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 ext4_release_system_zone(sb);
615 ext4_ext_release(sb);
616 ext4_xattr_put_super(sb);
617 if (sbi->s_journal) {
618 err = jbd2_journal_destroy(sbi->s_journal);
619 sbi->s_journal = NULL;
621 ext4_abort(sb, __func__,
622 "Couldn't clean up the journal");
624 if (!(sb->s_flags & MS_RDONLY)) {
625 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
626 es->s_state = cpu_to_le16(sbi->s_mount_state);
627 ext4_commit_super(sb, 1);
630 remove_proc_entry(sb->s_id, ext4_proc_root);
632 kobject_del(&sbi->s_kobj);
634 for (i = 0; i < sbi->s_gdb_count; i++)
635 brelse(sbi->s_group_desc[i]);
636 kfree(sbi->s_group_desc);
637 if (is_vmalloc_addr(sbi->s_flex_groups))
638 vfree(sbi->s_flex_groups);
640 kfree(sbi->s_flex_groups);
641 percpu_counter_destroy(&sbi->s_freeblocks_counter);
642 percpu_counter_destroy(&sbi->s_freeinodes_counter);
643 percpu_counter_destroy(&sbi->s_dirs_counter);
644 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
647 for (i = 0; i < MAXQUOTAS; i++)
648 kfree(sbi->s_qf_names[i]);
651 /* Debugging code just in case the in-memory inode orphan list
652 * isn't empty. The on-disk one can be non-empty if we've
653 * detected an error and taken the fs readonly, but the
654 * in-memory list had better be clean by this point. */
655 if (!list_empty(&sbi->s_orphan))
656 dump_orphan_list(sb, sbi);
657 J_ASSERT(list_empty(&sbi->s_orphan));
659 invalidate_bdev(sb->s_bdev);
660 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
662 * Invalidate the journal device's buffers. We don't want them
663 * floating about in memory - the physical journal device may
664 * hotswapped, and it breaks the `ro-after' testing code.
666 sync_blockdev(sbi->journal_bdev);
667 invalidate_bdev(sbi->journal_bdev);
668 ext4_blkdev_remove(sbi);
670 sb->s_fs_info = NULL;
672 * Now that we are completely done shutting down the
673 * superblock, we need to actually destroy the kobject.
677 kobject_put(&sbi->s_kobj);
678 wait_for_completion(&sbi->s_kobj_unregister);
679 kfree(sbi->s_blockgroup_lock);
683 static struct kmem_cache *ext4_inode_cachep;
686 * Called inside transaction, so use GFP_NOFS
688 static struct inode *ext4_alloc_inode(struct super_block *sb)
690 struct ext4_inode_info *ei;
692 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
696 ei->vfs_inode.i_version = 1;
697 ei->vfs_inode.i_data.writeback_index = 0;
698 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
699 INIT_LIST_HEAD(&ei->i_prealloc_list);
700 spin_lock_init(&ei->i_prealloc_lock);
702 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
703 * therefore it can be null here. Don't check it, just initialize
706 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
707 ei->i_reserved_data_blocks = 0;
708 ei->i_reserved_meta_blocks = 0;
709 ei->i_allocated_meta_blocks = 0;
710 ei->i_delalloc_reserved_flag = 0;
711 spin_lock_init(&(ei->i_block_reservation_lock));
712 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
713 ei->cur_aio_dio = NULL;
715 return &ei->vfs_inode;
718 static void ext4_destroy_inode(struct inode *inode)
720 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
721 ext4_msg(inode->i_sb, KERN_ERR,
722 "Inode %lu (%p): orphan list check failed!",
723 inode->i_ino, EXT4_I(inode));
724 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
725 EXT4_I(inode), sizeof(struct ext4_inode_info),
729 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
732 static void init_once(void *foo)
734 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
736 INIT_LIST_HEAD(&ei->i_orphan);
737 #ifdef CONFIG_EXT4_FS_XATTR
738 init_rwsem(&ei->xattr_sem);
740 init_rwsem(&ei->i_data_sem);
741 inode_init_once(&ei->vfs_inode);
744 static int init_inodecache(void)
746 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
747 sizeof(struct ext4_inode_info),
748 0, (SLAB_RECLAIM_ACCOUNT|
751 if (ext4_inode_cachep == NULL)
756 static void destroy_inodecache(void)
758 kmem_cache_destroy(ext4_inode_cachep);
761 static void ext4_clear_inode(struct inode *inode)
763 ext4_discard_preallocations(inode);
764 if (EXT4_JOURNAL(inode))
765 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
766 &EXT4_I(inode)->jinode);
769 static inline void ext4_show_quota_options(struct seq_file *seq,
770 struct super_block *sb)
772 #if defined(CONFIG_QUOTA)
773 struct ext4_sb_info *sbi = EXT4_SB(sb);
775 if (sbi->s_jquota_fmt)
776 seq_printf(seq, ",jqfmt=%s",
777 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
779 if (sbi->s_qf_names[USRQUOTA])
780 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
782 if (sbi->s_qf_names[GRPQUOTA])
783 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
785 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
786 seq_puts(seq, ",usrquota");
788 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
789 seq_puts(seq, ",grpquota");
795 * - it's set to a non-default value OR
796 * - if the per-sb default is different from the global default
798 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
801 unsigned long def_mount_opts;
802 struct super_block *sb = vfs->mnt_sb;
803 struct ext4_sb_info *sbi = EXT4_SB(sb);
804 struct ext4_super_block *es = sbi->s_es;
806 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
807 def_errors = le16_to_cpu(es->s_errors);
809 if (sbi->s_sb_block != 1)
810 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
811 if (test_opt(sb, MINIX_DF))
812 seq_puts(seq, ",minixdf");
813 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
814 seq_puts(seq, ",grpid");
815 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
816 seq_puts(seq, ",nogrpid");
817 if (sbi->s_resuid != EXT4_DEF_RESUID ||
818 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
819 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
821 if (sbi->s_resgid != EXT4_DEF_RESGID ||
822 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
823 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
825 if (test_opt(sb, ERRORS_RO)) {
826 if (def_errors == EXT4_ERRORS_PANIC ||
827 def_errors == EXT4_ERRORS_CONTINUE) {
828 seq_puts(seq, ",errors=remount-ro");
831 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
832 seq_puts(seq, ",errors=continue");
833 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
834 seq_puts(seq, ",errors=panic");
835 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
836 seq_puts(seq, ",nouid32");
837 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
838 seq_puts(seq, ",debug");
839 if (test_opt(sb, OLDALLOC))
840 seq_puts(seq, ",oldalloc");
841 #ifdef CONFIG_EXT4_FS_XATTR
842 if (test_opt(sb, XATTR_USER) &&
843 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
844 seq_puts(seq, ",user_xattr");
845 if (!test_opt(sb, XATTR_USER) &&
846 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
847 seq_puts(seq, ",nouser_xattr");
850 #ifdef CONFIG_EXT4_FS_POSIX_ACL
851 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
852 seq_puts(seq, ",acl");
853 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
854 seq_puts(seq, ",noacl");
856 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
857 seq_printf(seq, ",commit=%u",
858 (unsigned) (sbi->s_commit_interval / HZ));
860 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
861 seq_printf(seq, ",min_batch_time=%u",
862 (unsigned) sbi->s_min_batch_time);
864 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
865 seq_printf(seq, ",max_batch_time=%u",
866 (unsigned) sbi->s_min_batch_time);
870 * We're changing the default of barrier mount option, so
871 * let's always display its mount state so it's clear what its
874 seq_puts(seq, ",barrier=");
875 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
876 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
877 seq_puts(seq, ",journal_async_commit");
878 if (test_opt(sb, NOBH))
879 seq_puts(seq, ",nobh");
880 if (test_opt(sb, I_VERSION))
881 seq_puts(seq, ",i_version");
882 if (!test_opt(sb, DELALLOC))
883 seq_puts(seq, ",nodelalloc");
887 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
889 * journal mode get enabled in different ways
890 * So just print the value even if we didn't specify it
892 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
893 seq_puts(seq, ",data=journal");
894 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
895 seq_puts(seq, ",data=ordered");
896 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
897 seq_puts(seq, ",data=writeback");
899 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
900 seq_printf(seq, ",inode_readahead_blks=%u",
901 sbi->s_inode_readahead_blks);
903 if (test_opt(sb, DATA_ERR_ABORT))
904 seq_puts(seq, ",data_err=abort");
906 if (test_opt(sb, NO_AUTO_DA_ALLOC))
907 seq_puts(seq, ",noauto_da_alloc");
909 ext4_show_quota_options(seq, sb);
914 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
915 u64 ino, u32 generation)
919 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
920 return ERR_PTR(-ESTALE);
921 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
922 return ERR_PTR(-ESTALE);
924 /* iget isn't really right if the inode is currently unallocated!!
926 * ext4_read_inode will return a bad_inode if the inode had been
927 * deleted, so we should be safe.
929 * Currently we don't know the generation for parent directory, so
930 * a generation of 0 means "accept any"
932 inode = ext4_iget(sb, ino);
934 return ERR_CAST(inode);
935 if (generation && inode->i_generation != generation) {
937 return ERR_PTR(-ESTALE);
943 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
944 int fh_len, int fh_type)
946 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
950 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
951 int fh_len, int fh_type)
953 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
958 * Try to release metadata pages (indirect blocks, directories) which are
959 * mapped via the block device. Since these pages could have journal heads
960 * which would prevent try_to_free_buffers() from freeing them, we must use
961 * jbd2 layer's try_to_free_buffers() function to release them.
