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 flush_workqueue(sbi->dio_unwritten_wq);
650 destroy_workqueue(sbi->dio_unwritten_wq);
655 ext4_commit_super(sb, 1);
657 if (sbi->s_journal) {
658 err = jbd2_journal_destroy(sbi->s_journal);
659 sbi->s_journal = NULL;
661 ext4_abort(sb, __func__,
662 "Couldn't clean up the journal");
665 ext4_release_system_zone(sb);
667 ext4_ext_release(sb);
668 ext4_xattr_put_super(sb);
670 if (!(sb->s_flags & MS_RDONLY)) {
671 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
672 es->s_state = cpu_to_le16(sbi->s_mount_state);
673 ext4_commit_super(sb, 1);
676 remove_proc_entry(sb->s_id, ext4_proc_root);
678 kobject_del(&sbi->s_kobj);
680 for (i = 0; i < sbi->s_gdb_count; i++)
681 brelse(sbi->s_group_desc[i]);
682 kfree(sbi->s_group_desc);
683 if (is_vmalloc_addr(sbi->s_flex_groups))
684 vfree(sbi->s_flex_groups);
686 kfree(sbi->s_flex_groups);
687 percpu_counter_destroy(&sbi->s_freeblocks_counter);
688 percpu_counter_destroy(&sbi->s_freeinodes_counter);
689 percpu_counter_destroy(&sbi->s_dirs_counter);
690 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
693 for (i = 0; i < MAXQUOTAS; i++)
694 kfree(sbi->s_qf_names[i]);
697 /* Debugging code just in case the in-memory inode orphan list
698 * isn't empty. The on-disk one can be non-empty if we've
699 * detected an error and taken the fs readonly, but the
700 * in-memory list had better be clean by this point. */
701 if (!list_empty(&sbi->s_orphan))
702 dump_orphan_list(sb, sbi);
703 J_ASSERT(list_empty(&sbi->s_orphan));
705 invalidate_bdev(sb->s_bdev);
706 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
708 * Invalidate the journal device's buffers. We don't want them
709 * floating about in memory - the physical journal device may
710 * hotswapped, and it breaks the `ro-after' testing code.
712 sync_blockdev(sbi->journal_bdev);
713 invalidate_bdev(sbi->journal_bdev);
714 ext4_blkdev_remove(sbi);
716 sb->s_fs_info = NULL;
718 * Now that we are completely done shutting down the
719 * superblock, we need to actually destroy the kobject.
723 kobject_put(&sbi->s_kobj);
724 wait_for_completion(&sbi->s_kobj_unregister);
725 kfree(sbi->s_blockgroup_lock);
729 static struct kmem_cache *ext4_inode_cachep;
732 * Called inside transaction, so use GFP_NOFS
734 static struct inode *ext4_alloc_inode(struct super_block *sb)
736 struct ext4_inode_info *ei;
738 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
742 ei->vfs_inode.i_version = 1;
743 ei->vfs_inode.i_data.writeback_index = 0;
744 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
745 INIT_LIST_HEAD(&ei->i_prealloc_list);
746 spin_lock_init(&ei->i_prealloc_lock);
748 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
749 * therefore it can be null here. Don't check it, just initialize
752 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
753 ei->i_reserved_data_blocks = 0;
754 ei->i_reserved_meta_blocks = 0;
755 ei->i_allocated_meta_blocks = 0;
756 ei->i_da_metadata_calc_len = 0;
757 ei->i_delalloc_reserved_flag = 0;
758 spin_lock_init(&(ei->i_block_reservation_lock));
760 ei->i_reserved_quota = 0;
762 INIT_LIST_HEAD(&ei->i_completed_io_list);
763 spin_lock_init(&ei->i_completed_io_lock);
764 ei->cur_aio_dio = NULL;
766 ei->i_datasync_tid = 0;
768 return &ei->vfs_inode;
771 static void ext4_destroy_inode(struct inode *inode)
773 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
774 ext4_msg(inode->i_sb, KERN_ERR,
775 "Inode %lu (%p): orphan list check failed!",
776 inode->i_ino, EXT4_I(inode));
777 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
778 EXT4_I(inode), sizeof(struct ext4_inode_info),
782 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
785 static void init_once(void *foo)
787 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
789 INIT_LIST_HEAD(&ei->i_orphan);
790 #ifdef CONFIG_EXT4_FS_XATTR
791 init_rwsem(&ei->xattr_sem);
793 init_rwsem(&ei->i_data_sem);
794 inode_init_once(&ei->vfs_inode);
797 static int init_inodecache(void)
799 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
800 sizeof(struct ext4_inode_info),
801 0, (SLAB_RECLAIM_ACCOUNT|
804 if (ext4_inode_cachep == NULL)
809 static void destroy_inodecache(void)
811 kmem_cache_destroy(ext4_inode_cachep);
814 static void ext4_clear_inode(struct inode *inode)
817 ext4_discard_preallocations(inode);
818 if (EXT4_JOURNAL(inode))
819 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
820 &EXT4_I(inode)->jinode);
823 static inline void ext4_show_quota_options(struct seq_file *seq,
824 struct super_block *sb)
826 #if defined(CONFIG_QUOTA)
827 struct ext4_sb_info *sbi = EXT4_SB(sb);
829 if (sbi->s_jquota_fmt) {
832 switch (sbi->s_jquota_fmt) {
843 seq_printf(seq, ",jqfmt=%s", fmtname);
846 if (sbi->s_qf_names[USRQUOTA])
847 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
849 if (sbi->s_qf_names[GRPQUOTA])
850 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
852 if (test_opt(sb, USRQUOTA))
853 seq_puts(seq, ",usrquota");
855 if (test_opt(sb, GRPQUOTA))
856 seq_puts(seq, ",grpquota");
862 * - it's set to a non-default value OR
863 * - if the per-sb default is different from the global default
865 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
868 unsigned long def_mount_opts;
869 struct super_block *sb = vfs->mnt_sb;
870 struct ext4_sb_info *sbi = EXT4_SB(sb);
871 struct ext4_super_block *es = sbi->s_es;
873 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
874 def_errors = le16_to_cpu(es->s_errors);
876 if (sbi->s_sb_block != 1)
877 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
878 if (test_opt(sb, MINIX_DF))
879 seq_puts(seq, ",minixdf");
880 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
881 seq_puts(seq, ",grpid");
882 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
883 seq_puts(seq, ",nogrpid");
884 if (sbi->s_resuid != EXT4_DEF_RESUID ||
885 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
886 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
888 if (sbi->s_resgid != EXT4_DEF_RESGID ||
889 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
890 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
892 if (test_opt(sb, ERRORS_RO)) {
893 if (def_errors == EXT4_ERRORS_PANIC ||
894 def_errors == EXT4_ERRORS_CONTINUE) {
895 seq_puts(seq, ",errors=remount-ro");
898 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
899 seq_puts(seq, ",errors=continue");
900 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
901 seq_puts(seq, ",errors=panic");
902 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
903 seq_puts(seq, ",nouid32");
904 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
905 seq_puts(seq, ",debug");
906 if (test_opt(sb, OLDALLOC))
907 seq_puts(seq, ",oldalloc");
908 #ifdef CONFIG_EXT4_FS_XATTR
909 if (test_opt(sb, XATTR_USER) &&
910 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
911 seq_puts(seq, ",user_xattr");
912 if (!test_opt(sb, XATTR_USER) &&
913 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
914 seq_puts(seq, ",nouser_xattr");
917 #ifdef CONFIG_EXT4_FS_POSIX_ACL
918 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
919 seq_puts(seq, ",acl");
920 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
921 seq_puts(seq, ",noacl");
923 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
924 seq_printf(seq, ",commit=%u",
925 (unsigned) (sbi->s_commit_interval / HZ));
927 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
928 seq_printf(seq, ",min_batch_time=%u",
929 (unsigned) sbi->s_min_batch_time);
931 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
932 seq_printf(seq, ",max_batch_time=%u",
933 (unsigned) sbi->s_min_batch_time);
937 * We're changing the default of barrier mount option, so
938 * let's always display its mount state so it's clear what its
941 seq_puts(seq, ",barrier=");
942 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
943 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
944 seq_puts(seq, ",journal_async_commit");
945 if (test_opt(sb, NOBH))
946 seq_puts(seq, ",nobh");
947 if (test_opt(sb, I_VERSION))
948 seq_puts(seq, ",i_version");
949 if (!test_opt(sb, DELALLOC))
950 seq_puts(seq, ",nodelalloc");
954 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
956 * journal mode get enabled in different ways
957 * So just print the value even if we didn't specify it
959 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
960 seq_puts(seq, ",data=journal");
961 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
962 seq_puts(seq, ",data=ordered");
963 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
964 seq_puts(seq, ",data=writeback");
966 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
967 seq_printf(seq, ",inode_readahead_blks=%u",
968 sbi->s_inode_readahead_blks);
970 if (test_opt(sb, DATA_ERR_ABORT))
971 seq_puts(seq, ",data_err=abort");
973 if (test_opt(sb, NO_AUTO_DA_ALLOC))
974 seq_puts(seq, ",noauto_da_alloc");
976 if (test_opt(sb, DISCARD))
977 seq_puts(seq, ",discard");
979 if (test_opt(sb, NOLOAD))
980 seq_puts(seq, ",norecovery");
982 if (test_opt(sb, DIOREAD_NOLOCK))
983 seq_puts(seq, ",dioread_nolock");
985 ext4_show_quota_options(seq, sb);
990 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
991 u64 ino, u32 generation)
995 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
996 return ERR_PTR(-ESTALE);
997 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
998 return ERR_PTR(-ESTALE);
1000 /* iget isn't really right if the inode is currently unallocated!!
1002 * ext4_read_inode will return a bad_inode if the inode had been
1003 * deleted, so we should be safe.
1005 * Currently we don't know the generation for parent directory, so
1006 * a generation of 0 means "accept any"
1008 inode = ext4_iget(sb, ino);
1010 return ERR_CAST(inode);
1011 if (generation && inode->i_generation != generation) {
1013 return ERR_PTR(-ESTALE);
1019 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1020 int fh_len, int fh_type)
1022 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1023 ext4_nfs_get_inode);
1026 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1027 int fh_len, int fh_type)
1029 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1030 ext4_nfs_get_inode);
1034 * Try to release metadata pages (indirect blocks, directories) which are
1035 * mapped via the block device. Since these pages could have journal heads
1036 * which would prevent try_to_free_buffers() from freeing them, we must use
1037 * jbd2 layer's try_to_free_buffers() function to release them.
