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 * Wrappers for jbd2_journal_start/end.
195 * The only special thing we need to do here is to make sure that all
196 * journal_end calls result in the superblock being marked dirty, so
197 * that sync() will call the filesystem's write_super callback if
200 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
204 if (sb->s_flags & MS_RDONLY)
205 return ERR_PTR(-EROFS);
207 /* Special case here: if the journal has aborted behind our
208 * backs (eg. EIO in the commit thread), then we still need to
209 * take the FS itself readonly cleanly. */
210 journal = EXT4_SB(sb)->s_journal;
212 if (is_journal_aborted(journal)) {
213 ext4_abort(sb, __func__, "Detected aborted journal");
214 return ERR_PTR(-EROFS);
216 return jbd2_journal_start(journal, nblocks);
219 * We're not journaling, return the appropriate indication.
221 current->journal_info = EXT4_NOJOURNAL_HANDLE;
222 return current->journal_info;
226 * The only special thing we need to do here is to make sure that all
227 * jbd2_journal_stop calls result in the superblock being marked dirty, so
228 * that sync() will call the filesystem's write_super callback if
231 int __ext4_journal_stop(const char *where, handle_t *handle)
233 struct super_block *sb;
237 if (!ext4_handle_valid(handle)) {
239 * Do this here since we don't call jbd2_journal_stop() in
242 current->journal_info = NULL;
245 sb = handle->h_transaction->t_journal->j_private;
247 rc = jbd2_journal_stop(handle);
252 __ext4_std_error(sb, where, err);
256 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
257 struct buffer_head *bh, handle_t *handle, int err)
260 const char *errstr = ext4_decode_error(NULL, err, nbuf);
262 BUG_ON(!ext4_handle_valid(handle));
265 BUFFER_TRACE(bh, "abort");
270 if (is_handle_aborted(handle))
273 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
274 caller, errstr, err_fn);
276 jbd2_journal_abort_handle(handle);
279 /* Deal with the reporting of failure conditions on a filesystem such as
280 * inconsistencies detected or read IO failures.
282 * On ext2, we can store the error state of the filesystem in the
283 * superblock. That is not possible on ext4, because we may have other
284 * write ordering constraints on the superblock which prevent us from
285 * writing it out straight away; and given that the journal is about to
286 * be aborted, we can't rely on the current, or future, transactions to
287 * write out the superblock safely.
289 * We'll just use the jbd2_journal_abort() error code to record an error in
290 * the journal instead. On recovery, the journal will compain about
291 * that error until we've noted it down and cleared it.
294 static void ext4_handle_error(struct super_block *sb)
296 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
301 if (sb->s_flags & MS_RDONLY)
304 if (!test_opt(sb, ERRORS_CONT)) {
305 journal_t *journal = EXT4_SB(sb)->s_journal;
307 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
309 jbd2_journal_abort(journal, -EIO);
311 if (test_opt(sb, ERRORS_RO)) {
312 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
313 sb->s_flags |= MS_RDONLY;
315 ext4_commit_super(sb, 1);
316 if (test_opt(sb, ERRORS_PANIC))
317 panic("EXT4-fs (device %s): panic forced after error\n",
321 void ext4_error(struct super_block *sb, const char *function,
322 const char *fmt, ...)
327 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
332 ext4_handle_error(sb);
335 static const char *ext4_decode_error(struct super_block *sb, int errno,
342 errstr = "IO failure";
345 errstr = "Out of memory";
348 if (!sb || (EXT4_SB(sb)->s_journal &&
349 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
350 errstr = "Journal has aborted";
352 errstr = "Readonly filesystem";
355 /* If the caller passed in an extra buffer for unknown
356 * errors, textualise them now. Else we just return
359 /* Check for truncated error codes... */
360 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
369 /* __ext4_std_error decodes expected errors from journaling functions
370 * automatically and invokes the appropriate error response. */
372 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
377 /* Special case: if the error is EROFS, and we're not already
378 * inside a transaction, then there's really no point in logging
380 if (errno == -EROFS && journal_current_handle() == NULL &&
381 (sb->s_flags & MS_RDONLY))
384 errstr = ext4_decode_error(sb, errno, nbuf);
385 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
386 sb->s_id, function, errstr);
388 ext4_handle_error(sb);
392 * ext4_abort is a much stronger failure handler than ext4_error. The
393 * abort function may be used to deal with unrecoverable failures such
394 * as journal IO errors or ENOMEM at a critical moment in log management.
396 * We unconditionally force the filesystem into an ABORT|READONLY state,
397 * unless the error response on the fs has been set to panic in which
398 * case we take the easy way out and panic immediately.
401 void ext4_abort(struct super_block *sb, const char *function,
402 const char *fmt, ...)
407 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
412 if (test_opt(sb, ERRORS_PANIC))
413 panic("EXT4-fs panic from previous error\n");
415 if (sb->s_flags & MS_RDONLY)
418 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
420 sb->s_flags |= MS_RDONLY;
421 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
422 if (EXT4_SB(sb)->s_journal)
423 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
426 void ext4_msg (struct super_block * sb, const char *prefix,
427 const char *fmt, ...)
432 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
438 void ext4_warning(struct super_block *sb, const char *function,
439 const char *fmt, ...)
444 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
451 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
452 const char *function, const char *fmt, ...)
457 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
460 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
465 if (test_opt(sb, ERRORS_CONT)) {
466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
468 ext4_commit_super(sb, 0);
471 ext4_unlock_group(sb, grp);
472 ext4_handle_error(sb);
474 * We only get here in the ERRORS_RO case; relocking the group
475 * may be dangerous, but nothing bad will happen since the
476 * filesystem will have already been marked read/only and the
477 * journal has been aborted. We return 1 as a hint to callers
478 * who might what to use the return value from
479 * ext4_grp_locked_error() to distinguish beween the
480 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
481 * aggressively from the ext4 function in question, with a
482 * more appropriate error code.
484 ext4_lock_group(sb, grp);
488 void ext4_update_dynamic_rev(struct super_block *sb)
490 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
492 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
495 ext4_warning(sb, __func__,
496 "updating to rev %d because of new feature flag, "
497 "running e2fsck is recommended",
500 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
501 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
502 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
503 /* leave es->s_feature_*compat flags alone */
504 /* es->s_uuid will be set by e2fsck if empty */
507 * The rest of the superblock fields should be zero, and if not it
508 * means they are likely already in use, so leave them alone. We
509 * can leave it up to e2fsck to clean up any inconsistencies there.
514 * Open the external journal device
516 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
518 struct block_device *bdev;
519 char b[BDEVNAME_SIZE];
521 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
527 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
528 __bdevname(dev, b), PTR_ERR(bdev));
533 * Release the journal device
535 static int ext4_blkdev_put(struct block_device *bdev)
538 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
541 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
543 struct block_device *bdev;
546 bdev = sbi->journal_bdev;
548 ret = ext4_blkdev_put(bdev);
549 sbi->journal_bdev = NULL;
554 static inline struct inode *orphan_list_entry(struct list_head *l)
556 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
559 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
563 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
564 le32_to_cpu(sbi->s_es->s_last_orphan));
566 printk(KERN_ERR "sb_info orphan list:\n");
567 list_for_each(l, &sbi->s_orphan) {
568 struct inode *inode = orphan_list_entry(l);
570 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
571 inode->i_sb->s_id, inode->i_ino, inode,
572 inode->i_mode, inode->i_nlink,
577 static void ext4_put_super(struct super_block *sb)
579 struct ext4_sb_info *sbi = EXT4_SB(sb);
580 struct ext4_super_block *es = sbi->s_es;
583 flush_workqueue(sbi->dio_unwritten_wq);
584 destroy_workqueue(sbi->dio_unwritten_wq);
589 ext4_commit_super(sb, 1);
591 ext4_release_system_zone(sb);
593 ext4_ext_release(sb);
594 ext4_xattr_put_super(sb);
595 if (sbi->s_journal) {
596 err = jbd2_journal_destroy(sbi->s_journal);
597 sbi->s_journal = NULL;
599 ext4_abort(sb, __func__,
600 "Couldn't clean up the journal");
602 if (!(sb->s_flags & MS_RDONLY)) {
603 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
604 es->s_state = cpu_to_le16(sbi->s_mount_state);
605 ext4_commit_super(sb, 1);
608 remove_proc_entry(sb->s_id, ext4_proc_root);
610 kobject_del(&sbi->s_kobj);
612 for (i = 0; i < sbi->s_gdb_count; i++)
613 brelse(sbi->s_group_desc[i]);
614 kfree(sbi->s_group_desc);
615 if (is_vmalloc_addr(sbi->s_flex_groups))
616 vfree(sbi->s_flex_groups);
618 kfree(sbi->s_flex_groups);
619 percpu_counter_destroy(&sbi->s_freeblocks_counter);
620 percpu_counter_destroy(&sbi->s_freeinodes_counter);
621 percpu_counter_destroy(&sbi->s_dirs_counter);
622 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
625 for (i = 0; i < MAXQUOTAS; i++)
626 kfree(sbi->s_qf_names[i]);
629 /* Debugging code just in case the in-memory inode orphan list
630 * isn't empty. The on-disk one can be non-empty if we've
631 * detected an error and taken the fs readonly, but the
632 * in-memory list had better be clean by this point. */
633 if (!list_empty(&sbi->s_orphan))
634 dump_orphan_list(sb, sbi);
635 J_ASSERT(list_empty(&sbi->s_orphan));
637 invalidate_bdev(sb->s_bdev);
638 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
640 * Invalidate the journal device's buffers. We don't want them
641 * floating about in memory - the physical journal device may
642 * hotswapped, and it breaks the `ro-after' testing code.