963 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
966 journal_t *journal = EXT4_SB(sb)->s_journal;
968 WARN_ON(PageChecked(page));
969 if (!page_has_buffers(page))
972 return jbd2_journal_try_to_free_buffers(journal, page,
974 return try_to_free_buffers(page);
978 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
979 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
981 static int ext4_write_dquot(struct dquot *dquot);
982 static int ext4_acquire_dquot(struct dquot *dquot);
983 static int ext4_release_dquot(struct dquot *dquot);
984 static int ext4_mark_dquot_dirty(struct dquot *dquot);
985 static int ext4_write_info(struct super_block *sb, int type);
986 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
987 char *path, int remount);
988 static int ext4_quota_on_mount(struct super_block *sb, int type);
989 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
990 size_t len, loff_t off);
991 static ssize_t ext4_quota_write(struct super_block *sb, int type,
992 const char *data, size_t len, loff_t off);
994 static const struct dquot_operations ext4_quota_operations = {
995 .initialize = dquot_initialize,
997 .alloc_space = dquot_alloc_space,
998 .reserve_space = dquot_reserve_space,
999 .claim_space = dquot_claim_space,
1000 .release_rsv = dquot_release_reserved_space,
1001 .get_reserved_space = ext4_get_reserved_space,
1002 .alloc_inode = dquot_alloc_inode,
1003 .free_space = dquot_free_space,
1004 .free_inode = dquot_free_inode,
1005 .transfer = dquot_transfer,
1006 .write_dquot = ext4_write_dquot,
1007 .acquire_dquot = ext4_acquire_dquot,
1008 .release_dquot = ext4_release_dquot,
1009 .mark_dirty = ext4_mark_dquot_dirty,
1010 .write_info = ext4_write_info,
1011 .alloc_dquot = dquot_alloc,
1012 .destroy_dquot = dquot_destroy,
1015 static const struct quotactl_ops ext4_qctl_operations = {
1016 .quota_on = ext4_quota_on,
1017 .quota_off = vfs_quota_off,
1018 .quota_sync = vfs_quota_sync,
1019 .get_info = vfs_get_dqinfo,
1020 .set_info = vfs_set_dqinfo,
1021 .get_dqblk = vfs_get_dqblk,
1022 .set_dqblk = vfs_set_dqblk
1026 static const struct super_operations ext4_sops = {
1027 .alloc_inode = ext4_alloc_inode,
1028 .destroy_inode = ext4_destroy_inode,
1029 .write_inode = ext4_write_inode,
1030 .dirty_inode = ext4_dirty_inode,
1031 .delete_inode = ext4_delete_inode,
1032 .put_super = ext4_put_super,
1033 .sync_fs = ext4_sync_fs,
1034 .freeze_fs = ext4_freeze,
1035 .unfreeze_fs = ext4_unfreeze,
1036 .statfs = ext4_statfs,
1037 .remount_fs = ext4_remount,
1038 .clear_inode = ext4_clear_inode,
1039 .show_options = ext4_show_options,
1041 .quota_read = ext4_quota_read,
1042 .quota_write = ext4_quota_write,
1044 .bdev_try_to_free_page = bdev_try_to_free_page,
1047 static const struct super_operations ext4_nojournal_sops = {
1048 .alloc_inode = ext4_alloc_inode,
1049 .destroy_inode = ext4_destroy_inode,
1050 .write_inode = ext4_write_inode,
1051 .dirty_inode = ext4_dirty_inode,
1052 .delete_inode = ext4_delete_inode,
1053 .write_super = ext4_write_super,
1054 .put_super = ext4_put_super,
1055 .statfs = ext4_statfs,
1056 .remount_fs = ext4_remount,
1057 .clear_inode = ext4_clear_inode,
1058 .show_options = ext4_show_options,
1060 .quota_read = ext4_quota_read,
1061 .quota_write = ext4_quota_write,
1063 .bdev_try_to_free_page = bdev_try_to_free_page,
1066 static const struct export_operations ext4_export_ops = {
1067 .fh_to_dentry = ext4_fh_to_dentry,
1068 .fh_to_parent = ext4_fh_to_parent,
1069 .get_parent = ext4_get_parent,
1073 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1074 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1075 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1076 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1077 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1078 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1079 Opt_journal_update, Opt_journal_dev,
1080 Opt_journal_checksum, Opt_journal_async_commit,
1081 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1082 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1083 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1084 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1085 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1086 Opt_usrquota, Opt_grpquota, Opt_i_version,
1087 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1088 Opt_block_validity, Opt_noblock_validity,
1089 Opt_inode_readahead_blks, Opt_journal_ioprio
1092 static const match_table_t tokens = {
1093 {Opt_bsd_df, "bsddf"},
1094 {Opt_minix_df, "minixdf"},
1095 {Opt_grpid, "grpid"},
1096 {Opt_grpid, "bsdgroups"},
1097 {Opt_nogrpid, "nogrpid"},
1098 {Opt_nogrpid, "sysvgroups"},
1099 {Opt_resgid, "resgid=%u"},
1100 {Opt_resuid, "resuid=%u"},
1102 {Opt_err_cont, "errors=continue"},
1103 {Opt_err_panic, "errors=panic"},
1104 {Opt_err_ro, "errors=remount-ro"},
1105 {Opt_nouid32, "nouid32"},
1106 {Opt_debug, "debug"},
1107 {Opt_oldalloc, "oldalloc"},
1108 {Opt_orlov, "orlov"},
1109 {Opt_user_xattr, "user_xattr"},
1110 {Opt_nouser_xattr, "nouser_xattr"},
1112 {Opt_noacl, "noacl"},
1113 {Opt_noload, "noload"},
1116 {Opt_commit, "commit=%u"},
1117 {Opt_min_batch_time, "min_batch_time=%u"},
1118 {Opt_max_batch_time, "max_batch_time=%u"},
1119 {Opt_journal_update, "journal=update"},
1120 {Opt_journal_dev, "journal_dev=%u"},
1121 {Opt_journal_checksum, "journal_checksum"},
1122 {Opt_journal_async_commit, "journal_async_commit"},
1123 {Opt_abort, "abort"},
1124 {Opt_data_journal, "data=journal"},
1125 {Opt_data_ordered, "data=ordered"},
1126 {Opt_data_writeback, "data=writeback"},
1127 {Opt_data_err_abort, "data_err=abort"},
1128 {Opt_data_err_ignore, "data_err=ignore"},
1129 {Opt_mb_history_length, "mb_history_length=%u"},
1130 {Opt_offusrjquota, "usrjquota="},
1131 {Opt_usrjquota, "usrjquota=%s"},
1132 {Opt_offgrpjquota, "grpjquota="},
1133 {Opt_grpjquota, "grpjquota=%s"},
1134 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1135 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1136 {Opt_grpquota, "grpquota"},
1137 {Opt_noquota, "noquota"},
1138 {Opt_quota, "quota"},
1139 {Opt_usrquota, "usrquota"},
1140 {Opt_barrier, "barrier=%u"},
1141 {Opt_barrier, "barrier"},
1142 {Opt_nobarrier, "nobarrier"},
1143 {Opt_i_version, "i_version"},
1144 {Opt_stripe, "stripe=%u"},
1145 {Opt_resize, "resize"},
1146 {Opt_delalloc, "delalloc"},
1147 {Opt_nodelalloc, "nodelalloc"},
1148 {Opt_block_validity, "block_validity"},
1149 {Opt_noblock_validity, "noblock_validity"},
1150 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1151 {Opt_journal_ioprio, "journal_ioprio=%u"},
1152 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1153 {Opt_auto_da_alloc, "auto_da_alloc"},
1154 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1158 static ext4_fsblk_t get_sb_block(void **data)
1160 ext4_fsblk_t sb_block;
1161 char *options = (char *) *data;
1163 if (!options || strncmp(options, "sb=", 3) != 0)
1164 return 1; /* Default location */
1167 /* TODO: use simple_strtoll with >32bit ext4 */
1168 sb_block = simple_strtoul(options, &options, 0);
1169 if (*options && *options != ',') {
1170 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1174 if (*options == ',')
1176 *data = (void *) options;
1181 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1183 static int parse_options(char *options, struct super_block *sb,
1184 unsigned long *journal_devnum,
1185 unsigned int *journal_ioprio,
1186 ext4_fsblk_t *n_blocks_count, int is_remount)
1188 struct ext4_sb_info *sbi = EXT4_SB(sb);
1190 substring_t args[MAX_OPT_ARGS];
1201 while ((p = strsep(&options, ",")) != NULL) {
1206 token = match_token(p, tokens, args);
1209 clear_opt(sbi->s_mount_opt, MINIX_DF);
1212 set_opt(sbi->s_mount_opt, MINIX_DF);
1215 set_opt(sbi->s_mount_opt, GRPID);
1218 clear_opt(sbi->s_mount_opt, GRPID);
1221 if (match_int(&args[0], &option))
1223 sbi->s_resuid = option;
1226 if (match_int(&args[0], &option))
1228 sbi->s_resgid = option;
1231 /* handled by get_sb_block() instead of here */
1232 /* *sb_block = match_int(&args[0]); */
1235 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1236 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1237 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1240 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1241 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1242 set_opt(sbi->s_mount_opt, ERRORS_RO);
1245 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1246 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1247 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1250 set_opt(sbi->s_mount_opt, NO_UID32);
1253 set_opt(sbi->s_mount_opt, DEBUG);
1256 set_opt(sbi->s_mount_opt, OLDALLOC);
1259 clear_opt(sbi->s_mount_opt, OLDALLOC);
1261 #ifdef CONFIG_EXT4_FS_XATTR
1262 case Opt_user_xattr:
1263 set_opt(sbi->s_mount_opt, XATTR_USER);
1265 case Opt_nouser_xattr:
1266 clear_opt(sbi->s_mount_opt, XATTR_USER);
1269 case Opt_user_xattr:
1270 case Opt_nouser_xattr:
1271 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1274 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1276 set_opt(sbi->s_mount_opt, POSIX_ACL);
1279 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1284 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1287 case Opt_journal_update:
1289 /* Eventually we will want to be able to create
1290 a journal file here. For now, only allow the
1291 user to specify an existing inode to be the
1294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
1298 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1300 case Opt_journal_dev:
1302 ext4_msg(sb, KERN_ERR,
1303 "Cannot specify journal on remount");
1306 if (match_int(&args[0], &option))
1308 *journal_devnum = option;
1310 case Opt_journal_checksum:
1311 break; /* Kept for backwards compatibility */
1312 case Opt_journal_async_commit:
1313 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1316 set_opt(sbi->s_mount_opt, NOLOAD);
1319 if (match_int(&args[0], &option))
1324 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1325 sbi->s_commit_interval = HZ * option;
1327 case Opt_max_batch_time:
1328 if (match_int(&args[0], &option))
1333 option = EXT4_DEF_MAX_BATCH_TIME;
1334 sbi->s_max_batch_time = option;
1336 case Opt_min_batch_time:
1337 if (match_int(&args[0], &option))
1341 sbi->s_min_batch_time = option;
1343 case Opt_data_journal:
1344 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1346 case Opt_data_ordered:
1347 data_opt = EXT4_MOUNT_ORDERED_DATA;
1349 case Opt_data_writeback:
1350 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1353 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1355 ext4_msg(sb, KERN_ERR,
1356 "Cannot change data mode on remount");
1360 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1361 sbi->s_mount_opt |= data_opt;
1364 case Opt_data_err_abort:
1365 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1367 case Opt_data_err_ignore:
1368 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1370 case Opt_mb_history_length:
1371 if (match_int(&args[0], &option))
1375 sbi->s_mb_history_max = option;
1384 if (sb_any_quota_loaded(sb) &&
1385 !sbi->s_qf_names[qtype]) {
1386 ext4_msg(sb, KERN_ERR,
1387 "Cannot change journaled "
1388 "quota options when quota turned on");