1039 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1042 journal_t *journal = EXT4_SB(sb)->s_journal;
1044 WARN_ON(PageChecked(page));
1045 if (!page_has_buffers(page))
1048 return jbd2_journal_try_to_free_buffers(journal, page,
1049 wait & ~__GFP_WAIT);
1050 return try_to_free_buffers(page);
1054 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1055 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1057 static int ext4_write_dquot(struct dquot *dquot);
1058 static int ext4_acquire_dquot(struct dquot *dquot);
1059 static int ext4_release_dquot(struct dquot *dquot);
1060 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1061 static int ext4_write_info(struct super_block *sb, int type);
1062 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1063 char *path, int remount);
1064 static int ext4_quota_on_mount(struct super_block *sb, int type);
1065 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1066 size_t len, loff_t off);
1067 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1068 const char *data, size_t len, loff_t off);
1070 static const struct dquot_operations ext4_quota_operations = {
1072 .get_reserved_space = ext4_get_reserved_space,
1074 .write_dquot = ext4_write_dquot,
1075 .acquire_dquot = ext4_acquire_dquot,
1076 .release_dquot = ext4_release_dquot,
1077 .mark_dirty = ext4_mark_dquot_dirty,
1078 .write_info = ext4_write_info,
1079 .alloc_dquot = dquot_alloc,
1080 .destroy_dquot = dquot_destroy,
1083 static const struct quotactl_ops ext4_qctl_operations = {
1084 .quota_on = ext4_quota_on,
1085 .quota_off = vfs_quota_off,
1086 .quota_sync = vfs_quota_sync,
1087 .get_info = vfs_get_dqinfo,
1088 .set_info = vfs_set_dqinfo,
1089 .get_dqblk = vfs_get_dqblk,
1090 .set_dqblk = vfs_set_dqblk
1094 static const struct super_operations ext4_sops = {
1095 .alloc_inode = ext4_alloc_inode,
1096 .destroy_inode = ext4_destroy_inode,
1097 .write_inode = ext4_write_inode,
1098 .dirty_inode = ext4_dirty_inode,
1099 .delete_inode = ext4_delete_inode,
1100 .put_super = ext4_put_super,
1101 .sync_fs = ext4_sync_fs,
1102 .freeze_fs = ext4_freeze,
1103 .unfreeze_fs = ext4_unfreeze,
1104 .statfs = ext4_statfs,
1105 .remount_fs = ext4_remount,
1106 .clear_inode = ext4_clear_inode,
1107 .show_options = ext4_show_options,
1109 .quota_read = ext4_quota_read,
1110 .quota_write = ext4_quota_write,
1112 .bdev_try_to_free_page = bdev_try_to_free_page,
1115 static const struct super_operations ext4_nojournal_sops = {
1116 .alloc_inode = ext4_alloc_inode,
1117 .destroy_inode = ext4_destroy_inode,
1118 .write_inode = ext4_write_inode,
1119 .dirty_inode = ext4_dirty_inode,
1120 .delete_inode = ext4_delete_inode,
1121 .write_super = ext4_write_super,
1122 .put_super = ext4_put_super,
1123 .statfs = ext4_statfs,
1124 .remount_fs = ext4_remount,
1125 .clear_inode = ext4_clear_inode,
1126 .show_options = ext4_show_options,
1128 .quota_read = ext4_quota_read,
1129 .quota_write = ext4_quota_write,
1131 .bdev_try_to_free_page = bdev_try_to_free_page,
1134 static const struct export_operations ext4_export_ops = {
1135 .fh_to_dentry = ext4_fh_to_dentry,
1136 .fh_to_parent = ext4_fh_to_parent,
1137 .get_parent = ext4_get_parent,
1141 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1142 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1143 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1144 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1145 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1146 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1147 Opt_journal_update, Opt_journal_dev,
1148 Opt_journal_checksum, Opt_journal_async_commit,
1149 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1150 Opt_data_err_abort, Opt_data_err_ignore,
1151 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1152 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1153 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1154 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1155 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1156 Opt_block_validity, Opt_noblock_validity,
1157 Opt_inode_readahead_blks, Opt_journal_ioprio,
1158 Opt_dioread_nolock, Opt_dioread_lock,
1159 Opt_discard, Opt_nodiscard,
1162 static const match_table_t tokens = {
1163 {Opt_bsd_df, "bsddf"},
1164 {Opt_minix_df, "minixdf"},
1165 {Opt_grpid, "grpid"},
1166 {Opt_grpid, "bsdgroups"},
1167 {Opt_nogrpid, "nogrpid"},
1168 {Opt_nogrpid, "sysvgroups"},
1169 {Opt_resgid, "resgid=%u"},
1170 {Opt_resuid, "resuid=%u"},
1172 {Opt_err_cont, "errors=continue"},
1173 {Opt_err_panic, "errors=panic"},
1174 {Opt_err_ro, "errors=remount-ro"},
1175 {Opt_nouid32, "nouid32"},
1176 {Opt_debug, "debug"},
1177 {Opt_oldalloc, "oldalloc"},
1178 {Opt_orlov, "orlov"},
1179 {Opt_user_xattr, "user_xattr"},
1180 {Opt_nouser_xattr, "nouser_xattr"},
1182 {Opt_noacl, "noacl"},
1183 {Opt_noload, "noload"},
1184 {Opt_noload, "norecovery"},
1187 {Opt_commit, "commit=%u"},
1188 {Opt_min_batch_time, "min_batch_time=%u"},
1189 {Opt_max_batch_time, "max_batch_time=%u"},
1190 {Opt_journal_update, "journal=update"},
1191 {Opt_journal_dev, "journal_dev=%u"},
1192 {Opt_journal_checksum, "journal_checksum"},
1193 {Opt_journal_async_commit, "journal_async_commit"},
1194 {Opt_abort, "abort"},
1195 {Opt_data_journal, "data=journal"},
1196 {Opt_data_ordered, "data=ordered"},
1197 {Opt_data_writeback, "data=writeback"},
1198 {Opt_data_err_abort, "data_err=abort"},
1199 {Opt_data_err_ignore, "data_err=ignore"},
1200 {Opt_offusrjquota, "usrjquota="},
1201 {Opt_usrjquota, "usrjquota=%s"},
1202 {Opt_offgrpjquota, "grpjquota="},
1203 {Opt_grpjquota, "grpjquota=%s"},
1204 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1205 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1206 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1207 {Opt_grpquota, "grpquota"},
1208 {Opt_noquota, "noquota"},
1209 {Opt_quota, "quota"},
1210 {Opt_usrquota, "usrquota"},
1211 {Opt_barrier, "barrier=%u"},
1212 {Opt_barrier, "barrier"},
1213 {Opt_nobarrier, "nobarrier"},
1214 {Opt_i_version, "i_version"},
1215 {Opt_stripe, "stripe=%u"},
1216 {Opt_resize, "resize"},
1217 {Opt_delalloc, "delalloc"},
1218 {Opt_nodelalloc, "nodelalloc"},
1219 {Opt_block_validity, "block_validity"},
1220 {Opt_noblock_validity, "noblock_validity"},
1221 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1222 {Opt_journal_ioprio, "journal_ioprio=%u"},
1223 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1224 {Opt_auto_da_alloc, "auto_da_alloc"},
1225 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1226 {Opt_dioread_nolock, "dioread_nolock"},
1227 {Opt_dioread_lock, "dioread_lock"},
1228 {Opt_discard, "discard"},
1229 {Opt_nodiscard, "nodiscard"},
1233 static ext4_fsblk_t get_sb_block(void **data)
1235 ext4_fsblk_t sb_block;
1236 char *options = (char *) *data;
1238 if (!options || strncmp(options, "sb=", 3) != 0)
1239 return 1; /* Default location */
1242 /* TODO: use simple_strtoll with >32bit ext4 */
1243 sb_block = simple_strtoul(options, &options, 0);
1244 if (*options && *options != ',') {
1245 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1249 if (*options == ',')
1251 *data = (void *) options;
1256 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1257 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1258 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1261 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1263 struct ext4_sb_info *sbi = EXT4_SB(sb);
1266 if (sb_any_quota_loaded(sb) &&
1267 !sbi->s_qf_names[qtype]) {
1268 ext4_msg(sb, KERN_ERR,
1269 "Cannot change journaled "
1270 "quota options when quota turned on");
1273 qname = match_strdup(args);
1275 ext4_msg(sb, KERN_ERR,
1276 "Not enough memory for storing quotafile name");
1279 if (sbi->s_qf_names[qtype] &&
1280 strcmp(sbi->s_qf_names[qtype], qname)) {
1281 ext4_msg(sb, KERN_ERR,
1282 "%s quota file already specified", QTYPE2NAME(qtype));
1286 sbi->s_qf_names[qtype] = qname;
1287 if (strchr(sbi->s_qf_names[qtype], '/')) {
1288 ext4_msg(sb, KERN_ERR,
1289 "quotafile must be on filesystem root");
1290 kfree(sbi->s_qf_names[qtype]);
1291 sbi->s_qf_names[qtype] = NULL;
1294 set_opt(sbi->s_mount_opt, QUOTA);
1298 static int clear_qf_name(struct super_block *sb, int qtype)
1301 struct ext4_sb_info *sbi = EXT4_SB(sb);
1303 if (sb_any_quota_loaded(sb) &&
1304 sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1306 " when quota turned on");
1310 * The space will be released later when all options are confirmed
1313 sbi->s_qf_names[qtype] = NULL;
1318 static int parse_options(char *options, struct super_block *sb,
1319 unsigned long *journal_devnum,
1320 unsigned int *journal_ioprio,
1321 ext4_fsblk_t *n_blocks_count, int is_remount)
1323 struct ext4_sb_info *sbi = EXT4_SB(sb);
1325 substring_t args[MAX_OPT_ARGS];
1335 while ((p = strsep(&options, ",")) != NULL) {
1341 * Initialize args struct so we know whether arg was
1342 * found; some options take optional arguments.