644 sync_blockdev(sbi->journal_bdev);
645 invalidate_bdev(sbi->journal_bdev);
646 ext4_blkdev_remove(sbi);
648 sb->s_fs_info = NULL;
650 * Now that we are completely done shutting down the
651 * superblock, we need to actually destroy the kobject.
655 kobject_put(&sbi->s_kobj);
656 wait_for_completion(&sbi->s_kobj_unregister);
657 kfree(sbi->s_blockgroup_lock);
661 static struct kmem_cache *ext4_inode_cachep;
664 * Called inside transaction, so use GFP_NOFS
666 static struct inode *ext4_alloc_inode(struct super_block *sb)
668 struct ext4_inode_info *ei;
670 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
674 ei->vfs_inode.i_version = 1;
675 ei->vfs_inode.i_data.writeback_index = 0;
676 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
677 INIT_LIST_HEAD(&ei->i_prealloc_list);
678 spin_lock_init(&ei->i_prealloc_lock);
680 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
681 * therefore it can be null here. Don't check it, just initialize
684 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
685 ei->i_reserved_data_blocks = 0;
686 ei->i_reserved_meta_blocks = 0;
687 ei->i_allocated_meta_blocks = 0;
688 ei->i_delalloc_reserved_flag = 0;
689 spin_lock_init(&(ei->i_block_reservation_lock));
691 return &ei->vfs_inode;
694 static void ext4_destroy_inode(struct inode *inode)
696 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
697 ext4_msg(inode->i_sb, KERN_ERR,
698 "Inode %lu (%p): orphan list check failed!",
699 inode->i_ino, EXT4_I(inode));
700 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
701 EXT4_I(inode), sizeof(struct ext4_inode_info),
705 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
708 static void init_once(void *foo)
710 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
712 INIT_LIST_HEAD(&ei->i_orphan);
713 #ifdef CONFIG_EXT4_FS_XATTR
714 init_rwsem(&ei->xattr_sem);
716 init_rwsem(&ei->i_data_sem);
717 inode_init_once(&ei->vfs_inode);
720 static int init_inodecache(void)
722 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
723 sizeof(struct ext4_inode_info),
724 0, (SLAB_RECLAIM_ACCOUNT|
727 if (ext4_inode_cachep == NULL)
732 static void destroy_inodecache(void)
734 kmem_cache_destroy(ext4_inode_cachep);
737 static void ext4_clear_inode(struct inode *inode)
739 ext4_discard_preallocations(inode);
740 if (EXT4_JOURNAL(inode))
741 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
742 &EXT4_I(inode)->jinode);
745 static inline void ext4_show_quota_options(struct seq_file *seq,
746 struct super_block *sb)
748 #if defined(CONFIG_QUOTA)
749 struct ext4_sb_info *sbi = EXT4_SB(sb);
751 if (sbi->s_jquota_fmt)
752 seq_printf(seq, ",jqfmt=%s",
753 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
755 if (sbi->s_qf_names[USRQUOTA])
756 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
758 if (sbi->s_qf_names[GRPQUOTA])
759 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
761 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
762 seq_puts(seq, ",usrquota");
764 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
765 seq_puts(seq, ",grpquota");
771 * - it's set to a non-default value OR
772 * - if the per-sb default is different from the global default
774 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
777 unsigned long def_mount_opts;
778 struct super_block *sb = vfs->mnt_sb;
779 struct ext4_sb_info *sbi = EXT4_SB(sb);
780 struct ext4_super_block *es = sbi->s_es;
782 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
783 def_errors = le16_to_cpu(es->s_errors);
785 if (sbi->s_sb_block != 1)
786 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
787 if (test_opt(sb, MINIX_DF))
788 seq_puts(seq, ",minixdf");
789 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
790 seq_puts(seq, ",grpid");
791 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
792 seq_puts(seq, ",nogrpid");
793 if (sbi->s_resuid != EXT4_DEF_RESUID ||
794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
795 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
797 if (sbi->s_resgid != EXT4_DEF_RESGID ||
798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
799 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
801 if (test_opt(sb, ERRORS_RO)) {
802 if (def_errors == EXT4_ERRORS_PANIC ||
803 def_errors == EXT4_ERRORS_CONTINUE) {
804 seq_puts(seq, ",errors=remount-ro");
807 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
808 seq_puts(seq, ",errors=continue");
809 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
810 seq_puts(seq, ",errors=panic");
811 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
812 seq_puts(seq, ",nouid32");
813 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
814 seq_puts(seq, ",debug");
815 if (test_opt(sb, OLDALLOC))
816 seq_puts(seq, ",oldalloc");
817 #ifdef CONFIG_EXT4_FS_XATTR
818 if (test_opt(sb, XATTR_USER) &&
819 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
820 seq_puts(seq, ",user_xattr");
821 if (!test_opt(sb, XATTR_USER) &&
822 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
823 seq_puts(seq, ",nouser_xattr");
826 #ifdef CONFIG_EXT4_FS_POSIX_ACL
827 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
828 seq_puts(seq, ",acl");
829 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
830 seq_puts(seq, ",noacl");
832 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
833 seq_printf(seq, ",commit=%u",
834 (unsigned) (sbi->s_commit_interval / HZ));
836 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
837 seq_printf(seq, ",min_batch_time=%u",
838 (unsigned) sbi->s_min_batch_time);
840 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
841 seq_printf(seq, ",max_batch_time=%u",
842 (unsigned) sbi->s_min_batch_time);
846 * We're changing the default of barrier mount option, so
847 * let's always display its mount state so it's clear what its
850 seq_puts(seq, ",barrier=");
851 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
852 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
853 seq_puts(seq, ",journal_async_commit");
854 if (test_opt(sb, NOBH))
855 seq_puts(seq, ",nobh");
856 if (test_opt(sb, I_VERSION))
857 seq_puts(seq, ",i_version");
858 if (!test_opt(sb, DELALLOC))
859 seq_puts(seq, ",nodelalloc");
863 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
865 * journal mode get enabled in different ways
866 * So just print the value even if we didn't specify it
868 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
869 seq_puts(seq, ",data=journal");
870 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
871 seq_puts(seq, ",data=ordered");
872 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
873 seq_puts(seq, ",data=writeback");
875 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
876 seq_printf(seq, ",inode_readahead_blks=%u",
877 sbi->s_inode_readahead_blks);
879 if (test_opt(sb, DATA_ERR_ABORT))
880 seq_puts(seq, ",data_err=abort");
882 if (test_opt(sb, NO_AUTO_DA_ALLOC))
883 seq_puts(seq, ",noauto_da_alloc");
885 ext4_show_quota_options(seq, sb);
890 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
891 u64 ino, u32 generation)
895 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
896 return ERR_PTR(-ESTALE);
897 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
898 return ERR_PTR(-ESTALE);
900 /* iget isn't really right if the inode is currently unallocated!!
902 * ext4_read_inode will return a bad_inode if the inode had been
903 * deleted, so we should be safe.
905 * Currently we don't know the generation for parent directory, so
906 * a generation of 0 means "accept any"
908 inode = ext4_iget(sb, ino);
910 return ERR_CAST(inode);
911 if (generation && inode->i_generation != generation) {
913 return ERR_PTR(-ESTALE);
919 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
920 int fh_len, int fh_type)
922 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
926 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
927 int fh_len, int fh_type)
929 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
934 * Try to release metadata pages (indirect blocks, directories) which are
935 * mapped via the block device. Since these pages could have journal heads
936 * which would prevent try_to_free_buffers() from freeing them, we must use
937 * jbd2 layer's try_to_free_buffers() function to release them.