1391 qname = match_strdup(&args[0]);
1393 ext4_msg(sb, KERN_ERR,
1394 "Not enough memory for "
1395 "storing quotafile name");
1398 if (sbi->s_qf_names[qtype] &&
1399 strcmp(sbi->s_qf_names[qtype], qname)) {
1400 ext4_msg(sb, KERN_ERR,
1401 "%s quota file already "
1402 "specified", QTYPE2NAME(qtype));
1406 sbi->s_qf_names[qtype] = qname;
1407 if (strchr(sbi->s_qf_names[qtype], '/')) {
1408 ext4_msg(sb, KERN_ERR,
1409 "quotafile must be on "
1411 kfree(sbi->s_qf_names[qtype]);
1412 sbi->s_qf_names[qtype] = NULL;
1415 set_opt(sbi->s_mount_opt, QUOTA);
1417 case Opt_offusrjquota:
1420 case Opt_offgrpjquota:
1423 if (sb_any_quota_loaded(sb) &&
1424 sbi->s_qf_names[qtype]) {
1425 ext4_msg(sb, KERN_ERR, "Cannot change "
1426 "journaled quota options when "
1431 * The space will be released later when all options
1432 * are confirmed to be correct
1434 sbi->s_qf_names[qtype] = NULL;
1436 case Opt_jqfmt_vfsold:
1437 qfmt = QFMT_VFS_OLD;
1439 case Opt_jqfmt_vfsv0:
1442 if (sb_any_quota_loaded(sb) &&
1443 sbi->s_jquota_fmt != qfmt) {
1444 ext4_msg(sb, KERN_ERR, "Cannot change "
1445 "journaled quota options when "
1449 sbi->s_jquota_fmt = qfmt;
1453 set_opt(sbi->s_mount_opt, QUOTA);
1454 set_opt(sbi->s_mount_opt, USRQUOTA);
1457 set_opt(sbi->s_mount_opt, QUOTA);
1458 set_opt(sbi->s_mount_opt, GRPQUOTA);
1461 if (sb_any_quota_loaded(sb)) {
1462 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1463 "options when quota turned on");
1466 clear_opt(sbi->s_mount_opt, QUOTA);
1467 clear_opt(sbi->s_mount_opt, USRQUOTA);
1468 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1474 ext4_msg(sb, KERN_ERR,
1475 "quota options not supported");
1479 case Opt_offusrjquota:
1480 case Opt_offgrpjquota:
1481 case Opt_jqfmt_vfsold:
1482 case Opt_jqfmt_vfsv0:
1483 ext4_msg(sb, KERN_ERR,
1484 "journaled quota options not supported");
1490 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1493 clear_opt(sbi->s_mount_opt, BARRIER);
1496 if (match_int(&args[0], &option)) {
1497 set_opt(sbi->s_mount_opt, BARRIER);
1501 set_opt(sbi->s_mount_opt, BARRIER);
1503 clear_opt(sbi->s_mount_opt, BARRIER);
1509 ext4_msg(sb, KERN_ERR,
1510 "resize option only available "
1514 if (match_int(&args[0], &option) != 0)
1516 *n_blocks_count = option;
1519 set_opt(sbi->s_mount_opt, NOBH);
1522 clear_opt(sbi->s_mount_opt, NOBH);
1525 set_opt(sbi->s_mount_opt, I_VERSION);
1526 sb->s_flags |= MS_I_VERSION;
1528 case Opt_nodelalloc:
1529 clear_opt(sbi->s_mount_opt, DELALLOC);
1532 if (match_int(&args[0], &option))
1536 sbi->s_stripe = option;
1539 set_opt(sbi->s_mount_opt, DELALLOC);
1541 case Opt_block_validity:
1542 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1544 case Opt_noblock_validity:
1545 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1547 case Opt_inode_readahead_blks:
1548 if (match_int(&args[0], &option))
1550 if (option < 0 || option > (1 << 30))
1552 if (!is_power_of_2(option)) {
1553 ext4_msg(sb, KERN_ERR,
1554 "EXT4-fs: inode_readahead_blks"
1555 " must be a power of 2");
1558 sbi->s_inode_readahead_blks = option;
1560 case Opt_journal_ioprio:
1561 if (match_int(&args[0], &option))
1563 if (option < 0 || option > 7)
1565 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1568 case Opt_noauto_da_alloc:
1569 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1571 case Opt_auto_da_alloc:
1572 if (match_int(&args[0], &option)) {
1573 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1577 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1579 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1582 ext4_msg(sb, KERN_ERR,
1583 "Unrecognized mount option \"%s\" "
1584 "or missing value", p);
1589 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1590 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1591 sbi->s_qf_names[USRQUOTA])
1592 clear_opt(sbi->s_mount_opt, USRQUOTA);
1594 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1595 sbi->s_qf_names[GRPQUOTA])
1596 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1598 if ((sbi->s_qf_names[USRQUOTA] &&
1599 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1600 (sbi->s_qf_names[GRPQUOTA] &&
1601 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1602 ext4_msg(sb, KERN_ERR, "old and new quota "
1607 if (!sbi->s_jquota_fmt) {
1608 ext4_msg(sb, KERN_ERR, "journaled quota format "
1613 if (sbi->s_jquota_fmt) {
1614 ext4_msg(sb, KERN_ERR, "journaled quota format "
1615 "specified with no journaling "
1624 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1627 struct ext4_sb_info *sbi = EXT4_SB(sb);
1630 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1631 ext4_msg(sb, KERN_ERR, "revision level too high, "
1632 "forcing read-only mode");
1637 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1638 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1639 "running e2fsck is recommended");
1640 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1641 ext4_msg(sb, KERN_WARNING,
1642 "warning: mounting fs with errors, "
1643 "running e2fsck is recommended");
1644 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1645 le16_to_cpu(es->s_mnt_count) >=
1646 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1647 ext4_msg(sb, KERN_WARNING,
1648 "warning: maximal mount count reached, "
1649 "running e2fsck is recommended");
1650 else if (le32_to_cpu(es->s_checkinterval) &&
1651 (le32_to_cpu(es->s_lastcheck) +
1652 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1653 ext4_msg(sb, KERN_WARNING,
1654 "warning: checktime reached, "
1655 "running e2fsck is recommended");
1656 if (!sbi->s_journal)
1657 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1658 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1659 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1660 le16_add_cpu(&es->s_mnt_count, 1);
1661 es->s_mtime = cpu_to_le32(get_seconds());
1662 ext4_update_dynamic_rev(sb);
1664 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1666 ext4_commit_super(sb, 1);
1667 if (test_opt(sb, DEBUG))
1668 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1669 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1671 sbi->s_groups_count,
1672 EXT4_BLOCKS_PER_GROUP(sb),
1673 EXT4_INODES_PER_GROUP(sb),
1679 static int ext4_fill_flex_info(struct super_block *sb)
1681 struct ext4_sb_info *sbi = EXT4_SB(sb);
1682 struct ext4_group_desc *gdp = NULL;
1683 ext4_group_t flex_group_count;
1684 ext4_group_t flex_group;
1685 int groups_per_flex = 0;
1689 if (!sbi->s_es->s_log_groups_per_flex) {
1690 sbi->s_log_groups_per_flex = 0;
1694 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1695 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1697 /* We allocate both existing and potentially added groups */
1698 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1699 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1700 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1701 size = flex_group_count * sizeof(struct flex_groups);
1702 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1703 if (sbi->s_flex_groups == NULL) {
1704 sbi->s_flex_groups = vmalloc(size);
1705 if (sbi->s_flex_groups)
1706 memset(sbi->s_flex_groups, 0, size);
1708 if (sbi->s_flex_groups == NULL) {
1709 ext4_msg(sb, KERN_ERR, "not enough memory for "
1710 "%u flex groups", flex_group_count);
1714 for (i = 0; i < sbi->s_groups_count; i++) {
1715 gdp = ext4_get_group_desc(sb, i, NULL);
1717 flex_group = ext4_flex_group(sbi, i);
1718 atomic_add(ext4_free_inodes_count(sb, gdp),
1719 &sbi->s_flex_groups[flex_group].free_inodes);
1720 atomic_add(ext4_free_blks_count(sb, gdp),
1721 &sbi->s_flex_groups[flex_group].free_blocks);
1722 atomic_add(ext4_used_dirs_count(sb, gdp),
1723 &sbi->s_flex_groups[flex_group].used_dirs);
1731 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1732 struct ext4_group_desc *gdp)
1736 if (sbi->s_es->s_feature_ro_compat &
1737 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1738 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1739 __le32 le_group = cpu_to_le32(block_group);
1741 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1742 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1743 crc = crc16(crc, (__u8 *)gdp, offset);
1744 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1745 /* for checksum of struct ext4_group_desc do the rest...*/
1746 if ((sbi->s_es->s_feature_incompat &
1747 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1748 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1749 crc = crc16(crc, (__u8 *)gdp + offset,
1750 le16_to_cpu(sbi->s_es->s_desc_size) -
1754 return cpu_to_le16(crc);
1757 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1758 struct ext4_group_desc *gdp)
1760 if ((sbi->s_es->s_feature_ro_compat &
1761 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1762 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1768 /* Called at mount-time, super-block is locked */
1769 static int ext4_check_descriptors(struct super_block *sb)
1771 struct ext4_sb_info *sbi = EXT4_SB(sb);
1772 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1773 ext4_fsblk_t last_block;
1774 ext4_fsblk_t block_bitmap;
1775 ext4_fsblk_t inode_bitmap;
1776 ext4_fsblk_t inode_table;
1777 int flexbg_flag = 0;
1780 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1783 ext4_debug("Checking group descriptors");
1785 for (i = 0; i < sbi->s_groups_count; i++) {
1786 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1788 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1789 last_block = ext4_blocks_count(sbi->s_es) - 1;
1791 last_block = first_block +
1792 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1794 block_bitmap = ext4_block_bitmap(sb, gdp);
1795 if (block_bitmap < first_block || block_bitmap > last_block) {
1796 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1797 "Block bitmap for group %u not in group "
1798 "(block %llu)!", i, block_bitmap);
1801 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1802 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1804 "Inode bitmap for group %u not in group "
1805 "(block %llu)!", i, inode_bitmap);
1808 inode_table = ext4_inode_table(sb, gdp);
1809 if (inode_table < first_block ||
1810 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1811 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1812 "Inode table for group %u not in group "
1813 "(block %llu)!", i, inode_table);
1816 ext4_lock_group(sb, i);
1817 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1818 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1819 "Checksum for group %u failed (%u!=%u)",
1820 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1821 gdp)), le16_to_cpu(gdp->bg_checksum));
1822 if (!(sb->s_flags & MS_RDONLY)) {
1823 ext4_unlock_group(sb, i);
1827 ext4_unlock_group(sb, i);
1829 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1832 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1833 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1837 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1838 * the superblock) which were deleted from all directories, but held open by
1839 * a process at the time of a crash. We walk the list and try to delete these
1840 * inodes at recovery time (only with a read-write filesystem).
