_quota_write,
1133 .bdev_try_to_free_page = bdev_try_to_free_page,
1136 static const struct export_operations ext4_export_ops = {
1137 .fh_to_dentry = ext4_fh_to_dentry,
1138 .fh_to_parent = ext4_fh_to_parent,
1139 .get_parent = ext4_get_parent,
1143 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1144 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1145 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1146 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1147 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1148 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1149 Opt_journal_update, Opt_journal_dev,
1150 Opt_journal_checksum, Opt_journal_async_commit,
1151 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1152 Opt_data_err_abort, Opt_data_err_ignore,
1153 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1154 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1155 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1156 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1157 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1158 Opt_block_validity, Opt_noblock_validity,
1159 Opt_inode_readahead_blks, Opt_journal_ioprio,
1160 Opt_dioread_nolock, Opt_dioread_lock,
1161 Opt_discard, Opt_nodiscard,
1164 static const match_table_t tokens = {
1165 {Opt_bsd_df, "bsddf"},
1166 {Opt_minix_df, "minixdf"},
1167 {Opt_grpid, "grpid"},
1168 {Opt_grpid, "bsdgroups"},
1169 {Opt_nogrpid, "nogrpid"},
1170 {Opt_nogrpid, "sysvgroups"},
1171 {Opt_resgid, "resgid=%u"},
1172 {Opt_resuid, "resuid=%u"},
1174 {Opt_err_cont, "errors=continue"},
1175 {Opt_err_panic, "errors=panic"},
1176 {Opt_err_ro, "errors=remount-ro"},
1177 {Opt_nouid32, "nouid32"},
1178 {Opt_debug, "debug"},
1179 {Opt_oldalloc, "oldalloc"},
1180 {Opt_orlov, "orlov"},
1181 {Opt_user_xattr, "user_xattr"},
1182 {Opt_nouser_xattr, "nouser_xattr"},
1184 {Opt_noacl, "noacl"},
1185 {Opt_noload, "noload"},
1186 {Opt_noload, "norecovery"},
1189 {Opt_commit, "commit=%u"},
1190 {Opt_min_batch_time, "min_batch_time=%u"},
1191 {Opt_max_batch_time, "max_batch_time=%u"},
1192 {Opt_journal_update, "journal=update"},
1193 {Opt_journal_dev, "journal_dev=%u"},
1194 {Opt_journal_checksum, "journal_checksum"},
1195 {Opt_journal_async_commit, "journal_async_commit"},
1196 {Opt_abort, "abort"},
1197 {Opt_data_journal, "data=journal"},
1198 {Opt_data_ordered, "data=ordered"},
1199 {Opt_data_writeback, "data=writeback"},
1200 {Opt_data_err_abort, "data_err=abort"},
1201 {Opt_data_err_ignore, "data_err=ignore"},
1202 {Opt_offusrjquota, "usrjquota="},
1203 {Opt_usrjquota, "usrjquota=%s"},
1204 {Opt_offgrpjquota, "grpjquota="},
1205 {Opt_grpjquota, "grpjquota=%s"},
1206 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1207 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1208 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1209 {Opt_grpquota, "grpquota"},
1210 {Opt_noquota, "noquota"},
1211 {Opt_quota, "quota"},
1212 {Opt_usrquota, "usrquota"},
1213 {Opt_barrier, "barrier=%u"},
1214 {Opt_barrier, "barrier"},
1215 {Opt_nobarrier, "nobarrier"},
1216 {Opt_i_version, "i_version"},
1217 {Opt_stripe, "stripe=%u"},
1218 {Opt_resize, "resize"},
1219 {Opt_delalloc, "delalloc"},
1220 {Opt_nodelalloc, "nodelalloc"},
1221 {Opt_block_validity, "block_validity"},
1222 {Opt_noblock_validity, "noblock_validity"},
1223 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1224 {Opt_journal_ioprio, "journal_ioprio=%u"},
1225 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1226 {Opt_auto_da_alloc, "auto_da_alloc"},
1227 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1228 {Opt_dioread_nolock, "dioread_nolock"},
1229 {Opt_dioread_lock, "dioread_lock"},
1230 {Opt_discard, "discard"},
1231 {Opt_nodiscard, "nodiscard"},
1235 static ext4_fsblk_t get_sb_block(void **data)
1237 ext4_fsblk_t sb_block;
1238 char *options = (char *) *data;
1240 if (!options || strncmp(options, "sb=", 3) != 0)
1241 return 1; /* Default location */
1244 /* TODO: use simple_strtoll with >32bit ext4 */
1245 sb_block = simple_strtoul(options, &options, 0);
1246 if (*options && *options != ',') {
1247 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1251 if (*options == ',')
1253 *data = (void *) options;
1258 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1259 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1260 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1263 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1265 struct ext4_sb_info *sbi = EXT4_SB(sb);
1268 if (sb_any_quota_loaded(sb) &&
1269 !sbi->s_qf_names[qtype]) {
1270 ext4_msg(sb, KERN_ERR,
1271 "Cannot change journaled "
1272 "quota options when quota turned on");
1275 qname = match_strdup(args);
1277 ext4_msg(sb, KERN_ERR,
1278 "Not enough memory for storing quotafile name");
1281 if (sbi->s_qf_names[qtype] &&
1282 strcmp(sbi->s_qf_names[qtype], qname)) {
1283 ext4_msg(sb, KERN_ERR,
1284 "%s quota file already specified", QTYPE2NAME(qtype));
1288 sbi->s_qf_names[qtype] = qname;
1289 if (strchr(sbi->s_qf_names[qtype], '/')) {
1290 ext4_msg(sb, KERN_ERR,
1291 "quotafile must be on filesystem root");
1292 kfree(sbi->s_qf_names[qtype]);
1293 sbi->s_qf_names[qtype] = NULL;
1296 set_opt(sbi->s_mount_opt, QUOTA);
1300 static int clear_qf_name(struct super_block *sb, int qtype)
1303 struct ext4_sb_info *sbi = EXT4_SB(sb);
1305 if (sb_any_quota_loaded(sb) &&
1306 sbi->s_qf_names[qtype]) {
1307 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1308 " when quota turned on");
1312 * The space will be released later when all options are confirmed
1315 sbi->s_qf_names[qtype] = NULL;
1320 static int parse_options(char *options, struct super_block *sb,
1321 unsigned long *journal_devnum,
1322 unsigned int *journal_ioprio,
1323 ext4_fsblk_t *n_blocks_count, int is_remount)
1325 struct ext4_sb_info *sbi = EXT4_SB(sb);
1327 substring_t args[MAX_OPT_ARGS];
1337 while ((p = strsep(&options, ",")) != NULL) {
1343 * Initialize args struct so we know whether arg was
1344 * found; some options take optional arguments.