1344 args[0].to = args[0].from = 0;
1345 token = match_token(p, tokens, args);
1348 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1349 clear_opt(sbi->s_mount_opt, MINIX_DF);
1352 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1353 set_opt(sbi->s_mount_opt, MINIX_DF);
1357 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1358 set_opt(sbi->s_mount_opt, GRPID);
1362 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1363 clear_opt(sbi->s_mount_opt, GRPID);
1367 if (match_int(&args[0], &option))
1369 sbi->s_resuid = option;
1372 if (match_int(&args[0], &option))
1374 sbi->s_resgid = option;
1377 /* handled by get_sb_block() instead of here */
1378 /* *sb_block = match_int(&args[0]); */
1381 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1382 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1383 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1386 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1387 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1388 set_opt(sbi->s_mount_opt, ERRORS_RO);
1391 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1392 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1393 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1396 set_opt(sbi->s_mount_opt, NO_UID32);
1399 set_opt(sbi->s_mount_opt, DEBUG);
1402 set_opt(sbi->s_mount_opt, OLDALLOC);
1405 clear_opt(sbi->s_mount_opt, OLDALLOC);
1407 #ifdef CONFIG_EXT4_FS_XATTR
1408 case Opt_user_xattr:
1409 set_opt(sbi->s_mount_opt, XATTR_USER);
1411 case Opt_nouser_xattr:
1412 clear_opt(sbi->s_mount_opt, XATTR_USER);
1415 case Opt_user_xattr:
1416 case Opt_nouser_xattr:
1417 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1420 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1422 set_opt(sbi->s_mount_opt, POSIX_ACL);
1425 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1430 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1433 case Opt_journal_update:
1435 /* Eventually we will want to be able to create
1436 a journal file here. For now, only allow the
1437 user to specify an existing inode to be the
1440 ext4_msg(sb, KERN_ERR,
1441 "Cannot specify journal on remount");
1444 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1446 case Opt_journal_dev:
1448 ext4_msg(sb, KERN_ERR,
1449 "Cannot specify journal on remount");
1452 if (match_int(&args[0], &option))
1454 *journal_devnum = option;
1456 case Opt_journal_checksum:
1457 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1459 case Opt_journal_async_commit:
1460 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1461 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1464 set_opt(sbi->s_mount_opt, NOLOAD);
1467 if (match_int(&args[0], &option))
1472 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1473 sbi->s_commit_interval = HZ * option;
1475 case Opt_max_batch_time:
1476 if (match_int(&args[0], &option))
1481 option = EXT4_DEF_MAX_BATCH_TIME;
1482 sbi->s_max_batch_time = option;
1484 case Opt_min_batch_time:
1485 if (match_int(&args[0], &option))
1489 sbi->s_min_batch_time = option;
1491 case Opt_data_journal:
1492 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1494 case Opt_data_ordered:
1495 data_opt = EXT4_MOUNT_ORDERED_DATA;
1497 case Opt_data_writeback:
1498 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1501 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Cannot change data mode on remount");
1507 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1508 sbi->s_mount_opt |= data_opt;
1511 case Opt_data_err_abort:
1512 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1514 case Opt_data_err_ignore:
1515 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1519 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1523 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1526 case Opt_offusrjquota:
1527 if (!clear_qf_name(sb, USRQUOTA))
1530 case Opt_offgrpjquota:
1531 if (!clear_qf_name(sb, GRPQUOTA))
1535 case Opt_jqfmt_vfsold:
1536 qfmt = QFMT_VFS_OLD;
1538 case Opt_jqfmt_vfsv0:
1541 case Opt_jqfmt_vfsv1:
1544 if (sb_any_quota_loaded(sb) &&
1545 sbi->s_jquota_fmt != qfmt) {
1546 ext4_msg(sb, KERN_ERR, "Cannot change "
1547 "journaled quota options when "
1551 sbi->s_jquota_fmt = qfmt;
1555 set_opt(sbi->s_mount_opt, QUOTA);
1556 set_opt(sbi->s_mount_opt, USRQUOTA);
1559 set_opt(sbi->s_mount_opt, QUOTA);
1560 set_opt(sbi->s_mount_opt, GRPQUOTA);
1563 if (sb_any_quota_loaded(sb)) {
1564 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1565 "options when quota turned on");
1568 clear_opt(sbi->s_mount_opt, QUOTA);
1569 clear_opt(sbi->s_mount_opt, USRQUOTA);
1570 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1576 ext4_msg(sb, KERN_ERR,
1577 "quota options not supported");
1581 case Opt_offusrjquota:
1582 case Opt_offgrpjquota:
1583 case Opt_jqfmt_vfsold:
1584 case Opt_jqfmt_vfsv0:
1585 case Opt_jqfmt_vfsv1:
1586 ext4_msg(sb, KERN_ERR,
1587 "journaled quota options not supported");
1593 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1596 clear_opt(sbi->s_mount_opt, BARRIER);
1600 if (match_int(&args[0], &option))
1603 option = 1; /* No argument, default to 1 */
1605 set_opt(sbi->s_mount_opt, BARRIER);
1607 clear_opt(sbi->s_mount_opt, BARRIER);
1613 ext4_msg(sb, KERN_ERR,
1614 "resize option only available "
1618 if (match_int(&args[0], &option) != 0)
1620 *n_blocks_count = option;
1623 set_opt(sbi->s_mount_opt, NOBH);
1626 clear_opt(sbi->s_mount_opt, NOBH);
1629 set_opt(sbi->s_mount_opt, I_VERSION);
1630 sb->s_flags |= MS_I_VERSION;
1632 case Opt_nodelalloc:
1633 clear_opt(sbi->s_mount_opt, DELALLOC);
1636 if (match_int(&args[0], &option))
1640 sbi->s_stripe = option;
1643 set_opt(sbi->s_mount_opt, DELALLOC);
1645 case Opt_block_validity:
1646 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1648 case Opt_noblock_validity:
1649 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1651 case Opt_inode_readahead_blks:
1652 if (match_int(&args[0], &option))
1654 if (option < 0 || option > (1 << 30))
1656 if (!is_power_of_2(option)) {
1657 ext4_msg(sb, KERN_ERR,
1658 "EXT4-fs: inode_readahead_blks"
1659 " must be a power of 2");
1662 sbi->s_inode_readahead_blks = option;
1664 case Opt_journal_ioprio:
1665 if (match_int(&args[0], &option))
1667 if (option < 0 || option > 7)
1669 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1672 case Opt_noauto_da_alloc:
1673 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1675 case Opt_auto_da_alloc:
1677 if (match_int(&args[0], &option))
1680 option = 1; /* No argument, default to 1 */
1682 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1684 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1687 set_opt(sbi->s_mount_opt, DISCARD);
1690 clear_opt(sbi->s_mount_opt, DISCARD);
1692 case Opt_dioread_nolock:
1693 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1695 case Opt_dioread_lock:
1696 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1699 ext4_msg(sb, KERN_ERR,
1700 "Unrecognized mount option \"%s\" "
1701 "or missing value", p);
1706 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1707 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1708 clear_opt(sbi->s_mount_opt, USRQUOTA);
1710 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1711 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1713 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1714 ext4_msg(sb, KERN_ERR, "old and new quota "
1719 if (!sbi->s_jquota_fmt) {
1720 ext4_msg(sb, KERN_ERR, "journaled quota format "
1725 if (sbi->s_jquota_fmt) {
1726 ext4_msg(sb, KERN_ERR, "journaled quota format "
1727 "specified with no journaling "
1736 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1739 struct ext4_sb_info *sbi = EXT4_SB(sb);
1742 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1743 ext4_msg(sb, KERN_ERR, "revision level too high, "
1744 "forcing read-only mode");
1749 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1750 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1751 "running e2fsck is recommended");
1752 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1753 ext4_msg(sb, KERN_WARNING,
1754 "warning: mounting fs with errors, "
1755 "running e2fsck is recommended");
1756 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1757 le16_to_cpu(es->s_mnt_count) >=
1758 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1759 ext4_msg(sb, KERN_WARNING,
1760 "warning: maximal mount count reached, "
1761 "running e2fsck is recommended");
1762 else if (le32_to_cpu(es->s_checkinterval) &&
1763 (le32_to_cpu(es->s_lastcheck) +
1764 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1765 ext4_msg(sb, KERN_WARNING,
1766 "warning: checktime reached, "
1767 "running e2fsck is recommended");
1768 if (!sbi->s_journal)
1769 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1770 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1771 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1772 le16_add_cpu(&es->s_mnt_count, 1);
1773 es->s_mtime = cpu_to_le32(get_seconds());
1774 ext4_update_dynamic_rev(sb);
1776 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1778 ext4_commit_super(sb, 1);
1779 if (test_opt(sb, DEBUG))
1780 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1781 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1783 sbi->s_groups_count,
1784 EXT4_BLOCKS_PER_GROUP(sb),
1785 EXT4_INODES_PER_GROUP(sb),
1791 static int ext4_fill_flex_info(struct super_block *sb)
1793 struct ext4_sb_info *sbi = EXT4_SB(sb);
1794 struct ext4_group_desc *gdp = NULL;
1795 ext4_group_t flex_group_count;
1796 ext4_group_t flex_group;
1797 int groups_per_flex = 0;
1801 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1802 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1804 if (groups_per_flex < 2) {
1805 sbi->s_log_groups_per_flex = 0;
1809 /* We allocate both existing and potentially added groups */
1810 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1811 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1812 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1813 size = flex_group_count * sizeof(struct flex_groups);
1814 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1815 if (sbi->s_flex_groups == NULL) {
1816 sbi->s_flex_groups = vmalloc(size);
1817 if (sbi->s_flex_groups)
1818 memset(sbi->s_flex_groups, 0, size);
1820 if (sbi->s_flex_groups == NULL) {
1821 ext4_msg(sb, KERN_ERR, "not enough memory for "
1822 "%u flex groups", flex_group_count);
1826 for (i = 0; i < sbi->s_groups_count; i++) {
1827 gdp = ext4_get_group_desc(sb, i, NULL);
1829 flex_group = ext4_flex_group(sbi, i);
1830 atomic_add(ext4_free_inodes_count(sb, gdp),
1831 &sbi->s_flex_groups[flex_group].free_inodes);
1832 atomic_add(ext4_free_blks_count(sb, gdp),
1833 &sbi->s_flex_groups[flex_group].free_blocks);
1834 atomic_add(ext4_used_dirs_count(sb, gdp),
1835 &sbi->s_flex_groups[flex_group].used_dirs);
1843 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1844 struct ext4_group_desc *gdp)
1848 if (sbi->s_es->s_feature_ro_compat &
1849 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1850 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1851 __le32 le_group = cpu_to_le32(block_group);
1853 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1854 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1855 crc = crc16(crc, (__u8 *)gdp, offset);
1856 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1857 /* for checksum of struct ext4_group_desc do the rest...*/
1858 if ((sbi->s_es->s_feature_incompat &
1859 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1860 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1861 crc = crc16(crc, (__u8 *)gdp + offset,
1862 le16_to_cpu(sbi->s_es->s_desc_size) -
1866 return cpu_to_le16(crc);
1869 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1870 struct ext4_group_desc *gdp)
1872 if ((sbi->s_es->s_feature_ro_compat &
1873 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1874 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1880 /* Called at mount-time, super-block is locked */
1881 static int ext4_check_descriptors(struct super_block *sb)
1883 struct ext4_sb_info *sbi = EXT4_SB(sb);
1884 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1885 ext4_fsblk_t last_block;
1886 ext4_fsblk_t block_bitmap;
1887 ext4_fsblk_t inode_bitmap;
1888 ext4_fsblk_t inode_table;
1889 int flexbg_flag = 0;
1892 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1895 ext4_debug("Checking group descriptors");
1897 for (i = 0; i < sbi->s_groups_count; i++) {
1898 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1900 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1901 last_block = ext4_blocks_count(sbi->s_es) - 1;
1903 last_block = first_block +
1904 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1906 block_bitmap = ext4_block_bitmap(sb, gdp);
1907 if (block_bitmap < first_block || block_bitmap > last_block) {
1908 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1909 "Block bitmap for group %u not in group "
1910 "(block %llu)!", i, block_bitmap);
1913 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1914 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1915 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1916 "Inode bitmap for group %u not in group "
1917 "(block %llu)!", i, inode_bitmap);
1920 inode_table = ext4_inode_table(sb, gdp);
1921 if (inode_table < first_block ||
1922 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1923 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1924 "Inode table for group %u not in group "
1925 "(block %llu)!", i, inode_table);
1928 ext4_lock_group(sb, i);
1929 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1930 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1931 "Checksum for group %u failed (%u!=%u)",
1932 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1933 gdp)), le16_to_cpu(gdp->bg_checksum));
1934 if (!(sb->s_flags & MS_RDONLY)) {
1935 ext4_unlock_group(sb, i);
1939 ext4_unlock_group(sb, i);
1941 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1944 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1945 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1949 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1950 * the superblock) which were deleted from all directories, but held open by
1951 * a process at the time of a crash. We walk the list and try to delete these
1952 * inodes at recovery time (only with a read-write filesystem).
