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 static int default_mb_history_length = 1000;
55 module_param_named(default_mb_history_length, default_mb_history_length,
57 MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
60 struct proc_dir_entry *ext4_proc_root;
61 static struct kset *ext4_kset;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
80 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
88 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
96 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
99 return le32_to_cpu(bg->bg_inode_table_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
104 __u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
112 __u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
120 __u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
128 __u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
136 void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
144 void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
152 void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
160 void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
168 void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
176 void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
184 void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
193 /* Just increment the non-pointer handle value */
194 static handle_t *ext4_get_nojournal(void)
196 handle_t *handle = current->journal_info;
197 unsigned long ref_cnt = (unsigned long)handle;
199 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
202 handle = (handle_t *)ref_cnt;
204 current->journal_info = handle;
209 /* Decrement the non-pointer handle value */
210 static void ext4_put_nojournal(handle_t *handle)
212 unsigned long ref_cnt = (unsigned long)handle;
214 BUG_ON(ref_cnt == 0);
217 handle = (handle_t *)ref_cnt;
219 current->journal_info = handle;
223 * Wrappers for jbd2_journal_start/end.
225 * The only special thing we need to do here is to make sure that all
226 * journal_end calls result in the superblock being marked dirty, so
227 * that sync() will call the filesystem's write_super callback if
230 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
234 if (sb->s_flags & MS_RDONLY)
235 return ERR_PTR(-EROFS);
237 /* Special case here: if the journal has aborted behind our
238 * backs (eg. EIO in the commit thread), then we still need to
239 * take the FS itself readonly cleanly. */
240 journal = EXT4_SB(sb)->s_journal;
242 if (is_journal_aborted(journal)) {
243 ext4_abort(sb, __func__, "Detected aborted journal");
244 return ERR_PTR(-EROFS);
246 return jbd2_journal_start(journal, nblocks);
248 return ext4_get_nojournal();
252 * The only special thing we need to do here is to make sure that all
253 * jbd2_journal_stop calls result in the superblock being marked dirty, so
254 * that sync() will call the filesystem's write_super callback if
257 int __ext4_journal_stop(const char *where, handle_t *handle)
259 struct super_block *sb;
263 if (!ext4_handle_valid(handle)) {
264 ext4_put_nojournal(handle);
267 sb = handle->h_transaction->t_journal->j_private;
269 rc = jbd2_journal_stop(handle);
274 __ext4_std_error(sb, where, err);
278 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
279 struct buffer_head *bh, handle_t *handle, int err)
282 const char *errstr = ext4_decode_error(NULL, err, nbuf);
284 BUG_ON(!ext4_handle_valid(handle));
287 BUFFER_TRACE(bh, "abort");
292 if (is_handle_aborted(handle))
295 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
296 caller, errstr, err_fn);
298 jbd2_journal_abort_handle(handle);
301 /* Deal with the reporting of failure conditions on a filesystem such as
302 * inconsistencies detected or read IO failures.
304 * On ext2, we can store the error state of the filesystem in the
305 * superblock. That is not possible on ext4, because we may have other
306 * write ordering constraints on the superblock which prevent us from
307 * writing it out straight away; and given that the journal is about to
308 * be aborted, we can't rely on the current, or future, transactions to
309 * write out the superblock safely.
311 * We'll just use the jbd2_journal_abort() error code to record an error in
312 * the journal instead. On recovery, the journal will compain about
313 * that error until we've noted it down and cleared it.
316 static void ext4_handle_error(struct super_block *sb)
318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 if (sb->s_flags & MS_RDONLY)
326 if (!test_opt(sb, ERRORS_CONT)) {
327 journal_t *journal = EXT4_SB(sb)->s_journal;
329 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
331 jbd2_journal_abort(journal, -EIO);
333 if (test_opt(sb, ERRORS_RO)) {
334 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
335 sb->s_flags |= MS_RDONLY;
337 ext4_commit_super(sb, 1);
338 if (test_opt(sb, ERRORS_PANIC))
339 panic("EXT4-fs (device %s): panic forced after error\n",
343 void ext4_error(struct super_block *sb, const char *function,
344 const char *fmt, ...)
349 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
354 ext4_handle_error(sb);
357 static const char *ext4_decode_error(struct super_block *sb, int errno,
364 errstr = "IO failure";
367 errstr = "Out of memory";
370 if (!sb || (EXT4_SB(sb)->s_journal &&
371 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
372 errstr = "Journal has aborted";
374 errstr = "Readonly filesystem";
377 /* If the caller passed in an extra buffer for unknown
378 * errors, textualise them now. Else we just return
381 /* Check for truncated error codes... */
382 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
391 /* __ext4_std_error decodes expected errors from journaling functions
392 * automatically and invokes the appropriate error response. */
394 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
399 /* Special case: if the error is EROFS, and we're not already
400 * inside a transaction, then there's really no point in logging
402 if (errno == -EROFS && journal_current_handle() == NULL &&
403 (sb->s_flags & MS_RDONLY))
406 errstr = ext4_decode_error(sb, errno, nbuf);
407 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
408 sb->s_id, function, errstr);
410 ext4_handle_error(sb);
414 * ext4_abort is a much stronger failure handler than ext4_error. The
415 * abort function may be used to deal with unrecoverable failures such
416 * as journal IO errors or ENOMEM at a critical moment in log management.
418 * We unconditionally force the filesystem into an ABORT|READONLY state,
419 * unless the error response on the fs has been set to panic in which
420 * case we take the easy way out and panic immediately.
423 void ext4_abort(struct super_block *sb, const char *function,
424 const char *fmt, ...)
429 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
434 if (test_opt(sb, ERRORS_PANIC))
435 panic("EXT4-fs panic from previous error\n");
437 if (sb->s_flags & MS_RDONLY)
440 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
441 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
442 sb->s_flags |= MS_RDONLY;
443 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
444 if (EXT4_SB(sb)->s_journal)
445 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
448 void ext4_msg (struct super_block * sb, const char *prefix,
449 const char *fmt, ...)
454 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
460 void ext4_warning(struct super_block *sb, const char *function,
461 const char *fmt, ...)
466 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
473 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
474 const char *function, const char *fmt, ...)
479 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
482 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
487 if (test_opt(sb, ERRORS_CONT)) {
488 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
489 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
490 ext4_commit_super(sb, 0);
493 ext4_unlock_group(sb, grp);
494 ext4_handle_error(sb);
496 * We only get here in the ERRORS_RO case; relocking the group
497 * may be dangerous, but nothing bad will happen since the
498 * filesystem will have already been marked read/only and the
499 * journal has been aborted. We return 1 as a hint to callers
500 * who might what to use the return value from
501 * ext4_grp_locked_error() to distinguish beween the
502 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
503 * aggressively from the ext4 function in question, with a
504 * more appropriate error code.
506 ext4_lock_group(sb, grp);
510 void ext4_update_dynamic_rev(struct super_block *sb)
512 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
514 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
517 ext4_warning(sb, __func__,
518 "updating to rev %d because of new feature flag, "
519 "running e2fsck is recommended",
522 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
523 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
524 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
525 /* leave es->s_feature_*compat flags alone */
526 /* es->s_uuid will be set by e2fsck if empty */
529 * The rest of the superblock fields should be zero, and if not it
530 * means they are likely already in use, so leave them alone. We
531 * can leave it up to e2fsck to clean up any inconsistencies there.
536 * Open the external journal device
538 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
540 struct block_device *bdev;
541 char b[BDEVNAME_SIZE];
543 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
549 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
550 __bdevname(dev, b), PTR_ERR(bdev));
555 * Release the journal device
557 static int ext4_blkdev_put(struct block_device *bdev)
560 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
563 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
565 struct block_device *bdev;
568 bdev = sbi->journal_bdev;
570 ret = ext4_blkdev_put(bdev);
571 sbi->journal_bdev = NULL;
576 static inline struct inode *orphan_list_entry(struct list_head *l)
578 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
581 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
585 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
586 le32_to_cpu(sbi->s_es->s_last_orphan));
588 printk(KERN_ERR "sb_info orphan list:\n");
589 list_for_each(l, &sbi->s_orphan) {
590 struct inode *inode = orphan_list_entry(l);
592 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
593 inode->i_sb->s_id, inode->i_ino, inode,
594 inode->i_mode, inode->i_nlink,
599 static void ext4_put_super(struct super_block *sb)
601 struct ext4_sb_info *sbi = EXT4_SB(sb);
602 struct ext4_super_block *es = sbi->s_es;
605 flush_workqueue(sbi->dio_unwritten_wq);
606 destroy_workqueue(sbi->dio_unwritten_wq);
611 ext4_commit_super(sb, 1);
613 if (sbi->s_journal) {
614 err = jbd2_journal_destroy(sbi->s_journal);
615 sbi->s_journal = NULL;
617 ext4_abort(sb, __func__,
618 "Couldn't clean up the journal");
621 ext4_release_system_zone(sb);
623 ext4_ext_release(sb);
624 ext4_xattr_put_super(sb);
626 if (!(sb->s_flags & MS_RDONLY)) {
627 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
628 es->s_state = cpu_to_le16(sbi->s_mount_state);
629 ext4_commit_super(sb, 1);
632 remove_proc_entry(sb->s_id, ext4_proc_root);
634 kobject_del(&sbi->s_kobj);
636 for (i = 0; i < sbi->s_gdb_count; i++)
637 brelse(sbi->s_group_desc[i]);
638 kfree(sbi->s_group_desc);
639 if (is_vmalloc_addr(sbi->s_flex_groups))
640 vfree(sbi->s_flex_groups);
642 kfree(sbi->s_flex_groups);
643 percpu_counter_destroy(&sbi->s_freeblocks_counter);
644 percpu_counter_destroy(&sbi->s_freeinodes_counter);
645 percpu_counter_destroy(&sbi->s_dirs_counter);
646 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
649 for (i = 0; i < MAXQUOTAS; i++)
650 kfree(sbi->s_qf_names[i]);
653 /* Debugging code just in case the in-memory inode orphan list
654 * isn't empty. The on-disk one can be non-empty if we've
655 * detected an error and taken the fs readonly, but the
656 * in-memory list had better be clean by this point. */
657 if (!list_empty(&sbi->s_orphan))
658 dump_orphan_list(sb, sbi);
659 J_ASSERT(list_empty(&sbi->s_orphan));
661 invalidate_bdev(sb->s_bdev);
662 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
664 * Invalidate the journal device's buffers. We don't want them
665 * floating about in memory - the physical journal device may
666 * hotswapped, and it breaks the `ro-after' testing code.
668 sync_blockdev(sbi->journal_bdev);
669 invalidate_bdev(sbi->journal_bdev);
670 ext4_blkdev_remove(sbi);
672 sb->s_fs_info = NULL;
674 * Now that we are completely done shutting down the
675 * superblock, we need to actually destroy the kobject.