1346 args[0].to = args[0].from = 0;
1347 token = match_token(p, tokens, args);
1350 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1351 clear_opt(sbi->s_mount_opt, MINIX_DF);
1354 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1355 set_opt(sbi->s_mount_opt, MINIX_DF);
1359 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1360 set_opt(sbi->s_mount_opt, GRPID);
1364 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1365 clear_opt(sbi->s_mount_opt, GRPID);
1369 if (match_int(&args[0], &option))
1371 sbi->s_resuid = option;
1374 if (match_int(&args[0], &option))
1376 sbi->s_resgid = option;
1379 /* handled by get_sb_block() instead of here */
1380 /* *sb_block = match_int(&args[0]); */
1383 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1384 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1385 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1388 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1389 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1390 set_opt(sbi->s_mount_opt, ERRORS_RO);
1393 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1394 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1395 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1398 set_opt(sbi->s_mount_opt, NO_UID32);
1401 set_opt(sbi->s_mount_opt, DEBUG);
1404 set_opt(sbi->s_mount_opt, OLDALLOC);
1407 clear_opt(sbi->s_mount_opt, OLDALLOC);
1409 #ifdef CONFIG_EXT4_FS_XATTR
1410 case Opt_user_xattr:
1411 set_opt(sbi->s_mount_opt, XATTR_USER);
1413 case Opt_nouser_xattr:
1414 clear_opt(sbi->s_mount_opt, XATTR_USER);
1417 case Opt_user_xattr:
1418 case Opt_nouser_xattr:
1419 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1424 set_opt(sbi->s_mount_opt, POSIX_ACL);
1427 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1432 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1435 case Opt_journal_update:
1437 /* Eventually we will want to be able to create
1438 a journal file here. For now, only allow the
1439 user to specify an existing inode to be the
1442 ext4_msg(sb, KERN_ERR,
1443 "Cannot specify journal on remount");
1446 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1448 case Opt_journal_dev:
1450 ext4_msg(sb, KERN_ERR,
1451 "Cannot specify journal on remount");
1454 if (match_int(&args[0], &option))
1456 *journal_devnum = option;
1458 case Opt_journal_checksum:
1459 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1461 case Opt_journal_async_commit:
1462 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1463 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1466 set_opt(sbi->s_mount_opt, NOLOAD);
1469 if (match_int(&args[0], &option))
1474 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1475 sbi->s_commit_interval = HZ * option;
1477 case Opt_max_batch_time:
1478 if (match_int(&args[0], &option))
1483 option = EXT4_DEF_MAX_BATCH_TIME;
1484 sbi->s_max_batch_time = option;
1486 case Opt_min_batch_time:
1487 if (match_int(&args[0], &option))
1491 sbi->s_min_batch_time = option;
1493 case Opt_data_journal:
1494 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1496 case Opt_data_ordered:
1497 data_opt = EXT4_MOUNT_ORDERED_DATA;
1499 case Opt_data_writeback:
1500 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1503 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1504 ext4_msg(sb, KERN_ERR,
1505 "Cannot change data mode on remount");
1509 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1510 sbi->s_mount_opt |= data_opt;
1513 case Opt_data_err_abort:
1514 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1516 case Opt_data_err_ignore:
1517 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1521 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1525 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1528 case Opt_offusrjquota:
1529 if (!clear_qf_name(sb, USRQUOTA))
1532 case Opt_offgrpjquota:
1533 if (!clear_qf_name(sb, GRPQUOTA))
1537 case Opt_jqfmt_vfsold:
1538 qfmt = QFMT_VFS_OLD;
1540 case Opt_jqfmt_vfsv0:
1543 case Opt_jqfmt_vfsv1:
1546 if (sb_any_quota_loaded(sb) &&
1547 sbi->s_jquota_fmt != qfmt) {
1548 ext4_msg(sb, KERN_ERR, "Cannot change "
1549 "journaled quota options when "
1553 sbi->s_jquota_fmt = qfmt;
1557 set_opt(sbi->s_mount_opt, QUOTA);
1558 set_opt(sbi->s_mount_opt, USRQUOTA);
1561 set_opt(sbi->s_mount_opt, QUOTA);
1562 set_opt(sbi->s_mount_opt, GRPQUOTA);
1565 if (sb_any_quota_loaded(sb)) {
1566 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1567 "options when quota turned on");
1570 clear_opt(sbi->s_mount_opt, QUOTA);
1571 clear_opt(sbi->s_mount_opt, USRQUOTA);
1572 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1578 ext4_msg(sb, KERN_ERR,
1579 "quota options not supported");
1583 case Opt_offusrjquota:
1584 case Opt_offgrpjquota:
1585 case Opt_jqfmt_vfsold:
1586 case Opt_jqfmt_vfsv0:
1587 case Opt_jqfmt_vfsv1:
1588 ext4_msg(sb, KERN_ERR,
1589 "journaled quota options not supported");
1595 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1598 clear_opt(sbi->s_mount_opt, BARRIER);
1602 if (match_int(&args[0], &option))
1605 option = 1; /* No argument, default to 1 */
1607 set_opt(sbi->s_mount_opt, BARRIER);
1609 clear_opt(sbi->s_mount_opt, BARRIER);
1615 ext4_msg(sb, KERN_ERR,
1616 "resize option only available "
1620 if (match_int(&args[0], &option) != 0)
1622 *n_blocks_count = option;
1625 ext4_msg(sb, KERN_WARNING,
1626 "Ignoring deprecated nobh option");
1629 ext4_msg(sb, KERN_WARNING,
1630 "Ignoring deprecated bh option");
1633 set_opt(sbi->s_mount_opt, I_VERSION);
1634 sb->s_flags |= MS_I_VERSION;
1636 case Opt_nodelalloc:
1637 clear_opt(sbi->s_mount_opt, DELALLOC);
1640 if (match_int(&args[0], &option))
1644 sbi->s_stripe = option;
1647 set_opt(sbi->s_mount_opt, DELALLOC);
1649 case Opt_block_validity:
1650 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1652 case Opt_noblock_validity:
1653 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1655 case Opt_inode_readahead_blks:
1656 if (match_int(&args[0], &option))
1658 if (option < 0 || option > (1 << 30))
1660 if (!is_power_of_2(option)) {
1661 ext4_msg(sb, KERN_ERR,
1662 "EXT4-fs: inode_readahead_blks"
1663 " must be a power of 2");
1666 sbi->s_inode_readahead_blks = option;
1668 case Opt_journal_ioprio:
1669 if (match_int(&args[0], &option))
1671 if (option < 0 || option > 7)
1673 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1676 case Opt_noauto_da_alloc:
1677 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1679 case Opt_auto_da_alloc:
1681 if (match_int(&args[0], &option))
1684 option = 1; /* No argument, default to 1 */
1686 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1688 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1691 set_opt(sbi->s_mount_opt, DISCARD);
1694 clear_opt(sbi->s_mount_opt, DISCARD);
1696 case Opt_dioread_nolock:
1697 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1699 case Opt_dioread_lock:
1700 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1703 ext4_msg(sb, KERN_ERR,
1704 "Unrecognized mount option \"%s\" "
1705 "or missing value", p);
1710 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1711 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1712 clear_opt(sbi->s_mount_opt, USRQUOTA);
1714 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1715 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1717 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1718 ext4_msg(sb, KERN_ERR, "old and new quota "
1723 if (!sbi->s_jquota_fmt) {
1724 ext4_msg(sb, KERN_ERR, "journaled quota format "
1729 if (sbi->s_jquota_fmt) {
1730 ext4_msg(sb, KERN_ERR, "journaled quota format "
1731 "specified with no journaling "
1740 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1743 struct ext4_sb_info *sbi = EXT4_SB(sb);
1746 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1747 ext4_msg(sb, KERN_ERR, "revision level too high, "
1748 "forcing read-only mode");
1753 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1754 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1755 "running e2fsck is recommended");
1756 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1757 ext4_msg(sb, KERN_WARNING,
1758 "warning: mounting fs with errors, "
1759 "running e2fsck is recommended");
1760 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1761 le16_to_cpu(es->s_mnt_count) >=
1762 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1763 ext4_msg(sb, KERN_WARNING,
1764 "warning: maximal mount count reached, "
1765 "running e2fsck is recommended");
1766 else if (le32_to_cpu(es->s_checkinterval) &&
1767 (le32_to_cpu(es->s_lastcheck) +
1768 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1769 ext4_msg(sb, KERN_WARNING,
1770 "warning: checktime reached, "
1771 "running e2fsck is recommended");
1772 if (!sbi->s_journal)
1773 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1774 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1775 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1776 le16_add_cpu(&es->s_mnt_count, 1);
1777 es->s_mtime = cpu_to_le32(get_seconds());
1778 ext4_update_dynamic_rev(sb);
1780 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1782 ext4_commit_super(sb, 1);
1783 if (test_opt(sb, DEBUG))
1784 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1785 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1787 sbi->s_groups_count,
1788 EXT4_BLOCKS_PER_GROUP(sb),
1789 EXT4_INODES_PER_GROUP(sb),
1795 static int ext4_fill_flex_info(struct super_block *sb)
1797 struct ext4_sb_info *sbi = EXT4_SB(sb);
1798 struct ext4_group_desc *gdp = NULL;
1799 ext4_group_t flex_group_count;
1800 ext4_group_t flex_group;
1801 int groups_per_flex = 0;
1805 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1806 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1808 if (groups_per_flex < 2) {
1809 sbi->s_log_groups_per_flex = 0;
1813 /* We allocate both existing and potentially added groups */
1814 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1815 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1816 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1817 size = flex_group_count * sizeof(struct flex_groups);
1818 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1819 if (sbi->s_flex_groups == NULL) {
1820 sbi->s_flex_groups = vmalloc(size);
1821 if (sbi->s_flex_groups)
1822 memset(sbi->s_flex_groups, 0, size);
1824 if (sbi->s_flex_groups == NULL) {
1825 ext4_msg(sb, KERN_ERR, "not enough memory for "
1826 "%u flex groups", flex_group_count);
1830 for (i = 0; i < sbi->s_groups_count; i++) {
1831 gdp = ext4_get_group_desc(sb, i, NULL);
1833 flex_group = ext4_flex_group(sbi, i);
1834 atomic_add(ext4_free_inodes_count(sb, gdp),
1835 &sbi->s_flex_groups[flex_group].free_inodes);
1836 atomic_add(ext4_free_blks_count(sb, gdp),
1837 &sbi->s_flex_groups[flex_group].free_blocks);
1838 atomic_add(ext4_used_dirs_count(sb, gdp),
1839 &sbi->s_flex_groups[flex_group].used_dirs);
1847 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1848 struct ext4_group_desc *gdp)
1852 if (sbi->s_es->s_feature_ro_compat &
1853 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1854 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1855 __le32 le_group = cpu_to_le32(block_group);
1857 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1858 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1859 crc = crc16(crc, (__u8 *)gdp, offset);
1860 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1861 /* for checksum of struct ext4_group_desc do the rest...*/
1862 if ((sbi->s_es->s_feature_incompat &
1863 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1864 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1865 crc = crc16(crc, (__u8 *)gdp + offset,
1866 le16_to_cpu(sbi->s_es->s_desc_size) -
1870 return cpu_to_le16(crc);
1873 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1874 struct ext4_group_desc *gdp)
1876 if ((sbi->s_es->s_feature_ro_compat &
1877 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1878 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1884 /* Called at mount-time, super-block is locked */
1885 static int ext4_check_descriptors(struct super_block *sb)
1887 struct ext4_sb_info *sbi = EXT4_SB(sb);
1888 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1889 ext4_fsblk_t last_block;
1890 ext4_fsblk_t block_bitmap;
1891 ext4_fsblk_t inode_bitmap;
1892 ext4_fsblk_t inode_table;
1893 int flexbg_flag = 0;
1896 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1899 ext4_debug("Checking group descriptors");
1901 for (i = 0; i < sbi->s_groups_count; i++) {
1902 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1904 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1905 last_block = ext4_blocks_count(sbi->s_es) - 1;
1907 last_block = first_block +
1908 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1910 block_bitmap = ext4_block_bitmap(sb, gdp);
1911 if (block_bitmap < first_block || block_bitmap > last_block) {
1912 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1913 "Block bitmap for group %u not in group "
1914 "(block %llu)!", i, block_bitmap);
1917 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1918 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1919 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1920 "Inode bitmap for group %u not in group "
1921 "(block %llu)!", i, inode_bitmap);
1924 inode_table = ext4_inode_table(sb, gdp);
1925 if (inode_table < first_block ||
1926 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1927 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1928 "Inode table for group %u not in group "
1929 "(block %llu)!", i, inode_table);
1932 ext4_lock_group(sb, i);
1933 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1934 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1935 "Checksum for group %u failed (%u!=%u)",
1936 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1937 gdp)), le16_to_cpu(gdp->bg_checksum));
1938 if (!(sb->s_flags & MS_RDONLY)) {
1939 ext4_unlock_group(sb, i);
1943 ext4_unlock_group(sb, i);
1945 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1948 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1949 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1953 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1954 * the superblock) which were deleted from all directories, but held open by
1955 * a process at the time of a crash. We walk the list and try to delete these
1956 * inodes at recovery time (only with a read-write filesystem).