1954 * In order to keep the orphan inode chain consistent during traversal (in
1955 * case of crash during recovery), we link each inode into the superblock
1956 * orphan list_head and handle it the same way as an inode deletion during
1957 * normal operation (which journals the operations for us).
1959 * We only do an iget() and an iput() on each inode, which is very safe if we
1960 * accidentally point at an in-use or already deleted inode. The worst that
1961 * can happen in this case is that we get a "bit already cleared" message from
1962 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1963 * e2fsck was run on this filesystem, and it must have already done the orphan
1964 * inode cleanup for us, so we can safely abort without any further action.
1966 static void ext4_orphan_cleanup(struct super_block *sb,
1967 struct ext4_super_block *es)
1969 unsigned int s_flags = sb->s_flags;
1970 int nr_orphans = 0, nr_truncates = 0;
1974 if (!es->s_last_orphan) {
1975 jbd_debug(4, "no orphan inodes to clean up\n");
1979 if (bdev_read_only(sb->s_bdev)) {
1980 ext4_msg(sb, KERN_ERR, "write access "
1981 "unavailable, skipping orphan cleanup");
1985 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1986 if (es->s_last_orphan)
1987 jbd_debug(1, "Errors on filesystem, "
1988 "clearing orphan list.\n");
1989 es->s_last_orphan = 0;
1990 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1994 if (s_flags & MS_RDONLY) {
1995 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1996 sb->s_flags &= ~MS_RDONLY;
1999 /* Needed for iput() to work correctly and not trash data */
2000 sb->s_flags |= MS_ACTIVE;
2001 /* Turn on quotas so that they are updated correctly */
2002 for (i = 0; i < MAXQUOTAS; i++) {
2003 if (EXT4_SB(sb)->s_qf_names[i]) {
2004 int ret = ext4_quota_on_mount(sb, i);
2006 ext4_msg(sb, KERN_ERR,
2007 "Cannot turn on journaled "
2008 "quota: error %d", ret);
2013 while (es->s_last_orphan) {
2014 struct inode *inode;
2016 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2017 if (IS_ERR(inode)) {
2018 es->s_last_orphan = 0;
2022 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2023 dquot_initialize(inode);
2024 if (inode->i_nlink) {
2025 ext4_msg(sb, KERN_DEBUG,
2026 "%s: truncating inode %lu to %lld bytes",
2027 __func__, inode->i_ino, inode->i_size);
2028 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2029 inode->i_ino, inode->i_size);
2030 ext4_truncate(inode);
2033 ext4_msg(sb, KERN_DEBUG,
2034 "%s: deleting unreferenced inode %lu",
2035 __func__, inode->i_ino);
2036 jbd_debug(2, "deleting unreferenced inode %lu\n",
2040 iput(inode); /* The delete magic happens here! */
2043 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2046 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2047 PLURAL(nr_orphans));
2049 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2050 PLURAL(nr_truncates));
2052 /* Turn quotas off */
2053 for (i = 0; i < MAXQUOTAS; i++) {
2054 if (sb_dqopt(sb)->files[i])
2055 vfs_quota_off(sb, i, 0);
2058 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2062 * Maximal extent format file size.
2063 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2064 * extent format containers, within a sector_t, and within i_blocks
2065 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2066 * so that won't be a limiting factor.
2068 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2070 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2073 loff_t upper_limit = MAX_LFS_FILESIZE;
2075 /* small i_blocks in vfs inode? */
2076 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2078 * CONFIG_LBDAF is not enabled implies the inode
2079 * i_block represent total blocks in 512 bytes
2080 * 32 == size of vfs inode i_blocks * 8
2082 upper_limit = (1LL << 32) - 1;
2084 /* total blocks in file system block size */
2085 upper_limit >>= (blkbits - 9);
2086 upper_limit <<= blkbits;
2089 /* 32-bit extent-start container, ee_block */
2094 /* Sanity check against vm- & vfs- imposed limits */
2095 if (res > upper_limit)
2102 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2103 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2104 * We need to be 1 filesystem block less than the 2^48 sector limit.
2106 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2108 loff_t res = EXT4_NDIR_BLOCKS;
2111 /* This is calculated to be the largest file size for a dense, block
2112 * mapped file such that the file's total number of 512-byte sectors,
2113 * including data and all indirect blocks, does not exceed (2^48 - 1).
2115 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2116 * number of 512-byte sectors of the file.
2119 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2121 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2122 * the inode i_block field represents total file blocks in
2123 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2125 upper_limit = (1LL << 32) - 1;
2127 /* total blocks in file system block size */
2128 upper_limit >>= (bits - 9);
2132 * We use 48 bit ext4_inode i_blocks
2133 * With EXT4_HUGE_FILE_FL set the i_blocks
2134 * represent total number of blocks in
2135 * file system block size
2137 upper_limit = (1LL << 48) - 1;
2141 /* indirect blocks */
2143 /* double indirect blocks */
2144 meta_blocks += 1 + (1LL << (bits-2));
2145 /* tripple indirect blocks */
2146 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2148 upper_limit -= meta_blocks;
2149 upper_limit <<= bits;
2151 res += 1LL << (bits-2);
2152 res += 1LL << (2*(bits-2));
2153 res += 1LL << (3*(bits-2));
2155 if (res > upper_limit)
2158 if (res > MAX_LFS_FILESIZE)
2159 res = MAX_LFS_FILESIZE;
2164 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2165 ext4_fsblk_t logical_sb_block, int nr)
2167 struct ext4_sb_info *sbi = EXT4_SB(sb);
2168 ext4_group_t bg, first_meta_bg;
2171 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2173 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2175 return logical_sb_block + nr + 1;
2176 bg = sbi->s_desc_per_block * nr;
2177 if (ext4_bg_has_super(sb, bg))
2180 return (has_super + ext4_group_first_block_no(sb, bg));
2184 * ext4_get_stripe_size: Get the stripe size.
2185 * @sbi: In memory super block info
2187 * If we have specified it via mount option, then
2188 * use the mount option value. If the value specified at mount time is
2189 * greater than the blocks per group use the super block value.
2190 * If the super block value is greater than blocks per group return 0.
2191 * Allocator needs it be less than blocks per group.
2194 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2196 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2197 unsigned long stripe_width =
2198 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2200 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2201 return sbi->s_stripe;
2203 if (stripe_width <= sbi->s_blocks_per_group)
2204 return stripe_width;
2206 if (stride <= sbi->s_blocks_per_group)
2215 struct attribute attr;
2216 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2217 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2218 const char *, size_t);
2222 static int parse_strtoul(const char *buf,
2223 unsigned long max, unsigned long *value)
2227 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2228 endp = skip_spaces(endp);
2229 if (*endp || *value > max)
2235 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2236 struct ext4_sb_info *sbi,
2239 return snprintf(buf, PAGE_SIZE, "%llu\n",
2240 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2243 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2244 struct ext4_sb_info *sbi, char *buf)
2246 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2248 return snprintf(buf, PAGE_SIZE, "%lu\n",
2249 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2250 sbi->s_sectors_written_start) >> 1);
2253 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2254 struct ext4_sb_info *sbi, char *buf)
2256 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2258 return snprintf(buf, PAGE_SIZE, "%llu\n",
2259 (unsigned long long)(sbi->s_kbytes_written +
2260 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2261 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2264 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2265 struct ext4_sb_info *sbi,
2266 const char *buf, size_t count)
2270 if (parse_strtoul(buf, 0x40000000, &t))
2273 if (!is_power_of_2(t))
2276 sbi->s_inode_readahead_blks = t;
2280 static ssize_t sbi_ui_show(struct ext4_attr *a,
2281 struct ext4_sb_info *sbi, char *buf)
2283 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2285 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2288 static ssize_t sbi_ui_store(struct ext4_attr *a,
2289 struct ext4_sb_info *sbi,
2290 const char *buf, size_t count)
2292 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2295 if (parse_strtoul(buf, 0xffffffff, &t))
2301 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2302 static struct ext4_attr ext4_attr_##_name = { \
2303 .attr = {.name = __stringify(_name), .mode = _mode }, \
2306 .offset = offsetof(struct ext4_sb_info, _elname), \
2308 #define EXT4_ATTR(name, mode, show, store) \
2309 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2311 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2312 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2313 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2314 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2315 #define ATTR_LIST(name) &ext4_attr_##name.attr
2317 EXT4_RO_ATTR(delayed_allocation_blocks);
2318 EXT4_RO_ATTR(session_write_kbytes);
2319 EXT4_RO_ATTR(lifetime_write_kbytes);
2320 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2321 inode_readahead_blks_store, s_inode_readahead_blks);
2322 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2323 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2324 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2325 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2326 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2327 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2328 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2329 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2331 static struct attribute *ext4_attrs[] = {
2332 ATTR_LIST(delayed_allocation_blocks),
2333 ATTR_LIST(session_write_kbytes),
2334 ATTR_LIST(lifetime_write_kbytes),
2335 ATTR_LIST(inode_readahead_blks),
2336 ATTR_LIST(inode_goal),
2337 ATTR_LIST(mb_stats),
2338 ATTR_LIST(mb_max_to_scan),
2339 ATTR_LIST(mb_min_to_scan),
2340 ATTR_LIST(mb_order2_req),
2341 ATTR_LIST(mb_stream_req),
2342 ATTR_LIST(mb_group_prealloc),
2343 ATTR_LIST(max_writeback_mb_bump),
2347 static ssize_t ext4_attr_show(struct kobject *kobj,
2348 struct attribute *attr, char *buf)
2350 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2352 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2354 return a->show ? a->show(a, sbi, buf) : 0;
2357 static ssize_t ext4_attr_store(struct kobject *kobj,
2358 struct attribute *attr,
2359 const char *buf, size_t len)
2361 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2363 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2365 return a->store ? a->store(a, sbi, buf, len) : 0;
2368 static void ext4_sb_release(struct kobject *kobj)
2370 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2372 complete(&sbi->s_kobj_unregister);
2376 static const struct sysfs_ops ext4_attr_ops = {
2377 .show = ext4_attr_show,
2378 .store = ext4_attr_store,
2381 static struct kobj_type ext4_ktype = {
2382 .default_attrs = ext4_attrs,
2383 .sysfs_ops = &ext4_attr_ops,
2384 .release = ext4_sb_release,
2388 * Check whether this filesystem can be mounted based on
2389 * the features present and the RDONLY/RDWR mount requested.