939 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
942 journal_t *journal = EXT4_SB(sb)->s_journal;
944 WARN_ON(PageChecked(page));
945 if (!page_has_buffers(page))
948 return jbd2_journal_try_to_free_buffers(journal, page,
950 return try_to_free_buffers(page);
954 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
955 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
957 static int ext4_write_dquot(struct dquot *dquot);
958 static int ext4_acquire_dquot(struct dquot *dquot);
959 static int ext4_release_dquot(struct dquot *dquot);
960 static int ext4_mark_dquot_dirty(struct dquot *dquot);
961 static int ext4_write_info(struct super_block *sb, int type);
962 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
963 char *path, int remount);
964 static int ext4_quota_on_mount(struct super_block *sb, int type);
965 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
966 size_t len, loff_t off);
967 static ssize_t ext4_quota_write(struct super_block *sb, int type,
968 const char *data, size_t len, loff_t off);
970 static const struct dquot_operations ext4_quota_operations = {
971 .initialize = dquot_initialize,
973 .alloc_space = dquot_alloc_space,
974 .reserve_space = dquot_reserve_space,
975 .claim_space = dquot_claim_space,
976 .release_rsv = dquot_release_reserved_space,
977 .get_reserved_space = ext4_get_reserved_space,
978 .alloc_inode = dquot_alloc_inode,
979 .free_space = dquot_free_space,
980 .free_inode = dquot_free_inode,
981 .transfer = dquot_transfer,
982 .write_dquot = ext4_write_dquot,
983 .acquire_dquot = ext4_acquire_dquot,
984 .release_dquot = ext4_release_dquot,
985 .mark_dirty = ext4_mark_dquot_dirty,
986 .write_info = ext4_write_info,
987 .alloc_dquot = dquot_alloc,
988 .destroy_dquot = dquot_destroy,
991 static const struct quotactl_ops ext4_qctl_operations = {
992 .quota_on = ext4_quota_on,
993 .quota_off = vfs_quota_off,
994 .quota_sync = vfs_quota_sync,
995 .get_info = vfs_get_dqinfo,
996 .set_info = vfs_set_dqinfo,
997 .get_dqblk = vfs_get_dqblk,
998 .set_dqblk = vfs_set_dqblk
1002 static const struct super_operations ext4_sops = {
1003 .alloc_inode = ext4_alloc_inode,
1004 .destroy_inode = ext4_destroy_inode,
1005 .write_inode = ext4_write_inode,
1006 .dirty_inode = ext4_dirty_inode,
1007 .delete_inode = ext4_delete_inode,
1008 .put_super = ext4_put_super,
1009 .sync_fs = ext4_sync_fs,
1010 .freeze_fs = ext4_freeze,
1011 .unfreeze_fs = ext4_unfreeze,
1012 .statfs = ext4_statfs,
1013 .remount_fs = ext4_remount,
1014 .clear_inode = ext4_clear_inode,
1015 .show_options = ext4_show_options,
1017 .quota_read = ext4_quota_read,
1018 .quota_write = ext4_quota_write,
1020 .bdev_try_to_free_page = bdev_try_to_free_page,
1023 static const struct super_operations ext4_nojournal_sops = {
1024 .alloc_inode = ext4_alloc_inode,
1025 .destroy_inode = ext4_destroy_inode,
1026 .write_inode = ext4_write_inode,
1027 .dirty_inode = ext4_dirty_inode,
1028 .delete_inode = ext4_delete_inode,
1029 .write_super = ext4_write_super,
1030 .put_super = ext4_put_super,
1031 .statfs = ext4_statfs,
1032 .remount_fs = ext4_remount,
1033 .clear_inode = ext4_clear_inode,
1034 .show_options = ext4_show_options,
1036 .quota_read = ext4_quota_read,
1037 .quota_write = ext4_quota_write,
1039 .bdev_try_to_free_page = bdev_try_to_free_page,
1042 static const struct export_operations ext4_export_ops = {
1043 .fh_to_dentry = ext4_fh_to_dentry,
1044 .fh_to_parent = ext4_fh_to_parent,
1045 .get_parent = ext4_get_parent,
1049 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1050 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1051 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1052 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1053 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1054 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1055 Opt_journal_update, Opt_journal_dev,
1056 Opt_journal_checksum, Opt_journal_async_commit,
1057 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1058 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1059 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1060 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1061 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1062 Opt_usrquota, Opt_grpquota, Opt_i_version,
1063 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1064 Opt_block_validity, Opt_noblock_validity,
1065 Opt_inode_readahead_blks, Opt_journal_ioprio
1068 static const match_table_t tokens = {
1069 {Opt_bsd_df, "bsddf"},
1070 {Opt_minix_df, "minixdf"},
1071 {Opt_grpid, "grpid"},
1072 {Opt_grpid, "bsdgroups"},
1073 {Opt_nogrpid, "nogrpid"},
1074 {Opt_nogrpid, "sysvgroups"},
1075 {Opt_resgid, "resgid=%u"},
1076 {Opt_resuid, "resuid=%u"},
1078 {Opt_err_cont, "errors=continue"},
1079 {Opt_err_panic, "errors=panic"},
1080 {Opt_err_ro, "errors=remount-ro"},
1081 {Opt_nouid32, "nouid32"},
1082 {Opt_debug, "debug"},
1083 {Opt_oldalloc, "oldalloc"},
1084 {Opt_orlov, "orlov"},
1085 {Opt_user_xattr, "user_xattr"},
1086 {Opt_nouser_xattr, "nouser_xattr"},
1088 {Opt_noacl, "noacl"},
1089 {Opt_noload, "noload"},
1092 {Opt_commit, "commit=%u"},
1093 {Opt_min_batch_time, "min_batch_time=%u"},
1094 {Opt_max_batch_time, "max_batch_time=%u"},
1095 {Opt_journal_update, "journal=update"},
1096 {Opt_journal_dev, "journal_dev=%u"},
1097 {Opt_journal_checksum, "journal_checksum"},
1098 {Opt_journal_async_commit, "journal_async_commit"},
1099 {Opt_abort, "abort"},
1100 {Opt_data_journal, "data=journal"},
1101 {Opt_data_ordered, "data=ordered"},
1102 {Opt_data_writeback, "data=writeback"},
1103 {Opt_data_err_abort, "data_err=abort"},
1104 {Opt_data_err_ignore, "data_err=ignore"},
1105 {Opt_mb_history_length, "mb_history_length=%u"},
1106 {Opt_offusrjquota, "usrjquota="},
1107 {Opt_usrjquota, "usrjquota=%s"},
1108 {Opt_offgrpjquota, "grpjquota="},
1109 {Opt_grpjquota, "grpjquota=%s"},
1110 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1111 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1112 {Opt_grpquota, "grpquota"},
1113 {Opt_noquota, "noquota"},
1114 {Opt_quota, "quota"},
1115 {Opt_usrquota, "usrquota"},
1116 {Opt_barrier, "barrier=%u"},
1117 {Opt_barrier, "barrier"},
1118 {Opt_nobarrier, "nobarrier"},
1119 {Opt_i_version, "i_version"},
1120 {Opt_stripe, "stripe=%u"},
1121 {Opt_resize, "resize"},
1122 {Opt_delalloc, "delalloc"},
1123 {Opt_nodelalloc, "nodelalloc"},
1124 {Opt_block_validity, "block_validity"},
1125 {Opt_noblock_validity, "noblock_validity"},
1126 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1127 {Opt_journal_ioprio, "journal_ioprio=%u"},
1128 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1129 {Opt_auto_da_alloc, "auto_da_alloc"},
1130 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1134 static ext4_fsblk_t get_sb_block(void **data)
1136 ext4_fsblk_t sb_block;
1137 char *options = (char *) *data;
1139 if (!options || strncmp(options, "sb=", 3) != 0)
1140 return 1; /* Default location */
1143 /* TODO: use simple_strtoll with >32bit ext4 */
1144 sb_block = simple_strtoul(options, &options, 0);
1145 if (*options && *options != ',') {
1146 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1150 if (*options == ',')
1152 *data = (void *) options;
1157 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1159 static int parse_options(char *options, struct super_block *sb,
1160 unsigned long *journal_devnum,
1161 unsigned int *journal_ioprio,
1162 ext4_fsblk_t *n_blocks_count, int is_remount)
1164 struct ext4_sb_info *sbi = EXT4_SB(sb);
1166 substring_t args[MAX_OPT_ARGS];
1177 while ((p = strsep(&options, ",")) != NULL) {
1182 token = match_token(p, tokens, args);
1185 clear_opt(sbi->s_mount_opt, MINIX_DF);
1188 set_opt(sbi->s_mount_opt, MINIX_DF);
1191 set_opt(sbi->s_mount_opt, GRPID);
1194 clear_opt(sbi->s_mount_opt, GRPID);
1197 if (match_int(&args[0], &option))
1199 sbi->s_resuid = option;
1202 if (match_int(&args[0], &option))
1204 sbi->s_resgid = option;
1207 /* handled by get_sb_block() instead of here */
1208 /* *sb_block = match_int(&args[0]); */
1211 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1212 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1213 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1216 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1217 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1218 set_opt(sbi->s_mount_opt, ERRORS_RO);
1221 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1222 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1223 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1226 set_opt(sbi->s_mount_opt, NO_UID32);
1229 set_opt(sbi->s_mount_opt, DEBUG);
1232 set_opt(sbi->s_mount_opt, OLDALLOC);
1235 clear_opt(sbi->s_mount_opt, OLDALLOC);
1237 #ifdef CONFIG_EXT4_FS_XATTR
1238 case Opt_user_xattr:
1239 set_opt(sbi->s_mount_opt, XATTR_USER);
1241 case Opt_nouser_xattr:
1242 clear_opt(sbi->s_mount_opt, XATTR_USER);
1245 case Opt_user_xattr:
1246 case Opt_nouser_xattr:
1247 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1250 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1252 set_opt(sbi->s_mount_opt, POSIX_ACL);
1255 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1260 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1263 case Opt_journal_update:
1265 /* Eventually we will want to be able to create
1266 a journal file here. For now, only allow the
1267 user to specify an existing inode to be the
1270 ext4_msg(sb, KERN_ERR,
1271 "Cannot specify journal on remount");
1274 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1276 case Opt_journal_dev:
1278 ext4_msg(sb, KERN_ERR,
1279 "Cannot specify journal on remount");