679 kobject_put(&sbi->s_kobj);
680 wait_for_completion(&sbi->s_kobj_unregister);
681 kfree(sbi->s_blockgroup_lock);
685 static struct kmem_cache *ext4_inode_cachep;
688 * Called inside transaction, so use GFP_NOFS
690 static struct inode *ext4_alloc_inode(struct super_block *sb)
692 struct ext4_inode_info *ei;
694 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
698 ei->vfs_inode.i_version = 1;
699 ei->vfs_inode.i_data.writeback_index = 0;
700 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
701 INIT_LIST_HEAD(&ei->i_prealloc_list);
702 spin_lock_init(&ei->i_prealloc_lock);
704 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
705 * therefore it can be null here. Don't check it, just initialize
708 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
709 ei->i_reserved_data_blocks = 0;
710 ei->i_reserved_meta_blocks = 0;
711 ei->i_allocated_meta_blocks = 0;
712 ei->i_delalloc_reserved_flag = 0;
713 spin_lock_init(&(ei->i_block_reservation_lock));
714 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
715 ei->cur_aio_dio = NULL;
717 return &ei->vfs_inode;
720 static void ext4_destroy_inode(struct inode *inode)
722 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
723 ext4_msg(inode->i_sb, KERN_ERR,
724 "Inode %lu (%p): orphan list check failed!",
725 inode->i_ino, EXT4_I(inode));
726 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
727 EXT4_I(inode), sizeof(struct ext4_inode_info),
731 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
734 static void init_once(void *foo)
736 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
738 INIT_LIST_HEAD(&ei->i_orphan);
739 #ifdef CONFIG_EXT4_FS_XATTR
740 init_rwsem(&ei->xattr_sem);
742 init_rwsem(&ei->i_data_sem);
743 inode_init_once(&ei->vfs_inode);
746 static int init_inodecache(void)
748 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
749 sizeof(struct ext4_inode_info),
750 0, (SLAB_RECLAIM_ACCOUNT|
753 if (ext4_inode_cachep == NULL)
758 static void destroy_inodecache(void)
760 kmem_cache_destroy(ext4_inode_cachep);
763 static void ext4_clear_inode(struct inode *inode)
765 ext4_discard_preallocations(inode);
766 if (EXT4_JOURNAL(inode))
767 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
768 &EXT4_I(inode)->jinode);
771 static inline void ext4_show_quota_options(struct seq_file *seq,
772 struct super_block *sb)
774 #if defined(CONFIG_QUOTA)
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 if (sbi->s_jquota_fmt)
778 seq_printf(seq, ",jqfmt=%s",
779 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
781 if (sbi->s_qf_names[USRQUOTA])
782 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
784 if (sbi->s_qf_names[GRPQUOTA])
785 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
787 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
788 seq_puts(seq, ",usrquota");
790 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
791 seq_puts(seq, ",grpquota");
797 * - it's set to a non-default value OR
798 * - if the per-sb default is different from the global default
800 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
803 unsigned long def_mount_opts;
804 struct super_block *sb = vfs->mnt_sb;
805 struct ext4_sb_info *sbi = EXT4_SB(sb);
806 struct ext4_super_block *es = sbi->s_es;
808 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
809 def_errors = le16_to_cpu(es->s_errors);
811 if (sbi->s_sb_block != 1)
812 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
813 if (test_opt(sb, MINIX_DF))
814 seq_puts(seq, ",minixdf");
815 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
816 seq_puts(seq, ",grpid");
817 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
818 seq_puts(seq, ",nogrpid");
819 if (sbi->s_resuid != EXT4_DEF_RESUID ||
820 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
821 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
823 if (sbi->s_resgid != EXT4_DEF_RESGID ||
824 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
825 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
827 if (test_opt(sb, ERRORS_RO)) {
828 if (def_errors == EXT4_ERRORS_PANIC ||
829 def_errors == EXT4_ERRORS_CONTINUE) {
830 seq_puts(seq, ",errors=remount-ro");
833 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
834 seq_puts(seq, ",errors=continue");
835 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
836 seq_puts(seq, ",errors=panic");
837 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
838 seq_puts(seq, ",nouid32");
839 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
840 seq_puts(seq, ",debug");
841 if (test_opt(sb, OLDALLOC))
842 seq_puts(seq, ",oldalloc");
843 #ifdef CONFIG_EXT4_FS_XATTR
844 if (test_opt(sb, XATTR_USER) &&
845 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
846 seq_puts(seq, ",user_xattr");
847 if (!test_opt(sb, XATTR_USER) &&
848 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
849 seq_puts(seq, ",nouser_xattr");
852 #ifdef CONFIG_EXT4_FS_POSIX_ACL
853 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
854 seq_puts(seq, ",acl");
855 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
856 seq_puts(seq, ",noacl");
858 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
859 seq_printf(seq, ",commit=%u",
860 (unsigned) (sbi->s_commit_interval / HZ));
862 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
863 seq_printf(seq, ",min_batch_time=%u",
864 (unsigned) sbi->s_min_batch_time);
866 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
867 seq_printf(seq, ",max_batch_time=%u",
868 (unsigned) sbi->s_min_batch_time);
872 * We're changing the default of barrier mount option, so
873 * let's always display its mount state so it's clear what its
876 seq_puts(seq, ",barrier=");
877 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
878 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
879 seq_puts(seq, ",journal_async_commit");
880 if (test_opt(sb, NOBH))
881 seq_puts(seq, ",nobh");
882 if (test_opt(sb, I_VERSION))
883 seq_puts(seq, ",i_version");
884 if (!test_opt(sb, DELALLOC))
885 seq_puts(seq, ",nodelalloc");
889 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
891 * journal mode get enabled in different ways
892 * So just print the value even if we didn't specify it
894 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
895 seq_puts(seq, ",data=journal");
896 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
897 seq_puts(seq, ",data=ordered");
898 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
899 seq_puts(seq, ",data=writeback");
901 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
902 seq_printf(seq, ",inode_readahead_blks=%u",
903 sbi->s_inode_readahead_blks);
905 if (test_opt(sb, DATA_ERR_ABORT))
906 seq_puts(seq, ",data_err=abort");
908 if (test_opt(sb, NO_AUTO_DA_ALLOC))
909 seq_puts(seq, ",noauto_da_alloc");
911 if (test_opt(sb, DISCARD))
912 seq_puts(seq, ",discard");
914 if (test_opt(sb, NOLOAD))
915 seq_puts(seq, ",norecovery");
917 ext4_show_quota_options(seq, sb);
922 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
923 u64 ino, u32 generation)
927 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
928 return ERR_PTR(-ESTALE);
929 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
930 return ERR_PTR(-ESTALE);
932 /* iget isn't really right if the inode is currently unallocated!!
934 * ext4_read_inode will return a bad_inode if the inode had been
935 * deleted, so we should be safe.
937 * Currently we don't know the generation for parent directory, so
938 * a generation of 0 means "accept any"
940 inode = ext4_iget(sb, ino);
942 return ERR_CAST(inode);
943 if (generation && inode->i_generation != generation) {
945 return ERR_PTR(-ESTALE);
951 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
952 int fh_len, int fh_type)
954 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
958 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
959 int fh_len, int fh_type)
961 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
966 * Try to release metadata pages (indirect blocks, directories) which are
967 * mapped via the block device. Since these pages could have journal heads
968 * which would prevent try_to_free_buffers() from freeing them, we must use
969 * jbd2 layer's try_to_free_buffers() function to release them.
971 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
974 journal_t *journal = EXT4_SB(sb)->s_journal;
976 WARN_ON(PageChecked(page));
977 if (!page_has_buffers(page))
980 return jbd2_journal_try_to_free_buffers(journal, page,
982 return try_to_free_buffers(page);
986 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
987 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
989 static int ext4_write_dquot(struct dquot *dquot);
990 static int ext4_acquire_dquot(struct dquot *dquot);
991 static int ext4_release_dquot(struct dquot *dquot);
992 static int ext4_mark_dquot_dirty(struct dquot *dquot);
993 static int ext4_write_info(struct super_block *sb, int type);
994 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
995 char *path, int remount);
996 static int ext4_quota_on_mount(struct super_block *sb, int type);
997 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
998 size_t len, loff_t off);
999 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1000 const char *data, size_t len, loff_t off);
1002 static struct dquot_operations ext4_quota_operations = {
1003 .initialize = dquot_initialize,
1005 .alloc_space = dquot_alloc_space,
1006 .reserve_space = dquot_reserve_space,
1007 .claim_space = dquot_claim_space,
1008 .release_rsv = dquot_release_reserved_space,
1009 .get_reserved_space = ext4_get_reserved_space,
1010 .alloc_inode = dquot_alloc_inode,
1011 .free_space = dquot_free_space,
1012 .free_inode = dquot_free_inode,
1013 .transfer = dquot_transfer,
1014 .write_dquot = ext4_write_dquot,
1015 .acquire_dquot = ext4_acquire_dquot,
1016 .release_dquot = ext4_release_dquot,
1017 .mark_dirty = ext4_mark_dquot_dirty,
1018 .write_info = ext4_write_info,
1019 .alloc_dquot = dquot_alloc,
1020 .destroy_dquot = dquot_destroy,
1023 static struct quotactl_ops ext4_qctl_operations = {
1024 .quota_on = ext4_quota_on,
1025 .quota_off = vfs_quota_off,
1026 .quota_sync = vfs_quota_sync,
1027 .get_info = vfs_get_dqinfo,
1028 .set_info = vfs_set_dqinfo,
1029 .get_dqblk = vfs_get_dqblk,
1030 .set_dqblk = vfs_set_dqblk
1034 static const struct super_operations ext4_sops = {
1035 .alloc_inode = ext4_alloc_inode,
1036 .destroy_inode = ext4_destroy_inode,
1037 .write_inode = ext4_write_inode,
1038 .dirty_inode = ext4_dirty_inode,
1039 .delete_inode = ext4_delete_inode,
1040 .put_super = ext4_put_super,
1041 .sync_fs = ext4_sync_fs,
1042 .freeze_fs = ext4_freeze,
1043 .unfreeze_fs = ext4_unfreeze,
1044 .statfs = ext4_statfs,
1045 .remount_fs = ext4_remount,
1046 .clear_inode = ext4_clear_inode,
1047 .show_options = ext4_show_options,
1049 .quota_read = ext4_quota_read,
1050 .quota_write = ext4_quota_write,
1052 .bdev_try_to_free_page = bdev_try_to_free_page,
1055 static const struct super_operations ext4_nojournal_sops = {
1056 .alloc_inode = ext4_alloc_inode,
1057 .destroy_inode = ext4_destroy_inode,
1058 .write_inode = ext4_write_inode,
1059 .dirty_inode = ext4_dirty_inode,
1060 .delete_inode = ext4_delete_inode,
1061 .write_super = ext4_write_super,
1062 .put_super = ext4_put_super,
1063 .statfs = ext4_statfs,
1064 .remount_fs = ext4_remount,
1065 .clear_inode = ext4_clear_inode,
1066 .show_options = ext4_show_options,
1068 .quota_read = ext4_quota_read,
1069 .quota_write = ext4_quota_write,
1071 .bdev_try_to_free_page = bdev_try_to_free_page,
1074 static const struct export_operations ext4_export_ops = {
1075 .fh_to_dentry = ext4_fh_to_dentry,
1076 .fh_to_parent = ext4_fh_to_parent,
1077 .get_parent = ext4_get_parent,
1081 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1082 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1083 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1084 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1085 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1086 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1087 Opt_journal_update, Opt_journal_dev,