1958 * In order to keep the orphan inode chain consistent during traversal (in
1959 * case of crash during recovery), we link each inode into the superblock
1960 * orphan list_head and handle it the same way as an inode deletion during
1961 * normal operation (which journals the operations for us).
1963 * We only do an iget() and an iput() on each inode, which is very safe if we
1964 * accidentally point at an in-use or already deleted inode. The worst that
1965 * can happen in this case is that we get a "bit already cleared" message from
1966 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1967 * e2fsck was run on this filesystem, and it must have already done the orphan
1968 * inode cleanup for us, so we can safely abort without any further action.
1970 static void ext4_orphan_cleanup(struct super_block *sb,
1971 struct ext4_super_block *es)
1973 unsigned int s_flags = sb->s_flags;
1974 int nr_orphans = 0, nr_truncates = 0;
1978 if (!es->s_last_orphan) {
1979 jbd_debug(4, "no orphan inodes to clean up\n");
1983 if (bdev_read_only(sb->s_bdev)) {
1984 ext4_msg(sb, KERN_ERR, "write access "
1985 "unavailable, skipping orphan cleanup");
1989 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1990 if (es->s_last_orphan)
1991 jbd_debug(1, "Errors on filesystem, "
1992 "clearing orphan list.\n");
1993 es->s_last_orphan = 0;
1994 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1998 if (s_flags & MS_RDONLY) {
1999 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2000 sb->s_flags &= ~MS_RDONLY;
2003 /* Needed for iput() to work correctly and not trash data */
2004 sb->s_flags |= MS_ACTIVE;
2005 /* Turn on quotas so that they are updated correctly */
2006 for (i = 0; i < MAXQUOTAS; i++) {
2007 if (EXT4_SB(sb)->s_qf_names[i]) {
2008 int ret = ext4_quota_on_mount(sb, i);
2010 ext4_msg(sb, KERN_ERR,
2011 "Cannot turn on journaled "
2012 "quota: error %d", ret);
2017 while (es->s_last_orphan) {
2018 struct inode *inode;
2020 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2021 if (IS_ERR(inode)) {
2022 es->s_last_orphan = 0;
2026 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2027 dquot_initialize(inode);
2028 if (inode->i_nlink) {
2029 ext4_msg(sb, KERN_DEBUG,
2030 "%s: truncating inode %lu to %lld bytes",
2031 __func__, inode->i_ino, inode->i_size);
2032 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2033 inode->i_ino, inode->i_size);
2034 ext4_truncate(inode);
2037 ext4_msg(sb, KERN_DEBUG,
2038 "%s: deleting unreferenced inode %lu",
2039 __func__, inode->i_ino);
2040 jbd_debug(2, "deleting unreferenced inode %lu\n",
2044 iput(inode); /* The delete magic happens here! */
2047 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2050 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2051 PLURAL(nr_orphans));
2053 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2054 PLURAL(nr_truncates));
2056 /* Turn quotas off */
2057 for (i = 0; i < MAXQUOTAS; i++) {
2058 if (sb_dqopt(sb)->files[i])
2059 dquot_quota_off(sb, i);
2062 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2066 * Maximal extent format file size.
2067 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2068 * extent format containers, within a sector_t, and within i_blocks
2069 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2070 * so that won't be a limiting factor.
2072 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2074 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2077 loff_t upper_limit = MAX_LFS_FILESIZE;
2079 /* small i_blocks in vfs inode? */
2080 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2082 * CONFIG_LBDAF is not enabled implies the inode
2083 * i_block represent total blocks in 512 bytes
2084 * 32 == size of vfs inode i_blocks * 8
2086 upper_limit = (1LL << 32) - 1;
2088 /* total blocks in file system block size */
2089 upper_limit >>= (blkbits - 9);
2090 upper_limit <<= blkbits;
2093 /* 32-bit extent-start container, ee_block */
2098 /* Sanity check against vm- & vfs- imposed limits */
2099 if (res > upper_limit)
2106 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2107 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2108 * We need to be 1 filesystem block less than the 2^48 sector limit.
2110 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2112 loff_t res = EXT4_NDIR_BLOCKS;
2115 /* This is calculated to be the largest file size for a dense, block
2116 * mapped file such that the file's total number of 512-byte sectors,
2117 * including data and all indirect blocks, does not exceed (2^48 - 1).
2119 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2120 * number of 512-byte sectors of the file.
2123 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2125 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2126 * the inode i_block field represents total file blocks in
2127 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2129 upper_limit = (1LL << 32) - 1;
2131 /* total blocks in file system block size */
2132 upper_limit >>= (bits - 9);
2136 * We use 48 bit ext4_inode i_blocks
2137 * With EXT4_HUGE_FILE_FL set the i_blocks
2138 * represent total number of blocks in
2139 * file system block size
2141 upper_limit = (1LL << 48) - 1;
2145 /* indirect blocks */
2147 /* double indirect blocks */
2148 meta_blocks += 1 + (1LL << (bits-2));
2149 /* tripple indirect blocks */
2150 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2152 upper_limit -= meta_blocks;
2153 upper_limit <<= bits;
2155 res += 1LL << (bits-2);
2156 res += 1LL << (2*(bits-2));
2157 res += 1LL << (3*(bits-2));
2159 if (res > upper_limit)
2162 if (res > MAX_LFS_FILESIZE)
2163 res = MAX_LFS_FILESIZE;
2168 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2169 ext4_fsblk_t logical_sb_block, int nr)
2171 struct ext4_sb_info *sbi = EXT4_SB(sb);
2172 ext4_group_t bg, first_meta_bg;
2175 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2177 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2179 return logical_sb_block + nr + 1;
2180 bg = sbi->s_desc_per_block * nr;
2181 if (ext4_bg_has_super(sb, bg))
2184 return (has_super + ext4_group_first_block_no(sb, bg));
2188 * ext4_get_stripe_size: Get the stripe size.
2189 * @sbi: In memory super block info
2191 * If we have specified it via mount option, then
2192 * use the mount option value. If the value specified at mount time is
2193 * greater than the blocks per group use the super block value.
2194 * If the super block value is greater than blocks per group return 0.
2195 * Allocator needs it be less than blocks per group.