1282 if (match_int(&args[0], &option))
1284 *journal_devnum = option;
1286 case Opt_journal_checksum:
1287 break; /* Kept for backwards compatibility */
1288 case Opt_journal_async_commit:
1289 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1292 set_opt(sbi->s_mount_opt, NOLOAD);
1295 if (match_int(&args[0], &option))
1300 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1301 sbi->s_commit_interval = HZ * option;
1303 case Opt_max_batch_time:
1304 if (match_int(&args[0], &option))
1309 option = EXT4_DEF_MAX_BATCH_TIME;
1310 sbi->s_max_batch_time = option;
1312 case Opt_min_batch_time:
1313 if (match_int(&args[0], &option))
1317 sbi->s_min_batch_time = option;
1319 case Opt_data_journal:
1320 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1322 case Opt_data_ordered:
1323 data_opt = EXT4_MOUNT_ORDERED_DATA;
1325 case Opt_data_writeback:
1326 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1329 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1331 ext4_msg(sb, KERN_ERR,
1332 "Cannot change data mode on remount");
1336 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1337 sbi->s_mount_opt |= data_opt;
1340 case Opt_data_err_abort:
1341 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1343 case Opt_data_err_ignore:
1344 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1346 case Opt_mb_history_length:
1347 if (match_int(&args[0], &option))
1351 sbi->s_mb_history_max = option;
1360 if (sb_any_quota_loaded(sb) &&
1361 !sbi->s_qf_names[qtype]) {
1362 ext4_msg(sb, KERN_ERR,
1363 "Cannot change journaled "
1364 "quota options when quota turned on");
1367 qname = match_strdup(&args[0]);
1369 ext4_msg(sb, KERN_ERR,
1370 "Not enough memory for "
1371 "storing quotafile name");
1374 if (sbi->s_qf_names[qtype] &&
1375 strcmp(sbi->s_qf_names[qtype], qname)) {
1376 ext4_msg(sb, KERN_ERR,
1377 "%s quota file already "
1378 "specified", QTYPE2NAME(qtype));
1382 sbi->s_qf_names[qtype] = qname;
1383 if (strchr(sbi->s_qf_names[qtype], '/')) {
1384 ext4_msg(sb, KERN_ERR,
1385 "quotafile must be on "
1387 kfree(sbi->s_qf_names[qtype]);
1388 sbi->s_qf_names[qtype] = NULL;
1391 set_opt(sbi->s_mount_opt, QUOTA);
1393 case Opt_offusrjquota:
1396 case Opt_offgrpjquota:
1399 if (sb_any_quota_loaded(sb) &&
1400 sbi->s_qf_names[qtype]) {
1401 ext4_msg(sb, KERN_ERR, "Cannot change "
1402 "journaled quota options when "
1407 * The space will be released later when all options
1408 * are confirmed to be correct
1410 sbi->s_qf_names[qtype] = NULL;
1412 case Opt_jqfmt_vfsold:
1413 qfmt = QFMT_VFS_OLD;
1415 case Opt_jqfmt_vfsv0:
1418 if (sb_any_quota_loaded(sb) &&
1419 sbi->s_jquota_fmt != qfmt) {
1420 ext4_msg(sb, KERN_ERR, "Cannot change "
1421 "journaled quota options when "
1425 sbi->s_jquota_fmt = qfmt;
1429 set_opt(sbi->s_mount_opt, QUOTA);
1430 set_opt(sbi->s_mount_opt, USRQUOTA);
1433 set_opt(sbi->s_mount_opt, QUOTA);
1434 set_opt(sbi->s_mount_opt, GRPQUOTA);
1437 if (sb_any_quota_loaded(sb)) {
1438 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1439 "options when quota turned on");
1442 clear_opt(sbi->s_mount_opt, QUOTA);
1443 clear_opt(sbi->s_mount_opt, USRQUOTA);
1444 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1450 ext4_msg(sb, KERN_ERR,
1451 "quota options not supported");
1455 case Opt_offusrjquota:
1456 case Opt_offgrpjquota:
1457 case Opt_jqfmt_vfsold:
1458 case Opt_jqfmt_vfsv0:
1459 ext4_msg(sb, KERN_ERR,
1460 "journaled quota options not supported");
1466 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1469 clear_opt(sbi->s_mount_opt, BARRIER);
1472 if (match_int(&args[0], &option)) {
1473 set_opt(sbi->s_mount_opt, BARRIER);
1477 set_opt(sbi->s_mount_opt, BARRIER);
1479 clear_opt(sbi->s_mount_opt, BARRIER);
1485 ext4_msg(sb, KERN_ERR,
1486 "resize option only available "
1490 if (match_int(&args[0], &option) != 0)
1492 *n_blocks_count = option;
1495 set_opt(sbi->s_mount_opt, NOBH);
1498 clear_opt(sbi->s_mount_opt, NOBH);
1501 set_opt(sbi->s_mount_opt, I_VERSION);
1502 sb->s_flags |= MS_I_VERSION;
1504 case Opt_nodelalloc:
1505 clear_opt(sbi->s_mount_opt, DELALLOC);
1508 if (match_int(&args[0], &option))
1512 sbi->s_stripe = option;
1515 set_opt(sbi->s_mount_opt, DELALLOC);
1517 case Opt_block_validity:
1518 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1520 case Opt_noblock_validity:
1521 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1523 case Opt_inode_readahead_blks:
1524 if (match_int(&args[0], &option))
1526 if (option < 0 || option > (1 << 30))
1528 if (!is_power_of_2(option)) {
1529 ext4_msg(sb, KERN_ERR,
1530 "EXT4-fs: inode_readahead_blks"
1531 " must be a power of 2");
1534 sbi->s_inode_readahead_blks = option;
1536 case Opt_journal_ioprio:
1537 if (match_int(&args[0], &option))
1539 if (option < 0 || option > 7)
1541 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1544 case Opt_noauto_da_alloc:
1545 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1547 case Opt_auto_da_alloc:
1548 if (match_int(&args[0], &option)) {
1549 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1553 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1555 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1558 ext4_msg(sb, KERN_ERR,
1559 "Unrecognized mount option \"%s\" "
1560 "or missing value", p);
1565 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1566 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1567 sbi->s_qf_names[USRQUOTA])
1568 clear_opt(sbi->s_mount_opt, USRQUOTA);
1570 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1571 sbi->s_qf_names[GRPQUOTA])
1572 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1574 if ((sbi->s_qf_names[USRQUOTA] &&
1575 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1576 (sbi->s_qf_names[GRPQUOTA] &&
1577 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1578 ext4_msg(sb, KERN_ERR, "old and new quota "
1583 if (!sbi->s_jquota_fmt) {
1584 ext4_msg(sb, KERN_ERR, "journaled quota format "
1589 if (sbi->s_jquota_fmt) {
1590 ext4_msg(sb, KERN_ERR, "journaled quota format "
1591 "specified with no journaling "
1600 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1603 struct ext4_sb_info *sbi = EXT4_SB(sb);
1606 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1607 ext4_msg(sb, KERN_ERR, "revision level too high, "
1608 "forcing read-only mode");
1613 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1614 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1615 "running e2fsck is recommended");
1616 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1617 ext4_msg(sb, KERN_WARNING,
1618 "warning: mounting fs with errors, "
1619 "running e2fsck is recommended");
1620 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1621 le16_to_cpu(es->s_mnt_count) >=
1622 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1623 ext4_msg(sb, KERN_WARNING,
1624 "warning: maximal mount count reached, "
1625 "running e2fsck is recommended");
1626 else if (le32_to_cpu(es->s_checkinterval) &&
1627 (le32_to_cpu(es->s_lastcheck) +
1628 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1629 ext4_msg(sb, KERN_WARNING,
1630 "warning: checktime reached, "
1631 "running e2fsck is recommended");
1632 if (!sbi->s_journal)
1633 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1634 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1635 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1636 le16_add_cpu(&es->s_mnt_count, 1);
1637 es->s_mtime = cpu_to_le32(get_seconds());
1638 ext4_update_dynamic_rev(sb);
1640 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1642 ext4_commit_super(sb, 1);
1643 if (test_opt(sb, DEBUG))
1644 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1645 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1647 sbi->s_groups_count,
1648 EXT4_BLOCKS_PER_GROUP(sb),
1649 EXT4_INODES_PER_GROUP(sb),
1652 if (EXT4_SB(sb)->s_journal) {
1653 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1654 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1655 "external", EXT4_SB(sb)->s_journal->j_devname);
1657 ext4_msg(sb, KERN_INFO, "no journal");
1662 static int ext4_fill_flex_info(struct super_block *sb)
1664 struct ext4_sb_info *sbi = EXT4_SB(sb);
1665 struct ext4_group_desc *gdp = NULL;
1666 ext4_group_t flex_group_count;
1667 ext4_group_t flex_group;
1668 int groups_per_flex = 0;
1672 if (!sbi->s_es->s_log_groups_per_flex) {
1673 sbi->s_log_groups_per_flex = 0;
1677 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1678 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1680 /* We allocate both existing and potentially added groups */
1681 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1682 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1683 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1684 size = flex_group_count * sizeof(struct flex_groups);
1685 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1686 if (sbi->s_flex_groups == NULL) {
1687 sbi->s_flex_groups = vmalloc(size);
1688 if (sbi->s_flex_groups)
1689 memset(sbi->s_flex_groups, 0, size);
1691 if (sbi->s_flex_groups == NULL) {
1692 ext4_msg(sb, KERN_ERR, "not enough memory for "
1693 "%u flex groups", flex_group_count);
1697 for (i = 0; i < sbi->s_groups_count; i++) {
1698 gdp = ext4_get_group_desc(sb, i, NULL);
1700 flex_group = ext4_flex_group(sbi, i);
1701 atomic_add(ext4_free_inodes_count(sb, gdp),
1702 &sbi->s_flex_groups[flex_group].free_inodes);
1703 atomic_add(ext4_free_blks_count(sb, gdp),
1704 &sbi->s_flex_groups[flex_group].free_blocks);
1705 atomic_add(ext4_used_dirs_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].used_dirs);
1714 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1715 struct ext4_group_desc *gdp)