1088 Opt_journal_checksum, Opt_journal_async_commit,
1089 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1090 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1091 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1092 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1093 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1094 Opt_usrquota, Opt_grpquota, Opt_i_version,
1095 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1096 Opt_block_validity, Opt_noblock_validity,
1097 Opt_inode_readahead_blks, Opt_journal_ioprio,
1098 Opt_discard, Opt_nodiscard,
1101 static const match_table_t tokens = {
1102 {Opt_bsd_df, "bsddf"},
1103 {Opt_minix_df, "minixdf"},
1104 {Opt_grpid, "grpid"},
1105 {Opt_grpid, "bsdgroups"},
1106 {Opt_nogrpid, "nogrpid"},
1107 {Opt_nogrpid, "sysvgroups"},
1108 {Opt_resgid, "resgid=%u"},
1109 {Opt_resuid, "resuid=%u"},
1111 {Opt_err_cont, "errors=continue"},
1112 {Opt_err_panic, "errors=panic"},
1113 {Opt_err_ro, "errors=remount-ro"},
1114 {Opt_nouid32, "nouid32"},
1115 {Opt_debug, "debug"},
1116 {Opt_oldalloc, "oldalloc"},
1117 {Opt_orlov, "orlov"},
1118 {Opt_user_xattr, "user_xattr"},
1119 {Opt_nouser_xattr, "nouser_xattr"},
1121 {Opt_noacl, "noacl"},
1122 {Opt_noload, "noload"},
1123 {Opt_noload, "norecovery"},
1126 {Opt_commit, "commit=%u"},
1127 {Opt_min_batch_time, "min_batch_time=%u"},
1128 {Opt_max_batch_time, "max_batch_time=%u"},
1129 {Opt_journal_update, "journal=update"},
1130 {Opt_journal_dev, "journal_dev=%u"},
1131 {Opt_journal_checksum, "journal_checksum"},
1132 {Opt_journal_async_commit, "journal_async_commit"},
1133 {Opt_abort, "abort"},
1134 {Opt_data_journal, "data=journal"},
1135 {Opt_data_ordered, "data=ordered"},
1136 {Opt_data_writeback, "data=writeback"},
1137 {Opt_data_err_abort, "data_err=abort"},
1138 {Opt_data_err_ignore, "data_err=ignore"},
1139 {Opt_mb_history_length, "mb_history_length=%u"},
1140 {Opt_offusrjquota, "usrjquota="},
1141 {Opt_usrjquota, "usrjquota=%s"},
1142 {Opt_offgrpjquota, "grpjquota="},
1143 {Opt_grpjquota, "grpjquota=%s"},
1144 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1145 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1146 {Opt_grpquota, "grpquota"},
1147 {Opt_noquota, "noquota"},
1148 {Opt_quota, "quota"},
1149 {Opt_usrquota, "usrquota"},
1150 {Opt_barrier, "barrier=%u"},
1151 {Opt_barrier, "barrier"},
1152 {Opt_nobarrier, "nobarrier"},
1153 {Opt_i_version, "i_version"},
1154 {Opt_stripe, "stripe=%u"},
1155 {Opt_resize, "resize"},
1156 {Opt_delalloc, "delalloc"},
1157 {Opt_nodelalloc, "nodelalloc"},
1158 {Opt_block_validity, "block_validity"},
1159 {Opt_noblock_validity, "noblock_validity"},
1160 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1161 {Opt_journal_ioprio, "journal_ioprio=%u"},
1162 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1163 {Opt_auto_da_alloc, "auto_da_alloc"},
1164 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1165 {Opt_discard, "discard"},
1166 {Opt_nodiscard, "nodiscard"},
1170 static ext4_fsblk_t get_sb_block(void **data)
1172 ext4_fsblk_t sb_block;
1173 char *options = (char *) *data;
1175 if (!options || strncmp(options, "sb=", 3) != 0)
1176 return 1; /* Default location */
1179 /* TODO: use simple_strtoll with >32bit ext4 */
1180 sb_block = simple_strtoul(options, &options, 0);
1181 if (*options && *options != ',') {
1182 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1186 if (*options == ',')
1188 *data = (void *) options;
1193 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1195 static int parse_options(char *options, struct super_block *sb,
1196 unsigned long *journal_devnum,
1197 unsigned int *journal_ioprio,
1198 ext4_fsblk_t *n_blocks_count, int is_remount)
1200 struct ext4_sb_info *sbi = EXT4_SB(sb);
1202 substring_t args[MAX_OPT_ARGS];
1213 while ((p = strsep(&options, ",")) != NULL) {
1218 token = match_token(p, tokens, args);
1221 clear_opt(sbi->s_mount_opt, MINIX_DF);
1224 set_opt(sbi->s_mount_opt, MINIX_DF);
1227 set_opt(sbi->s_mount_opt, GRPID);
1230 clear_opt(sbi->s_mount_opt, GRPID);
1233 if (match_int(&args[0], &option))
1235 sbi->s_resuid = option;
1238 if (match_int(&args[0], &option))
1240 sbi->s_resgid = option;
1243 /* handled by get_sb_block() instead of here */
1244 /* *sb_block = match_int(&args[0]); */
1247 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1248 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1249 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1252 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1253 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1254 set_opt(sbi->s_mount_opt, ERRORS_RO);
1257 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1258 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1259 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1262 set_opt(sbi->s_mount_opt, NO_UID32);
1265 set_opt(sbi->s_mount_opt, DEBUG);
1268 set_opt(sbi->s_mount_opt, OLDALLOC);
1271 clear_opt(sbi->s_mount_opt, OLDALLOC);
1273 #ifdef CONFIG_EXT4_FS_XATTR
1274 case Opt_user_xattr:
1275 set_opt(sbi->s_mount_opt, XATTR_USER);
1277 case Opt_nouser_xattr:
1278 clear_opt(sbi->s_mount_opt, XATTR_USER);
1281 case Opt_user_xattr:
1282 case Opt_nouser_xattr:
1283 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1286 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1288 set_opt(sbi->s_mount_opt, POSIX_ACL);
1291 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1296 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1299 case Opt_journal_update:
1301 /* Eventually we will want to be able to create
1302 a journal file here. For now, only allow the
1303 user to specify an existing inode to be the
1306 ext4_msg(sb, KERN_ERR,
1307 "Cannot specify journal on remount");
1310 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1312 case Opt_journal_dev:
1314 ext4_msg(sb, KERN_ERR,
1315 "Cannot specify journal on remount");
1318 if (match_int(&args[0], &option))
1320 *journal_devnum = option;
1322 case Opt_journal_checksum:
1323 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1325 case Opt_journal_async_commit:
1326 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1327 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1330 set_opt(sbi->s_mount_opt, NOLOAD);
1333 if (match_int(&args[0], &option))
1338 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1339 sbi->s_commit_interval = HZ * option;
1341 case Opt_max_batch_time:
1342 if (match_int(&args[0], &option))
1347 option = EXT4_DEF_MAX_BATCH_TIME;
1348 sbi->s_max_batch_time = option;
1350 case Opt_min_batch_time:
1351 if (match_int(&args[0], &option))
1355 sbi->s_min_batch_time = option;
1357 case Opt_data_journal:
1358 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1360 case Opt_data_ordered:
1361 data_opt = EXT4_MOUNT_ORDERED_DATA;
1363 case Opt_data_writeback:
1364 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1367 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1369 ext4_msg(sb, KERN_ERR,
1370 "Cannot change data mode on remount");
1374 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1375 sbi->s_mount_opt |= data_opt;
1378 case Opt_data_err_abort:
1379 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1381 case Opt_data_err_ignore:
1382 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1384 case Opt_mb_history_length:
1385 if (match_int(&args[0], &option))
1389 sbi->s_mb_history_max = option;
1398 if (sb_any_quota_loaded(sb) &&
1399 !sbi->s_qf_names[qtype]) {
1400 ext4_msg(sb, KERN_ERR,
1401 "Cannot change journaled "
1402 "quota options when quota turned on");
1405 qname = match_strdup(&args[0]);
1407 ext4_msg(sb, KERN_ERR,
1408 "Not enough memory for "
1409 "storing quotafile name");
1412 if (sbi->s_qf_names[qtype] &&
1413 strcmp(sbi->s_qf_names[qtype], qname)) {
1414 ext4_msg(sb, KERN_ERR,
1415 "%s quota file already "
1416 "specified", QTYPE2NAME(qtype));
1420 sbi->s_qf_names[qtype] = qname;
1421 if (strchr(sbi->s_qf_names[qtype], '/')) {
1422 ext4_msg(sb, KERN_ERR,
1423 "quotafile must be on "
1425 kfree(sbi->s_qf_names[qtype]);
1426 sbi->s_qf_names[qtype] = NULL;
1429 set_opt(sbi->s_mount_opt, QUOTA);
1431 case Opt_offusrjquota:
1434 case Opt_offgrpjquota:
1437 if (sb_any_quota_loaded(sb) &&
1438 sbi->s_qf_names[qtype]) {
1439 ext4_msg(sb, KERN_ERR, "Cannot change "
1440 "journaled quota options when "
1445 * The space will be released later when all options
1446 * are confirmed to be correct
1448 sbi->s_qf_names[qtype] = NULL;
1450 case Opt_jqfmt_vfsold:
1451 qfmt = QFMT_VFS_OLD;
1453 case Opt_jqfmt_vfsv0:
1456 if (sb_any_quota_loaded(sb) &&
1457 sbi->s_jquota_fmt != qfmt) {
1458 ext4_msg(sb, KERN_ERR, "Cannot change "
1459 "journaled quota options when "
1463 sbi->s_jquota_fmt = qfmt;
1467 set_opt(sbi->s_mount_opt, QUOTA);
1468 set_opt(sbi->s_mount_opt, USRQUOTA);
1471 set_opt(sbi->s_mount_opt, QUOTA);
1472 set_opt(sbi->s_mount_opt, GRPQUOTA);
1475 if (sb_any_quota_loaded(sb)) {
1476 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1477 "options when quota turned on");
1480 clear_opt(sbi->s_mount_opt, QUOTA);
1481 clear_opt(sbi->s_mount_opt, USRQUOTA);
1482 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1488 ext4_msg(sb, KERN_ERR,
1489 "quota options not supported");
1493 case Opt_offusrjquota:
1494 case Opt_offgrpjquota:
1495 case Opt_jqfmt_vfsold:
1496 case Opt_jqfmt_vfsv0:
1497 ext4_msg(sb, KERN_ERR,
1498 "journaled quota options not supported");
1504 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1507 clear_opt(sbi->s_mount_opt, BARRIER);
1510 if (match_int(&args[0], &option)) {
1511 set_opt(sbi->s_mount_opt, BARRIER);
1515 set_opt(sbi->s_mount_opt, BARRIER);
1517 clear_opt(sbi->s_mount_opt, BARRIER);
1523 ext4_msg(sb, KERN_ERR,
1524 "resize option only available "
1528 if (match_int(&args[0], &option) != 0)
1530 *n_blocks_count = option;
1533 set_opt(sbi->s_mount_opt, NOBH);
1536 clear_opt(sbi->s_mount_opt, NOBH);
1539 set_opt(sbi->s_mount_opt, I_VERSION);
1540 sb->s_flags |= MS_I_VERSION;
1542 case Opt_nodelalloc:
1543 clear_opt(sbi->s_mount_opt, DELALLOC);
1546 if (match_int(&args[0], &option))
1550 sbi->s_stripe = option;
1553 set_opt(sbi->s_mount_opt, DELALLOC);
1555 case Opt_block_validity:
1556 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1558 case Opt_noblock_validity:
1559 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1561 case Opt_inode_readahead_blks:
1562 if (match_int(&args[0], &option))
1564 if (option < 0 || option > (1 << 30))
1566 if (!is_power_of_2(option)) {
1567 ext4_msg(sb, KERN_ERR,
1568 "EXT4-fs: inode_readahead_blks"
1569 " must be a power of 2");
1572 sbi->s_inode_readahead_blks = option;
1574 case Opt_journal_ioprio:
1575 if (match_int(&args[0], &option))
1577 if (option < 0 || option > 7)
1579 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1582 case Opt_noauto_da_alloc:
1583 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1585 case Opt_auto_da_alloc:
1586 if (match_int(&args[0], &option)) {
1587 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1591 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1593 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1596 set_opt(sbi->s_mount_opt, DISCARD);
1599 clear_opt(sbi->s_mount_opt, DISCARD);
1602 ext4_msg(sb, KERN_ERR,
1603 "Unrecognized mount option \"%s\" "
1604 "or missing value", p);
1609 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1610 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1611 sbi->s_qf_names[USRQUOTA])
1612 clear_opt(sbi->s_mount_opt, USRQUOTA);
1614 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1615 sbi->s_qf_names[GRPQUOTA])
1616 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1618 if ((sbi->s_qf_names[USRQUOTA] &&
1619 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1620 (sbi->s_qf_names[GRPQUOTA] &&