1842 * In order to keep the orphan inode chain consistent during traversal (in
1843 * case of crash during recovery), we link each inode into the superblock
1844 * orphan list_head and handle it the same way as an inode deletion during
1845 * normal operation (which journals the operations for us).
1847 * We only do an iget() and an iput() on each inode, which is very safe if we
1848 * accidentally point at an in-use or already deleted inode. The worst that
1849 * can happen in this case is that we get a "bit already cleared" message from
1850 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1851 * e2fsck was run on this filesystem, and it must have already done the orphan
1852 * inode cleanup for us, so we can safely abort without any further action.
1854 static void ext4_orphan_cleanup(struct super_block *sb,
1855 struct ext4_super_block *es)
1857 unsigned int s_flags = sb->s_flags;
1858 int nr_orphans = 0, nr_truncates = 0;
1862 if (!es->s_last_orphan) {
1863 jbd_debug(4, "no orphan inodes to clean up\n");
1867 if (bdev_read_only(sb->s_bdev)) {
1868 ext4_msg(sb, KERN_ERR, "write access "
1869 "unavailable, skipping orphan cleanup");
1873 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1874 if (es->s_last_orphan)
1875 jbd_debug(1, "Errors on filesystem, "
1876 "clearing orphan list.\n");
1877 es->s_last_orphan = 0;
1878 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1882 if (s_flags & MS_RDONLY) {
1883 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1884 sb->s_flags &= ~MS_RDONLY;
1887 /* Needed for iput() to work correctly and not trash data */
1888 sb->s_flags |= MS_ACTIVE;
1889 /* Turn on quotas so that they are updated correctly */
1890 for (i = 0; i < MAXQUOTAS; i++) {
1891 if (EXT4_SB(sb)->s_qf_names[i]) {
1892 int ret = ext4_quota_on_mount(sb, i);
1894 ext4_msg(sb, KERN_ERR,
1895 "Cannot turn on journaled "
1896 "quota: error %d", ret);
1901 while (es->s_last_orphan) {
1902 struct inode *inode;
1904 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1905 if (IS_ERR(inode)) {
1906 es->s_last_orphan = 0;
1910 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1912 if (inode->i_nlink) {
1913 ext4_msg(sb, KERN_DEBUG,
1914 "%s: truncating inode %lu to %lld bytes",
1915 __func__, inode->i_ino, inode->i_size);
1916 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1917 inode->i_ino, inode->i_size);
1918 ext4_truncate(inode);
1921 ext4_msg(sb, KERN_DEBUG,
1922 "%s: deleting unreferenced inode %lu",
1923 __func__, inode->i_ino);
1924 jbd_debug(2, "deleting unreferenced inode %lu\n",
1928 iput(inode); /* The delete magic happens here! */
1931 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1934 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1935 PLURAL(nr_orphans));
1937 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1938 PLURAL(nr_truncates));
1940 /* Turn quotas off */
1941 for (i = 0; i < MAXQUOTAS; i++) {
1942 if (sb_dqopt(sb)->files[i])
1943 vfs_quota_off(sb, i, 0);
1946 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1950 * Maximal extent format file size.
1951 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1952 * extent format containers, within a sector_t, and within i_blocks
1953 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1954 * so that won't be a limiting factor.
1956 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1958 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1961 loff_t upper_limit = MAX_LFS_FILESIZE;
1963 /* small i_blocks in vfs inode? */
1964 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1966 * CONFIG_LBDAF is not enabled implies the inode
1967 * i_block represent total blocks in 512 bytes
1968 * 32 == size of vfs inode i_blocks * 8
1970 upper_limit = (1LL << 32) - 1;
1972 /* total blocks in file system block size */
1973 upper_limit >>= (blkbits - 9);
1974 upper_limit <<= blkbits;
1977 /* 32-bit extent-start container, ee_block */
1982 /* Sanity check against vm- & vfs- imposed limits */
1983 if (res > upper_limit)
1990 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1991 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1992 * We need to be 1 filesystem block less than the 2^48 sector limit.
1994 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1996 loff_t res = EXT4_NDIR_BLOCKS;
1999 /* This is calculated to be the largest file size for a dense, block
2000 * mapped file such that the file's total number of 512-byte sectors,
2001 * including data and all indirect blocks, does not exceed (2^48 - 1).
2003 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2004 * number of 512-byte sectors of the file.
2007 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2009 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2010 * the inode i_block field represents total file blocks in
2011 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2013 upper_limit = (1LL << 32) - 1;
2015 /* total blocks in file system block size */
2016 upper_limit >>= (bits - 9);
2020 * We use 48 bit ext4_inode i_blocks
2021 * With EXT4_HUGE_FILE_FL set the i_blocks
2022 * represent total number of blocks in
2023 * file system block size
2025 upper_limit = (1LL << 48) - 1;
2029 /* indirect blocks */
2031 /* double indirect blocks */
2032 meta_blocks += 1 + (1LL << (bits-2));
2033 /* tripple indirect blocks */
2034 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2036 upper_limit -= meta_blocks;
2037 upper_limit <<= bits;
2039 res += 1LL << (bits-2);
2040 res += 1LL << (2*(bits-2));
2041 res += 1LL << (3*(bits-2));
2043 if (res > upper_limit)
2046 if (res > MAX_LFS_FILESIZE)
2047 res = MAX_LFS_FILESIZE;
2052 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2053 ext4_fsblk_t logical_sb_block, int nr)
2055 struct ext4_sb_info *sbi = EXT4_SB(sb);
2056 ext4_group_t bg, first_meta_bg;
2059 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2061 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2063 return logical_sb_block + nr + 1;
2064 bg = sbi->s_desc_per_block * nr;
2065 if (ext4_bg_has_super(sb, bg))
2068 return (has_super + ext4_group_first_block_no(sb, bg));
2072 * ext4_get_stripe_size: Get the stripe size.
2073 * @sbi: In memory super block info
2075 * If we have specified it via mount option, then
2076 * use the mount option value. If the value specified at mount time is
2077 * greater than the blocks per group use the super block value.
2078 * If the super block value is greater than blocks per group return 0.
2079 * Allocator needs it be less than blocks per group.
2082 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2084 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2085 unsigned long stripe_width =
2086 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2088 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2089 return sbi->s_stripe;
2091 if (stripe_width <= sbi->s_blocks_per_group)
2092 return stripe_width;
2094 if (stride <= sbi->s_blocks_per_group)
2103 struct attribute attr;
2104 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2105 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2106 const char *, size_t);
2110 static int parse_strtoul(const char *buf,
2111 unsigned long max, unsigned long *value)
2115 while (*buf && isspace(*buf))
2117 *value = simple_strtoul(buf, &endp, 0);
2118 while (*endp && isspace(*endp))
2120 if (*endp || *value > max)
2126 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2127 struct ext4_sb_info *sbi,
2130 return snprintf(buf, PAGE_SIZE, "%llu\n",
2131 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2134 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2135 struct ext4_sb_info *sbi, char *buf)
2137 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2139 return snprintf(buf, PAGE_SIZE, "%lu\n",
2140 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2141 sbi->s_sectors_written_start) >> 1);
2144 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2145 struct ext4_sb_info *sbi, char *buf)
2147 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2149 return snprintf(buf, PAGE_SIZE, "%llu\n",
2150 sbi->s_kbytes_written +
2151 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2152 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2155 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2156 struct ext4_sb_info *sbi,
2157 const char *buf, size_t count)
2161 if (parse_strtoul(buf, 0x40000000, &t))
2164 if (!is_power_of_2(t))
2167 sbi->s_inode_readahead_blks = t;
2171 static ssize_t sbi_ui_show(struct ext4_attr *a,
2172 struct ext4_sb_info *sbi, char *buf)
2174 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2176 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2179 static ssize_t sbi_ui_store(struct ext4_attr *a,
2180 struct ext4_sb_info *sbi,
2181 const char *buf, size_t count)
2183 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2186 if (parse_strtoul(buf, 0xffffffff, &t))
2192 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2193 static struct ext4_attr ext4_attr_##_name = { \
2194 .attr = {.name = __stringify(_name), .mode = _mode }, \
2197 .offset = offsetof(struct ext4_sb_info, _elname), \
2199 #define EXT4_ATTR(name, mode, show, store) \
2200 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2202 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2203 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2204 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2205 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2206 #define ATTR_LIST(name) &ext4_attr_##name.attr
2208 EXT4_RO_ATTR(delayed_allocation_blocks);
2209 EXT4_RO_ATTR(session_write_kbytes);
2210 EXT4_RO_ATTR(lifetime_write_kbytes);
2211 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2212 inode_readahead_blks_store, s_inode_readahead_blks);
2213 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2214 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2215 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2216 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2217 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2218 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2219 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2220 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2222 static struct attribute *ext4_attrs[] = {
2223 ATTR_LIST(delayed_allocation_blocks),
2224 ATTR_LIST(session_write_kbytes),
2225 ATTR_LIST(lifetime_write_kbytes),
2226 ATTR_LIST(inode_readahead_blks),
2227 ATTR_LIST(inode_goal),
2228 ATTR_LIST(mb_stats),
2229 ATTR_LIST(mb_max_to_scan),
2230 ATTR_LIST(mb_min_to_scan),
2231 ATTR_LIST(mb_order2_req),
2232 ATTR_LIST(mb_stream_req),
2233 ATTR_LIST(mb_group_prealloc),
2234 ATTR_LIST(max_writeback_mb_bump),
2238 static ssize_t ext4_attr_show(struct kobject *kobj,
2239 struct attribute *attr, char *buf)
2241 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2243 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2245 return a->show ? a->show(a, sbi, buf) : 0;
2248 static ssize_t ext4_attr_store(struct kobject *kobj,
2249 struct attribute *attr,
2250 const char *buf, size_t len)
2252 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2254 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2256 return a->store ? a->store(a, sbi, buf, len) : 0;
2259 static void ext4_sb_release(struct kobject *kobj)
2261 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2263 complete(&sbi->s_kobj_unregister);
2267 static struct sysfs_ops ext4_attr_ops = {
2268 .show = ext4_attr_show,
2269 .store = ext4_attr_store,
2272 static struct kobj_type ext4_ktype = {
2273 .default_attrs = ext4_attrs,
2274 .sysfs_ops = &ext4_attr_ops,
2275 .release = ext4_sb_release,
2279 * Check whether this filesystem can be mounted based on
2280 * the features present and the RDONLY/RDWR mount requested.