2390 * Returns 1 if this filesystem can be mounted as requested,
2391 * 0 if it cannot be.
2393 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2395 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2396 ext4_msg(sb, KERN_ERR,
2397 "Couldn't mount because of "
2398 "unsupported optional features (%x)",
2399 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2400 ~EXT4_FEATURE_INCOMPAT_SUPP));
2407 /* Check that feature set is OK for a read-write mount */
2408 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2409 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2410 "unsupported optional features (%x)",
2411 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2412 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2416 * Large file size enabled file system can only be mounted
2417 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2419 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2420 if (sizeof(blkcnt_t) < sizeof(u64)) {
2421 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2422 "cannot be mounted RDWR without "
2430 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2431 __releases(kernel_lock)
2432 __acquires(kernel_lock)
2434 struct buffer_head *bh;
2435 struct ext4_super_block *es = NULL;
2436 struct ext4_sb_info *sbi;
2438 ext4_fsblk_t sb_block = get_sb_block(&data);
2439 ext4_fsblk_t logical_sb_block;
2440 unsigned long offset = 0;
2441 unsigned long journal_devnum = 0;
2442 unsigned long def_mount_opts;
2448 unsigned int db_count;
2450 int needs_recovery, has_huge_files;
2453 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2455 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2459 sbi->s_blockgroup_lock =
2460 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2461 if (!sbi->s_blockgroup_lock) {
2465 sb->s_fs_info = sbi;
2466 sbi->s_mount_opt = 0;
2467 sbi->s_resuid = EXT4_DEF_RESUID;
2468 sbi->s_resgid = EXT4_DEF_RESGID;
2469 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2470 sbi->s_sb_block = sb_block;
2471 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2476 /* Cleanup superblock name */
2477 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2480 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2482 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2487 * The ext4 superblock will not be buffer aligned for other than 1kB
2488 * block sizes. We need to calculate the offset from buffer start.
2490 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2491 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2492 offset = do_div(logical_sb_block, blocksize);
2494 logical_sb_block = sb_block;
2497 if (!(bh = sb_bread(sb, logical_sb_block))) {
2498 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2502 * Note: s_es must be initialized as soon as possible because
2503 * some ext4 macro-instructions depend on its value
2505 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2507 sb->s_magic = le16_to_cpu(es->s_magic);
2508 if (sb->s_magic != EXT4_SUPER_MAGIC)
2510 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2512 /* Set defaults before we parse the mount options */
2513 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2514 if (def_mount_opts & EXT4_DEFM_DEBUG)
2515 set_opt(sbi->s_mount_opt, DEBUG);
2516 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2517 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2519 set_opt(sbi->s_mount_opt, GRPID);
2521 if (def_mount_opts & EXT4_DEFM_UID16)
2522 set_opt(sbi->s_mount_opt, NO_UID32);
2523 #ifdef CONFIG_EXT4_FS_XATTR
2524 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2525 set_opt(sbi->s_mount_opt, XATTR_USER);
2527 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2528 if (def_mount_opts & EXT4_DEFM_ACL)
2529 set_opt(sbi->s_mount_opt, POSIX_ACL);
2531 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2532 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2533 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2534 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2535 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2536 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2538 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2539 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2540 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2541 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2543 set_opt(sbi->s_mount_opt, ERRORS_RO);
2545 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2546 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2547 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2548 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2549 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2551 set_opt(sbi->s_mount_opt, BARRIER);
2554 * enable delayed allocation by default
2555 * Use -o nodelalloc to turn it off
2557 if (!IS_EXT3_SB(sb))
2558 set_opt(sbi->s_mount_opt, DELALLOC);
2560 if (!parse_options((char *) data, sb, &journal_devnum,
2561 &journal_ioprio, NULL, 0))
2564 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2565 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2567 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2568 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2569 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2570 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2571 ext4_msg(sb, KERN_WARNING,
2572 "feature flags set on rev 0 fs, "
2573 "running e2fsck is recommended");
2576 * Check feature flags regardless of the revision level, since we
2577 * previously didn't change the revision level when setting the flags,
2578 * so there is a chance incompat flags are set on a rev 0 filesystem.
2580 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2583 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2585 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2586 blocksize > EXT4_MAX_BLOCK_SIZE) {
2587 ext4_msg(sb, KERN_ERR,
2588 "Unsupported filesystem blocksize %d", blocksize);
2592 if (sb->s_blocksize != blocksize) {
2593 /* Validate the filesystem blocksize */
2594 if (!sb_set_blocksize(sb, blocksize)) {
2595 ext4_msg(sb, KERN_ERR, "bad block size %d",
2601 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2602 offset = do_div(logical_sb_block, blocksize);
2603 bh = sb_bread(sb, logical_sb_block);
2605 ext4_msg(sb, KERN_ERR,
2606 "Can't read superblock on 2nd try");
2609 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2611 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2612 ext4_msg(sb, KERN_ERR,
2613 "Magic mismatch, very weird!");
2618 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2619 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2620 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2622 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2624 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2625 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2626 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2628 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2629 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2630 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2631 (!is_power_of_2(sbi->s_inode_size)) ||
2632 (sbi->s_inode_size > blocksize)) {
2633 ext4_msg(sb, KERN_ERR,
2634 "unsupported inode size: %d",
2638 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2639 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2642 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2643 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2644 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2645 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2646 !is_power_of_2(sbi->s_desc_size)) {
2647 ext4_msg(sb, KERN_ERR,
2648 "unsupported descriptor size %lu",
2653 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2655 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2656 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2657 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2660 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2661 if (sbi->s_inodes_per_block == 0)
2663 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2664 sbi->s_inodes_per_block;
2665 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2667 sbi->s_mount_state = le16_to_cpu(es->s_state);
2668 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2669 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2671 for (i = 0; i < 4; i++)
2672 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2673 sbi->s_def_hash_version = es->s_def_hash_version;
2674 i = le32_to_cpu(es->s_flags);
2675 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2676 sbi->s_hash_unsigned = 3;
2677 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2678 #ifdef __CHAR_UNSIGNED__
2679 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2680 sbi->s_hash_unsigned = 3;
2682 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2687 if (sbi->s_blocks_per_group > blocksize * 8) {
2688 ext4_msg(sb, KERN_ERR,
2689 "#blocks per group too big: %lu",
2690 sbi->s_blocks_per_group);
2693 if (sbi->s_inodes_per_group > blocksize * 8) {
2694 ext4_msg(sb, KERN_ERR,
2695 "#inodes per group too big: %lu",
2696 sbi->s_inodes_per_group);
2701 * Test whether we have more sectors than will fit in sector_t,
2702 * and whether the max offset is addressable by the page cache.
2704 if ((ext4_blocks_count(es) >
2705 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2706 (ext4_blocks_count(es) >
2707 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2708 ext4_msg(sb, KERN_ERR, "filesystem"
2709 " too large to mount safely on this system");
2710 if (sizeof(sector_t) < 8)
2711 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2716 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2719 /* check blocks count against device size */
2720 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2721 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2722 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2723 "exceeds size of device (%llu blocks)",
2724 ext4_blocks_count(es), blocks_count);
2729 * It makes no sense for the first data block to be beyond the end
2730 * of the filesystem.
2732 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2733 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2734 "block %u is beyond end of filesystem (%llu)",
2735 le32_to_cpu(es->s_first_data_block),
2736 ext4_blocks_count(es));
2739 blocks_count = (ext4_blocks_count(es) -
2740 le32_to_cpu(es->s_first_data_block) +
2741 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2742 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2743 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2744 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2745 "(block count %llu, first data block %u, "
2746 "blocks per group %lu)", sbi->s_groups_count,
2747 ext4_blocks_count(es),
2748 le32_to_cpu(es->s_first_data_block),
2749 EXT4_BLOCKS_PER_GROUP(sb));
2752 sbi->s_groups_count = blocks_count;
2753 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2754 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2755 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2756 EXT4_DESC_PER_BLOCK(sb);
2757 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2759 if (sbi->s_group_desc == NULL) {
2760 ext4_msg(sb, KERN_ERR, "not enough memory");
2764 #ifdef CONFIG_PROC_FS
2766 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2769 bgl_lock_init(sbi->s_blockgroup_lock);
2771 for (i = 0; i < db_count; i++) {
2772 block = descriptor_loc(sb, logical_sb_block, i);
2773 sbi->s_group_desc[i] = sb_bread(sb, block);
2774 if (!sbi->s_group_desc[i]) {
2775 ext4_msg(sb, KERN_ERR,
2776 "can't read group descriptor %d", i);
2781 if (!ext4_check_descriptors(sb)) {
2782 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2785 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2786 if (!ext4_fill_flex_info(sb)) {
2787 ext4_msg(sb, KERN_ERR,
2788 "unable to initialize "
2789 "flex_bg meta info!");
2793 sbi->s_gdb_count = db_count;
2794 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2795 spin_lock_init(&sbi->s_next_gen_lock);
2797 sbi->s_stripe = ext4_get_stripe_size(sbi);
2798 sbi->s_max_writeback_mb_bump = 128;
2801 * set up enough so that it can read an inode
2803 if (!test_opt(sb, NOLOAD) &&
2804 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2805 sb->s_op = &ext4_sops;
2807 sb->s_op = &ext4_nojournal_sops;
2808 sb->s_export_op = &ext4_export_ops;
2809 sb->s_xattr = ext4_xattr_handlers;
2811 sb->s_qcop = &ext4_qctl_operations;
2812 sb->dq_op = &ext4_quota_operations;
2814 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2815 mutex_init(&sbi->s_orphan_lock);
2816 mutex_init(&sbi->s_resize_lock);
2820 needs_recovery = (es->s_last_orphan != 0 ||
2821 EXT4_HAS_INCOMPAT_FEATURE(sb,
2822 EXT4_FEATURE_INCOMPAT_RECOVER));
2825 * The first inode we look at is the journal inode. Don't try
2826 * root first: it may be modified in the journal!