1621 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1622 ext4_msg(sb, KERN_ERR, "old and new quota "
1627 if (!sbi->s_jquota_fmt) {
1628 ext4_msg(sb, KERN_ERR, "journaled quota format "
1633 if (sbi->s_jquota_fmt) {
1634 ext4_msg(sb, KERN_ERR, "journaled quota format "
1635 "specified with no journaling "
1644 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1647 struct ext4_sb_info *sbi = EXT4_SB(sb);
1650 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1651 ext4_msg(sb, KERN_ERR, "revision level too high, "
1652 "forcing read-only mode");
1657 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1658 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1659 "running e2fsck is recommended");
1660 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1661 ext4_msg(sb, KERN_WARNING,
1662 "warning: mounting fs with errors, "
1663 "running e2fsck is recommended");
1664 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1665 le16_to_cpu(es->s_mnt_count) >=
1666 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1667 ext4_msg(sb, KERN_WARNING,
1668 "warning: maximal mount count reached, "
1669 "running e2fsck is recommended");
1670 else if (le32_to_cpu(es->s_checkinterval) &&
1671 (le32_to_cpu(es->s_lastcheck) +
1672 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1673 ext4_msg(sb, KERN_WARNING,
1674 "warning: checktime reached, "
1675 "running e2fsck is recommended");
1676 if (!sbi->s_journal)
1677 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1678 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1679 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1680 le16_add_cpu(&es->s_mnt_count, 1);
1681 es->s_mtime = cpu_to_le32(get_seconds());
1682 ext4_update_dynamic_rev(sb);
1684 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1686 ext4_commit_super(sb, 1);
1687 if (test_opt(sb, DEBUG))
1688 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1689 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1691 sbi->s_groups_count,
1692 EXT4_BLOCKS_PER_GROUP(sb),
1693 EXT4_INODES_PER_GROUP(sb),
1696 if (EXT4_SB(sb)->s_journal) {
1697 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1698 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1699 "external", EXT4_SB(sb)->s_journal->j_devname);
1701 ext4_msg(sb, KERN_INFO, "no journal");
1706 static int ext4_fill_flex_info(struct super_block *sb)
1708 struct ext4_sb_info *sbi = EXT4_SB(sb);
1709 struct ext4_group_desc *gdp = NULL;
1710 ext4_group_t flex_group_count;
1711 ext4_group_t flex_group;
1712 int groups_per_flex = 0;
1716 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1717 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1719 if (groups_per_flex < 2) {
1720 sbi->s_log_groups_per_flex = 0;
1724 /* We allocate both existing and potentially added groups */
1725 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1726 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1727 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1728 size = flex_group_count * sizeof(struct flex_groups);
1729 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1730 if (sbi->s_flex_groups == NULL) {
1731 sbi->s_flex_groups = vmalloc(size);
1732 if (sbi->s_flex_groups)
1733 memset(sbi->s_flex_groups, 0, size);
1735 if (sbi->s_flex_groups == NULL) {
1736 ext4_msg(sb, KERN_ERR, "not enough memory for "
1737 "%u flex groups", flex_group_count);
1741 for (i = 0; i < sbi->s_groups_count; i++) {
1742 gdp = ext4_get_group_desc(sb, i, NULL);
1744 flex_group = ext4_flex_group(sbi, i);
1745 atomic_add(ext4_free_inodes_count(sb, gdp),
1746 &sbi->s_flex_groups[flex_group].free_inodes);
1747 atomic_add(ext4_free_blks_count(sb, gdp),
1748 &sbi->s_flex_groups[flex_group].free_blocks);
1749 atomic_add(ext4_used_dirs_count(sb, gdp),
1750 &sbi->s_flex_groups[flex_group].used_dirs);
1758 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1759 struct ext4_group_desc *gdp)
1763 if (sbi->s_es->s_feature_ro_compat &
1764 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1765 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1766 __le32 le_group = cpu_to_le32(block_group);
1768 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1769 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1770 crc = crc16(crc, (__u8 *)gdp, offset);
1771 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1772 /* for checksum of struct ext4_group_desc do the rest...*/
1773 if ((sbi->s_es->s_feature_incompat &
1774 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1775 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1776 crc = crc16(crc, (__u8 *)gdp + offset,
1777 le16_to_cpu(sbi->s_es->s_desc_size) -
1781 return cpu_to_le16(crc);
1784 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1785 struct ext4_group_desc *gdp)
1787 if ((sbi->s_es->s_feature_ro_compat &
1788 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1789 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1795 /* Called at mount-time, super-block is locked */
1796 static int ext4_check_descriptors(struct super_block *sb)
1798 struct ext4_sb_info *sbi = EXT4_SB(sb);
1799 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1800 ext4_fsblk_t last_block;
1801 ext4_fsblk_t block_bitmap;
1802 ext4_fsblk_t inode_bitmap;
1803 ext4_fsblk_t inode_table;
1804 int flexbg_flag = 0;
1807 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1810 ext4_debug("Checking group descriptors");
1812 for (i = 0; i < sbi->s_groups_count; i++) {
1813 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1815 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1816 last_block = ext4_blocks_count(sbi->s_es) - 1;
1818 last_block = first_block +
1819 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1821 block_bitmap = ext4_block_bitmap(sb, gdp);
1822 if (block_bitmap < first_block || block_bitmap > last_block) {
1823 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1824 "Block bitmap for group %u not in group "
1825 "(block %llu)!", i, block_bitmap);
1828 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1829 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1830 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1831 "Inode bitmap for group %u not in group "
1832 "(block %llu)!", i, inode_bitmap);
1835 inode_table = ext4_inode_table(sb, gdp);
1836 if (inode_table < first_block ||
1837 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1838 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1839 "Inode table for group %u not in group "
1840 "(block %llu)!", i, inode_table);
1843 ext4_lock_group(sb, i);
1844 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1845 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1846 "Checksum for group %u failed (%u!=%u)",
1847 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1848 gdp)), le16_to_cpu(gdp->bg_checksum));
1849 if (!(sb->s_flags & MS_RDONLY)) {
1850 ext4_unlock_group(sb, i);
1854 ext4_unlock_group(sb, i);
1856 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1859 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1860 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1864 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1865 * the superblock) which were deleted from all directories, but held open by
1866 * a process at the time of a crash. We walk the list and try to delete these
1867 * inodes at recovery time (only with a read-write filesystem).
1869 * In order to keep the orphan inode chain consistent during traversal (in
1870 * case of crash during recovery), we link each inode into the superblock
1871 * orphan list_head and handle it the same way as an inode deletion during
1872 * normal operation (which journals the operations for us).
1874 * We only do an iget() and an iput() on each inode, which is very safe if we
1875 * accidentally point at an in-use or already deleted inode. The worst that
1876 * can happen in this case is that we get a "bit already cleared" message from
1877 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1878 * e2fsck was run on this filesystem, and it must have already done the orphan
1879 * inode cleanup for us, so we can safely abort without any further action.
1881 static void ext4_orphan_cleanup(struct super_block *sb,
1882 struct ext4_super_block *es)
1884 unsigned int s_flags = sb->s_flags;
1885 int nr_orphans = 0, nr_truncates = 0;
1889 if (!es->s_last_orphan) {
1890 jbd_debug(4, "no orphan inodes to clean up\n");
1894 if (bdev_read_only(sb->s_bdev)) {
1895 ext4_msg(sb, KERN_ERR, "write access "
1896 "unavailable, skipping orphan cleanup");
1900 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1901 if (es->s_last_orphan)
1902 jbd_debug(1, "Errors on filesystem, "
1903 "clearing orphan list.\n");
1904 es->s_last_orphan = 0;
1905 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1909 if (s_flags & MS_RDONLY) {
1910 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1911 sb->s_flags &= ~MS_RDONLY;
1914 /* Needed for iput() to work correctly and not trash data */
1915 sb->s_flags |= MS_ACTIVE;
1916 /* Turn on quotas so that they are updated correctly */
1917 for (i = 0; i < MAXQUOTAS; i++) {
1918 if (EXT4_SB(sb)->s_qf_names[i]) {
1919 int ret = ext4_quota_on_mount(sb, i);
1921 ext4_msg(sb, KERN_ERR,
1922 "Cannot turn on journaled "
1923 "quota: error %d", ret);
1928 while (es->s_last_orphan) {
1929 struct inode *inode;
1931 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1932 if (IS_ERR(inode)) {
1933 es->s_last_orphan = 0;
1937 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1939 if (inode->i_nlink) {
1940 ext4_msg(sb, KERN_DEBUG,
1941 "%s: truncating inode %lu to %lld bytes",
1942 __func__, inode->i_ino, inode->i_size);
1943 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1944 inode->i_ino, inode->i_size);
1945 ext4_truncate(inode);
1948 ext4_msg(sb, KERN_DEBUG,
1949 "%s: deleting unreferenced inode %lu",
1950 __func__, inode->i_ino);
1951 jbd_debug(2, "deleting unreferenced inode %lu\n",
1955 iput(inode); /* The delete magic happens here! */
1958 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1961 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1962 PLURAL(nr_orphans));
1964 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1965 PLURAL(nr_truncates));
1967 /* Turn quotas off */
1968 for (i = 0; i < MAXQUOTAS; i++) {
1969 if (sb_dqopt(sb)->files[i])
1970 vfs_quota_off(sb, i, 0);
1973 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1977 * Maximal extent format file size.
1978 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1979 * extent format containers, within a sector_t, and within i_blocks
1980 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1981 * so that won't be a limiting factor.
1983 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1985 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1988 loff_t upper_limit = MAX_LFS_FILESIZE;
1990 /* small i_blocks in vfs inode? */
1991 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1993 * CONFIG_LBDAF is not enabled implies the inode
1994 * i_block represent total blocks in 512 bytes
1995 * 32 == size of vfs inode i_blocks * 8
1997 upper_limit = (1LL << 32) - 1;
1999 /* total blocks in file system block size */
2000 upper_limit >>= (blkbits - 9);
2001 upper_limit <<= blkbits;
2004 /* 32-bit extent-start container, ee_block */
2009 /* Sanity check against vm- & vfs- imposed limits */
2010 if (res > upper_limit)
2017 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2018 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2019 * We need to be 1 filesystem block less than the 2^48 sector limit.
2021 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2023 loff_t res = EXT4_NDIR_BLOCKS;
2026 /* This is calculated to be the largest file size for a dense, block
2027 * mapped file such that the file's total number of 512-byte sectors,
2028 * including data and all indirect blocks, does not exceed (2^48 - 1).
2030 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2031 * number of 512-byte sectors of the file.