2281 * Returns 1 if this filesystem can be mounted as requested,
2282 * 0 if it cannot be.
2284 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2286 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2287 ext4_msg(sb, KERN_ERR,
2288 "Couldn't mount because of "
2289 "unsupported optional features (%x)",
2290 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2291 ~EXT4_FEATURE_INCOMPAT_SUPP));
2298 /* Check that feature set is OK for a read-write mount */
2299 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2300 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2301 "unsupported optional features (%x)",
2302 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2303 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2307 * Large file size enabled file system can only be mounted
2308 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2310 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2311 if (sizeof(blkcnt_t) < sizeof(u64)) {
2312 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2313 "cannot be mounted RDWR without "
2321 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2322 __releases(kernel_lock)
2323 __acquires(kernel_lock)
2325 struct buffer_head *bh;
2326 struct ext4_super_block *es = NULL;
2327 struct ext4_sb_info *sbi;
2329 ext4_fsblk_t sb_block = get_sb_block(&data);
2330 ext4_fsblk_t logical_sb_block;
2331 unsigned long offset = 0;
2332 unsigned long journal_devnum = 0;
2333 unsigned long def_mount_opts;
2339 unsigned int db_count;
2341 int needs_recovery, has_huge_files;
2344 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2346 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2350 sbi->s_blockgroup_lock =
2351 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2352 if (!sbi->s_blockgroup_lock) {
2356 sb->s_fs_info = sbi;
2357 sbi->s_mount_opt = 0;
2358 sbi->s_resuid = EXT4_DEF_RESUID;
2359 sbi->s_resgid = EXT4_DEF_RESGID;
2360 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2361 sbi->s_sb_block = sb_block;
2362 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2367 /* Cleanup superblock name */
2368 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2371 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2373 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2378 * The ext4 superblock will not be buffer aligned for other than 1kB
2379 * block sizes. We need to calculate the offset from buffer start.
2381 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2382 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2383 offset = do_div(logical_sb_block, blocksize);
2385 logical_sb_block = sb_block;
2388 if (!(bh = sb_bread(sb, logical_sb_block))) {
2389 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2393 * Note: s_es must be initialized as soon as possible because
2394 * some ext4 macro-instructions depend on its value
2396 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2398 sb->s_magic = le16_to_cpu(es->s_magic);
2399 if (sb->s_magic != EXT4_SUPER_MAGIC)
2401 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2403 /* Set defaults before we parse the mount options */
2404 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2405 if (def_mount_opts & EXT4_DEFM_DEBUG)
2406 set_opt(sbi->s_mount_opt, DEBUG);
2407 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2408 set_opt(sbi->s_mount_opt, GRPID);
2409 if (def_mount_opts & EXT4_DEFM_UID16)
2410 set_opt(sbi->s_mount_opt, NO_UID32);
2411 #ifdef CONFIG_EXT4_FS_XATTR
2412 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2413 set_opt(sbi->s_mount_opt, XATTR_USER);
2415 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2416 if (def_mount_opts & EXT4_DEFM_ACL)
2417 set_opt(sbi->s_mount_opt, POSIX_ACL);
2419 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2420 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2421 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2422 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2423 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2424 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2426 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2427 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2428 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2429 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2431 set_opt(sbi->s_mount_opt, ERRORS_RO);
2433 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2434 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2435 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2436 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2437 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2438 sbi->s_mb_history_max = default_mb_history_length;
2440 set_opt(sbi->s_mount_opt, BARRIER);
2443 * enable delayed allocation by default
2444 * Use -o nodelalloc to turn it off
2446 set_opt(sbi->s_mount_opt, DELALLOC);
2448 if (!parse_options((char *) data, sb, &journal_devnum,
2449 &journal_ioprio, NULL, 0))
2452 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2453 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2455 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2456 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2457 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2458 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2459 ext4_msg(sb, KERN_WARNING,
2460 "feature flags set on rev 0 fs, "
2461 "running e2fsck is recommended");
2464 * Check feature flags regardless of the revision level, since we
2465 * previously didn't change the revision level when setting the flags,
2466 * so there is a chance incompat flags are set on a rev 0 filesystem.
2468 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2471 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2473 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2474 blocksize > EXT4_MAX_BLOCK_SIZE) {
2475 ext4_msg(sb, KERN_ERR,
2476 "Unsupported filesystem blocksize %d", blocksize);
2480 if (sb->s_blocksize != blocksize) {
2481 /* Validate the filesystem blocksize */
2482 if (!sb_set_blocksize(sb, blocksize)) {
2483 ext4_msg(sb, KERN_ERR, "bad block size %d",
2489 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2490 offset = do_div(logical_sb_block, blocksize);
2491 bh = sb_bread(sb, logical_sb_block);
2493 ext4_msg(sb, KERN_ERR,
2494 "Can't read superblock on 2nd try");
2497 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2499 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2500 ext4_msg(sb, KERN_ERR,
2501 "Magic mismatch, very weird!");
2506 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2507 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2508 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2510 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2512 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2513 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2514 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2516 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2517 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2518 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2519 (!is_power_of_2(sbi->s_inode_size)) ||
2520 (sbi->s_inode_size > blocksize)) {
2521 ext4_msg(sb, KERN_ERR,
2522 "unsupported inode size: %d",
2526 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2527 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2530 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2531 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2532 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2533 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2534 !is_power_of_2(sbi->s_desc_size)) {
2535 ext4_msg(sb, KERN_ERR,
2536 "unsupported descriptor size %lu",
2541 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2543 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2544 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2545 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2548 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2549 if (sbi->s_inodes_per_block == 0)
2551 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2552 sbi->s_inodes_per_block;
2553 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2555 sbi->s_mount_state = le16_to_cpu(es->s_state);
2556 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2557 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2559 for (i = 0; i < 4; i++)
2560 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2561 sbi->s_def_hash_version = es->s_def_hash_version;
2562 i = le32_to_cpu(es->s_flags);
2563 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2564 sbi->s_hash_unsigned = 3;
2565 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2566 #ifdef __CHAR_UNSIGNED__
2567 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2568 sbi->s_hash_unsigned = 3;
2570 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2575 if (sbi->s_blocks_per_group > blocksize * 8) {
2576 ext4_msg(sb, KERN_ERR,
2577 "#blocks per group too big: %lu",
2578 sbi->s_blocks_per_group);
2581 if (sbi->s_inodes_per_group > blocksize * 8) {
2582 ext4_msg(sb, KERN_ERR,
2583 "#inodes per group too big: %lu",
2584 sbi->s_inodes_per_group);
2589 * Test whether we have more sectors than will fit in sector_t,
2590 * and whether the max offset is addressable by the page cache.
2592 if ((ext4_blocks_count(es) >
2593 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2594 (ext4_blocks_count(es) >
2595 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2596 ext4_msg(sb, KERN_ERR, "filesystem"
2597 " too large to mount safely on this system");
2598 if (sizeof(sector_t) < 8)
2599 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2604 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2607 /* check blocks count against device size */
2608 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2609 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2610 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2611 "exceeds size of device (%llu blocks)",
2612 ext4_blocks_count(es), blocks_count);
2617 * It makes no sense for the first data block to be beyond the end
2618 * of the filesystem.
2620 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2621 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2622 "block %u is beyond end of filesystem (%llu)",
2623 le32_to_cpu(es->s_first_data_block),
2624 ext4_blocks_count(es));
2627 blocks_count = (ext4_blocks_count(es) -
2628 le32_to_cpu(es->s_first_data_block) +
2629 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2630 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2631 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2632 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2633 "(block count %llu, first data block %u, "
2634 "blocks per group %lu)", sbi->s_groups_count,
2635 ext4_blocks_count(es),
2636 le32_to_cpu(es->s_first_data_block),
2637 EXT4_BLOCKS_PER_GROUP(sb));
2640 sbi->s_groups_count = blocks_count;
2641 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2642 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2643 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2644 EXT4_DESC_PER_BLOCK(sb);
2645 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2647 if (sbi->s_group_desc == NULL) {
2648 ext4_msg(sb, KERN_ERR, "not enough memory");
2652 #ifdef CONFIG_PROC_FS
2654 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2657 bgl_lock_init(sbi->s_blockgroup_lock);
2659 for (i = 0; i < db_count; i++) {
2660 block = descriptor_loc(sb, logical_sb_block, i);
2661 sbi->s_group_desc[i] = sb_bread(sb, block);
2662 if (!sbi->s_group_desc[i]) {
2663 ext4_msg(sb, KERN_ERR,
2664 "can't read group descriptor %d", i);
2669 if (!ext4_check_descriptors(sb)) {
2670 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2673 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2674 if (!ext4_fill_flex_info(sb)) {
2675 ext4_msg(sb, KERN_ERR,
2676 "unable to initialize "
2677 "flex_bg meta info!");
2681 sbi->s_gdb_count = db_count;
2682 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2683 spin_lock_init(&sbi->s_next_gen_lock);
2685 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2686 ext4_count_free_blocks(sb));
2688 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2689 ext4_count_free_inodes(sb));
2692 err = percpu_counter_init(&sbi->s_dirs_counter,
2693 ext4_count_dirs(sb));
2696 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2699 ext4_msg(sb, KERN_ERR, "insufficient memory");
2703 sbi->s_stripe = ext4_get_stripe_size(sbi);
2704 sbi->s_max_writeback_mb_bump = 128;
2707 * set up enough so that it can read an inode
2709 if (!test_opt(sb, NOLOAD) &&
2710 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2711 sb->s_op = &ext4_sops;
2713 sb->s_op = &ext4_nojournal_sops;
2714 sb->s_export_op = &ext4_export_ops;
2715 sb->s_xattr = ext4_xattr_handlers;
2717 sb->s_qcop = &ext4_qctl_operations;
2718 sb->dq_op = &ext4_quota_operations;
2720 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2721 mutex_init(&sbi->s_orphan_lock);
2722 mutex_init(&sbi->s_resize_lock);
2726 needs_recovery = (es->s_last_orphan != 0 ||
2727 EXT4_HAS_INCOMPAT_FEATURE(sb,
2728 EXT4_FEATURE_INCOMPAT_RECOVER));
2731 * The first inode we look at is the journal inode. Don't try
2732 * root first: it may be modified in the journal!