2198 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2200 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2201 unsigned long stripe_width =
2202 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2204 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2205 return sbi->s_stripe;
2207 if (stripe_width <= sbi->s_blocks_per_group)
2208 return stripe_width;
2210 if (stride <= sbi->s_blocks_per_group)
2219 struct attribute attr;
2220 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2221 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2222 const char *, size_t);
2226 static int parse_strtoul(const char *buf,
2227 unsigned long max, unsigned long *value)
2231 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2232 endp = skip_spaces(endp);
2233 if (*endp || *value > max)
2239 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2240 struct ext4_sb_info *sbi,
2243 return snprintf(buf, PAGE_SIZE, "%llu\n",
2244 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2247 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2248 struct ext4_sb_info *sbi, char *buf)
2250 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2252 return snprintf(buf, PAGE_SIZE, "%lu\n",
2253 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2254 sbi->s_sectors_written_start) >> 1);
2257 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2258 struct ext4_sb_info *sbi, char *buf)
2260 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2262 return snprintf(buf, PAGE_SIZE, "%llu\n",
2263 (unsigned long long)(sbi->s_kbytes_written +
2264 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2265 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2268 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2269 struct ext4_sb_info *sbi,
2270 const char *buf, size_t count)
2274 if (parse_strtoul(buf, 0x40000000, &t))
2277 if (!is_power_of_2(t))
2280 sbi->s_inode_readahead_blks = t;
2284 static ssize_t sbi_ui_show(struct ext4_attr *a,
2285 struct ext4_sb_info *sbi, char *buf)
2287 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2289 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2292 static ssize_t sbi_ui_store(struct ext4_attr *a,
2293 struct ext4_sb_info *sbi,
2294 const char *buf, size_t count)
2296 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2299 if (parse_strtoul(buf, 0xffffffff, &t))
2305 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2306 static struct ext4_attr ext4_attr_##_name = { \
2307 .attr = {.name = __stringify(_name), .mode = _mode }, \
2310 .offset = offsetof(struct ext4_sb_info, _elname), \
2312 #define EXT4_ATTR(name, mode, show, store) \
2313 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2315 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2316 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2317 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2318 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2319 #define ATTR_LIST(name) &ext4_attr_##name.attr
2321 EXT4_RO_ATTR(delayed_allocation_blocks);
2322 EXT4_RO_ATTR(session_write_kbytes);
2323 EXT4_RO_ATTR(lifetime_write_kbytes);
2324 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2325 inode_readahead_blks_store, s_inode_readahead_blks);
2326 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2327 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2328 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2329 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2330 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2331 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2332 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2333 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2335 static struct attribute *ext4_attrs[] = {
2336 ATTR_LIST(delayed_allocation_blocks),
2337 ATTR_LIST(session_write_kbytes),
2338 ATTR_LIST(lifetime_write_kbytes),
2339 ATTR_LIST(inode_readahead_blks),
2340 ATTR_LIST(inode_goal),
2341 ATTR_LIST(mb_stats),
2342 ATTR_LIST(mb_max_to_scan),
2343 ATTR_LIST(mb_min_to_scan),
2344 ATTR_LIST(mb_order2_req),
2345 ATTR_LIST(mb_stream_req),
2346 ATTR_LIST(mb_group_prealloc),
2347 ATTR_LIST(max_writeback_mb_bump),
2351 static ssize_t ext4_attr_show(struct kobject *kobj,
2352 struct attribute *attr, char *buf)
2354 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2356 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2358 return a->show ? a->show(a, sbi, buf) : 0;
2361 static ssize_t ext4_attr_store(struct kobject *kobj,
2362 struct attribute *attr,
2363 const char *buf, size_t len)
2365 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2367 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2369 return a->store ? a->store(a, sbi, buf, len) : 0;
2372 static void ext4_sb_release(struct kobject *kobj)
2374 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2376 complete(&sbi->s_kobj_unregister);
2380 static const struct sysfs_ops ext4_attr_ops = {
2381 .show = ext4_attr_show,
2382 .store = ext4_attr_store,
2385 static struct kobj_type ext4_ktype = {
2386 .default_attrs = ext4_attrs,
2387 .sysfs_ops = &ext4_attr_ops,
2388 .release = ext4_sb_release,
2392 * Check whether this filesystem can be mounted based on
2393 * the features present and the RDONLY/RDWR mount requested.
2394 * Returns 1 if this filesystem can be mounted as requested,
2395 * 0 if it cannot be.
2397 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2399 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2400 ext4_msg(sb, KERN_ERR,
2401 "Couldn't mount because of "
2402 "unsupported optional features (%x)",
2403 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2404 ~EXT4_FEATURE_INCOMPAT_SUPP));
2411 /* Check that feature set is OK for a read-write mount */
2412 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2413 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2414 "unsupported optional features (%x)",
2415 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2416 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2420 * Large file size enabled file system can only be mounted
2421 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2423 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2424 if (sizeof(blkcnt_t) < sizeof(u64)) {
2425 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2426 "cannot be mounted RDWR without "
2434 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2435 __releases(kernel_lock)
2436 __acquires(kernel_lock)
2438 char *orig_data = kstrdup(data, GFP_KERNEL);
2439 struct buffer_head *bh;
2440 struct ext4_super_block *es = NULL;
2441 struct ext4_sb_info *sbi;
2443 ext4_fsblk_t sb_block = get_sb_block(&data);
2444 ext4_fsblk_t logical_sb_block;
2445 unsigned long offset = 0;
2446 unsigned long journal_devnum = 0;
2447 unsigned long def_mount_opts;
2453 unsigned int db_count;
2455 int needs_recovery, has_huge_files;
2458 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2460 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2464 sbi->s_blockgroup_lock =
2465 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2466 if (!sbi->s_blockgroup_lock) {
2470 sb->s_fs_info = sbi;
2471 sbi->s_mount_opt = 0;
2472 sbi->s_resuid = EXT4_DEF_RESUID;
2473 sbi->s_resgid = EXT4_DEF_RESGID;
2474 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2475 sbi->s_sb_block = sb_block;
2476 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2481 /* Cleanup superblock name */
2482 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2485 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2487 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2492 * The ext4 superblock will not be buffer aligned for other than 1kB
2493 * block sizes. We need to calculate the offset from buffer start.
2495 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2496 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2497 offset = do_div(logical_sb_block, blocksize);
2499 logical_sb_block = sb_block;
2502 if (!(bh = sb_bread(sb, logical_sb_block))) {
2503 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2507 * Note: s_es must be initialized as soon as possible because
2508 * some ext4 macro-instructions depend on its value
2510 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2512 sb->s_magic = le16_to_cpu(es->s_magic);
2513 if (sb->s_magic != EXT4_SUPER_MAGIC)
2515 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2517 /* Set defaults before we parse the mount options */
2518 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2519 if (def_mount_opts & EXT4_DEFM_DEBUG)
2520 set_opt(sbi->s_mount_opt, DEBUG);
2521 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2522 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2524 set_opt(sbi->s_mount_opt, GRPID);
2526 if (def_mount_opts & EXT4_DEFM_UID16)
2527 set_opt(sbi->s_mount_opt, NO_UID32);
2528 #ifdef CONFIG_EXT4_FS_XATTR
2529 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2530 set_opt(sbi->s_mount_opt, XATTR_USER);
2532 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2533 if (def_mount_opts & EXT4_DEFM_ACL)
2534 set_opt(sbi->s_mount_opt, POSIX_ACL);
2536 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2537 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2538 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2539 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2540 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2541 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2543 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2544 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2545 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2546 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2548 set_opt(sbi->s_mount_opt, ERRORS_RO);
2550 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2551 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2552 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2553 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2554 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2556 set_opt(sbi->s_mount_opt, BARRIER);
2559 * enable delayed allocation by default
2560 * Use -o nodelalloc to turn it off
2562 if (!IS_EXT3_SB(sb))
2563 set_opt(sbi->s_mount_opt, DELALLOC);
2565 if (!parse_options((char *) data, sb, &journal_devnum,
2566 &journal_ioprio, NULL, 0))
2569 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2570 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2572 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2573 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2574 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2575 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2576 ext4_msg(sb, KERN_WARNING,
2577 "feature flags set on rev 0 fs, "
2578 "running e2fsck is recommended");
2581 * Check feature flags regardless of the revision level, since we
2582 * previously didn't change the revision level when setting the flags,
2583 * so there is a chance incompat flags are set on a rev 0 filesystem.