1719 if (sbi->s_es->s_feature_ro_compat &
1720 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1721 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1722 __le32 le_group = cpu_to_le32(block_group);
1724 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1725 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1726 crc = crc16(crc, (__u8 *)gdp, offset);
1727 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1728 /* for checksum of struct ext4_group_desc do the rest...*/
1729 if ((sbi->s_es->s_feature_incompat &
1730 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1731 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1732 crc = crc16(crc, (__u8 *)gdp + offset,
1733 le16_to_cpu(sbi->s_es->s_desc_size) -
1737 return cpu_to_le16(crc);
1740 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1741 struct ext4_group_desc *gdp)
1743 if ((sbi->s_es->s_feature_ro_compat &
1744 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1745 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1751 /* Called at mount-time, super-block is locked */
1752 static int ext4_check_descriptors(struct super_block *sb)
1754 struct ext4_sb_info *sbi = EXT4_SB(sb);
1755 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1756 ext4_fsblk_t last_block;
1757 ext4_fsblk_t block_bitmap;
1758 ext4_fsblk_t inode_bitmap;
1759 ext4_fsblk_t inode_table;
1760 int flexbg_flag = 0;
1763 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1766 ext4_debug("Checking group descriptors");
1768 for (i = 0; i < sbi->s_groups_count; i++) {
1769 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1771 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1772 last_block = ext4_blocks_count(sbi->s_es) - 1;
1774 last_block = first_block +
1775 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1777 block_bitmap = ext4_block_bitmap(sb, gdp);
1778 if (block_bitmap < first_block || block_bitmap > last_block) {
1779 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1780 "Block bitmap for group %u not in group "
1781 "(block %llu)!", i, block_bitmap);
1784 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1785 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1786 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1787 "Inode bitmap for group %u not in group "
1788 "(block %llu)!", i, inode_bitmap);
1791 inode_table = ext4_inode_table(sb, gdp);
1792 if (inode_table < first_block ||
1793 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1794 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1795 "Inode table for group %u not in group "
1796 "(block %llu)!", i, inode_table);
1799 ext4_lock_group(sb, i);
1800 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1801 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1802 "Checksum for group %u failed (%u!=%u)",
1803 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1804 gdp)), le16_to_cpu(gdp->bg_checksum));
1805 if (!(sb->s_flags & MS_RDONLY)) {
1806 ext4_unlock_group(sb, i);
1810 ext4_unlock_group(sb, i);
1812 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1815 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1816 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1820 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1821 * the superblock) which were deleted from all directories, but held open by
1822 * a process at the time of a crash. We walk the list and try to delete these
1823 * inodes at recovery time (only with a read-write filesystem).
1825 * In order to keep the orphan inode chain consistent during traversal (in
1826 * case of crash during recovery), we link each inode into the superblock
1827 * orphan list_head and handle it the same way as an inode deletion during
1828 * normal operation (which journals the operations for us).
1830 * We only do an iget() and an iput() on each inode, which is very safe if we
1831 * accidentally point at an in-use or already deleted inode. The worst that
1832 * can happen in this case is that we get a "bit already cleared" message from
1833 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1834 * e2fsck was run on this filesystem, and it must have already done the orphan
1835 * inode cleanup for us, so we can safely abort without any further action.
1837 static void ext4_orphan_cleanup(struct super_block *sb,
1838 struct ext4_super_block *es)
1840 unsigned int s_flags = sb->s_flags;
1841 int nr_orphans = 0, nr_truncates = 0;
1845 if (!es->s_last_orphan) {
1846 jbd_debug(4, "no orphan inodes to clean up\n");
1850 if (bdev_read_only(sb->s_bdev)) {
1851 ext4_msg(sb, KERN_ERR, "write access "
1852 "unavailable, skipping orphan cleanup");
1856 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1857 if (es->s_last_orphan)
1858 jbd_debug(1, "Errors on filesystem, "
1859 "clearing orphan list.\n");
1860 es->s_last_orphan = 0;
1861 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1865 if (s_flags & MS_RDONLY) {
1866 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1867 sb->s_flags &= ~MS_RDONLY;
1870 /* Needed for iput() to work correctly and not trash data */
1871 sb->s_flags |= MS_ACTIVE;
1872 /* Turn on quotas so that they are updated correctly */
1873 for (i = 0; i < MAXQUOTAS; i++) {
1874 if (EXT4_SB(sb)->s_qf_names[i]) {
1875 int ret = ext4_quota_on_mount(sb, i);
1877 ext4_msg(sb, KERN_ERR,
1878 "Cannot turn on journaled "
1879 "quota: error %d", ret);
1884 while (es->s_last_orphan) {
1885 struct inode *inode;
1887 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1888 if (IS_ERR(inode)) {
1889 es->s_last_orphan = 0;
1893 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1895 if (inode->i_nlink) {
1896 ext4_msg(sb, KERN_DEBUG,
1897 "%s: truncating inode %lu to %lld bytes",
1898 __func__, inode->i_ino, inode->i_size);
1899 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1900 inode->i_ino, inode->i_size);
1901 ext4_truncate(inode);
1904 ext4_msg(sb, KERN_DEBUG,
1905 "%s: deleting unreferenced inode %lu",
1906 __func__, inode->i_ino);
1907 jbd_debug(2, "deleting unreferenced inode %lu\n",
1911 iput(inode); /* The delete magic happens here! */
1914 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1917 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1918 PLURAL(nr_orphans));
1920 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1921 PLURAL(nr_truncates));
1923 /* Turn quotas off */
1924 for (i = 0; i < MAXQUOTAS; i++) {
1925 if (sb_dqopt(sb)->files[i])
1926 vfs_quota_off(sb, i, 0);
1929 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1933 * Maximal extent format file size.
1934 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1935 * extent format containers, within a sector_t, and within i_blocks
1936 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1937 * so that won't be a limiting factor.
1939 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1941 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1944 loff_t upper_limit = MAX_LFS_FILESIZE;
1946 /* small i_blocks in vfs inode? */
1947 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1949 * CONFIG_LBDAF is not enabled implies the inode
1950 * i_block represent total blocks in 512 bytes
1951 * 32 == size of vfs inode i_blocks * 8
1953 upper_limit = (1LL << 32) - 1;
1955 /* total blocks in file system block size */
1956 upper_limit >>= (blkbits - 9);
1957 upper_limit <<= blkbits;
1960 /* 32-bit extent-start container, ee_block */
1965 /* Sanity check against vm- & vfs- imposed limits */
1966 if (res > upper_limit)
1973 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1974 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1975 * We need to be 1 filesystem block less than the 2^48 sector limit.
1977 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1979 loff_t res = EXT4_NDIR_BLOCKS;
1982 /* This is calculated to be the largest file size for a dense, block
1983 * mapped file such that the file's total number of 512-byte sectors,
1984 * including data and all indirect blocks, does not exceed (2^48 - 1).
1986 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1987 * number of 512-byte sectors of the file.
1990 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1992 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1993 * the inode i_block field represents total file blocks in
1994 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1996 upper_limit = (1LL << 32) - 1;
1998 /* total blocks in file system block size */
1999 upper_limit >>= (bits - 9);
2003 * We use 48 bit ext4_inode i_blocks
2004 * With EXT4_HUGE_FILE_FL set the i_blocks
2005 * represent total number of blocks in
2006 * file system block size
2008 upper_limit = (1LL << 48) - 1;
2012 /* indirect blocks */
2014 /* double indirect blocks */
2015 meta_blocks += 1 + (1LL << (bits-2));
2016 /* tripple indirect blocks */
2017 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2019 upper_limit -= meta_blocks;
2020 upper_limit <<= bits;
2022 res += 1LL << (bits-2);
2023 res += 1LL << (2*(bits-2));
2024 res += 1LL << (3*(bits-2));
2026 if (res > upper_limit)
2029 if (res > MAX_LFS_FILESIZE)
2030 res = MAX_LFS_FILESIZE;
2035 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2036 ext4_fsblk_t logical_sb_block, int nr)
2038 struct ext4_sb_info *sbi = EXT4_SB(sb);
2039 ext4_group_t bg, first_meta_bg;
2042 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2044 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2046 return logical_sb_block + nr + 1;
2047 bg = sbi->s_desc_per_block * nr;
2048 if (ext4_bg_has_super(sb, bg))
2051 return (has_super + ext4_group_first_block_no(sb, bg));
2055 * ext4_get_stripe_size: Get the stripe size.