2734 if (!test_opt(sb, NOLOAD) &&
2735 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2736 if (ext4_load_journal(sb, es, journal_devnum))
2738 if (!(sb->s_flags & MS_RDONLY) &&
2739 EXT4_SB(sb)->s_journal->j_failed_commit) {
2740 ext4_msg(sb, KERN_CRIT, "error: "
2741 "ext4_fill_super: Journal transaction "
2743 EXT4_SB(sb)->s_journal->j_failed_commit);
2744 if (test_opt(sb, ERRORS_RO)) {
2745 ext4_msg(sb, KERN_CRIT,
2746 "Mounting filesystem read-only");
2747 sb->s_flags |= MS_RDONLY;
2748 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2749 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2751 if (test_opt(sb, ERRORS_PANIC)) {
2752 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2753 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2754 ext4_commit_super(sb, 1);
2758 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2759 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2760 ext4_msg(sb, KERN_ERR, "required journal recovery "
2761 "suppressed and not mounted read-only");
2764 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2765 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2766 sbi->s_journal = NULL;
2771 if (ext4_blocks_count(es) > 0xffffffffULL &&
2772 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2773 JBD2_FEATURE_INCOMPAT_64BIT)) {
2774 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2778 jbd2_journal_set_features(sbi->s_journal,
2779 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2780 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2781 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2782 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2784 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2785 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2787 /* We have now updated the journal if required, so we can
2788 * validate the data journaling mode. */
2789 switch (test_opt(sb, DATA_FLAGS)) {
2791 /* No mode set, assume a default based on the journal
2792 * capabilities: ORDERED_DATA if the journal can
2793 * cope, else JOURNAL_DATA
2795 if (jbd2_journal_check_available_features
2796 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2797 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2799 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2802 case EXT4_MOUNT_ORDERED_DATA:
2803 case EXT4_MOUNT_WRITEBACK_DATA:
2804 if (!jbd2_journal_check_available_features
2805 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2806 ext4_msg(sb, KERN_ERR, "Journal does not support "
2807 "requested data journaling mode");
2813 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2817 if (test_opt(sb, NOBH)) {
2818 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2819 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2820 "its supported only with writeback mode");
2821 clear_opt(sbi->s_mount_opt, NOBH);
2824 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2825 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2826 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2827 goto failed_mount_wq;
2831 * The jbd2_journal_load will have done any necessary log recovery,
2832 * so we can safely mount the rest of the filesystem now.
2835 root = ext4_iget(sb, EXT4_ROOT_INO);
2837 ext4_msg(sb, KERN_ERR, "get root inode failed");
2838 ret = PTR_ERR(root);
2841 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2843 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2846 sb->s_root = d_alloc_root(root);
2848 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2854 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2856 /* determine the minimum size of new large inodes, if present */
2857 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2858 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2859 EXT4_GOOD_OLD_INODE_SIZE;
2860 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2861 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2862 if (sbi->s_want_extra_isize <
2863 le16_to_cpu(es->s_want_extra_isize))
2864 sbi->s_want_extra_isize =
2865 le16_to_cpu(es->s_want_extra_isize);
2866 if (sbi->s_want_extra_isize <
2867 le16_to_cpu(es->s_min_extra_isize))
2868 sbi->s_want_extra_isize =
2869 le16_to_cpu(es->s_min_extra_isize);
2872 /* Check if enough inode space is available */
2873 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2874 sbi->s_inode_size) {
2875 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2876 EXT4_GOOD_OLD_INODE_SIZE;
2877 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2881 if (test_opt(sb, DELALLOC) &&
2882 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2883 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2884 "requested data journaling mode");
2885 clear_opt(sbi->s_mount_opt, DELALLOC);
2888 err = ext4_setup_system_zone(sb);
2890 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2891 "zone (%d)\n", err);
2896 err = ext4_mb_init(sb, needs_recovery);
2898 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2903 sbi->s_kobj.kset = ext4_kset;
2904 init_completion(&sbi->s_kobj_unregister);
2905 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2908 ext4_mb_release(sb);
2909 ext4_ext_release(sb);
2913 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2914 ext4_orphan_cleanup(sb, es);
2915 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2916 if (needs_recovery) {
2917 ext4_msg(sb, KERN_INFO, "recovery complete");
2918 ext4_mark_recovery_complete(sb, es);
2920 if (EXT4_SB(sb)->s_journal) {
2921 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2922 descr = " journalled data mode";
2923 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2924 descr = " ordered data mode";
2926 descr = " writeback data mode";
2928 descr = "out journal";
2930 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2937 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2941 ext4_msg(sb, KERN_ERR, "mount failed");
2942 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2944 ext4_release_system_zone(sb);
2945 if (sbi->s_journal) {
2946 jbd2_journal_destroy(sbi->s_journal);
2947 sbi->s_journal = NULL;
2950 if (sbi->s_flex_groups) {
2951 if (is_vmalloc_addr(sbi->s_flex_groups))
2952 vfree(sbi->s_flex_groups);
2954 kfree(sbi->s_flex_groups);
2956 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2957 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2958 percpu_counter_destroy(&sbi->s_dirs_counter);
2959 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2961 for (i = 0; i < db_count; i++)
2962 brelse(sbi->s_group_desc[i]);
2963 kfree(sbi->s_group_desc);
2966 remove_proc_entry(sb->s_id, ext4_proc_root);
2969 for (i = 0; i < MAXQUOTAS; i++)
2970 kfree(sbi->s_qf_names[i]);
2972 ext4_blkdev_remove(sbi);
2975 sb->s_fs_info = NULL;
2976 kfree(sbi->s_blockgroup_lock);
2983 * Setup any per-fs journal parameters now. We'll do this both on
2984 * initial mount, once the journal has been initialised but before we've
2985 * done any recovery; and again on any subsequent remount.
2987 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2989 struct ext4_sb_info *sbi = EXT4_SB(sb);
2991 journal->j_commit_interval = sbi->s_commit_interval;
2992 journal->j_min_batch_time = sbi->s_min_batch_time;
2993 journal->j_max_batch_time = sbi->s_max_batch_time;
2995 spin_lock(&journal->j_state_lock);
2996 if (test_opt(sb, BARRIER))
2997 journal->j_flags |= JBD2_BARRIER;
2999 journal->j_flags &= ~JBD2_BARRIER;
3000 if (test_opt(sb, DATA_ERR_ABORT))
3001 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3003 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3004 spin_unlock(&journal->j_state_lock);
3007 static journal_t *ext4_get_journal(struct super_block *sb,
3008 unsigned int journal_inum)
3010 struct inode *journal_inode;
3013 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3015 /* First, test for the existence of a valid inode on disk. Bad
3016 * things happen if we iget() an unused inode, as the subsequent
3017 * iput() will try to delete it. */
3019 journal_inode = ext4_iget(sb, journal_inum);
3020 if (IS_ERR(journal_inode)) {
3021 ext4_msg(sb, KERN_ERR, "no journal found");
3024 if (!journal_inode->i_nlink) {
3025 make_bad_inode(journal_inode);
3026 iput(journal_inode);
3027 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3031 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3032 journal_inode, journal_inode->i_size);
3033 if (!S_ISREG(journal_inode->i_mode)) {
3034 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3035 iput(journal_inode);
3039 journal = jbd2_journal_init_inode(journal_inode);
3041 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3042 iput(journal_inode);
3045 journal->j_private = sb;
3046 ext4_init_journal_params(sb, journal);
3050 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3053 struct buffer_head *bh;
3057 int hblock, blocksize;
3058 ext4_fsblk_t sb_block;
3059 unsigned long offset;
3060 struct ext4_super_block *es;
3061 struct block_device *bdev;
3063 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3065 bdev = ext4_blkdev_get(j_dev, sb);
3069 if (bd_claim(bdev, sb)) {
3070 ext4_msg(sb, KERN_ERR,
3071 "failed to claim external journal device");
3072 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3076 blocksize = sb->s_blocksize;
3077 hblock = bdev_logical_block_size(bdev);
3078 if (blocksize < hblock) {
3079 ext4_msg(sb, KERN_ERR,
3080 "blocksize too small for journal device");
3084 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3085 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3086 set_blocksize(bdev, blocksize);
3087 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3088 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3089 "external journal");
3093 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3094 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3095 !(le32_to_cpu(es->s_feature_incompat) &
3096 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3097 ext4_msg(sb, KERN_ERR, "external journal has "
3103 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3104 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3109 len = ext4_blocks_count(es);
3110 start = sb_block + 1;
3111 brelse(bh); /* we're done with the superblock */
3113 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3114 start, len, blocksize);
3116 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3119 journal->j_private = sb;
3120 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3121 wait_on_buffer(journal->j_sb_buffer);
3122 if (!buffer_uptodate(journal->j_sb_buffer)) {
3123 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3126 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3127 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3128 "user (unsupported) - %d",
3129 be32_to_cpu(journal->j_superblock->s_nr_users));
3132 EXT4_SB(sb)->journal_bdev = bdev;
3133 ext4_init_journal_params(sb, journal);
3137 jbd2_journal_destroy(journal);
3139 ext4_blkdev_put(bdev);
3143 static int ext4_load_journal(struct super_block *sb,
3144 struct ext4_super_block *es,
3145 unsigned long journal_devnum)
3148 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3151 int really_read_only;
3153 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3155 if (journal_devnum &&
3156 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3157 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3158 "numbers have changed");
3159 journal_dev = new_decode_dev(journal_devnum);
3161 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3163 really_read_only = bdev_read_only(sb->s_bdev);
3166 * Are we loading a blank journal or performing recovery after a
3167 * crash? For recovery, we need to check in advance whether we
3168 * can get read-write access to the device.