2585 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2588 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2590 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2591 blocksize > EXT4_MAX_BLOCK_SIZE) {
2592 ext4_msg(sb, KERN_ERR,
2593 "Unsupported filesystem blocksize %d", blocksize);
2597 if (sb->s_blocksize != blocksize) {
2598 /* Validate the filesystem blocksize */
2599 if (!sb_set_blocksize(sb, blocksize)) {
2600 ext4_msg(sb, KERN_ERR, "bad block size %d",
2606 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2607 offset = do_div(logical_sb_block, blocksize);
2608 bh = sb_bread(sb, logical_sb_block);
2610 ext4_msg(sb, KERN_ERR,
2611 "Can't read superblock on 2nd try");
2614 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2616 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2617 ext4_msg(sb, KERN_ERR,
2618 "Magic mismatch, very weird!");
2623 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2624 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2625 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2627 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2629 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2630 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2631 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2633 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2634 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2635 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2636 (!is_power_of_2(sbi->s_inode_size)) ||
2637 (sbi->s_inode_size > blocksize)) {
2638 ext4_msg(sb, KERN_ERR,
2639 "unsupported inode size: %d",
2643 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2644 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2647 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2648 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2649 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2650 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2651 !is_power_of_2(sbi->s_desc_size)) {
2652 ext4_msg(sb, KERN_ERR,
2653 "unsupported descriptor size %lu",
2658 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2660 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2661 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2662 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2665 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2666 if (sbi->s_inodes_per_block == 0)
2668 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2669 sbi->s_inodes_per_block;
2670 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2672 sbi->s_mount_state = le16_to_cpu(es->s_state);
2673 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2674 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2676 for (i = 0; i < 4; i++)
2677 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2678 sbi->s_def_hash_version = es->s_def_hash_version;
2679 i = le32_to_cpu(es->s_flags);
2680 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2681 sbi->s_hash_unsigned = 3;
2682 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2683 #ifdef __CHAR_UNSIGNED__
2684 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2685 sbi->s_hash_unsigned = 3;
2687 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2692 if (sbi->s_blocks_per_group > blocksize * 8) {
2693 ext4_msg(sb, KERN_ERR,
2694 "#blocks per group too big: %lu",
2695 sbi->s_blocks_per_group);
2698 if (sbi->s_inodes_per_group > blocksize * 8) {
2699 ext4_msg(sb, KERN_ERR,
2700 "#inodes per group too big: %lu",
2701 sbi->s_inodes_per_group);
2706 * Test whether we have more sectors than will fit in sector_t,
2707 * and whether the max offset is addressable by the page cache.
2709 if ((ext4_blocks_count(es) >
2710 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2711 (ext4_blocks_count(es) >
2712 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2713 ext4_msg(sb, KERN_ERR, "filesystem"
2714 " too large to mount safely on this system");
2715 if (sizeof(sector_t) < 8)
2716 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2721 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2724 /* check blocks count against device size */
2725 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2726 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2727 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2728 "exceeds size of device (%llu blocks)",
2729 ext4_blocks_count(es), blocks_count);
2734 * It makes no sense for the first data block to be beyond the end
2735 * of the filesystem.
2737 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2738 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2739 "block %u is beyond end of filesystem (%llu)",
2740 le32_to_cpu(es->s_first_data_block),
2741 ext4_blocks_count(es));
2744 blocks_count = (ext4_blocks_count(es) -
2745 le32_to_cpu(es->s_first_data_block) +
2746 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2747 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2748 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2749 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2750 "(block count %llu, first data block %u, "
2751 "blocks per group %lu)", sbi->s_groups_count,
2752 ext4_blocks_count(es),
2753 le32_to_cpu(es->s_first_data_block),
2754 EXT4_BLOCKS_PER_GROUP(sb));
2757 sbi->s_groups_count = blocks_count;
2758 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2759 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2760 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2761 EXT4_DESC_PER_BLOCK(sb);
2762 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2764 if (sbi->s_group_desc == NULL) {
2765 ext4_msg(sb, KERN_ERR, "not enough memory");
2769 #ifdef CONFIG_PROC_FS
2771 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2774 bgl_lock_init(sbi->s_blockgroup_lock);
2776 for (i = 0; i < db_count; i++) {
2777 block = descriptor_loc(sb, logical_sb_block, i);
2778 sbi->s_group_desc[i] = sb_bread(sb, block);
2779 if (!sbi->s_group_desc[i]) {
2780 ext4_msg(sb, KERN_ERR,
2781 "can't read group descriptor %d", i);
2786 if (!ext4_check_descriptors(sb)) {
2787 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2790 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2791 if (!ext4_fill_flex_info(sb)) {
2792 ext4_msg(sb, KERN_ERR,
2793 "unable to initialize "
2794 "flex_bg meta info!");
2798 sbi->s_gdb_count = db_count;
2799 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2800 spin_lock_init(&sbi->s_next_gen_lock);
2802 sbi->s_stripe = ext4_get_stripe_size(sbi);
2803 sbi->s_max_writeback_mb_bump = 128;
2806 * set up enough so that it can read an inode
2808 if (!test_opt(sb, NOLOAD) &&
2809 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2810 sb->s_op = &ext4_sops;
2812 sb->s_op = &ext4_nojournal_sops;
2813 sb->s_export_op = &ext4_export_ops;
2814 sb->s_xattr = ext4_xattr_handlers;
2816 sb->s_qcop = &ext4_qctl_operations;
2817 sb->dq_op = &ext4_quota_operations;
2819 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2820 mutex_init(&sbi->s_orphan_lock);
2821 mutex_init(&sbi->s_resize_lock);
2825 needs_recovery = (es->s_last_orphan != 0 ||
2826 EXT4_HAS_INCOMPAT_FEATURE(sb,
2827 EXT4_FEATURE_INCOMPAT_RECOVER));
2830 * The first inode we look at is the journal inode. Don't try
2831 * root first: it may be modified in the journal!
2833 if (!test_opt(sb, NOLOAD) &&
2834 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2835 if (ext4_load_journal(sb, es, journal_devnum))
2837 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2838 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2839 ext4_msg(sb, KERN_ERR, "required journal recovery "
2840 "suppressed and not mounted read-only");
2841 goto failed_mount_wq;
2843 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2844 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2845 sbi->s_journal = NULL;
2850 if (ext4_blocks_count(es) > 0xffffffffULL &&
2851 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2852 JBD2_FEATURE_INCOMPAT_64BIT)) {
2853 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2854 goto failed_mount_wq;
2857 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2858 jbd2_journal_set_features(sbi->s_journal,
2859 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2860 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2861 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2862 jbd2_journal_set_features(sbi->s_journal,
2863 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2864 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2865 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2867 jbd2_journal_clear_features(sbi->s_journal,
2868 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2869 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2872 /* We have now updated the journal if required, so we can
2873 * validate the data journaling mode. */
2874 switch (test_opt(sb, DATA_FLAGS)) {
2876 /* No mode set, assume a default based on the journal
2877 * capabilities: ORDERED_DATA if the journal can
2878 * cope, else JOURNAL_DATA
2880 if (jbd2_journal_check_available_features
2881 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2882 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2884 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2887 case EXT4_MOUNT_ORDERED_DATA:
2888 case EXT4_MOUNT_WRITEBACK_DATA:
2889 if (!jbd2_journal_check_available_features
2890 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2891 ext4_msg(sb, KERN_ERR, "Journal does not support "
2892 "requested data journaling mode");
2893 goto failed_mount_wq;
2898 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2901 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2902 ext4_count_free_blocks(sb));
2904 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2905 ext4_count_free_inodes(sb));
2907 err = percpu_counter_init(&sbi->s_dirs_counter,
2908 ext4_count_dirs(sb));
2910 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2912 ext4_msg(sb, KERN_ERR, "insufficient memory");
2913 goto failed_mount_wq;
2916 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2917 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2918 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2919 goto failed_mount_wq;
2923 * The jbd2_journal_load will have done any necessary log recovery,
2924 * so we can safely mount the rest of the filesystem now.