2034 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2036 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2037 * the inode i_block field represents total file blocks in
2038 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2040 upper_limit = (1LL << 32) - 1;
2042 /* total blocks in file system block size */
2043 upper_limit >>= (bits - 9);
2047 * We use 48 bit ext4_inode i_blocks
2048 * With EXT4_HUGE_FILE_FL set the i_blocks
2049 * represent total number of blocks in
2050 * file system block size
2052 upper_limit = (1LL << 48) - 1;
2056 /* indirect blocks */
2058 /* double indirect blocks */
2059 meta_blocks += 1 + (1LL << (bits-2));
2060 /* tripple indirect blocks */
2061 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2063 upper_limit -= meta_blocks;
2064 upper_limit <<= bits;
2066 res += 1LL << (bits-2);
2067 res += 1LL << (2*(bits-2));
2068 res += 1LL << (3*(bits-2));
2070 if (res > upper_limit)
2073 if (res > MAX_LFS_FILESIZE)
2074 res = MAX_LFS_FILESIZE;
2079 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2080 ext4_fsblk_t logical_sb_block, int nr)
2082 struct ext4_sb_info *sbi = EXT4_SB(sb);
2083 ext4_group_t bg, first_meta_bg;
2086 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2088 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2090 return logical_sb_block + nr + 1;
2091 bg = sbi->s_desc_per_block * nr;
2092 if (ext4_bg_has_super(sb, bg))
2095 return (has_super + ext4_group_first_block_no(sb, bg));
2099 * ext4_get_stripe_size: Get the stripe size.
2100 * @sbi: In memory super block info
2102 * If we have specified it via mount option, then
2103 * use the mount option value. If the value specified at mount time is
2104 * greater than the blocks per group use the super block value.
2105 * If the super block value is greater than blocks per group return 0.
2106 * Allocator needs it be less than blocks per group.
2109 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2111 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2112 unsigned long stripe_width =
2113 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2115 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2116 return sbi->s_stripe;
2118 if (stripe_width <= sbi->s_blocks_per_group)
2119 return stripe_width;
2121 if (stride <= sbi->s_blocks_per_group)
2130 struct attribute attr;
2131 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2132 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2133 const char *, size_t);
2137 static int parse_strtoul(const char *buf,
2138 unsigned long max, unsigned long *value)
2142 while (*buf && isspace(*buf))
2144 *value = simple_strtoul(buf, &endp, 0);
2145 while (*endp && isspace(*endp))
2147 if (*endp || *value > max)
2153 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2154 struct ext4_sb_info *sbi,
2157 return snprintf(buf, PAGE_SIZE, "%llu\n",
2158 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2161 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2162 struct ext4_sb_info *sbi, char *buf)
2164 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2166 return snprintf(buf, PAGE_SIZE, "%lu\n",
2167 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2168 sbi->s_sectors_written_start) >> 1);
2171 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2172 struct ext4_sb_info *sbi, char *buf)
2174 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2176 return snprintf(buf, PAGE_SIZE, "%llu\n",
2177 sbi->s_kbytes_written +
2178 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2179 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2182 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2183 struct ext4_sb_info *sbi,
2184 const char *buf, size_t count)
2188 if (parse_strtoul(buf, 0x40000000, &t))
2191 if (!is_power_of_2(t))
2194 sbi->s_inode_readahead_blks = t;
2198 static ssize_t sbi_ui_show(struct ext4_attr *a,
2199 struct ext4_sb_info *sbi, char *buf)
2201 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2203 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2206 static ssize_t sbi_ui_store(struct ext4_attr *a,
2207 struct ext4_sb_info *sbi,
2208 const char *buf, size_t count)
2210 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2213 if (parse_strtoul(buf, 0xffffffff, &t))
2219 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2220 static struct ext4_attr ext4_attr_##_name = { \
2221 .attr = {.name = __stringify(_name), .mode = _mode }, \
2224 .offset = offsetof(struct ext4_sb_info, _elname), \
2226 #define EXT4_ATTR(name, mode, show, store) \
2227 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2229 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2230 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2231 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2232 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2233 #define ATTR_LIST(name) &ext4_attr_##name.attr
2235 EXT4_RO_ATTR(delayed_allocation_blocks);
2236 EXT4_RO_ATTR(session_write_kbytes);
2237 EXT4_RO_ATTR(lifetime_write_kbytes);
2238 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2239 inode_readahead_blks_store, s_inode_readahead_blks);
2240 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2241 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2242 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2243 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2244 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2245 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2246 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2247 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2249 static struct attribute *ext4_attrs[] = {
2250 ATTR_LIST(delayed_allocation_blocks),
2251 ATTR_LIST(session_write_kbytes),
2252 ATTR_LIST(lifetime_write_kbytes),
2253 ATTR_LIST(inode_readahead_blks),
2254 ATTR_LIST(inode_goal),
2255 ATTR_LIST(mb_stats),
2256 ATTR_LIST(mb_max_to_scan),
2257 ATTR_LIST(mb_min_to_scan),
2258 ATTR_LIST(mb_order2_req),
2259 ATTR_LIST(mb_stream_req),
2260 ATTR_LIST(mb_group_prealloc),
2261 ATTR_LIST(max_writeback_mb_bump),
2265 static ssize_t ext4_attr_show(struct kobject *kobj,
2266 struct attribute *attr, char *buf)
2268 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2270 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2272 return a->show ? a->show(a, sbi, buf) : 0;
2275 static ssize_t ext4_attr_store(struct kobject *kobj,
2276 struct attribute *attr,
2277 const char *buf, size_t len)
2279 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2281 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2283 return a->store ? a->store(a, sbi, buf, len) : 0;
2286 static void ext4_sb_release(struct kobject *kobj)
2288 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2290 complete(&sbi->s_kobj_unregister);
2294 static struct sysfs_ops ext4_attr_ops = {
2295 .show = ext4_attr_show,
2296 .store = ext4_attr_store,
2299 static struct kobj_type ext4_ktype = {
2300 .default_attrs = ext4_attrs,
2301 .sysfs_ops = &ext4_attr_ops,
2302 .release = ext4_sb_release,
2306 * Check whether this filesystem can be mounted based on
2307 * the features present and the RDONLY/RDWR mount requested.
2308 * Returns 1 if this filesystem can be mounted as requested,
2309 * 0 if it cannot be.
2311 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2313 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2314 ext4_msg(sb, KERN_ERR,
2315 "Couldn't mount because of "
2316 "unsupported optional features (%x)",
2317 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2318 ~EXT4_FEATURE_INCOMPAT_SUPP));
2325 /* Check that feature set is OK for a read-write mount */
2326 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2327 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2328 "unsupported optional features (%x)",
2329 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2330 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2334 * Large file size enabled file system can only be mounted
2335 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2338 if (sizeof(blkcnt_t) < sizeof(u64)) {
2339 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2340 "cannot be mounted RDWR without "
2348 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2349 __releases(kernel_lock)
2350 __acquires(kernel_lock)
2352 struct buffer_head *bh;
2353 struct ext4_super_block *es = NULL;
2354 struct ext4_sb_info *sbi;
2356 ext4_fsblk_t sb_block = get_sb_block(&data);
2357 ext4_fsblk_t logical_sb_block;
2358 unsigned long offset = 0;
2359 unsigned long journal_devnum = 0;
2360 unsigned long def_mount_opts;
2366 unsigned int db_count;
2368 int needs_recovery, has_huge_files;
2371 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2373 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2377 sbi->s_blockgroup_lock =
2378 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2379 if (!sbi->s_blockgroup_lock) {
2383 sb->s_fs_info = sbi;
2384 sbi->s_mount_opt = 0;
2385 sbi->s_resuid = EXT4_DEF_RESUID;
2386 sbi->s_resgid = EXT4_DEF_RESGID;
2387 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2388 sbi->s_sb_block = sb_block;
2389 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2394 /* Cleanup superblock name */
2395 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2398 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2400 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2405 * The ext4 superblock will not be buffer aligned for other than 1kB
2406 * block sizes. We need to calculate the offset from buffer start.
2408 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2409 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2410 offset = do_div(logical_sb_block, blocksize);
2412 logical_sb_block = sb_block;
2415 if (!(bh = sb_bread(sb, logical_sb_block))) {
2416 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2420 * Note: s_es must be initialized as soon as possible because
2421 * some ext4 macro-instructions depend on its value
2423 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2425 sb->s_magic = le16_to_cpu(es->s_magic);
2426 if (sb->s_magic != EXT4_SUPER_MAGIC)
2428 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2430 /* Set defaults before we parse the mount options */
2431 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2432 if (def_mount_opts & EXT4_DEFM_DEBUG)
2433 set_opt(sbi->s_mount_opt, DEBUG);
2434 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2435 set_opt(sbi->s_mount_opt, GRPID);
2436 if (def_mount_opts & EXT4_DEFM_UID16)
2437 set_opt(sbi->s_mount_opt, NO_UID32);
2438 #ifdef CONFIG_EXT4_FS_XATTR
2439 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2440 set_opt(sbi->s_mount_opt, XATTR_USER);
2442 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2443 if (def_mount_opts & EXT4_DEFM_ACL)
2444 set_opt(sbi->s_mount_opt, POSIX_ACL);
2446 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2447 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2448 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2449 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2450 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2451 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2453 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2454 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2455 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2456 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2458 set_opt(sbi->s_mount_opt, ERRORS_RO);
2460 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2461 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2462 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2463 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2464 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2465 sbi->s_mb_history_max = default_mb_history_length;
2467 set_opt(sbi->s_mount_opt, BARRIER);
2470 * enable delayed allocation by default
2471 * Use -o nodelalloc to turn it off
2473 set_opt(sbi->s_mount_opt, DELALLOC);
2475 if (!parse_options((char *) data, sb, &journal_devnum,
2476 &journal_ioprio, NULL, 0))
2479 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2480 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2482 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2483 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2484 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2485 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2486 ext4_msg(sb, KERN_WARNING,
2487 "feature flags set on rev 0 fs, "
2488 "running e2fsck is recommended");
2491 * Check feature flags regardless of the revision level, since we
2492 * previously didn't change the revision level when setting the flags,
2493 * so there is a chance incompat flags are set on a rev 0 filesystem.
2495 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2498 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2500 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2501 blocksize > EXT4_MAX_BLOCK_SIZE) {
2502 ext4_msg(sb, KERN_ERR,
2503 "Unsupported filesystem blocksize %d", blocksize);
2507 if (sb->s_blocksize != blocksize) {
2508 /* Validate the filesystem blocksize */
2509 if (!sb_set_blocksize(sb, blocksize)) {
2510 ext4_msg(sb, KERN_ERR, "bad block size %d",
2516 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2517 offset = do_div(logical_sb_block, blocksize);
2518 bh = sb_bread(sb, logical_sb_block);
2520 ext4_msg(sb, KERN_ERR,
2521 "Can't read superblock on 2nd try");
2524 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2526 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2527 ext4_msg(sb, KERN_ERR,
2528 "Magic mismatch, very weird!");
2533 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2534 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2535 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2537 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2539 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2540 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2541 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2543 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2544 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2545 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2546 (!is_power_of_2(sbi->s_inode_size)) ||
2547 (sbi->s_inode_size > blocksize)) {
2548 ext4_msg(sb, KERN_ERR,
2549 "unsupported inode size: %d",
2553 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2554 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2557 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2558 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2559 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2560 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2561 !is_power_of_2(sbi->s_desc_size)) {
2562 ext4_msg(sb, KERN_ERR,
2563 "unsupported descriptor size %lu",
2568 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2570 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2571 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2572 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2575 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2576 if (sbi->s_inodes_per_block == 0)
2578 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2579 sbi->s_inodes_per_block;
2580 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2582 sbi->s_mount_state = le16_to_cpu(es->s_state);
2583 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2584 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2586 for (i = 0; i < 4; i++)
2587 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2588 sbi->s_def_hash_version = es->s_def_hash_version;
2589 i = le32_to_cpu(es->s_flags);
2590 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2591 sbi->s_hash_unsigned = 3;
2592 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2593 #ifdef __CHAR_UNSIGNED__
2594 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2595 sbi->s_hash_unsigned = 3;
2597 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2602 if (sbi->s_blocks_per_group > blocksize * 8) {
2603 ext4_msg(sb, KERN_ERR,
2604 "#blocks per group too big: %lu",
2605 sbi->s_blocks_per_group);
2608 if (sbi->s_inodes_per_group > blocksize * 8) {
2609 ext4_msg(sb, KERN_ERR,
2610 "#inodes per group too big: %lu",
2611 sbi->s_inodes_per_group);
2616 * Test whether we have more sectors than will fit in sector_t,
2617 * and whether the max offset is addressable by the page cache.