2056 * @sbi: In memory super block info
2058 * If we have specified it via mount option, then
2059 * use the mount option value. If the value specified at mount time is
2060 * greater than the blocks per group use the super block value.
2061 * If the super block value is greater than blocks per group return 0.
2062 * Allocator needs it be less than blocks per group.
2065 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2067 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2068 unsigned long stripe_width =
2069 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2071 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2072 return sbi->s_stripe;
2074 if (stripe_width <= sbi->s_blocks_per_group)
2075 return stripe_width;
2077 if (stride <= sbi->s_blocks_per_group)
2086 struct attribute attr;
2087 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2088 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2089 const char *, size_t);
2093 static int parse_strtoul(const char *buf,
2094 unsigned long max, unsigned long *value)
2098 while (*buf && isspace(*buf))
2100 *value = simple_strtoul(buf, &endp, 0);
2101 while (*endp && isspace(*endp))
2103 if (*endp || *value > max)
2109 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2110 struct ext4_sb_info *sbi,
2113 return snprintf(buf, PAGE_SIZE, "%llu\n",
2114 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2117 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2118 struct ext4_sb_info *sbi, char *buf)
2120 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2122 return snprintf(buf, PAGE_SIZE, "%lu\n",
2123 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2124 sbi->s_sectors_written_start) >> 1);
2127 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2128 struct ext4_sb_info *sbi, char *buf)
2130 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2132 return snprintf(buf, PAGE_SIZE, "%llu\n",
2133 sbi->s_kbytes_written +
2134 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2135 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2138 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2139 struct ext4_sb_info *sbi,
2140 const char *buf, size_t count)
2144 if (parse_strtoul(buf, 0x40000000, &t))
2147 if (!is_power_of_2(t))
2150 sbi->s_inode_readahead_blks = t;
2154 static ssize_t sbi_ui_show(struct ext4_attr *a,
2155 struct ext4_sb_info *sbi, char *buf)
2157 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2159 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2162 static ssize_t sbi_ui_store(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi,
2164 const char *buf, size_t count)
2166 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2169 if (parse_strtoul(buf, 0xffffffff, &t))
2175 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2176 static struct ext4_attr ext4_attr_##_name = { \
2177 .attr = {.name = __stringify(_name), .mode = _mode }, \
2180 .offset = offsetof(struct ext4_sb_info, _elname), \
2182 #define EXT4_ATTR(name, mode, show, store) \
2183 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2185 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2186 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2187 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2188 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2189 #define ATTR_LIST(name) &ext4_attr_##name.attr
2191 EXT4_RO_ATTR(delayed_allocation_blocks);
2192 EXT4_RO_ATTR(session_write_kbytes);
2193 EXT4_RO_ATTR(lifetime_write_kbytes);
2194 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2195 inode_readahead_blks_store, s_inode_readahead_blks);
2196 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2197 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2198 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2199 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2200 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2201 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2202 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2203 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2205 static struct attribute *ext4_attrs[] = {
2206 ATTR_LIST(delayed_allocation_blocks),
2207 ATTR_LIST(session_write_kbytes),
2208 ATTR_LIST(lifetime_write_kbytes),
2209 ATTR_LIST(inode_readahead_blks),
2210 ATTR_LIST(inode_goal),
2211 ATTR_LIST(mb_stats),
2212 ATTR_LIST(mb_max_to_scan),
2213 ATTR_LIST(mb_min_to_scan),
2214 ATTR_LIST(mb_order2_req),
2215 ATTR_LIST(mb_stream_req),
2216 ATTR_LIST(mb_group_prealloc),
2217 ATTR_LIST(max_writeback_mb_bump),
2221 static ssize_t ext4_attr_show(struct kobject *kobj,
2222 struct attribute *attr, char *buf)
2224 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2226 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2228 return a->show ? a->show(a, sbi, buf) : 0;
2231 static ssize_t ext4_attr_store(struct kobject *kobj,
2232 struct attribute *attr,
2233 const char *buf, size_t len)
2235 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2237 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2239 return a->store ? a->store(a, sbi, buf, len) : 0;
2242 static void ext4_sb_release(struct kobject *kobj)
2244 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2246 complete(&sbi->s_kobj_unregister);
2250 static struct sysfs_ops ext4_attr_ops = {
2251 .show = ext4_attr_show,
2252 .store = ext4_attr_store,
2255 static struct kobj_type ext4_ktype = {
2256 .default_attrs = ext4_attrs,
2257 .sysfs_ops = &ext4_attr_ops,
2258 .release = ext4_sb_release,
2262 * Check whether this filesystem can be mounted based on
2263 * the features present and the RDONLY/RDWR mount requested.
2264 * Returns 1 if this filesystem can be mounted as requested,
2265 * 0 if it cannot be.
2267 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2269 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2270 ext4_msg(sb, KERN_ERR,
2271 "Couldn't mount because of "
2272 "unsupported optional features (%x)",
2273 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2274 ~EXT4_FEATURE_INCOMPAT_SUPP));
2281 /* Check that feature set is OK for a read-write mount */
2282 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2283 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2284 "unsupported optional features (%x)",
2285 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2286 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2290 * Large file size enabled file system can only be mounted
2291 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2293 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2294 if (sizeof(blkcnt_t) < sizeof(u64)) {
2295 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2296 "cannot be mounted RDWR without "
2304 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2305 __releases(kernel_lock)
2306 __acquires(kernel_lock)
2308 struct buffer_head *bh;
2309 struct ext4_super_block *es = NULL;
2310 struct ext4_sb_info *sbi;
2312 ext4_fsblk_t sb_block = get_sb_block(&data);
2313 ext4_fsblk_t logical_sb_block;
2314 unsigned long offset = 0;
2315 unsigned long journal_devnum = 0;
2316 unsigned long def_mount_opts;
2322 unsigned int db_count;
2324 int needs_recovery, has_huge_files;
2327 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2329 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2333 sbi->s_blockgroup_lock =
2334 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2335 if (!sbi->s_blockgroup_lock) {
2339 sb->s_fs_info = sbi;
2340 sbi->s_mount_opt = 0;
2341 sbi->s_resuid = EXT4_DEF_RESUID;
2342 sbi->s_resgid = EXT4_DEF_RESGID;
2343 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2344 sbi->s_sb_block = sb_block;
2345 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2350 /* Cleanup superblock name */
2351 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2354 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2356 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2361 * The ext4 superblock will not be buffer aligned for other than 1kB
2362 * block sizes. We need to calculate the offset from buffer start.
2364 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2365 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2366 offset = do_div(logical_sb_block, blocksize);
2368 logical_sb_block = sb_block;
2371 if (!(bh = sb_bread(sb, logical_sb_block))) {
2372 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2376 * Note: s_es must be initialized as soon as possible because
2377 * some ext4 macro-instructions depend on its value
2379 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2381 sb->s_magic = le16_to_cpu(es->s_magic);
2382 if (sb->s_magic != EXT4_SUPER_MAGIC)
2384 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2386 /* Set defaults before we parse the mount options */
2387 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2388 if (def_mount_opts & EXT4_DEFM_DEBUG)
2389 set_opt(sbi->s_mount_opt, DEBUG);
2390 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2391 set_opt(sbi->s_mount_opt, GRPID);
2392 if (def_mount_opts & EXT4_DEFM_UID16)
2393 set_opt(sbi->s_mount_opt, NO_UID32);
2394 #ifdef CONFIG_EXT4_FS_XATTR
2395 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2396 set_opt(sbi->s_mount_opt, XATTR_USER);
2398 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2399 if (def_mount_opts & EXT4_DEFM_ACL)
2400 set_opt(sbi->s_mount_opt, POSIX_ACL);
2402 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2403 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2404 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2405 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2406 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2407 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2409 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2410 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2411 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2412 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2414 set_opt(sbi->s_mount_opt, ERRORS_RO);
2416 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2417 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2418 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2419 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2420 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2421 sbi->s_mb_history_max = default_mb_history_length;
2423 set_opt(sbi->s_mount_opt, BARRIER);
2426 * enable delayed allocation by default
2427 * Use -o nodelalloc to turn it off
2429 set_opt(sbi->s_mount_opt, DELALLOC);
2431 if (!parse_options((char *) data, sb, &journal_devnum,
2432 &journal_ioprio, NULL, 0))
2435 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2436 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2438 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2439 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2440 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2441 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2442 ext4_msg(sb, KERN_WARNING,
2443 "feature flags set on rev 0 fs, "
2444 "running e2fsck is recommended");
2447 * Check feature flags regardless of the revision level, since we
2448 * previously didn't change the revision level when setting the flags,
2449 * so there is a chance incompat flags are set on a rev 0 filesystem.