2828 if (!test_opt(sb, NOLOAD) &&
2829 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2830 if (ext4_load_journal(sb, es, journal_devnum))
2832 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2833 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2834 ext4_msg(sb, KERN_ERR, "required journal recovery "
2835 "suppressed and not mounted read-only");
2836 goto failed_mount_wq;
2838 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2839 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2840 sbi->s_journal = NULL;
2845 if (ext4_blocks_count(es) > 0xffffffffULL &&
2846 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2847 JBD2_FEATURE_INCOMPAT_64BIT)) {
2848 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2849 goto failed_mount_wq;
2852 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2853 jbd2_journal_set_features(sbi->s_journal,
2854 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2855 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2856 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2857 jbd2_journal_set_features(sbi->s_journal,
2858 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2859 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2860 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2862 jbd2_journal_clear_features(sbi->s_journal,
2863 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2864 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2867 /* We have now updated the journal if required, so we can
2868 * validate the data journaling mode. */
2869 switch (test_opt(sb, DATA_FLAGS)) {
2871 /* No mode set, assume a default based on the journal
2872 * capabilities: ORDERED_DATA if the journal can
2873 * cope, else JOURNAL_DATA
2875 if (jbd2_journal_check_available_features
2876 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2877 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2879 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2882 case EXT4_MOUNT_ORDERED_DATA:
2883 case EXT4_MOUNT_WRITEBACK_DATA:
2884 if (!jbd2_journal_check_available_features
2885 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2886 ext4_msg(sb, KERN_ERR, "Journal does not support "
2887 "requested data journaling mode");
2888 goto failed_mount_wq;
2893 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2896 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2897 ext4_count_free_blocks(sb));
2899 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2900 ext4_count_free_inodes(sb));
2902 err = percpu_counter_init(&sbi->s_dirs_counter,
2903 ext4_count_dirs(sb));
2905 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2907 ext4_msg(sb, KERN_ERR, "insufficient memory");
2908 goto failed_mount_wq;
2910 if (test_opt(sb, NOBH)) {
2911 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2912 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2913 "its supported only with writeback mode");
2914 clear_opt(sbi->s_mount_opt, NOBH);
2916 if (test_opt(sb, DIOREAD_NOLOCK)) {
2917 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2918 "not supported with nobh mode");
2919 goto failed_mount_wq;
2922 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2923 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2924 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2925 goto failed_mount_wq;
2929 * The jbd2_journal_load will have done any necessary log recovery,
2930 * so we can safely mount the rest of the filesystem now.
2933 root = ext4_iget(sb, EXT4_ROOT_INO);
2935 ext4_msg(sb, KERN_ERR, "get root inode failed");
2936 ret = PTR_ERR(root);
2939 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2941 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2944 sb->s_root = d_alloc_root(root);
2946 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2952 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2954 /* determine the minimum size of new large inodes, if present */
2955 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2956 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2957 EXT4_GOOD_OLD_INODE_SIZE;
2958 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2959 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2960 if (sbi->s_want_extra_isize <
2961 le16_to_cpu(es->s_want_extra_isize))
2962 sbi->s_want_extra_isize =
2963 le16_to_cpu(es->s_want_extra_isize);
2964 if (sbi->s_want_extra_isize <
2965 le16_to_cpu(es->s_min_extra_isize))
2966 sbi->s_want_extra_isize =
2967 le16_to_cpu(es->s_min_extra_isize);
2970 /* Check if enough inode space is available */
2971 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2972 sbi->s_inode_size) {
2973 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2974 EXT4_GOOD_OLD_INODE_SIZE;
2975 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2979 if (test_opt(sb, DELALLOC) &&
2980 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2981 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2982 "requested data journaling mode");
2983 clear_opt(sbi->s_mount_opt, DELALLOC);
2985 if (test_opt(sb, DIOREAD_NOLOCK)) {
2986 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2987 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2988 "option - requested data journaling mode");
2989 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2991 if (sb->s_blocksize < PAGE_SIZE) {
2992 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2993 "option - block size is too small");
2994 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2998 err = ext4_setup_system_zone(sb);
3000 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3001 "zone (%d)\n", err);
3006 err = ext4_mb_init(sb, needs_recovery);
3008 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3013 sbi->s_kobj.kset = ext4_kset;
3014 init_completion(&sbi->s_kobj_unregister);
3015 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3018 ext4_mb_release(sb);
3019 ext4_ext_release(sb);
3023 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3024 ext4_orphan_cleanup(sb, es);
3025 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3026 if (needs_recovery) {
3027 ext4_msg(sb, KERN_INFO, "recovery complete");
3028 ext4_mark_recovery_complete(sb, es);
3030 if (EXT4_SB(sb)->s_journal) {
3031 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3032 descr = " journalled data mode";
3033 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3034 descr = " ordered data mode";
3036 descr = " writeback data mode";
3038 descr = "out journal";
3040 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
3047 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3051 ext4_msg(sb, KERN_ERR, "mount failed");
3052 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3054 ext4_release_system_zone(sb);
3055 if (sbi->s_journal) {
3056 jbd2_journal_destroy(sbi->s_journal);
3057 sbi->s_journal = NULL;
3059 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3060 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3061 percpu_counter_destroy(&sbi->s_dirs_counter);
3062 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3064 if (sbi->s_flex_groups) {
3065 if (is_vmalloc_addr(sbi->s_flex_groups))
3066 vfree(sbi->s_flex_groups);
3068 kfree(sbi->s_flex_groups);
3071 for (i = 0; i < db_count; i++)
3072 brelse(sbi->s_group_desc[i]);
3073 kfree(sbi->s_group_desc);
3076 remove_proc_entry(sb->s_id, ext4_proc_root);
3079 for (i = 0; i < MAXQUOTAS; i++)
3080 kfree(sbi->s_qf_names[i]);
3082 ext4_blkdev_remove(sbi);
3085 sb->s_fs_info = NULL;
3086 kfree(sbi->s_blockgroup_lock);
3093 * Setup any per-fs journal parameters now. We'll do this both on
3094 * initial mount, once the journal has been initialised but before we've
3095 * done any recovery; and again on any subsequent remount.
3097 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3099 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101 journal->j_commit_interval = sbi->s_commit_interval;
3102 journal->j_min_batch_time = sbi->s_min_batch_time;
3103 journal->j_max_batch_time = sbi->s_max_batch_time;
3105 spin_lock(&journal->j_state_lock);
3106 if (test_opt(sb, BARRIER))
3107 journal->j_flags |= JBD2_BARRIER;
3109 journal->j_flags &= ~JBD2_BARRIER;
3110 if (test_opt(sb, DATA_ERR_ABORT))
3111 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3113 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3114 spin_unlock(&journal->j_state_lock);
3117 static journal_t *ext4_get_journal(struct super_block *sb,
3118 unsigned int journal_inum)
3120 struct inode *journal_inode;
3123 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3125 /* First, test for the existence of a valid inode on disk. Bad
3126 * things happen if we iget() an unused inode, as the subsequent
3127 * iput() will try to delete it. */
3129 journal_inode = ext4_iget(sb, journal_inum);
3130 if (IS_ERR(journal_inode)) {
3131 ext4_msg(sb, KERN_ERR, "no journal found");
3134 if (!journal_inode->i_nlink) {
3135 make_bad_inode(journal_inode);
3136 iput(journal_inode);
3137 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3141 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3142 journal_inode, journal_inode->i_size);
3143 if (!S_ISREG(journal_inode->i_mode)) {
3144 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3145 iput(journal_inode);
3149 journal = jbd2_journal_init_inode(journal_inode);
3151 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3152 iput(journal_inode);
3155 journal->j_private = sb;
3156 ext4_init_journal_params(sb, journal);
3160 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3163 struct buffer_head *bh;
3167 int hblock, blocksize;
3168 ext4_fsblk_t sb_block;
3169 unsigned long offset;
3170 struct ext4_super_block *es;
3171 struct block_device *bdev;
3173 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3175 bdev = ext4_blkdev_get(j_dev, sb);
3179 if (bd_claim(bdev, sb)) {
3180 ext4_msg(sb, KERN_ERR,
3181 "failed to claim external journal device");
3182 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3186 blocksize = sb->s_blocksize;
3187 hblock = bdev_logical_block_size(bdev);
3188 if (blocksize < hblock) {
3189 ext4_msg(sb, KERN_ERR,
3190 "blocksize too small for journal device");
3194 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3195 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3196 set_blocksize(bdev, blocksize);
3197 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3198 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3199 "external journal");
3203 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3204 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3205 !(le32_to_cpu(es->s_feature_incompat) &
3206 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3207 ext4_msg(sb, KERN_ERR, "external journal has "
3213 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3214 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3219 len = ext4_blocks_count(es);
3220 start = sb_block + 1;
3221 brelse(bh); /* we're done with the superblock */
3223 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3224 start, len, blocksize);
3226 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3229 journal->j_private = sb;
3230 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3231 wait_on_buffer(journal->j_sb_buffer);
3232 if (!buffer_uptodate(journal->j_sb_buffer)) {
3233 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3236 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3237 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3238 "user (unsupported) - %d",
3239 be32_to_cpu(journal->j_superblock->s_nr_users));
3242 EXT4_SB(sb)->journal_bdev = bdev;
3243 ext4_init_journal_params(sb, journal);
3247 jbd2_journal_destroy(journal);
3249 ext4_blkdev_put(bdev);
3253 static int ext4_load_journal(struct super_block *sb,
3254 struct ext4_super_block *es,
3255 unsigned long journal_devnum)
3258 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3261 int really_read_only;
3263 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3265 if (journal_devnum &&
3266 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3267 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3268 "numbers have changed");
3269 journal_dev = new_decode_dev(journal_devnum);
3271 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3273 really_read_only = bdev_read_only(sb->s_bdev);
3276 * Are we loading a blank journal or performing recovery after a
3277 * crash? For recovery, we need to check in advance whether we
3278 * can get read-write access to the device.