2619 if ((ext4_blocks_count(es) >
2620 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2621 (ext4_blocks_count(es) >
2622 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2623 ext4_msg(sb, KERN_ERR, "filesystem"
2624 " too large to mount safely on this system");
2625 if (sizeof(sector_t) < 8)
2626 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2631 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2634 /* check blocks count against device size */
2635 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2636 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2637 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2638 "exceeds size of device (%llu blocks)",
2639 ext4_blocks_count(es), blocks_count);
2644 * It makes no sense for the first data block to be beyond the end
2645 * of the filesystem.
2647 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2648 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2649 "block %u is beyond end of filesystem (%llu)",
2650 le32_to_cpu(es->s_first_data_block),
2651 ext4_blocks_count(es));
2654 blocks_count = (ext4_blocks_count(es) -
2655 le32_to_cpu(es->s_first_data_block) +
2656 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2657 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2658 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2659 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2660 "(block count %llu, first data block %u, "
2661 "blocks per group %lu)", sbi->s_groups_count,
2662 ext4_blocks_count(es),
2663 le32_to_cpu(es->s_first_data_block),
2664 EXT4_BLOCKS_PER_GROUP(sb));
2667 sbi->s_groups_count = blocks_count;
2668 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2669 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2670 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2671 EXT4_DESC_PER_BLOCK(sb);
2672 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2674 if (sbi->s_group_desc == NULL) {
2675 ext4_msg(sb, KERN_ERR, "not enough memory");
2679 #ifdef CONFIG_PROC_FS
2681 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2684 bgl_lock_init(sbi->s_blockgroup_lock);
2686 for (i = 0; i < db_count; i++) {
2687 block = descriptor_loc(sb, logical_sb_block, i);
2688 sbi->s_group_desc[i] = sb_bread(sb, block);
2689 if (!sbi->s_group_desc[i]) {
2690 ext4_msg(sb, KERN_ERR,
2691 "can't read group descriptor %d", i);
2696 if (!ext4_check_descriptors(sb)) {
2697 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2700 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2701 if (!ext4_fill_flex_info(sb)) {
2702 ext4_msg(sb, KERN_ERR,
2703 "unable to initialize "
2704 "flex_bg meta info!");
2708 sbi->s_gdb_count = db_count;
2709 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2710 spin_lock_init(&sbi->s_next_gen_lock);
2712 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2713 ext4_count_free_blocks(sb));
2715 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2716 ext4_count_free_inodes(sb));
2719 err = percpu_counter_init(&sbi->s_dirs_counter,
2720 ext4_count_dirs(sb));
2723 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2726 ext4_msg(sb, KERN_ERR, "insufficient memory");
2730 sbi->s_stripe = ext4_get_stripe_size(sbi);
2731 sbi->s_max_writeback_mb_bump = 128;
2734 * set up enough so that it can read an inode
2736 if (!test_opt(sb, NOLOAD) &&
2737 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2738 sb->s_op = &ext4_sops;
2740 sb->s_op = &ext4_nojournal_sops;
2741 sb->s_export_op = &ext4_export_ops;
2742 sb->s_xattr = ext4_xattr_handlers;
2744 sb->s_qcop = &ext4_qctl_operations;
2745 sb->dq_op = &ext4_quota_operations;
2747 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2748 mutex_init(&sbi->s_orphan_lock);
2749 mutex_init(&sbi->s_resize_lock);
2753 needs_recovery = (es->s_last_orphan != 0 ||
2754 EXT4_HAS_INCOMPAT_FEATURE(sb,
2755 EXT4_FEATURE_INCOMPAT_RECOVER));
2758 * The first inode we look at is the journal inode. Don't try
2759 * root first: it may be modified in the journal!
2761 if (!test_opt(sb, NOLOAD) &&
2762 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2763 if (ext4_load_journal(sb, es, journal_devnum))
2765 if (!(sb->s_flags & MS_RDONLY) &&
2766 EXT4_SB(sb)->s_journal->j_failed_commit) {
2767 ext4_msg(sb, KERN_CRIT, "error: "
2768 "ext4_fill_super: Journal transaction "
2770 EXT4_SB(sb)->s_journal->j_failed_commit);
2771 if (test_opt(sb, ERRORS_RO)) {
2772 ext4_msg(sb, KERN_CRIT,
2773 "Mounting filesystem read-only");
2774 sb->s_flags |= MS_RDONLY;
2775 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2776 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2778 if (test_opt(sb, ERRORS_PANIC)) {
2779 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2780 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2781 ext4_commit_super(sb, 1);
2785 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2786 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2787 ext4_msg(sb, KERN_ERR, "required journal recovery "
2788 "suppressed and not mounted read-only");
2791 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2792 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2793 sbi->s_journal = NULL;
2798 if (ext4_blocks_count(es) > 0xffffffffULL &&
2799 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2800 JBD2_FEATURE_INCOMPAT_64BIT)) {
2801 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2805 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2806 jbd2_journal_set_features(sbi->s_journal,
2807 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2808 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2809 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2810 jbd2_journal_set_features(sbi->s_journal,
2811 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2812 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2813 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2815 jbd2_journal_clear_features(sbi->s_journal,
2816 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2817 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2820 /* We have now updated the journal if required, so we can
2821 * validate the data journaling mode. */
2822 switch (test_opt(sb, DATA_FLAGS)) {
2824 /* No mode set, assume a default based on the journal
2825 * capabilities: ORDERED_DATA if the journal can
2826 * cope, else JOURNAL_DATA
2828 if (jbd2_journal_check_available_features
2829 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2830 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2832 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2835 case EXT4_MOUNT_ORDERED_DATA:
2836 case EXT4_MOUNT_WRITEBACK_DATA:
2837 if (!jbd2_journal_check_available_features
2838 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2839 ext4_msg(sb, KERN_ERR, "Journal does not support "
2840 "requested data journaling mode");
2846 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2850 if (test_opt(sb, NOBH)) {
2851 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2852 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2853 "its supported only with writeback mode");
2854 clear_opt(sbi->s_mount_opt, NOBH);
2857 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2858 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2859 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2860 goto failed_mount_wq;
2864 * The jbd2_journal_load will have done any necessary log recovery,
2865 * so we can safely mount the rest of the filesystem now.
2868 root = ext4_iget(sb, EXT4_ROOT_INO);
2870 ext4_msg(sb, KERN_ERR, "get root inode failed");
2871 ret = PTR_ERR(root);
2874 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2876 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2879 sb->s_root = d_alloc_root(root);
2881 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2887 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2889 /* determine the minimum size of new large inodes, if present */
2890 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2891 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2892 EXT4_GOOD_OLD_INODE_SIZE;
2893 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2894 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2895 if (sbi->s_want_extra_isize <
2896 le16_to_cpu(es->s_want_extra_isize))
2897 sbi->s_want_extra_isize =
2898 le16_to_cpu(es->s_want_extra_isize);
2899 if (sbi->s_want_extra_isize <
2900 le16_to_cpu(es->s_min_extra_isize))
2901 sbi->s_want_extra_isize =
2902 le16_to_cpu(es->s_min_extra_isize);
2905 /* Check if enough inode space is available */
2906 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2907 sbi->s_inode_size) {
2908 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2909 EXT4_GOOD_OLD_INODE_SIZE;
2910 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2914 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2915 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2916 "requested data journaling mode");
2917 clear_opt(sbi->s_mount_opt, DELALLOC);
2918 } else if (test_opt(sb, DELALLOC))
2919 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2921 err = ext4_setup_system_zone(sb);
2923 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2924 "zone (%d)\n", err);
2929 err = ext4_mb_init(sb, needs_recovery);
2931 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2936 sbi->s_kobj.kset = ext4_kset;
2937 init_completion(&sbi->s_kobj_unregister);
2938 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2941 ext4_mb_release(sb);
2942 ext4_ext_release(sb);
2946 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2947 ext4_orphan_cleanup(sb, es);
2948 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2949 if (needs_recovery) {
2950 ext4_msg(sb, KERN_INFO, "recovery complete");
2951 ext4_mark_recovery_complete(sb, es);
2953 if (EXT4_SB(sb)->s_journal) {
2954 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2955 descr = " journalled data mode";
2956 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2957 descr = " ordered data mode";
2959 descr = " writeback data mode";
2961 descr = "out journal";
2963 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2970 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2974 ext4_msg(sb, KERN_ERR, "mount failed");
2975 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2977 ext4_release_system_zone(sb);
2978 if (sbi->s_journal) {
2979 jbd2_journal_destroy(sbi->s_journal);
2980 sbi->s_journal = NULL;
2983 if (sbi->s_flex_groups) {
2984 if (is_vmalloc_addr(sbi->s_flex_groups))
2985 vfree(sbi->s_flex_groups);
2987 kfree(sbi->s_flex_groups);
2989 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2990 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2991 percpu_counter_destroy(&sbi->s_dirs_counter);
2992 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2994 for (i = 0; i < db_count; i++)
2995 brelse(sbi->s_group_desc[i]);
2996 kfree(sbi->s_group_desc);
2999 remove_proc_entry(sb->s_id, ext4_proc_root);
3002 for (i = 0; i < MAXQUOTAS; i++)
3003 kfree(sbi->s_qf_names[i]);
3005 ext4_blkdev_remove(sbi);
3008 sb->s_fs_info = NULL;
3009 kfree(sbi->s_blockgroup_lock);
3016 * Setup any per-fs journal parameters now. We'll do this both on
3017 * initial mount, once the journal has been initialised but before we've
3018 * done any recovery; and again on any subsequent remount.