2451 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2454 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2456 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2457 blocksize > EXT4_MAX_BLOCK_SIZE) {
2458 ext4_msg(sb, KERN_ERR,
2459 "Unsupported filesystem blocksize %d", blocksize);
2463 if (sb->s_blocksize != blocksize) {
2464 /* Validate the filesystem blocksize */
2465 if (!sb_set_blocksize(sb, blocksize)) {
2466 ext4_msg(sb, KERN_ERR, "bad block size %d",
2472 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2473 offset = do_div(logical_sb_block, blocksize);
2474 bh = sb_bread(sb, logical_sb_block);
2476 ext4_msg(sb, KERN_ERR,
2477 "Can't read superblock on 2nd try");
2480 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2482 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2483 ext4_msg(sb, KERN_ERR,
2484 "Magic mismatch, very weird!");
2489 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2490 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2491 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2493 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2495 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2496 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2497 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2499 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2500 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2501 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2502 (!is_power_of_2(sbi->s_inode_size)) ||
2503 (sbi->s_inode_size > blocksize)) {
2504 ext4_msg(sb, KERN_ERR,
2505 "unsupported inode size: %d",
2509 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2510 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2513 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2514 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2515 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2516 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2517 !is_power_of_2(sbi->s_desc_size)) {
2518 ext4_msg(sb, KERN_ERR,
2519 "unsupported descriptor size %lu",
2524 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2526 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2527 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2528 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2531 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2532 if (sbi->s_inodes_per_block == 0)
2534 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2535 sbi->s_inodes_per_block;
2536 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2538 sbi->s_mount_state = le16_to_cpu(es->s_state);
2539 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2540 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2542 for (i = 0; i < 4; i++)
2543 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2544 sbi->s_def_hash_version = es->s_def_hash_version;
2545 i = le32_to_cpu(es->s_flags);
2546 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2547 sbi->s_hash_unsigned = 3;
2548 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2549 #ifdef __CHAR_UNSIGNED__
2550 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2551 sbi->s_hash_unsigned = 3;
2553 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2558 if (sbi->s_blocks_per_group > blocksize * 8) {
2559 ext4_msg(sb, KERN_ERR,
2560 "#blocks per group too big: %lu",
2561 sbi->s_blocks_per_group);
2564 if (sbi->s_inodes_per_group > blocksize * 8) {
2565 ext4_msg(sb, KERN_ERR,
2566 "#inodes per group too big: %lu",
2567 sbi->s_inodes_per_group);
2572 * Test whether we have more sectors than will fit in sector_t,
2573 * and whether the max offset is addressable by the page cache.
2575 if ((ext4_blocks_count(es) >
2576 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2577 (ext4_blocks_count(es) >
2578 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2579 ext4_msg(sb, KERN_ERR, "filesystem"
2580 " too large to mount safely on this system");
2581 if (sizeof(sector_t) < 8)
2582 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2587 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2590 /* check blocks count against device size */
2591 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2592 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2593 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2594 "exceeds size of device (%llu blocks)",
2595 ext4_blocks_count(es), blocks_count);
2600 * It makes no sense for the first data block to be beyond the end
2601 * of the filesystem.
2603 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2604 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2605 "block %u is beyond end of filesystem (%llu)",
2606 le32_to_cpu(es->s_first_data_block),
2607 ext4_blocks_count(es));
2610 blocks_count = (ext4_blocks_count(es) -
2611 le32_to_cpu(es->s_first_data_block) +
2612 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2613 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2614 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2615 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2616 "(block count %llu, first data block %u, "
2617 "blocks per group %lu)", sbi->s_groups_count,
2618 ext4_blocks_count(es),
2619 le32_to_cpu(es->s_first_data_block),
2620 EXT4_BLOCKS_PER_GROUP(sb));
2623 sbi->s_groups_count = blocks_count;
2624 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2625 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2626 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2627 EXT4_DESC_PER_BLOCK(sb);
2628 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2630 if (sbi->s_group_desc == NULL) {
2631 ext4_msg(sb, KERN_ERR, "not enough memory");
2635 #ifdef CONFIG_PROC_FS
2637 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2640 bgl_lock_init(sbi->s_blockgroup_lock);
2642 for (i = 0; i < db_count; i++) {
2643 block = descriptor_loc(sb, logical_sb_block, i);
2644 sbi->s_group_desc[i] = sb_bread(sb, block);
2645 if (!sbi->s_group_desc[i]) {
2646 ext4_msg(sb, KERN_ERR,
2647 "can't read group descriptor %d", i);
2652 if (!ext4_check_descriptors(sb)) {
2653 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2657 if (!ext4_fill_flex_info(sb)) {
2658 ext4_msg(sb, KERN_ERR,
2659 "unable to initialize "
2660 "flex_bg meta info!");
2664 sbi->s_gdb_count = db_count;
2665 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2666 spin_lock_init(&sbi->s_next_gen_lock);
2668 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2669 ext4_count_free_blocks(sb));
2671 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2672 ext4_count_free_inodes(sb));
2675 err = percpu_counter_init(&sbi->s_dirs_counter,
2676 ext4_count_dirs(sb));
2679 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2682 ext4_msg(sb, KERN_ERR, "insufficient memory");
2686 sbi->s_stripe = ext4_get_stripe_size(sbi);
2687 sbi->s_max_writeback_mb_bump = 128;
2690 * set up enough so that it can read an inode
2692 if (!test_opt(sb, NOLOAD) &&
2693 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2694 sb->s_op = &ext4_sops;
2696 sb->s_op = &ext4_nojournal_sops;
2697 sb->s_export_op = &ext4_export_ops;
2698 sb->s_xattr = ext4_xattr_handlers;
2700 sb->s_qcop = &ext4_qctl_operations;
2701 sb->dq_op = &ext4_quota_operations;
2703 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2704 mutex_init(&sbi->s_orphan_lock);
2705 mutex_init(&sbi->s_resize_lock);
2709 needs_recovery = (es->s_last_orphan != 0 ||
2710 EXT4_HAS_INCOMPAT_FEATURE(sb,
2711 EXT4_FEATURE_INCOMPAT_RECOVER));
2714 * The first inode we look at is the journal inode. Don't try
2715 * root first: it may be modified in the journal!
2717 if (!test_opt(sb, NOLOAD) &&
2718 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2719 if (ext4_load_journal(sb, es, journal_devnum))
2721 if (!(sb->s_flags & MS_RDONLY) &&
2722 EXT4_SB(sb)->s_journal->j_failed_commit) {
2723 ext4_msg(sb, KERN_CRIT, "error: "
2724 "ext4_fill_super: Journal transaction "
2726 EXT4_SB(sb)->s_journal->j_failed_commit);
2727 if (test_opt(sb, ERRORS_RO)) {
2728 ext4_msg(sb, KERN_CRIT,
2729 "Mounting filesystem read-only");
2730 sb->s_flags |= MS_RDONLY;
2731 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2732 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2734 if (test_opt(sb, ERRORS_PANIC)) {
2735 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2736 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2737 ext4_commit_super(sb, 1);
2741 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2742 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2743 ext4_msg(sb, KERN_ERR, "required journal recovery "
2744 "suppressed and not mounted read-only");
2747 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2748 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2749 sbi->s_journal = NULL;
2754 if (ext4_blocks_count(es) > 0xffffffffULL &&
2755 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2756 JBD2_FEATURE_INCOMPAT_64BIT)) {
2757 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2761 jbd2_journal_set_features(sbi->s_journal,
2762 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2763 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2764 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2765 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2767 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2768 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2770 /* We have now updated the journal if required, so we can
2771 * validate the data journaling mode. */
2772 switch (test_opt(sb, DATA_FLAGS)) {
2774 /* No mode set, assume a default based on the journal
2775 * capabilities: ORDERED_DATA if the journal can
2776 * cope, else JOURNAL_DATA
2778 if (jbd2_journal_check_available_features
2779 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2780 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2782 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2785 case EXT4_MOUNT_ORDERED_DATA:
2786 case EXT4_MOUNT_WRITEBACK_DATA:
2787 if (!jbd2_journal_check_available_features
2788 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2789 ext4_msg(sb, KERN_ERR, "Journal does not support "
2790 "requested data journaling mode");
2796 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2800 if (test_opt(sb, NOBH)) {
2801 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2802 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2803 "its supported only with writeback mode");
2804 clear_opt(sbi->s_mount_opt, NOBH);
2807 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2808 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2809 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2810 goto failed_mount_wq;
2814 * The jbd2_journal_load will have done any necessary log recovery,
2815 * so we can safely mount the rest of the filesystem now.