3170 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3171 if (sb->s_flags & MS_RDONLY) {
3172 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3173 "required on readonly filesystem");
3174 if (really_read_only) {
3175 ext4_msg(sb, KERN_ERR, "write access "
3176 "unavailable, cannot proceed");
3179 ext4_msg(sb, KERN_INFO, "write access will "
3180 "be enabled during recovery");
3184 if (journal_inum && journal_dev) {
3185 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3186 "and inode journals!");
3191 if (!(journal = ext4_get_journal(sb, journal_inum)))
3194 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3198 if (!(journal->j_flags & JBD2_BARRIER))
3199 ext4_msg(sb, KERN_INFO, "barriers disabled");
3201 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3202 err = jbd2_journal_update_format(journal);
3204 ext4_msg(sb, KERN_ERR, "error updating journal");
3205 jbd2_journal_destroy(journal);
3210 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3211 err = jbd2_journal_wipe(journal, !really_read_only);
3213 err = jbd2_journal_load(journal);
3216 ext4_msg(sb, KERN_ERR, "error loading journal");
3217 jbd2_journal_destroy(journal);
3221 EXT4_SB(sb)->s_journal = journal;
3222 ext4_clear_journal_err(sb, es);
3224 if (journal_devnum &&
3225 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3226 es->s_journal_dev = cpu_to_le32(journal_devnum);
3228 /* Make sure we flush the recovery flag to disk. */
3229 ext4_commit_super(sb, 1);
3235 static int ext4_commit_super(struct super_block *sb, int sync)
3237 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3238 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3243 if (buffer_write_io_error(sbh)) {
3245 * Oh, dear. A previous attempt to write the
3246 * superblock failed. This could happen because the
3247 * USB device was yanked out. Or it could happen to
3248 * be a transient write error and maybe the block will
3249 * be remapped. Nothing we can do but to retry the
3250 * write and hope for the best.
3252 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3253 "superblock detected");
3254 clear_buffer_write_io_error(sbh);
3255 set_buffer_uptodate(sbh);
3258 * If the file system is mounted read-only, don't update the
3259 * superblock write time. This avoids updating the superblock
3260 * write time when we are mounting the root file system
3261 * read/only but we need to replay the journal; at that point,
3262 * for people who are east of GMT and who make their clock
3263 * tick in localtime for Windows bug-for-bug compatibility,
3264 * the clock is set in the future, and this will cause e2fsck
3265 * to complain and force a full file system check.
3267 if (!(sb->s_flags & MS_RDONLY))
3268 es->s_wtime = cpu_to_le32(get_seconds());
3269 es->s_kbytes_written =
3270 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3271 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3272 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3273 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3274 &EXT4_SB(sb)->s_freeblocks_counter));
3275 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3276 &EXT4_SB(sb)->s_freeinodes_counter));
3278 BUFFER_TRACE(sbh, "marking dirty");
3279 mark_buffer_dirty(sbh);
3281 error = sync_dirty_buffer(sbh);
3285 error = buffer_write_io_error(sbh);
3287 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3289 clear_buffer_write_io_error(sbh);
3290 set_buffer_uptodate(sbh);
3297 * Have we just finished recovery? If so, and if we are mounting (or
3298 * remounting) the filesystem readonly, then we will end up with a
3299 * consistent fs on disk. Record that fact.
3301 static void ext4_mark_recovery_complete(struct super_block *sb,
3302 struct ext4_super_block *es)
3304 journal_t *journal = EXT4_SB(sb)->s_journal;
3306 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3307 BUG_ON(journal != NULL);
3310 jbd2_journal_lock_updates(journal);
3311 if (jbd2_journal_flush(journal) < 0)
3314 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3315 sb->s_flags & MS_RDONLY) {
3316 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3317 ext4_commit_super(sb, 1);
3321 jbd2_journal_unlock_updates(journal);
3325 * If we are mounting (or read-write remounting) a filesystem whose journal
3326 * has recorded an error from a previous lifetime, move that error to the
3327 * main filesystem now.
3329 static void ext4_clear_journal_err(struct super_block *sb,
3330 struct ext4_super_block *es)
3336 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3338 journal = EXT4_SB(sb)->s_journal;
3341 * Now check for any error status which may have been recorded in the
3342 * journal by a prior ext4_error() or ext4_abort()
3345 j_errno = jbd2_journal_errno(journal);
3349 errstr = ext4_decode_error(sb, j_errno, nbuf);
3350 ext4_warning(sb, __func__, "Filesystem error recorded "
3351 "from previous mount: %s", errstr);
3352 ext4_warning(sb, __func__, "Marking fs in need of "
3353 "filesystem check.");
3355 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3356 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3357 ext4_commit_super(sb, 1);
3359 jbd2_journal_clear_err(journal);
3364 * Force the running and committing transactions to commit,
3365 * and wait on the commit.
3367 int ext4_force_commit(struct super_block *sb)
3372 if (sb->s_flags & MS_RDONLY)
3375 journal = EXT4_SB(sb)->s_journal;
3377 ret = ext4_journal_force_commit(journal);
3382 static void ext4_write_super(struct super_block *sb)
3385 ext4_commit_super(sb, 1);
3389 static int ext4_sync_fs(struct super_block *sb, int wait)
3393 struct ext4_sb_info *sbi = EXT4_SB(sb);
3395 trace_ext4_sync_fs(sb, wait);
3396 flush_workqueue(sbi->dio_unwritten_wq);
3397 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3399 jbd2_log_wait_commit(sbi->s_journal, target);
3405 * LVM calls this function before a (read-only) snapshot is created. This
3406 * gives us a chance to flush the journal completely and mark the fs clean.
3408 static int ext4_freeze(struct super_block *sb)
3413 if (sb->s_flags & MS_RDONLY)
3416 journal = EXT4_SB(sb)->s_journal;
3418 /* Now we set up the journal barrier. */
3419 jbd2_journal_lock_updates(journal);
3422 * Don't clear the needs_recovery flag if we failed to flush
3425 error = jbd2_journal_flush(journal);
3428 jbd2_journal_unlock_updates(journal);
3432 /* Journal blocked and flushed, clear needs_recovery flag. */
3433 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3434 error = ext4_commit_super(sb, 1);
3441 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3442 * flag here, even though the filesystem is not technically dirty yet.
3444 static int ext4_unfreeze(struct super_block *sb)
3446 if (sb->s_flags & MS_RDONLY)
3450 /* Reset the needs_recovery flag before the fs is unlocked. */
3451 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3452 ext4_commit_super(sb, 1);
3454 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3458 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3460 struct ext4_super_block *es;
3461 struct ext4_sb_info *sbi = EXT4_SB(sb);
3462 ext4_fsblk_t n_blocks_count = 0;
3463 unsigned long old_sb_flags;
3464 struct ext4_mount_options old_opts;
3466 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3474 /* Store the original options */
3476 old_sb_flags = sb->s_flags;
3477 old_opts.s_mount_opt = sbi->s_mount_opt;
3478 old_opts.s_resuid = sbi->s_resuid;
3479 old_opts.s_resgid = sbi->s_resgid;
3480 old_opts.s_commit_interval = sbi->s_commit_interval;
3481 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3482 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3484 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3485 for (i = 0; i < MAXQUOTAS; i++)
3486 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3488 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3489 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3492 * Allow the "check" option to be passed as a remount option.
3494 if (!parse_options(data, sb, NULL, &journal_ioprio,
3495 &n_blocks_count, 1)) {
3500 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3501 ext4_abort(sb, __func__, "Abort forced by user");
3503 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3504 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3508 if (sbi->s_journal) {
3509 ext4_init_journal_params(sb, sbi->s_journal);
3510 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3513 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3514 n_blocks_count > ext4_blocks_count(es)) {
3515 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3520 if (*flags & MS_RDONLY) {
3522 * First of all, the unconditional stuff we have to do
3523 * to disable replay of the journal when we next remount
3525 sb->s_flags |= MS_RDONLY;
3528 * OK, test if we are remounting a valid rw partition
3529 * readonly, and if so set the rdonly flag and then
3530 * mark the partition as valid again.
3532 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3533 (sbi->s_mount_state & EXT4_VALID_FS))
3534 es->s_state = cpu_to_le16(sbi->s_mount_state);
3537 ext4_mark_recovery_complete(sb, es);
3539 /* Make sure we can mount this feature set readwrite */
3540 if (!ext4_feature_set_ok(sb, 0)) {
3545 * Make sure the group descriptor checksums
3546 * are sane. If they aren't, refuse to remount r/w.
3548 for (g = 0; g < sbi->s_groups_count; g++) {
3549 struct ext4_group_desc *gdp =
3550 ext4_get_group_desc(sb, g, NULL);
3552 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3553 ext4_msg(sb, KERN_ERR,
3554 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3555 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3556 le16_to_cpu(gdp->bg_checksum));
3563 * If we have an unprocessed orphan list hanging
3564 * around from a previously readonly bdev mount,
3565 * require a full umount/remount for now.
3567 if (es->s_last_orphan) {
3568 ext4_msg(sb, KERN_WARNING, "Couldn't "
3569 "remount RDWR because of unprocessed "
3570 "orphan inode list. Please "
3571 "umount/remount instead");
3577 * Mounting a RDONLY partition read-write, so reread
3578 * and store the current valid flag. (It may have
3579 * been changed by e2fsck since we originally mounted
3583 ext4_clear_journal_err(sb, es);
3584 sbi->s_mount_state = le16_to_cpu(es->s_state);
3585 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3587 if (!ext4_setup_super(sb, es, 0))
3588 sb->s_flags &= ~MS_RDONLY;
3591 ext4_setup_system_zone(sb);
3592 if (sbi->s_journal == NULL)
3593 ext4_commit_super(sb, 1);
3596 /* Release old quota file names */
3597 for (i = 0; i < MAXQUOTAS; i++)
3598 if (old_opts.s_qf_names[i] &&
3599 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3600 kfree(old_opts.s_qf_names[i]);
3607 sb->s_flags = old_sb_flags;
3608 sbi->s_mount_opt = old_opts.s_mount_opt;
3609 sbi->s_resuid = old_opts.s_resuid;
3610 sbi->s_resgid = old_opts.s_resgid;
3611 sbi->s_commit_interval = old_opts.s_commit_interval;
3612 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3613 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3615 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3616 for (i = 0; i < MAXQUOTAS; i++) {
3617 if (sbi->s_qf_names[i] &&
3618 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3619 kfree(sbi->s_qf_names[i]);
3620 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3628 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3630 struct super_block *sb = dentry->d_sb;
3631 struct ext4_sb_info *sbi = EXT4_SB(sb);
3632 struct ext4_super_block *es = sbi->s_es;
3635 if (test_opt(sb, MINIX_DF)) {
3636 sbi->s_overhead_last = 0;
3637 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3638 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3639 ext4_fsblk_t overhead = 0;
3642 * Compute the overhead (FS structures). This is constant
3643 * for a given filesystem unless the number of block groups
3644 * changes so we cache the previous value until it does.