3020 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3022 struct ext4_sb_info *sbi = EXT4_SB(sb);
3024 journal->j_commit_interval = sbi->s_commit_interval;
3025 journal->j_min_batch_time = sbi->s_min_batch_time;
3026 journal->j_max_batch_time = sbi->s_max_batch_time;
3028 spin_lock(&journal->j_state_lock);
3029 if (test_opt(sb, BARRIER))
3030 journal->j_flags |= JBD2_BARRIER;
3032 journal->j_flags &= ~JBD2_BARRIER;
3033 if (test_opt(sb, DATA_ERR_ABORT))
3034 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3036 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3037 spin_unlock(&journal->j_state_lock);
3040 static journal_t *ext4_get_journal(struct super_block *sb,
3041 unsigned int journal_inum)
3043 struct inode *journal_inode;
3046 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3048 /* First, test for the existence of a valid inode on disk. Bad
3049 * things happen if we iget() an unused inode, as the subsequent
3050 * iput() will try to delete it. */
3052 journal_inode = ext4_iget(sb, journal_inum);
3053 if (IS_ERR(journal_inode)) {
3054 ext4_msg(sb, KERN_ERR, "no journal found");
3057 if (!journal_inode->i_nlink) {
3058 make_bad_inode(journal_inode);
3059 iput(journal_inode);
3060 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3064 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3065 journal_inode, journal_inode->i_size);
3066 if (!S_ISREG(journal_inode->i_mode)) {
3067 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3068 iput(journal_inode);
3072 journal = jbd2_journal_init_inode(journal_inode);
3074 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3075 iput(journal_inode);
3078 journal->j_private = sb;
3079 ext4_init_journal_params(sb, journal);
3083 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3086 struct buffer_head *bh;
3090 int hblock, blocksize;
3091 ext4_fsblk_t sb_block;
3092 unsigned long offset;
3093 struct ext4_super_block *es;
3094 struct block_device *bdev;
3096 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3098 bdev = ext4_blkdev_get(j_dev, sb);
3102 if (bd_claim(bdev, sb)) {
3103 ext4_msg(sb, KERN_ERR,
3104 "failed to claim external journal device");
3105 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3109 blocksize = sb->s_blocksize;
3110 hblock = bdev_logical_block_size(bdev);
3111 if (blocksize < hblock) {
3112 ext4_msg(sb, KERN_ERR,
3113 "blocksize too small for journal device");
3117 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3118 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3119 set_blocksize(bdev, blocksize);
3120 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3121 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3122 "external journal");
3126 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3127 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3128 !(le32_to_cpu(es->s_feature_incompat) &
3129 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3130 ext4_msg(sb, KERN_ERR, "external journal has "
3136 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3137 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3142 len = ext4_blocks_count(es);
3143 start = sb_block + 1;
3144 brelse(bh); /* we're done with the superblock */
3146 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3147 start, len, blocksize);
3149 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3152 journal->j_private = sb;
3153 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3154 wait_on_buffer(journal->j_sb_buffer);
3155 if (!buffer_uptodate(journal->j_sb_buffer)) {
3156 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3159 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3160 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3161 "user (unsupported) - %d",
3162 be32_to_cpu(journal->j_superblock->s_nr_users));
3165 EXT4_SB(sb)->journal_bdev = bdev;
3166 ext4_init_journal_params(sb, journal);
3170 jbd2_journal_destroy(journal);
3172 ext4_blkdev_put(bdev);
3176 static int ext4_load_journal(struct super_block *sb,
3177 struct ext4_super_block *es,
3178 unsigned long journal_devnum)
3181 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3184 int really_read_only;
3186 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3188 if (journal_devnum &&
3189 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3190 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3191 "numbers have changed");
3192 journal_dev = new_decode_dev(journal_devnum);
3194 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3196 really_read_only = bdev_read_only(sb->s_bdev);
3199 * Are we loading a blank journal or performing recovery after a
3200 * crash? For recovery, we need to check in advance whether we
3201 * can get read-write access to the device.
3203 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3204 if (sb->s_flags & MS_RDONLY) {
3205 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3206 "required on readonly filesystem");
3207 if (really_read_only) {
3208 ext4_msg(sb, KERN_ERR, "write access "
3209 "unavailable, cannot proceed");
3212 ext4_msg(sb, KERN_INFO, "write access will "
3213 "be enabled during recovery");
3217 if (journal_inum && journal_dev) {
3218 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3219 "and inode journals!");
3224 if (!(journal = ext4_get_journal(sb, journal_inum)))
3227 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3231 if (journal->j_flags & JBD2_BARRIER)
3232 ext4_msg(sb, KERN_INFO, "barriers enabled");
3234 ext4_msg(sb, KERN_INFO, "barriers disabled");
3236 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3237 err = jbd2_journal_update_format(journal);
3239 ext4_msg(sb, KERN_ERR, "error updating journal");
3240 jbd2_journal_destroy(journal);
3245 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3246 err = jbd2_journal_wipe(journal, !really_read_only);
3248 err = jbd2_journal_load(journal);
3251 ext4_msg(sb, KERN_ERR, "error loading journal");
3252 jbd2_journal_destroy(journal);
3256 EXT4_SB(sb)->s_journal = journal;
3257 ext4_clear_journal_err(sb, es);
3259 if (journal_devnum &&
3260 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3261 es->s_journal_dev = cpu_to_le32(journal_devnum);
3263 /* Make sure we flush the recovery flag to disk. */
3264 ext4_commit_super(sb, 1);
3270 static int ext4_commit_super(struct super_block *sb, int sync)
3272 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3273 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3278 if (buffer_write_io_error(sbh)) {
3280 * Oh, dear. A previous attempt to write the
3281 * superblock failed. This could happen because the
3282 * USB device was yanked out. Or it could happen to
3283 * be a transient write error and maybe the block will
3284 * be remapped. Nothing we can do but to retry the
3285 * write and hope for the best.
3287 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3288 "superblock detected");
3289 clear_buffer_write_io_error(sbh);
3290 set_buffer_uptodate(sbh);
3293 * If the file system is mounted read-only, don't update the
3294 * superblock write time. This avoids updating the superblock
3295 * write time when we are mounting the root file system
3296 * read/only but we need to replay the journal; at that point,
3297 * for people who are east of GMT and who make their clock
3298 * tick in localtime for Windows bug-for-bug compatibility,
3299 * the clock is set in the future, and this will cause e2fsck
3300 * to complain and force a full file system check.
3302 if (!(sb->s_flags & MS_RDONLY))
3303 es->s_wtime = cpu_to_le32(get_seconds());
3304 es->s_kbytes_written =
3305 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3306 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3307 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3308 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3309 &EXT4_SB(sb)->s_freeblocks_counter));
3310 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3311 &EXT4_SB(sb)->s_freeinodes_counter));
3313 BUFFER_TRACE(sbh, "marking dirty");
3314 mark_buffer_dirty(sbh);
3316 error = sync_dirty_buffer(sbh);
3320 error = buffer_write_io_error(sbh);
3322 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3324 clear_buffer_write_io_error(sbh);
3325 set_buffer_uptodate(sbh);
3332 * Have we just finished recovery? If so, and if we are mounting (or
3333 * remounting) the filesystem readonly, then we will end up with a
3334 * consistent fs on disk. Record that fact.
3336 static void ext4_mark_recovery_complete(struct super_block *sb,
3337 struct ext4_super_block *es)
3339 journal_t *journal = EXT4_SB(sb)->s_journal;
3341 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3342 BUG_ON(journal != NULL);
3345 jbd2_journal_lock_updates(journal);
3346 if (jbd2_journal_flush(journal) < 0)
3349 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3350 sb->s_flags & MS_RDONLY) {
3351 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3352 ext4_commit_super(sb, 1);
3356 jbd2_journal_unlock_updates(journal);
3360 * If we are mounting (or read-write remounting) a filesystem whose journal
3361 * has recorded an error from a previous lifetime, move that error to the
3362 * main filesystem now.
3364 static void ext4_clear_journal_err(struct super_block *sb,
3365 struct ext4_super_block *es)
3371 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3373 journal = EXT4_SB(sb)->s_journal;
3376 * Now check for any error status which may have been recorded in the
3377 * journal by a prior ext4_error() or ext4_abort()
3380 j_errno = jbd2_journal_errno(journal);
3384 errstr = ext4_decode_error(sb, j_errno, nbuf);
3385 ext4_warning(sb, __func__, "Filesystem error recorded "
3386 "from previous mount: %s", errstr);
3387 ext4_warning(sb, __func__, "Marking fs in need of "
3388 "filesystem check.");
3390 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3391 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3392 ext4_commit_super(sb, 1);
3394 jbd2_journal_clear_err(journal);
3399 * Force the running and committing transactions to commit,
3400 * and wait on the commit.
3402 int ext4_force_commit(struct super_block *sb)
3407 if (sb->s_flags & MS_RDONLY)
3410 journal = EXT4_SB(sb)->s_journal;
3412 ret = ext4_journal_force_commit(journal);
3417 static void ext4_write_super(struct super_block *sb)
3420 ext4_commit_super(sb, 1);
3424 static int ext4_sync_fs(struct super_block *sb, int wait)
3428 struct ext4_sb_info *sbi = EXT4_SB(sb);
3430 trace_ext4_sync_fs(sb, wait);
3431 flush_workqueue(sbi->dio_unwritten_wq);
3432 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3434 jbd2_log_wait_commit(sbi->s_journal, target);
3440 * LVM calls this function before a (read-only) snapshot is created. This
3441 * gives us a chance to flush the journal completely and mark the fs clean.
3443 static int ext4_freeze(struct super_block *sb)
3448 if (sb->s_flags & MS_RDONLY)
3451 journal = EXT4_SB(sb)->s_journal;
3453 /* Now we set up the journal barrier. */
3454 jbd2_journal_lock_updates(journal);
3457 * Don't clear the needs_recovery flag if we failed to flush
3460 error = jbd2_journal_flush(journal);
3463 jbd2_journal_unlock_updates(journal);
3467 /* Journal blocked and flushed, clear needs_recovery flag. */
3468 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3469 error = ext4_commit_super(sb, 1);
3476 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3477 * flag here, even though the filesystem is not technically dirty yet.
3479 static int ext4_unfreeze(struct super_block *sb)
3481 if (sb->s_flags & MS_RDONLY)
3485 /* Reset the needs_recovery flag before the fs is unlocked. */
3486 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3487 ext4_commit_super(sb, 1);
3489 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3493 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3495 struct ext4_super_block *es;
3496 struct ext4_sb_info *sbi = EXT4_SB(sb);
3497 ext4_fsblk_t n_blocks_count = 0;
3498 unsigned long old_sb_flags;
3499 struct ext4_mount_options old_opts;
3501 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3509 /* Store the original options */
3511 old_sb_flags = sb->s_flags;
3512 old_opts.s_mount_opt = sbi->s_mount_opt;
3513 old_opts.s_resuid = sbi->s_resuid;
3514 old_opts.s_resgid = sbi->s_resgid;
3515 old_opts.s_commit_interval = sbi->s_commit_interval;
3516 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3517 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3519 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3520 for (i = 0; i < MAXQUOTAS; i++)
3521 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3523 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3524 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3527 * Allow the "check" option to be passed as a remount option.
3529 if (!parse_options(data, sb, NULL, &journal_ioprio,
3530 &n_blocks_count, 1)) {
3535 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3536 ext4_abort(sb, __func__, "Abort forced by user");
3538 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3539 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3543 if (sbi->s_journal) {
3544 ext4_init_journal_params(sb, sbi->s_journal);
3545 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3548 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3549 n_blocks_count > ext4_blocks_count(es)) {
3550 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3555 if (*flags & MS_RDONLY) {
3557 * First of all, the unconditional stuff we have to do
3558 * to disable replay of the journal when we next remount
3560 sb->s_flags |= MS_RDONLY;
3563 * OK, test if we are remounting a valid rw partition
3564 * readonly, and if so set the rdonly flag and then
3565 * mark the partition as valid again.
3567 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3568 (sbi->s_mount_state & EXT4_VALID_FS))
3569 es->s_state = cpu_to_le16(sbi->s_mount_state);
3572 ext4_mark_recovery_complete(sb, es);
3574 /* Make sure we can mount this feature set readwrite */
3575 if (!ext4_feature_set_ok(sb, 0)) {
3580 * Make sure the group descriptor checksums
3581 * are sane. If they aren't, refuse to remount r/w.
3583 for (g = 0; g < sbi->s_groups_count; g++) {
3584 struct ext4_group_desc *gdp =
3585 ext4_get_group_desc(sb, g, NULL);
3587 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3588 ext4_msg(sb, KERN_ERR,
3589 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3590 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3591 le16_to_cpu(gdp->bg_checksum));
3598 * If we have an unprocessed orphan list hanging
3599 * around from a previously readonly bdev mount,
3600 * require a full umount/remount for now.