3280 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3281 if (sb->s_flags & MS_RDONLY) {
3282 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3283 "required on readonly filesystem");
3284 if (really_read_only) {
3285 ext4_msg(sb, KERN_ERR, "write access "
3286 "unavailable, cannot proceed");
3289 ext4_msg(sb, KERN_INFO, "write access will "
3290 "be enabled during recovery");
3294 if (journal_inum && journal_dev) {
3295 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3296 "and inode journals!");
3301 if (!(journal = ext4_get_journal(sb, journal_inum)))
3304 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3308 if (!(journal->j_flags & JBD2_BARRIER))
3309 ext4_msg(sb, KERN_INFO, "barriers disabled");
3311 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3312 err = jbd2_journal_update_format(journal);
3314 ext4_msg(sb, KERN_ERR, "error updating journal");
3315 jbd2_journal_destroy(journal);
3320 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3321 err = jbd2_journal_wipe(journal, !really_read_only);
3323 err = jbd2_journal_load(journal);
3326 ext4_msg(sb, KERN_ERR, "error loading journal");
3327 jbd2_journal_destroy(journal);
3331 EXT4_SB(sb)->s_journal = journal;
3332 ext4_clear_journal_err(sb, es);
3334 if (journal_devnum &&
3335 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3336 es->s_journal_dev = cpu_to_le32(journal_devnum);
3338 /* Make sure we flush the recovery flag to disk. */
3339 ext4_commit_super(sb, 1);
3345 static int ext4_commit_super(struct super_block *sb, int sync)
3347 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3348 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3353 if (buffer_write_io_error(sbh)) {
3355 * Oh, dear. A previous attempt to write the
3356 * superblock failed. This could happen because the
3357 * USB device was yanked out. Or it could happen to
3358 * be a transient write error and maybe the block will
3359 * be remapped. Nothing we can do but to retry the
3360 * write and hope for the best.
3362 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3363 "superblock detected");
3364 clear_buffer_write_io_error(sbh);
3365 set_buffer_uptodate(sbh);
3368 * If the file system is mounted read-only, don't update the
3369 * superblock write time. This avoids updating the superblock
3370 * write time when we are mounting the root file system
3371 * read/only but we need to replay the journal; at that point,
3372 * for people who are east of GMT and who make their clock
3373 * tick in localtime for Windows bug-for-bug compatibility,
3374 * the clock is set in the future, and this will cause e2fsck
3375 * to complain and force a full file system check.
3377 if (!(sb->s_flags & MS_RDONLY))
3378 es->s_wtime = cpu_to_le32(get_seconds());
3379 es->s_kbytes_written =
3380 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3381 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3382 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3383 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3384 &EXT4_SB(sb)->s_freeblocks_counter));
3385 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3386 &EXT4_SB(sb)->s_freeinodes_counter));
3388 BUFFER_TRACE(sbh, "marking dirty");
3389 mark_buffer_dirty(sbh);
3391 error = sync_dirty_buffer(sbh);
3395 error = buffer_write_io_error(sbh);
3397 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3399 clear_buffer_write_io_error(sbh);
3400 set_buffer_uptodate(sbh);
3407 * Have we just finished recovery? If so, and if we are mounting (or
3408 * remounting) the filesystem readonly, then we will end up with a
3409 * consistent fs on disk. Record that fact.
3411 static void ext4_mark_recovery_complete(struct super_block *sb,
3412 struct ext4_super_block *es)
3414 journal_t *journal = EXT4_SB(sb)->s_journal;
3416 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3417 BUG_ON(journal != NULL);
3420 jbd2_journal_lock_updates(journal);
3421 if (jbd2_journal_flush(journal) < 0)
3424 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3425 sb->s_flags & MS_RDONLY) {
3426 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3427 ext4_commit_super(sb, 1);
3431 jbd2_journal_unlock_updates(journal);
3435 * If we are mounting (or read-write remounting) a filesystem whose journal
3436 * has recorded an error from a previous lifetime, move that error to the
3437 * main filesystem now.
3439 static void ext4_clear_journal_err(struct super_block *sb,
3440 struct ext4_super_block *es)
3446 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3448 journal = EXT4_SB(sb)->s_journal;
3451 * Now check for any error status which may have been recorded in the
3452 * journal by a prior ext4_error() or ext4_abort()
3455 j_errno = jbd2_journal_errno(journal);
3459 errstr = ext4_decode_error(sb, j_errno, nbuf);
3460 ext4_warning(sb, "Filesystem error recorded "
3461 "from previous mount: %s", errstr);
3462 ext4_warning(sb, "Marking fs in need of filesystem check.");
3464 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3465 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3466 ext4_commit_super(sb, 1);
3468 jbd2_journal_clear_err(journal);
3473 * Force the running and committing transactions to commit,
3474 * and wait on the commit.
3476 int ext4_force_commit(struct super_block *sb)
3481 if (sb->s_flags & MS_RDONLY)
3484 journal = EXT4_SB(sb)->s_journal;
3486 vfs_check_frozen(sb, SB_FREEZE_WRITE);
3487 ret = ext4_journal_force_commit(journal);
3493 static void ext4_write_super(struct super_block *sb)
3496 ext4_commit_super(sb, 1);
3500 static int ext4_sync_fs(struct super_block *sb, int wait)
3504 struct ext4_sb_info *sbi = EXT4_SB(sb);
3506 trace_ext4_sync_fs(sb, wait);
3507 flush_workqueue(sbi->dio_unwritten_wq);
3508 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3510 jbd2_log_wait_commit(sbi->s_journal, target);
3516 * LVM calls this function before a (read-only) snapshot is created. This
3517 * gives us a chance to flush the journal completely and mark the fs clean.
3519 static int ext4_freeze(struct super_block *sb)
3524 if (sb->s_flags & MS_RDONLY)
3527 journal = EXT4_SB(sb)->s_journal;
3529 /* Now we set up the journal barrier. */
3530 jbd2_journal_lock_updates(journal);
3533 * Don't clear the needs_recovery flag if we failed to flush
3536 error = jbd2_journal_flush(journal);
3540 /* Journal blocked and flushed, clear needs_recovery flag. */
3541 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3542 error = ext4_commit_super(sb, 1);
3544 /* we rely on s_frozen to stop further updates */
3545 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3550 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3551 * flag here, even though the filesystem is not technically dirty yet.
3553 static int ext4_unfreeze(struct super_block *sb)
3555 if (sb->s_flags & MS_RDONLY)
3559 /* Reset the needs_recovery flag before the fs is unlocked. */
3560 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3561 ext4_commit_super(sb, 1);
3566 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3568 struct ext4_super_block *es;
3569 struct ext4_sb_info *sbi = EXT4_SB(sb);
3570 ext4_fsblk_t n_blocks_count = 0;
3571 unsigned long old_sb_flags;
3572 struct ext4_mount_options old_opts;
3574 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3582 /* Store the original options */
3584 old_sb_flags = sb->s_flags;
3585 old_opts.s_mount_opt = sbi->s_mount_opt;
3586 old_opts.s_resuid = sbi->s_resuid;
3587 old_opts.s_resgid = sbi->s_resgid;
3588 old_opts.s_commit_interval = sbi->s_commit_interval;
3589 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3590 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3592 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3593 for (i = 0; i < MAXQUOTAS; i++)
3594 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3596 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3597 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3600 * Allow the "check" option to be passed as a remount option.
3602 if (!parse_options(data, sb, NULL, &journal_ioprio,
3603 &n_blocks_count, 1)) {
3608 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3609 ext4_abort(sb, __func__, "Abort forced by user");
3611 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3612 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3616 if (sbi->s_journal) {
3617 ext4_init_journal_params(sb, sbi->s_journal);
3618 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3621 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3622 n_blocks_count > ext4_blocks_count(es)) {
3623 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3628 if (*flags & MS_RDONLY) {
3630 * First of all, the unconditional stuff we have to do
3631 * to disable replay of the journal when we next remount
3633 sb->s_flags |= MS_RDONLY;
3636 * OK, test if we are remounting a valid rw partition
3637 * readonly, and if so set the rdonly flag and then
3638 * mark the partition as valid again.
3640 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3641 (sbi->s_mount_state & EXT4_VALID_FS))
3642 es->s_state = cpu_to_le16(sbi->s_mount_state);
3645 ext4_mark_recovery_complete(sb, es);
3647 /* Make sure we can mount this feature set readwrite */
3648 if (!ext4_feature_set_ok(sb, 0)) {
3653 * Make sure the group descriptor checksums
3654 * are sane. If they aren't, refuse to remount r/w.
3656 for (g = 0; g < sbi->s_groups_count; g++) {
3657 struct ext4_group_desc *gdp =
3658 ext4_get_group_desc(sb, g, NULL);
3660 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3661 ext4_msg(sb, KERN_ERR,
3662 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3663 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3664 le16_to_cpu(gdp->bg_checksum));
3671 * If we have an unprocessed orphan list hanging
3672 * around from a previously readonly bdev mount,
3673 * require a full umount/remount for now.