3648 * All of the blocks before first_data_block are
3651 overhead = le32_to_cpu(es->s_first_data_block);
3654 * Add the overhead attributed to the superblock and
3655 * block group descriptors. If the sparse superblocks
3656 * feature is turned on, then not all groups have this.
3658 for (i = 0; i < ngroups; i++) {
3659 overhead += ext4_bg_has_super(sb, i) +
3660 ext4_bg_num_gdb(sb, i);
3665 * Every block group has an inode bitmap, a block
3666 * bitmap, and an inode table.
3668 overhead += ngroups * (2 + sbi->s_itb_per_group);
3669 sbi->s_overhead_last = overhead;
3671 sbi->s_blocks_last = ext4_blocks_count(es);
3674 buf->f_type = EXT4_SUPER_MAGIC;
3675 buf->f_bsize = sb->s_blocksize;
3676 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3677 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3678 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3679 ext4_free_blocks_count_set(es, buf->f_bfree);
3680 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3681 if (buf->f_bfree < ext4_r_blocks_count(es))
3683 buf->f_files = le32_to_cpu(es->s_inodes_count);
3684 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3685 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3686 buf->f_namelen = EXT4_NAME_LEN;
3687 fsid = le64_to_cpup((void *)es->s_uuid) ^
3688 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3689 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3690 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3695 /* Helper function for writing quotas on sync - we need to start transaction
3696 * before quota file is locked for write. Otherwise the are possible deadlocks:
3697 * Process 1 Process 2
3698 * ext4_create() quota_sync()
3699 * jbd2_journal_start() write_dquot()
3700 * vfs_dq_init() down(dqio_mutex)
3701 * down(dqio_mutex) jbd2_journal_start()
3707 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3709 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3712 static int ext4_write_dquot(struct dquot *dquot)
3716 struct inode *inode;
3718 inode = dquot_to_inode(dquot);
3719 handle = ext4_journal_start(inode,
3720 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3722 return PTR_ERR(handle);
3723 ret = dquot_commit(dquot);
3724 err = ext4_journal_stop(handle);
3730 static int ext4_acquire_dquot(struct dquot *dquot)
3735 handle = ext4_journal_start(dquot_to_inode(dquot),
3736 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3738 return PTR_ERR(handle);
3739 ret = dquot_acquire(dquot);
3740 err = ext4_journal_stop(handle);
3746 static int ext4_release_dquot(struct dquot *dquot)
3751 handle = ext4_journal_start(dquot_to_inode(dquot),
3752 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3753 if (IS_ERR(handle)) {
3754 /* Release dquot anyway to avoid endless cycle in dqput() */
3755 dquot_release(dquot);
3756 return PTR_ERR(handle);
3758 ret = dquot_release(dquot);
3759 err = ext4_journal_stop(handle);
3765 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3767 /* Are we journaling quotas? */
3768 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3769 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3770 dquot_mark_dquot_dirty(dquot);
3771 return ext4_write_dquot(dquot);
3773 return dquot_mark_dquot_dirty(dquot);
3777 static int ext4_write_info(struct super_block *sb, int type)
3782 /* Data block + inode block */
3783 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3785 return PTR_ERR(handle);
3786 ret = dquot_commit_info(sb, type);
3787 err = ext4_journal_stop(handle);
3794 * Turn on quotas during mount time - we need to find
3795 * the quota file and such...
3797 static int ext4_quota_on_mount(struct super_block *sb, int type)
3799 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3800 EXT4_SB(sb)->s_jquota_fmt, type);
3804 * Standard function to be called on quota_on
3806 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3807 char *name, int remount)
3812 if (!test_opt(sb, QUOTA))
3814 /* When remounting, no checks are needed and in fact, name is NULL */
3816 return vfs_quota_on(sb, type, format_id, name, remount);
3818 err = kern_path(name, LOOKUP_FOLLOW, &path);
3822 /* Quotafile not on the same filesystem? */
3823 if (path.mnt->mnt_sb != sb) {
3827 /* Journaling quota? */
3828 if (EXT4_SB(sb)->s_qf_names[type]) {
3829 /* Quotafile not in fs root? */
3830 if (path.dentry->d_parent != sb->s_root)
3831 ext4_msg(sb, KERN_WARNING,
3832 "Quota file not on filesystem root. "
3833 "Journaled quota will not work");
3837 * When we journal data on quota file, we have to flush journal to see
3838 * all updates to the file when we bypass pagecache...
3840 if (EXT4_SB(sb)->s_journal &&
3841 ext4_should_journal_data(path.dentry->d_inode)) {
3843 * We don't need to lock updates but journal_flush() could
3844 * otherwise be livelocked...
3846 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3847 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3848 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3855 err = vfs_quota_on_path(sb, type, format_id, &path);
3860 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3861 * acquiring the locks... As quota files are never truncated and quota code
3862 * itself serializes the operations (and noone else should touch the files)
3863 * we don't have to be afraid of races */
3864 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3865 size_t len, loff_t off)
3867 struct inode *inode = sb_dqopt(sb)->files[type];
3868 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3870 int offset = off & (sb->s_blocksize - 1);
3873 struct buffer_head *bh;
3874 loff_t i_size = i_size_read(inode);
3878 if (off+len > i_size)
3881 while (toread > 0) {
3882 tocopy = sb->s_blocksize - offset < toread ?
3883 sb->s_blocksize - offset : toread;
3884 bh = ext4_bread(NULL, inode, blk, 0, &err);
3887 if (!bh) /* A hole? */
3888 memset(data, 0, tocopy);
3890 memcpy(data, bh->b_data+offset, tocopy);
3900 /* Write to quotafile (we know the transaction is already started and has
3901 * enough credits) */
3902 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3903 const char *data, size_t len, loff_t off)
3905 struct inode *inode = sb_dqopt(sb)->files[type];
3906 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3908 int offset = off & (sb->s_blocksize - 1);
3910 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3911 size_t towrite = len;
3912 struct buffer_head *bh;
3913 handle_t *handle = journal_current_handle();
3915 if (EXT4_SB(sb)->s_journal && !handle) {
3916 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3917 " cancelled because transaction is not started",
3918 (unsigned long long)off, (unsigned long long)len);
3921 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3922 while (towrite > 0) {
3923 tocopy = sb->s_blocksize - offset < towrite ?
3924 sb->s_blocksize - offset : towrite;
3925 bh = ext4_bread(handle, inode, blk, 1, &err);
3928 if (journal_quota) {
3929 err = ext4_journal_get_write_access(handle, bh);
3936 memcpy(bh->b_data+offset, data, tocopy);
3937 flush_dcache_page(bh->b_page);
3940 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3942 /* Always do at least ordered writes for quotas */
3943 err = ext4_jbd2_file_inode(handle, inode);
3944 mark_buffer_dirty(bh);
3955 if (len == towrite) {
3956 mutex_unlock(&inode->i_mutex);
3959 if (inode->i_size < off+len-towrite) {
3960 i_size_write(inode, off+len-towrite);
3961 EXT4_I(inode)->i_disksize = inode->i_size;
3963 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3964 ext4_mark_inode_dirty(handle, inode);
3965 mutex_unlock(&inode->i_mutex);
3966 return len - towrite;
3971 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3972 const char *dev_name, void *data, struct vfsmount *mnt)
3974 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3977 static struct file_system_type ext4_fs_type = {
3978 .owner = THIS_MODULE,
3980 .get_sb = ext4_get_sb,
3981 .kill_sb = kill_block_super,
3982 .fs_flags = FS_REQUIRES_DEV,
3985 #ifdef CONFIG_EXT4DEV_COMPAT
3986 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3987 const char *dev_name, void *data,struct vfsmount *mnt)
3989 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3990 "to mount using ext4\n", dev_name);
3991 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3992 "will go away by 2.6.31\n", dev_name);
3993 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3996 static struct file_system_type ext4dev_fs_type = {
3997 .owner = THIS_MODULE,
3999 .get_sb = ext4dev_get_sb,
4000 .kill_sb = kill_block_super,
4001 .fs_flags = FS_REQUIRES_DEV,
4003 MODULE_ALIAS("ext4dev");
4006 static int __init init_ext4_fs(void)
4010 err = init_ext4_system_zone();
4013 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4016 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4017 err = init_ext4_mballoc();
4021 err = init_ext4_xattr();
4024 err = init_inodecache();
4027 err = register_filesystem(&ext4_fs_type);
4030 #ifdef CONFIG_EXT4DEV_COMPAT
4031 err = register_filesystem(&ext4dev_fs_type);
4033 unregister_filesystem(&ext4_fs_type);
4039 destroy_inodecache();
4043 exit_ext4_mballoc();
4045 remove_proc_entry("fs/ext4", NULL);
4046 kset_unregister(ext4_kset);
4048 exit_ext4_system_zone();
4052 static void __exit exit_ext4_fs(void)
4054 unregister_filesystem(&ext4_fs_type);
4055 #ifdef CONFIG_EXT4DEV_COMPAT
4056 unregister_filesystem(&ext4dev_fs_type);
4058 destroy_inodecache();
4060 exit_ext4_mballoc();
4061 remove_proc_entry("fs/ext4", NULL);
4062 kset_unregister(ext4_kset);
4063 exit_ext4_system_zone();
4066 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4067 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4068 MODULE_LICENSE("GPL");
4069 module_init(init_ext4_fs)
4070 module_exit(exit_ext4_fs)