3602 if (es->s_last_orphan) {
3603 ext4_msg(sb, KERN_WARNING, "Couldn't "
3604 "remount RDWR because of unprocessed "
3605 "orphan inode list. Please "
3606 "umount/remount instead");
3612 * Mounting a RDONLY partition read-write, so reread
3613 * and store the current valid flag. (It may have
3614 * been changed by e2fsck since we originally mounted
3618 ext4_clear_journal_err(sb, es);
3619 sbi->s_mount_state = le16_to_cpu(es->s_state);
3620 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3622 if (!ext4_setup_super(sb, es, 0))
3623 sb->s_flags &= ~MS_RDONLY;
3626 ext4_setup_system_zone(sb);
3627 if (sbi->s_journal == NULL)
3628 ext4_commit_super(sb, 1);
3631 /* Release old quota file names */
3632 for (i = 0; i < MAXQUOTAS; i++)
3633 if (old_opts.s_qf_names[i] &&
3634 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3635 kfree(old_opts.s_qf_names[i]);
3642 sb->s_flags = old_sb_flags;
3643 sbi->s_mount_opt = old_opts.s_mount_opt;
3644 sbi->s_resuid = old_opts.s_resuid;
3645 sbi->s_resgid = old_opts.s_resgid;
3646 sbi->s_commit_interval = old_opts.s_commit_interval;
3647 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3648 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3650 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3651 for (i = 0; i < MAXQUOTAS; i++) {
3652 if (sbi->s_qf_names[i] &&
3653 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3654 kfree(sbi->s_qf_names[i]);
3655 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3663 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3665 struct super_block *sb = dentry->d_sb;
3666 struct ext4_sb_info *sbi = EXT4_SB(sb);
3667 struct ext4_super_block *es = sbi->s_es;
3670 if (test_opt(sb, MINIX_DF)) {
3671 sbi->s_overhead_last = 0;
3672 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3673 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3674 ext4_fsblk_t overhead = 0;
3677 * Compute the overhead (FS structures). This is constant
3678 * for a given filesystem unless the number of block groups
3679 * changes so we cache the previous value until it does.
3683 * All of the blocks before first_data_block are
3686 overhead = le32_to_cpu(es->s_first_data_block);
3689 * Add the overhead attributed to the superblock and
3690 * block group descriptors. If the sparse superblocks
3691 * feature is turned on, then not all groups have this.
3693 for (i = 0; i < ngroups; i++) {
3694 overhead += ext4_bg_has_super(sb, i) +
3695 ext4_bg_num_gdb(sb, i);
3700 * Every block group has an inode bitmap, a block
3701 * bitmap, and an inode table.
3703 overhead += ngroups * (2 + sbi->s_itb_per_group);
3704 sbi->s_overhead_last = overhead;
3706 sbi->s_blocks_last = ext4_blocks_count(es);
3709 buf->f_type = EXT4_SUPER_MAGIC;
3710 buf->f_bsize = sb->s_blocksize;
3711 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3712 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3713 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3714 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3715 if (buf->f_bfree < ext4_r_blocks_count(es))
3717 buf->f_files = le32_to_cpu(es->s_inodes_count);
3718 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3719 buf->f_namelen = EXT4_NAME_LEN;
3720 fsid = le64_to_cpup((void *)es->s_uuid) ^
3721 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3722 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3723 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3728 /* Helper function for writing quotas on sync - we need to start transaction
3729 * before quota file is locked for write. Otherwise the are possible deadlocks:
3730 * Process 1 Process 2
3731 * ext4_create() quota_sync()
3732 * jbd2_journal_start() write_dquot()
3733 * vfs_dq_init() down(dqio_mutex)
3734 * down(dqio_mutex) jbd2_journal_start()
3740 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3742 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3745 static int ext4_write_dquot(struct dquot *dquot)
3749 struct inode *inode;
3751 inode = dquot_to_inode(dquot);
3752 handle = ext4_journal_start(inode,
3753 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3755 return PTR_ERR(handle);
3756 ret = dquot_commit(dquot);
3757 err = ext4_journal_stop(handle);
3763 static int ext4_acquire_dquot(struct dquot *dquot)
3768 handle = ext4_journal_start(dquot_to_inode(dquot),
3769 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3771 return PTR_ERR(handle);
3772 ret = dquot_acquire(dquot);
3773 err = ext4_journal_stop(handle);
3779 static int ext4_release_dquot(struct dquot *dquot)
3784 handle = ext4_journal_start(dquot_to_inode(dquot),
3785 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3786 if (IS_ERR(handle)) {
3787 /* Release dquot anyway to avoid endless cycle in dqput() */
3788 dquot_release(dquot);
3789 return PTR_ERR(handle);
3791 ret = dquot_release(dquot);
3792 err = ext4_journal_stop(handle);
3798 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3800 /* Are we journaling quotas? */
3801 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3802 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3803 dquot_mark_dquot_dirty(dquot);
3804 return ext4_write_dquot(dquot);
3806 return dquot_mark_dquot_dirty(dquot);
3810 static int ext4_write_info(struct super_block *sb, int type)
3815 /* Data block + inode block */
3816 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3818 return PTR_ERR(handle);
3819 ret = dquot_commit_info(sb, type);
3820 err = ext4_journal_stop(handle);
3827 * Turn on quotas during mount time - we need to find
3828 * the quota file and such...
3830 static int ext4_quota_on_mount(struct super_block *sb, int type)
3832 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3833 EXT4_SB(sb)->s_jquota_fmt, type);
3837 * Standard function to be called on quota_on
3839 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3840 char *name, int remount)
3845 if (!test_opt(sb, QUOTA))
3847 /* When remounting, no checks are needed and in fact, name is NULL */
3849 return vfs_quota_on(sb, type, format_id, name, remount);
3851 err = kern_path(name, LOOKUP_FOLLOW, &path);
3855 /* Quotafile not on the same filesystem? */
3856 if (path.mnt->mnt_sb != sb) {
3860 /* Journaling quota? */
3861 if (EXT4_SB(sb)->s_qf_names[type]) {
3862 /* Quotafile not in fs root? */
3863 if (path.dentry->d_parent != sb->s_root)
3864 ext4_msg(sb, KERN_WARNING,
3865 "Quota file not on filesystem root. "
3866 "Journaled quota will not work");
3870 * When we journal data on quota file, we have to flush journal to see
3871 * all updates to the file when we bypass pagecache...
3873 if (EXT4_SB(sb)->s_journal &&
3874 ext4_should_journal_data(path.dentry->d_inode)) {
3876 * We don't need to lock updates but journal_flush() could
3877 * otherwise be livelocked...
3879 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3880 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3881 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3888 err = vfs_quota_on_path(sb, type, format_id, &path);
3893 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3894 * acquiring the locks... As quota files are never truncated and quota code
3895 * itself serializes the operations (and noone else should touch the files)
3896 * we don't have to be afraid of races */
3897 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3898 size_t len, loff_t off)
3900 struct inode *inode = sb_dqopt(sb)->files[type];
3901 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3903 int offset = off & (sb->s_blocksize - 1);
3906 struct buffer_head *bh;
3907 loff_t i_size = i_size_read(inode);
3911 if (off+len > i_size)
3914 while (toread > 0) {
3915 tocopy = sb->s_blocksize - offset < toread ?
3916 sb->s_blocksize - offset : toread;
3917 bh = ext4_bread(NULL, inode, blk, 0, &err);
3920 if (!bh) /* A hole? */
3921 memset(data, 0, tocopy);
3923 memcpy(data, bh->b_data+offset, tocopy);
3933 /* Write to quotafile (we know the transaction is already started and has
3934 * enough credits) */
3935 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3936 const char *data, size_t len, loff_t off)
3938 struct inode *inode = sb_dqopt(sb)->files[type];
3939 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3941 int offset = off & (sb->s_blocksize - 1);
3943 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3944 size_t towrite = len;
3945 struct buffer_head *bh;
3946 handle_t *handle = journal_current_handle();
3948 if (EXT4_SB(sb)->s_journal && !handle) {
3949 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3950 " cancelled because transaction is not started",
3951 (unsigned long long)off, (unsigned long long)len);
3954 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3955 while (towrite > 0) {
3956 tocopy = sb->s_blocksize - offset < towrite ?
3957 sb->s_blocksize - offset : towrite;
3958 bh = ext4_bread(handle, inode, blk, 1, &err);
3961 if (journal_quota) {
3962 err = ext4_journal_get_write_access(handle, bh);
3969 memcpy(bh->b_data+offset, data, tocopy);
3970 flush_dcache_page(bh->b_page);
3973 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3975 /* Always do at least ordered writes for quotas */
3976 err = ext4_jbd2_file_inode(handle, inode);
3977 mark_buffer_dirty(bh);
3988 if (len == towrite) {
3989 mutex_unlock(&inode->i_mutex);
3992 if (inode->i_size < off+len-towrite) {
3993 i_size_write(inode, off+len-towrite);
3994 EXT4_I(inode)->i_disksize = inode->i_size;
3996 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3997 ext4_mark_inode_dirty(handle, inode);
3998 mutex_unlock(&inode->i_mutex);
3999 return len - towrite;
4004 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4005 const char *dev_name, void *data, struct vfsmount *mnt)
4007 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4010 static struct file_system_type ext4_fs_type = {
4011 .owner = THIS_MODULE,
4013 .get_sb = ext4_get_sb,
4014 .kill_sb = kill_block_super,
4015 .fs_flags = FS_REQUIRES_DEV,
4018 #ifdef CONFIG_EXT4DEV_COMPAT
4019 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
4020 const char *dev_name, void *data,struct vfsmount *mnt)
4022 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
4023 "to mount using ext4\n", dev_name);
4024 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
4025 "will go away by 2.6.31\n", dev_name);
4026 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4029 static struct file_system_type ext4dev_fs_type = {
4030 .owner = THIS_MODULE,
4032 .get_sb = ext4dev_get_sb,
4033 .kill_sb = kill_block_super,
4034 .fs_flags = FS_REQUIRES_DEV,
4036 MODULE_ALIAS("ext4dev");
4039 static int __init init_ext4_fs(void)
4043 err = init_ext4_system_zone();
4046 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4049 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4050 err = init_ext4_mballoc();
4054 err = init_ext4_xattr();
4057 err = init_inodecache();
4060 err = register_filesystem(&ext4_fs_type);
4063 #ifdef CONFIG_EXT4DEV_COMPAT
4064 err = register_filesystem(&ext4dev_fs_type);
4066 unregister_filesystem(&ext4_fs_type);
4072 destroy_inodecache();
4076 exit_ext4_mballoc();
4078 remove_proc_entry("fs/ext4", NULL);
4079 kset_unregister(ext4_kset);
4081 exit_ext4_system_zone();
4085 static void __exit exit_ext4_fs(void)
4087 unregister_filesystem(&ext4_fs_type);
4088 #ifdef CONFIG_EXT4DEV_COMPAT
4089 unregister_filesystem(&ext4dev_fs_type);
4091 destroy_inodecache();
4093 exit_ext4_mballoc();
4094 remove_proc_entry("fs/ext4", NULL);
4095 kset_unregister(ext4_kset);
4096 exit_ext4_system_zone();
4099 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4100 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4101 MODULE_LICENSE("GPL");
4102 module_init(init_ext4_fs)
4103 module_exit(exit_ext4_fs)