3675 if (es->s_last_orphan) {
3676 ext4_msg(sb, KERN_WARNING, "Couldn't "
3677 "remount RDWR because of unprocessed "
3678 "orphan inode list. Please "
3679 "umount/remount instead");
3685 * Mounting a RDONLY partition read-write, so reread
3686 * and store the current valid flag. (It may have
3687 * been changed by e2fsck since we originally mounted
3691 ext4_clear_journal_err(sb, es);
3692 sbi->s_mount_state = le16_to_cpu(es->s_state);
3693 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3695 if (!ext4_setup_super(sb, es, 0))
3696 sb->s_flags &= ~MS_RDONLY;
3699 ext4_setup_system_zone(sb);
3700 if (sbi->s_journal == NULL)
3701 ext4_commit_super(sb, 1);
3704 /* Release old quota file names */
3705 for (i = 0; i < MAXQUOTAS; i++)
3706 if (old_opts.s_qf_names[i] &&
3707 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3708 kfree(old_opts.s_qf_names[i]);
3715 sb->s_flags = old_sb_flags;
3716 sbi->s_mount_opt = old_opts.s_mount_opt;
3717 sbi->s_resuid = old_opts.s_resuid;
3718 sbi->s_resgid = old_opts.s_resgid;
3719 sbi->s_commit_interval = old_opts.s_commit_interval;
3720 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3721 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3723 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3724 for (i = 0; i < MAXQUOTAS; i++) {
3725 if (sbi->s_qf_names[i] &&
3726 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3727 kfree(sbi->s_qf_names[i]);
3728 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3736 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3738 struct super_block *sb = dentry->d_sb;
3739 struct ext4_sb_info *sbi = EXT4_SB(sb);
3740 struct ext4_super_block *es = sbi->s_es;
3743 if (test_opt(sb, MINIX_DF)) {
3744 sbi->s_overhead_last = 0;
3745 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3746 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3747 ext4_fsblk_t overhead = 0;
3750 * Compute the overhead (FS structures). This is constant
3751 * for a given filesystem unless the number of block groups
3752 * changes so we cache the previous value until it does.
3756 * All of the blocks before first_data_block are
3759 overhead = le32_to_cpu(es->s_first_data_block);
3762 * Add the overhead attributed to the superblock and
3763 * block group descriptors. If the sparse superblocks
3764 * feature is turned on, then not all groups have this.
3766 for (i = 0; i < ngroups; i++) {
3767 overhead += ext4_bg_has_super(sb, i) +
3768 ext4_bg_num_gdb(sb, i);
3773 * Every block group has an inode bitmap, a block
3774 * bitmap, and an inode table.
3776 overhead += ngroups * (2 + sbi->s_itb_per_group);
3777 sbi->s_overhead_last = overhead;
3779 sbi->s_blocks_last = ext4_blocks_count(es);
3782 buf->f_type = EXT4_SUPER_MAGIC;
3783 buf->f_bsize = sb->s_blocksize;
3784 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3785 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3786 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3787 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3788 if (buf->f_bfree < ext4_r_blocks_count(es))
3790 buf->f_files = le32_to_cpu(es->s_inodes_count);
3791 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3792 buf->f_namelen = EXT4_NAME_LEN;
3793 fsid = le64_to_cpup((void *)es->s_uuid) ^
3794 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3795 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3796 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3801 /* Helper function for writing quotas on sync - we need to start transaction
3802 * before quota file is locked for write. Otherwise the are possible deadlocks:
3803 * Process 1 Process 2
3804 * ext4_create() quota_sync()
3805 * jbd2_journal_start() write_dquot()
3806 * dquot_initialize() down(dqio_mutex)
3807 * down(dqio_mutex) jbd2_journal_start()
3813 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3815 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3818 static int ext4_write_dquot(struct dquot *dquot)
3822 struct inode *inode;
3824 inode = dquot_to_inode(dquot);
3825 handle = ext4_journal_start(inode,
3826 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3828 return PTR_ERR(handle);
3829 ret = dquot_commit(dquot);
3830 err = ext4_journal_stop(handle);
3836 static int ext4_acquire_dquot(struct dquot *dquot)
3841 handle = ext4_journal_start(dquot_to_inode(dquot),
3842 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3844 return PTR_ERR(handle);
3845 ret = dquot_acquire(dquot);
3846 err = ext4_journal_stop(handle);
3852 static int ext4_release_dquot(struct dquot *dquot)
3857 handle = ext4_journal_start(dquot_to_inode(dquot),
3858 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3859 if (IS_ERR(handle)) {
3860 /* Release dquot anyway to avoid endless cycle in dqput() */
3861 dquot_release(dquot);
3862 return PTR_ERR(handle);
3864 ret = dquot_release(dquot);
3865 err = ext4_journal_stop(handle);
3871 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3873 /* Are we journaling quotas? */
3874 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3875 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3876 dquot_mark_dquot_dirty(dquot);
3877 return ext4_write_dquot(dquot);
3879 return dquot_mark_dquot_dirty(dquot);
3883 static int ext4_write_info(struct super_block *sb, int type)
3888 /* Data block + inode block */
3889 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3891 return PTR_ERR(handle);
3892 ret = dquot_commit_info(sb, type);
3893 err = ext4_journal_stop(handle);
3900 * Turn on quotas during mount time - we need to find
3901 * the quota file and such...
3903 static int ext4_quota_on_mount(struct super_block *sb, int type)
3905 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3906 EXT4_SB(sb)->s_jquota_fmt, type);
3910 * Standard function to be called on quota_on
3912 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3913 char *name, int remount)
3918 if (!test_opt(sb, QUOTA))
3920 /* When remounting, no checks are needed and in fact, name is NULL */
3922 return vfs_quota_on(sb, type, format_id, name, remount);
3924 err = kern_path(name, LOOKUP_FOLLOW, &path);
3928 /* Quotafile not on the same filesystem? */
3929 if (path.mnt->mnt_sb != sb) {
3933 /* Journaling quota? */
3934 if (EXT4_SB(sb)->s_qf_names[type]) {
3935 /* Quotafile not in fs root? */
3936 if (path.dentry->d_parent != sb->s_root)
3937 ext4_msg(sb, KERN_WARNING,
3938 "Quota file not on filesystem root. "
3939 "Journaled quota will not work");
3943 * When we journal data on quota file, we have to flush journal to see
3944 * all updates to the file when we bypass pagecache...
3946 if (EXT4_SB(sb)->s_journal &&
3947 ext4_should_journal_data(path.dentry->d_inode)) {
3949 * We don't need to lock updates but journal_flush() could
3950 * otherwise be livelocked...
3952 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3953 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3954 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3961 err = vfs_quota_on_path(sb, type, format_id, &path);
3966 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3967 * acquiring the locks... As quota files are never truncated and quota code
3968 * itself serializes the operations (and noone else should touch the files)
3969 * we don't have to be afraid of races */
3970 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3971 size_t len, loff_t off)
3973 struct inode *inode = sb_dqopt(sb)->files[type];
3974 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3976 int offset = off & (sb->s_blocksize - 1);
3979 struct buffer_head *bh;
3980 loff_t i_size = i_size_read(inode);
3984 if (off+len > i_size)
3987 while (toread > 0) {
3988 tocopy = sb->s_blocksize - offset < toread ?
3989 sb->s_blocksize - offset : toread;
3990 bh = ext4_bread(NULL, inode, blk, 0, &err);
3993 if (!bh) /* A hole? */
3994 memset(data, 0, tocopy);
3996 memcpy(data, bh->b_data+offset, tocopy);
4006 /* Write to quotafile (we know the transaction is already started and has
4007 * enough credits) */
4008 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4009 const char *data, size_t len, loff_t off)
4011 struct inode *inode = sb_dqopt(sb)->files[type];
4012 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4014 int offset = off & (sb->s_blocksize - 1);
4015 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4016 struct buffer_head *bh;
4017 handle_t *handle = journal_current_handle();
4019 if (EXT4_SB(sb)->s_journal && !handle) {
4020 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4021 " cancelled because transaction is not started",
4022 (unsigned long long)off, (unsigned long long)len);
4026 * Since we account only one data block in transaction credits,
4027 * then it is impossible to cross a block boundary.
4029 if (sb->s_blocksize - offset < len) {
4030 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4031 " cancelled because not block aligned",
4032 (unsigned long long)off, (unsigned long long)len);
4036 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4037 bh = ext4_bread(handle, inode, blk, 1, &err);
4040 if (journal_quota) {
4041 err = ext4_journal_get_write_access(handle, bh);
4048 memcpy(bh->b_data+offset, data, len);
4049 flush_dcache_page(bh->b_page);
4052 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4054 /* Always do at least ordered writes for quotas */
4055 err = ext4_jbd2_file_inode(handle, inode);
4056 mark_buffer_dirty(bh);
4061 mutex_unlock(&inode->i_mutex);
4064 if (inode->i_size < off + len) {
4065 i_size_write(inode, off + len);
4066 EXT4_I(inode)->i_disksize = inode->i_size;
4068 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4069 ext4_mark_inode_dirty(handle, inode);
4070 mutex_unlock(&inode->i_mutex);
4076 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4077 const char *dev_name, void *data, struct vfsmount *mnt)
4079 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4082 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4083 static struct file_system_type ext2_fs_type = {
4084 .owner = THIS_MODULE,
4086 .get_sb = ext4_get_sb,
4087 .kill_sb = kill_block_super,
4088 .fs_flags = FS_REQUIRES_DEV,
4091 static inline void register_as_ext2(void)
4093 int err = register_filesystem(&ext2_fs_type);
4096 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4099 static inline void unregister_as_ext2(void)
4101 unregister_filesystem(&ext2_fs_type);
4103 MODULE_ALIAS("ext2");
4105 static inline void register_as_ext2(void) { }
4106 static inline void unregister_as_ext2(void) { }
4109 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4110 static inline void register_as_ext3(void)
4112 int err = register_filesystem(&ext3_fs_type);
4115 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4118 static inline void unregister_as_ext3(void)
4120 unregister_filesystem(&ext3_fs_type);
4122 MODULE_ALIAS("ext3");
4124 static inline void register_as_ext3(void) { }
4125 static inline void unregister_as_ext3(void) { }
4128 static struct file_system_type ext4_fs_type = {
4129 .owner = THIS_MODULE,
4131 .get_sb = ext4_get_sb,
4132 .kill_sb = kill_block_super,
4133 .fs_flags = FS_REQUIRES_DEV,
4136 static int __init init_ext4_fs(void)
4140 err = init_ext4_system_zone();
4143 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4146 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4147 err = init_ext4_mballoc();
4151 err = init_ext4_xattr();
4154 err = init_inodecache();
4159 err = register_filesystem(&ext4_fs_type);
4164 unregister_as_ext2();
4165 unregister_as_ext3();
4166 destroy_inodecache();
4170 exit_ext4_mballoc();
4172 remove_proc_entry("fs/ext4", NULL);
4173 kset_unregister(ext4_kset);
4175 exit_ext4_system_zone();
4179 static void __exit exit_ext4_fs(void)
4181 unregister_as_ext2();
4182 unregister_as_ext3();
4183 unregister_filesystem(&ext4_fs_type);
4184 destroy_inodecache();
4186 exit_ext4_mballoc();
4187 remove_proc_entry("fs/ext4", NULL);
4188 kset_unregister(ext4_kset);
4189 exit_ext4_system_zone();
4192 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4193 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4194 MODULE_LICENSE("GPL");
4195 module_init(init_ext4_fs)
4196 module_exit(exit_ext4_fs)