2927 root = ext4_iget(sb, EXT4_ROOT_INO);
2929 ext4_msg(sb, KERN_ERR, "get root inode failed");
2930 ret = PTR_ERR(root);
2933 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2935 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2938 sb->s_root = d_alloc_root(root);
2940 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2946 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2948 /* determine the minimum size of new large inodes, if present */
2949 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2950 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2951 EXT4_GOOD_OLD_INODE_SIZE;
2952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2953 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2954 if (sbi->s_want_extra_isize <
2955 le16_to_cpu(es->s_want_extra_isize))
2956 sbi->s_want_extra_isize =
2957 le16_to_cpu(es->s_want_extra_isize);
2958 if (sbi->s_want_extra_isize <
2959 le16_to_cpu(es->s_min_extra_isize))
2960 sbi->s_want_extra_isize =
2961 le16_to_cpu(es->s_min_extra_isize);
2964 /* Check if enough inode space is available */
2965 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2966 sbi->s_inode_size) {
2967 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2968 EXT4_GOOD_OLD_INODE_SIZE;
2969 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2973 if (test_opt(sb, DELALLOC) &&
2974 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2975 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2976 "requested data journaling mode");
2977 clear_opt(sbi->s_mount_opt, DELALLOC);
2979 if (test_opt(sb, DIOREAD_NOLOCK)) {
2980 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2981 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2982 "option - requested data journaling mode");
2983 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2985 if (sb->s_blocksize < PAGE_SIZE) {
2986 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2987 "option - block size is too small");
2988 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2992 err = ext4_setup_system_zone(sb);
2994 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3000 err = ext4_mb_init(sb, needs_recovery);
3002 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3007 sbi->s_kobj.kset = ext4_kset;
3008 init_completion(&sbi->s_kobj_unregister);
3009 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3012 ext4_mb_release(sb);
3013 ext4_ext_release(sb);
3017 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3018 ext4_orphan_cleanup(sb, es);
3019 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3020 if (needs_recovery) {
3021 ext4_msg(sb, KERN_INFO, "recovery complete");
3022 ext4_mark_recovery_complete(sb, es);
3024 if (EXT4_SB(sb)->s_journal) {
3025 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3026 descr = " journalled data mode";
3027 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3028 descr = " ordered data mode";
3030 descr = " writeback data mode";
3032 descr = "out journal";
3034 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3035 "Opts: %s", descr, orig_data);
3043 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3047 ext4_msg(sb, KERN_ERR, "mount failed");
3048 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3050 ext4_release_system_zone(sb);
3051 if (sbi->s_journal) {
3052 jbd2_journal_destroy(sbi->s_journal);
3053 sbi->s_journal = NULL;
3055 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3056 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3057 percpu_counter_destroy(&sbi->s_dirs_counter);
3058 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3060 if (sbi->s_flex_groups) {
3061 if (is_vmalloc_addr(sbi->s_flex_groups))
3062 vfree(sbi->s_flex_groups);
3064 kfree(sbi->s_flex_groups);
3067 for (i = 0; i < db_count; i++)
3068 brelse(sbi->s_group_desc[i]);
3069 kfree(sbi->s_group_desc);
3072 remove_proc_entry(sb->s_id, ext4_proc_root);
3075 for (i = 0; i < MAXQUOTAS; i++)
3076 kfree(sbi->s_qf_names[i]);
3078 ext4_blkdev_remove(sbi);
3081 sb->s_fs_info = NULL;
3082 kfree(sbi->s_blockgroup_lock);
3090 * Setup any per-fs journal parameters now. We'll do this both on
3091 * initial mount, once the journal has been initialised but before we've
3092 * done any recovery; and again on any subsequent remount.
3094 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3096 struct ext4_sb_info *sbi = EXT4_SB(sb);
3098 journal->j_commit_interval = sbi->s_commit_interval;
3099 journal->j_min_batch_time = sbi->s_min_batch_time;
3100 journal->j_max_batch_time = sbi->s_max_batch_time;
3102 spin_lock(&journal->j_state_lock);
3103 if (test_opt(sb, BARRIER))
3104 journal->j_flags |= JBD2_BARRIER;
3106 journal->j_flags &= ~JBD2_BARRIER;
3107 if (test_opt(sb, DATA_ERR_ABORT))
3108 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3110 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3111 spin_unlock(&journal->j_state_lock);
3114 static journal_t *ext4_get_journal(struct super_block *sb,
3115 unsigned int journal_inum)
3117 struct inode *journal_inode;
3120 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3122 /* First, test for the existence of a valid inode on disk. Bad
3123 * things happen if we iget() an unused inode, as the subsequent
3124 * iput() will try to delete it. */
3126 journal_inode = ext4_iget(sb, journal_inum);
3127 if (IS_ERR(journal_inode)) {
3128 ext4_msg(sb, KERN_ERR, "no journal found");
3131 if (!journal_inode->i_nlink) {
3132 make_bad_inode(journal_inode);
3133 iput(journal_inode);
3134 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3138 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3139 journal_inode, journal_inode->i_size);
3140 if (!S_ISREG(journal_inode->i_mode)) {
3141 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3142 iput(journal_inode);
3146 journal = jbd2_journal_init_inode(journal_inode);
3148 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3149 iput(journal_inode);
3152 journal->j_private = sb;
3153 ext4_init_journal_params(sb, journal);
3157 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3160 struct buffer_head *bh;
3164 int hblock, blocksize;
3165 ext4_fsblk_t sb_block;
3166 unsigned long offset;
3167 struct ext4_super_block *es;
3168 struct block_device *bdev;
3170 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3172 bdev = ext4_blkdev_get(j_dev, sb);
3176 if (bd_claim(bdev, sb)) {
3177 ext4_msg(sb, KERN_ERR,
3178 "failed to claim external journal device");
3179 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3183 blocksize = sb->s_blocksize;
3184 hblock = bdev_logical_block_size(bdev);
3185 if (blocksize < hblock) {
3186 ext4_msg(sb, KERN_ERR,
3187 "blocksize too small for journal device");
3191 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3192 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3193 set_blocksize(bdev, blocksize);
3194 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3195 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3196 "external journal");
3200 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3201 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3202 !(le32_to_cpu(es->s_feature_incompat) &
3203 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3204 ext4_msg(sb, KERN_ERR, "external journal has "
3210 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3211 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3216 len = ext4_blocks_count(es);
3217 start = sb_block + 1;
3218 brelse(bh); /* we're done with the superblock */
3220 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3221 start, len, blocksize);
3223 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3226 journal->j_private = sb;
3227 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3228 wait_on_buffer(journal->j_sb_buffer);
3229 if (!buffer_uptodate(journal->j_sb_buffer)) {
3230 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3233 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3234 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3235 "user (unsupported) - %d",
3236 be32_to_cpu(journal->j_superblock->s_nr_users));
3239 EXT4_SB(sb)->journal_bdev = bdev;
3240 ext4_init_journal_params(sb, journal);
3244 jbd2_journal_destroy(journal);
3246 ext4_blkdev_put(bdev);
3250 static int ext4_load_journal(struct super_block *sb,
3251 struct ext4_super_block *es,
3252 unsigned long journal_devnum)
3255 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3258 int really_read_only;
3260 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3262 if (journal_devnum &&
3263 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3264 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3265 "numbers have changed");
3266 journal_dev = new_decode_dev(journal_devnum);
3268 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3270 really_read_only = bdev_read_only(sb->s_bdev);
3273 * Are we loading a blank journal or performing recovery after a
3274 * crash? For recovery, we need to check in advance whether we
3275 * can get read-write access to the device.
3277 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3278 if (sb->s_flags & MS_RDONLY) {
3279 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3280 "required on readonly filesystem");
3281 if (really_read_only) {
3282 ext4_msg(sb, KERN_ERR, "write access "
3283 "unavailable, cannot proceed");
3286 ext4_msg(sb, KERN_INFO, "write access will "
3287 "be enabled during recovery");
3291 if (journal_inum && journal_dev) {
3292 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3293 "and inode journals!");
3298 if (!(journal = ext4_get_journal(sb, journal_inum)))
3301 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3305 if (!(journal->j_flags & JBD2_BARRIER))
3306 ext4_msg(sb, KERN_INFO, "barriers disabled");
3308 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3309 err = jbd2_journal_update_format(journal);
3311 ext4_msg(sb, KERN_ERR, "error updating journal");
3312 jbd2_journal_destroy(journal);
3317 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3318 err = jbd2_journal_wipe(journal, !really_read_only);
3320 err = jbd2_journal_load(journal);
3323 ext4_msg(sb, KERN_ERR, "error loading journal");
3324 jbd2_journal_destroy(journal);
3328 EXT4_SB(sb)->s_journal = journal;
3329 ext4_clear_journal_err(sb, es);
3331 if (journal_devnum &&
3332 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3333 es->s_journal_dev = cpu_to_le32(journal_devnum);
3335 /* Make sure we flush the recovery flag to disk. */
3336 ext4_commit_super(sb, 1);
3342 static int ext4_commit_super(struct super_block *sb, int sync)
3344 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3345 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3350 if (buffer_write_io_error(sbh)) {
3352 * Oh, dear. A previous attempt to write the
3353 * superblock failed. This could happen because the
3354 * USB device was yanked out. Or it could happen to
3355 * be a transient write error and maybe the block will
3356 * be remapped. Nothing we can do but to retry the
3357 * write and hope for the best.
3359 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3360 "superblock detected");
3361 clear_buffer_write_io_error(sbh);
3362 set_buffer_uptodate(sbh);
3365 * If the file system is mounted read-only, don't update the
3366 * superblock write time. This avoids updating the superblock
3367 * write time when we are mounting the root file system
3368 * read/only but we need to replay the journal; at that point,
3369 * for people who are east of GMT and who make their clock
3370 * tick in localtime for Windows bug-for-bug compatibility,
3371 * the clock is set in the future, and this will cause e2fsck
3372 * to complain and force a full file system check.
3374 if (!(sb->s_flags & MS_RDONLY))
3375 es->s_wtime = cpu_to_le32(get_seconds());
3376 es->s_kbytes_written =
3377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3378 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3379 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3380 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3381 &EXT4_SB(sb)->s_freeblocks_counter));
3382 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3383 &EXT4_SB(sb)->s_freeinodes_counter));
3385 BUFFER_TRACE(sbh, "marking dirty");
3386 mark_buffer_dirty(sbh);
3388 error = sync_dirty_buffer(sbh);
3392 error = buffer_write_io_error(sbh);
3394 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3396 clear_buffer_write_io_error(sbh);
3397 set_buffer_uptodate(sbh);
3404 * Have we just finished recovery? If so, and if we are mounting (or
3405 * remounting) the filesystem readonly, then we will end up with a
3406 * consistent fs on disk. Record that fact.
3408 static void ext4_mark_recovery_complete(struct super_block *sb,
3409 struct ext4_super_block *es)
3411 journal_t *journal = EXT4_SB(sb)->s_journal;
3413 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3414 BUG_ON(journal != NULL);
3417 jbd2_journal_lock_updates(journal);
3418 if (jbd2_journal_flush(journal) < 0)
3421 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3422 sb->s_flags & MS_RDONLY) {
3423 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3424 ext4_commit_super(sb, 1);
3428 jbd2_journal_unlock_updates(journal);
3432 * If we are mounting (or read-write remounting) a filesystem whose journal
3433 * has recorded an error from a previous lifetime, move that error to the
3434 * main filesystem now.
3436 static void ext4_clear_journal_err(struct super_block *sb,
3437 struct ext4_super_block *es)
3443 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3445 journal = EXT4_SB(sb)->s_journal;
3448 * Now check for any error status which may have been recorded in the
3449 * journal by a prior ext4_error() or ext4_abort()
3452 j_errno = jbd2_journal_errno(journal);
3456 errstr = ext4_decode_error(sb, j_errno, nbuf);
3457 ext4_warning(sb, "Filesystem error recorded "
3458 "from previous mount: %s", errstr);
3459 ext4_warning(sb, "Marking fs in need of filesystem check.");
3461 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3462 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3463 ext4_commit_super(sb, 1);
3465 jbd2_journal_clear_err(journal);
3470 * Force the running and committing transactions to commit,
3471 * and wait on the commit.
3473 int ext4_force_commit(struct super_block *sb)
3478 if (sb->s_flags & MS_RDONLY)
3481 journal = EXT4_SB(sb)->s_journal;
3483 vfs_check_frozen(sb, SB_FREEZE_WRITE);
3484 ret = ext4_journal_force_commit(journal);
3490 static void ext4_write_super(struct super_block *sb)
3493 ext4_commit_super(sb, 1);
3497 static int ext4_sync_fs(struct super_block *sb, int wait)
3501 struct ext4_sb_info *sbi = EXT4_SB(sb);
3503 trace_ext4_sync_fs(sb, wait);
3504 flush_workqueue(sbi->dio_unwritten_wq);
3505 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3507 jbd2_log_wait_commit(sbi->s_journal, target);
3513 * LVM calls this function before a (read-only) snapshot is created. This
3514 * gives us a chance to flush the journal completely and mark the fs clean.
3516 static int ext4_freeze(struct super_block *sb)
3521 if (sb->s_flags & MS_RDONLY)
3524 journal = EXT4_SB(sb)->s_journal;
3526 /* Now we set up the journal barrier. */
3527 jbd2_journal_lock_updates(journal);
3530 * Don't clear the needs_recovery flag if we failed to flush
3533 error = jbd2_journal_flush(journal);
3537 /* Journal blocked and flushed, clear needs_recovery flag. */
3538 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3539 error = ext4_commit_super(sb, 1);
3541 /* we rely on s_frozen to stop further updates */
3542 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3547 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3548 * flag here, even though the filesystem is not technically dirty yet.
3550 static int ext4_unfreeze(struct super_block *sb)
3552 if (sb->s_flags & MS_RDONLY)
3556 /* Reset the needs_recovery flag before the fs is unlocked. */
3557 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3558 ext4_commit_super(sb, 1);
3563 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3565 struct ext4_super_block *es;
3566 struct ext4_sb_info *sbi = EXT4_SB(sb);
3567 ext4_fsblk_t n_blocks_count = 0;
3568 unsigned long old_sb_flags;
3569 struct ext4_mount_options old_opts;
3570 int enable_quota = 0;
3572 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3577 char *orig_data = kstrdup(data, GFP_KERNEL);
3581 /* Store the original options */
3583 old_sb_flags = sb->s_flags;
3584 old_opts.s_mount_opt = sbi->s_mount_opt;
3585 old_opts.s_resuid = sbi->s_resuid;
3586 old_opts.s_resgid = sbi->s_resgid;
3587 old_opts.s_commit_interval = sbi->s_commit_interval;
3588 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3589 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3591 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3592 for (i = 0; i < MAXQUOTAS; i++)
3593 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3595 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3596 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3599 * Allow the "check" option to be passed as a remount option.
3601 if (!parse_options(data, sb, NULL, &journal_ioprio,
3602 &n_blocks_count, 1)) {
3607 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3608 ext4_abort(sb, __func__, "Abort forced by user");
3610 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3611 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3615 if (sbi->s_journal) {
3616 ext4_init_journal_params(sb, sbi->s_journal);
3617 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3620 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3621 n_blocks_count > ext4_blocks_count(es)) {
3622 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3627 if (*flags & MS_RDONLY) {
3628 err = dquot_suspend(sb, -1);
3633 * First of all, the unconditional stuff we have to do
3634 * to disable replay of the journal when we next remount
3636 sb->s_flags |= MS_RDONLY;
3639 * OK, test if we are remounting a valid rw partition
3640 * readonly, and if so set the rdonly flag and then
3641 * mark the partition as valid again.
3643 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3644 (sbi->s_mount_state & EXT4_VALID_FS))
3645 es->s_state = cpu_to_le16(sbi->s_mount_state);
3648 ext4_mark_recovery_complete(sb, es);
3650 /* Make sure we can mount this feature set readwrite */
3651 if (!ext4_feature_set_ok(sb, 0)) {
3656 * Make sure the group descriptor checksums
3657 * are sane. If they aren't, refuse to remount r/w.
3659 for (g = 0; g < sbi->s_groups_count; g++) {
3660 struct ext4_group_desc *gdp =
3661 ext4_get_group_desc(sb, g, NULL);
3663 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3664 ext4_msg(sb, KERN_ERR,
3665 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3666 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3667 le16_to_cpu(gdp->bg_checksum));
3674 * If we have an unprocessed orphan list hanging
3675 * around from a previously readonly bdev mount,
3676 * require a full umount/remount for now.
3678 if (es->s_last_orphan) {
3679 ext4_msg(sb, KERN_WARNING, "Couldn't "
3680 "remount RDWR because of unprocessed "
3681 "orphan inode list. Please "
3682 "umount/remount instead");
3688 * Mounting a RDONLY partition read-write, so reread
3689 * and store the current valid flag. (It may have
3690 * been changed by e2fsck since we originally mounted
3694 ext4_clear_journal_err(sb, es);
3695 sbi->s_mount_state = le16_to_cpu(es->s_state);
3696 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3698 if (!ext4_setup_super(sb, es, 0))
3699 sb->s_flags &= ~MS_RDONLY;
3703 ext4_setup_system_zone(sb);
3704 if (sbi->s_journal == NULL)
3705 ext4_commit_super(sb, 1);
3708 /* Release old quota file names */
3709 for (i = 0; i < MAXQUOTAS; i++)
3710 if (old_opts.s_qf_names[i] &&
3711 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3712 kfree(old_opts.s_qf_names[i]);
3717 dquot_resume(sb, -1);
3719 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3724 sb->s_flags = old_sb_flags;
3725 sbi->s_mount_opt = old_opts.s_mount_opt;
3726 sbi->s_resuid = old_opts.s_resuid;
3727 sbi->s_resgid = old_opts.s_resgid;
3728 sbi->s_commit_interval = old_opts.s_commit_interval;
3729 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3730 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3732 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3733 for (i = 0; i < MAXQUOTAS; i++) {
3734 if (sbi->s_qf_names[i] &&
3735 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3736 kfree(sbi->s_qf_names[i]);
3737 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3746 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3748 struct super_block *sb = dentry->d_sb;
3749 struct ext4_sb_info *sbi = EXT4_SB(sb);
3750 struct ext4_super_block *es = sbi->s_es;
3753 if (test_opt(sb, MINIX_DF)) {
3754 sbi->s_overhead_last = 0;
3755 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3756 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3757 ext4_fsblk_t overhead = 0;
3760 * Compute the overhead (FS structures). This is constant
3761 * for a given filesystem unless the number of block groups
3762 * changes so we cache the previous value until it does.
3766 * All of the blocks before first_data_block are
3769 overhead = le32_to_cpu(es->s_first_data_block);
3772 * Add the overhead attributed to the superblock and
3773 * block group descriptors. If the sparse superblocks
3774 * feature is turned on, then not all groups have this.
3776 for (i = 0; i < ngroups; i++) {
3777 overhead += ext4_bg_has_super(sb, i) +
3778 ext4_bg_num_gdb(sb, i);
3783 * Every block group has an inode bitmap, a block
3784 * bitmap, and an inode table.
3786 overhead += ngroups * (2 + sbi->s_itb_per_group);
3787 sbi->s_overhead_last = overhead;
3789 sbi->s_blocks_last = ext4_blocks_count(es);
3792 buf->f_type = EXT4_SUPER_MAGIC;
3793 buf->f_bsize = sb->s_blocksize;
3794 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3795 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3796 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3797 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3798 if (buf->f_bfree < ext4_r_blocks_count(es))
3800 buf->f_files = le32_to_cpu(es->s_inodes_count);
3801 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3802 buf->f_namelen = EXT4_NAME_LEN;
3803 fsid = le64_to_cpup((void *)es->s_uuid) ^
3804 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3805 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3806 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3811 /* Helper function for writing quotas on sync - we need to start transaction
3812 * before quota file is locked for write. Otherwise the are possible deadlocks:
3813 * Process 1 Process 2
3814 * ext4_create() quota_sync()
3815 * jbd2_journal_start() write_dquot()
3816 * dquot_initialize() down(dqio_mutex)
3817 * down(dqio_mutex) jbd2_journal_start()
3823 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3825 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3828 static int ext4_write_dquot(struct dquot *dquot)
3832 struct inode *inode;
3834 inode = dquot_to_inode(dquot);
3835 handle = ext4_journal_start(inode,
3836 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3838 return PTR_ERR(handle);
3839 ret = dquot_commit(dquot);
3840 err = ext4_journal_stop(handle);
3846 static int ext4_acquire_dquot(struct dquot *dquot)
3851 handle = ext4_journal_start(dquot_to_inode(dquot),
3852 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3854 return PTR_ERR(handle);
3855 ret = dquot_acquire(dquot);
3856 err = ext4_journal_stop(handle);
3862 static int ext4_release_dquot(struct dquot *dquot)
3867 handle = ext4_journal_start(dquot_to_inode(dquot),
3868 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3869 if (IS_ERR(handle)) {
3870 /* Release dquot anyway to avoid endless cycle in dqput() */
3871 dquot_release(dquot);
3872 return PTR_ERR(handle);
3874 ret = dquot_release(dquot);
3875 err = ext4_journal_stop(handle);
3881 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3883 /* Are we journaling quotas? */
3884 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3885 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3886 dquot_mark_dquot_dirty(dquot);
3887 return ext4_write_dquot(dquot);
3889 return dquot_mark_dquot_dirty(dquot);
3893 static int ext4_write_info(struct super_block *sb, int type)
3898 /* Data block + inode block */
3899 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3901 return PTR_ERR(handle);
3902 ret = dquot_commit_info(sb, type);
3903 err = ext4_journal_stop(handle);
3910 * Turn on quotas during mount time - we need to find
3911 * the quota file and such...
3913 static int ext4_quota_on_mount(struct super_block *sb, int type)
3915 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3916 EXT4_SB(sb)->s_jquota_fmt, type);
3920 * Standard function to be called on quota_on
3922 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3928 if (!test_opt(sb, QUOTA))
3931 err = kern_path(name, LOOKUP_FOLLOW, &path);
3935 /* Quotafile not on the same filesystem? */
3936 if (path.mnt->mnt_sb != sb) {
3940 /* Journaling quota? */
3941 if (EXT4_SB(sb)->s_qf_names[type]) {
3942 /* Quotafile not in fs root? */
3943 if (path.dentry->d_parent != sb->s_root)
3944 ext4_msg(sb, KERN_WARNING,
3945 "Quota file not on filesystem root. "
3946 "Journaled quota will not work");
3950 * When we journal data on quota file, we have to flush journal to see
3951 * all updates to the file when we bypass pagecache...
3953 if (EXT4_SB(sb)->s_journal &&
3954 ext4_should_journal_data(path.dentry->d_inode)) {
3956 * We don't need to lock updates but journal_flush() could
3957 * otherwise be livelocked...
3959 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3960 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3961 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3968 err = dquot_quota_on_path(sb, type, format_id, &path);
3973 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3974 * acquiring the locks... As quota files are never truncated and quota code
3975 * itself serializes the operations (and noone else should touch the files)
3976 * we don't have to be afraid of races */
3977 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3978 size_t len, loff_t off)
3980 struct inode *inode = sb_dqopt(sb)->files[type];
3981 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3983 int offset = off & (sb->s_blocksize - 1);
3986 struct buffer_head *bh;
3987 loff_t i_size = i_size_read(inode);
3991 if (off+len > i_size)
3994 while (toread > 0) {
3995 tocopy = sb->s_blocksize - offset < toread ?
3996 sb->s_blocksize - offset : toread;
3997 bh = ext4_bread(NULL, inode, blk, 0, &err);
4000 if (!bh) /* A hole? */
4001 memset(data, 0, tocopy);
4003 memcpy(data, bh->b_data+offset, tocopy);
4013 /* Write to quotafile (we know the transaction is already started and has
4014 * enough credits) */
4015 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4016 const char *data, size_t len, loff_t off)
4018 struct inode *inode = sb_dqopt(sb)->files[type];
4019 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4021 int offset = off & (sb->s_blocksize - 1);
4022 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4023 struct buffer_head *bh;
4024 handle_t *handle = journal_current_handle();
4026 if (EXT4_SB(sb)->s_journal && !handle) {
4027 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4028 " cancelled because transaction is not started",
4029 (unsigned long long)off, (unsigned long long)len);
4033 * Since we account only one data block in transaction credits,
4034 * then it is impossible to cross a block boundary.
4036 if (sb->s_blocksize - offset < len) {
4037 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4038 " cancelled because not block aligned",
4039 (unsigned long long)off, (unsigned long long)len);
4043 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4044 bh = ext4_bread(handle, inode, blk, 1, &err);
4047 if (journal_quota) {
4048 err = ext4_journal_get_write_access(handle, bh);
4055 memcpy(bh->b_data+offset, data, len);
4056 flush_dcache_page(bh->b_page);
4059 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4061 /* Always do at least ordered writes for quotas */
4062 err = ext4_jbd2_file_inode(handle, inode);
4063 mark_buffer_dirty(bh);
4068 mutex_unlock(&inode->i_mutex);
4071 if (inode->i_size < off + len) {
4072 i_size_write(inode, off + len);
4073 EXT4_I(inode)->i_disksize = inode->i_size;
4075 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4076 ext4_mark_inode_dirty(handle, inode);
4077 mutex_unlock(&inode->i_mutex);
4083 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4084 const char *dev_name, void *data, struct vfsmount *mnt)
4086 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4089 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4090 static struct file_system_type ext2_fs_type = {
4091 .owner = THIS_MODULE,
4093 .get_sb = ext4_get_sb,
4094 .kill_sb = kill_block_super,
4095 .fs_flags = FS_REQUIRES_DEV,
4098 static inline void register_as_ext2(void)
4100 int err = register_filesystem(&ext2_fs_type);
4103 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4106 static inline void unregister_as_ext2(void)
4108 unregister_filesystem(&ext2_fs_type);
4110 MODULE_ALIAS("ext2");
4112 static inline void register_as_ext2(void) { }
4113 static inline void unregister_as_ext2(void) { }
4116 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4117 static inline void register_as_ext3(void)
4119 int err = register_filesystem(&ext3_fs_type);
4122 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4125 static inline void unregister_as_ext3(void)
4127 unregister_filesystem(&ext3_fs_type);
4129 MODULE_ALIAS("ext3");
4131 static inline void register_as_ext3(void) { }
4132 static inline void unregister_as_ext3(void) { }
4135 static struct file_system_type ext4_fs_type = {
4136 .owner = THIS_MODULE,
4138 .get_sb = ext4_get_sb,
4139 .kill_sb = kill_block_super,
4140 .fs_flags = FS_REQUIRES_DEV,
4143 static int __init init_ext4_fs(void)
4147 ext4_check_flag_values();
4148 err = init_ext4_system_zone();
4151 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4154 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4155 err = init_ext4_mballoc();
4159 err = init_ext4_xattr();
4162 err = init_inodecache();
4167 err = register_filesystem(&ext4_fs_type);
4172 unregister_as_ext2();
4173 unregister_as_ext3();
4174 destroy_inodecache();
4178 exit_ext4_mballoc();
4180 remove_proc_entry("fs/ext4", NULL);
4181 kset_unregister(ext4_kset);
4183 exit_ext4_system_zone();
4187 static void __exit exit_ext4_fs(void)
4189 unregister_as_ext2();
4190 unregister_as_ext3();
4191 unregister_filesystem(&ext4_fs_type);
4192 destroy_inodecache();
4194 exit_ext4_mballoc();
4195 remove_proc_entry("fs/ext4", NULL);
4196 kset_unregister(ext4_kset);
4197 exit_ext4_system_zone();
4200 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4201 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4202 MODULE_LICENSE("GPL");
4203 module_init(init_ext4_fs)
4204 module_exit(exit_ext4_fs)