2818 root = ext4_iget(sb, EXT4_ROOT_INO);
2820 ext4_msg(sb, KERN_ERR, "get root inode failed");
2821 ret = PTR_ERR(root);
2824 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2826 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2829 sb->s_root = d_alloc_root(root);
2831 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2837 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2839 /* determine the minimum size of new large inodes, if present */
2840 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2841 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2842 EXT4_GOOD_OLD_INODE_SIZE;
2843 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2844 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2845 if (sbi->s_want_extra_isize <
2846 le16_to_cpu(es->s_want_extra_isize))
2847 sbi->s_want_extra_isize =
2848 le16_to_cpu(es->s_want_extra_isize);
2849 if (sbi->s_want_extra_isize <
2850 le16_to_cpu(es->s_min_extra_isize))
2851 sbi->s_want_extra_isize =
2852 le16_to_cpu(es->s_min_extra_isize);
2855 /* Check if enough inode space is available */
2856 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2857 sbi->s_inode_size) {
2858 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2859 EXT4_GOOD_OLD_INODE_SIZE;
2860 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2865 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2866 "requested data journaling mode");
2867 clear_opt(sbi->s_mount_opt, DELALLOC);
2868 } else if (test_opt(sb, DELALLOC))
2869 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2871 err = ext4_setup_system_zone(sb);
2873 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2874 "zone (%d)\n", err);
2879 err = ext4_mb_init(sb, needs_recovery);
2881 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2886 sbi->s_kobj.kset = ext4_kset;
2887 init_completion(&sbi->s_kobj_unregister);
2888 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2891 ext4_mb_release(sb);
2892 ext4_ext_release(sb);
2896 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2897 ext4_orphan_cleanup(sb, es);
2898 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2899 if (needs_recovery) {
2900 ext4_msg(sb, KERN_INFO, "recovery complete");
2901 ext4_mark_recovery_complete(sb, es);
2903 if (EXT4_SB(sb)->s_journal) {
2904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2905 descr = " journalled data mode";
2906 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2907 descr = " ordered data mode";
2909 descr = " writeback data mode";
2911 descr = "out journal";
2913 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2920 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2924 ext4_msg(sb, KERN_ERR, "mount failed");
2925 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2927 ext4_release_system_zone(sb);
2928 if (sbi->s_journal) {
2929 jbd2_journal_destroy(sbi->s_journal);
2930 sbi->s_journal = NULL;
2933 if (sbi->s_flex_groups) {
2934 if (is_vmalloc_addr(sbi->s_flex_groups))
2935 vfree(sbi->s_flex_groups);
2937 kfree(sbi->s_flex_groups);
2939 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2940 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2941 percpu_counter_destroy(&sbi->s_dirs_counter);
2942 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2944 for (i = 0; i < db_count; i++)
2945 brelse(sbi->s_group_desc[i]);
2946 kfree(sbi->s_group_desc);
2949 remove_proc_entry(sb->s_id, ext4_proc_root);
2952 for (i = 0; i < MAXQUOTAS; i++)
2953 kfree(sbi->s_qf_names[i]);
2955 ext4_blkdev_remove(sbi);
2958 sb->s_fs_info = NULL;
2959 kfree(sbi->s_blockgroup_lock);
2966 * Setup any per-fs journal parameters now. We'll do this both on
2967 * initial mount, once the journal has been initialised but before we've
2968 * done any recovery; and again on any subsequent remount.
2970 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2972 struct ext4_sb_info *sbi = EXT4_SB(sb);
2974 journal->j_commit_interval = sbi->s_commit_interval;
2975 journal->j_min_batch_time = sbi->s_min_batch_time;
2976 journal->j_max_batch_time = sbi->s_max_batch_time;
2978 spin_lock(&journal->j_state_lock);
2979 if (test_opt(sb, BARRIER))
2980 journal->j_flags |= JBD2_BARRIER;
2982 journal->j_flags &= ~JBD2_BARRIER;
2983 if (test_opt(sb, DATA_ERR_ABORT))
2984 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2986 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2987 spin_unlock(&journal->j_state_lock);
2990 static journal_t *ext4_get_journal(struct super_block *sb,
2991 unsigned int journal_inum)
2993 struct inode *journal_inode;
2996 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2998 /* First, test for the existence of a valid inode on disk. Bad
2999 * things happen if we iget() an unused inode, as the subsequent
3000 * iput() will try to delete it. */
3002 journal_inode = ext4_iget(sb, journal_inum);
3003 if (IS_ERR(journal_inode)) {
3004 ext4_msg(sb, KERN_ERR, "no journal found");
3007 if (!journal_inode->i_nlink) {
3008 make_bad_inode(journal_inode);
3009 iput(journal_inode);
3010 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3014 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3015 journal_inode, journal_inode->i_size);
3016 if (!S_ISREG(journal_inode->i_mode)) {
3017 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3018 iput(journal_inode);
3022 journal = jbd2_journal_init_inode(journal_inode);
3024 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3025 iput(journal_inode);
3028 journal->j_private = sb;
3029 ext4_init_journal_params(sb, journal);
3033 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3036 struct buffer_head *bh;
3040 int hblock, blocksize;
3041 ext4_fsblk_t sb_block;
3042 unsigned long offset;
3043 struct ext4_super_block *es;
3044 struct block_device *bdev;
3046 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3048 bdev = ext4_blkdev_get(j_dev, sb);
3052 if (bd_claim(bdev, sb)) {
3053 ext4_msg(sb, KERN_ERR,
3054 "failed to claim external journal device");
3055 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3059 blocksize = sb->s_blocksize;
3060 hblock = bdev_logical_block_size(bdev);
3061 if (blocksize < hblock) {
3062 ext4_msg(sb, KERN_ERR,
3063 "blocksize too small for journal device");
3067 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3068 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3069 set_blocksize(bdev, blocksize);
3070 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3071 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3072 "external journal");
3076 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3077 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3078 !(le32_to_cpu(es->s_feature_incompat) &
3079 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3080 ext4_msg(sb, KERN_ERR, "external journal has "
3086 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3087 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3092 len = ext4_blocks_count(es);
3093 start = sb_block + 1;
3094 brelse(bh); /* we're done with the superblock */
3096 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3097 start, len, blocksize);
3099 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3102 journal->j_private = sb;
3103 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3104 wait_on_buffer(journal->j_sb_buffer);
3105 if (!buffer_uptodate(journal->j_sb_buffer)) {
3106 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3109 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3110 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3111 "user (unsupported) - %d",
3112 be32_to_cpu(journal->j_superblock->s_nr_users));
3115 EXT4_SB(sb)->journal_bdev = bdev;
3116 ext4_init_journal_params(sb, journal);
3120 jbd2_journal_destroy(journal);
3122 ext4_blkdev_put(bdev);
3126 static int ext4_load_journal(struct super_block *sb,
3127 struct ext4_super_block *es,
3128 unsigned long journal_devnum)
3131 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3134 int really_read_only;
3136 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3138 if (journal_devnum &&
3139 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3140 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3141 "numbers have changed");
3142 journal_dev = new_decode_dev(journal_devnum);
3144 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3146 really_read_only = bdev_read_only(sb->s_bdev);
3149 * Are we loading a blank journal or performing recovery after a
3150 * crash? For recovery, we need to check in advance whether we
3151 * can get read-write access to the device.
3153 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3154 if (sb->s_flags & MS_RDONLY) {
3155 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3156 "required on readonly filesystem");
3157 if (really_read_only) {
3158 ext4_msg(sb, KERN_ERR, "write access "
3159 "unavailable, cannot proceed");
3162 ext4_msg(sb, KERN_INFO, "write access will "
3163 "be enabled during recovery");
3167 if (journal_inum && journal_dev) {
3168 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3169 "and inode journals!");
3174 if (!(journal = ext4_get_journal(sb, journal_inum)))
3177 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3181 if (journal->j_flags & JBD2_BARRIER)
3182 ext4_msg(sb, KERN_INFO, "barriers enabled");
3184 ext4_msg(sb, KERN_INFO, "barriers disabled");
3186 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3187 err = jbd2_journal_update_format(journal);
3189 ext4_msg(sb, KERN_ERR, "error updating journal");
3190 jbd2_journal_destroy(journal);
3195 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3196 err = jbd2_journal_wipe(journal, !really_read_only);
3198 err = jbd2_journal_load(journal);
3201 ext4_msg(sb, KERN_ERR, "error loading journal");
3202 jbd2_journal_destroy(journal);
3206 EXT4_SB(sb)->s_journal = journal;
3207 ext4_clear_journal_err(sb, es);
3209 if (journal_devnum &&
3210 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3211 es->s_journal_dev = cpu_to_le32(journal_devnum);
3213 /* Make sure we flush the recovery flag to disk. */
3214 ext4_commit_super(sb, 1);
3220 static int ext4_commit_super(struct super_block *sb, int sync)
3222 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3223 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3228 if (buffer_write_io_error(sbh)) {
3230 * Oh, dear. A previous attempt to write the
3231 * superblock failed. This could happen because the
3232 * USB device was yanked out. Or it could happen to
3233 * be a transient write error and maybe the block will
3234 * be remapped. Nothing we can do but to retry the
3235 * write and hope for the best.
3237 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3238 "superblock detected");
3239 clear_buffer_write_io_error(sbh);
3240 set_buffer_uptodate(sbh);
3243 * If the file system is mounted read-only, don't update the
3244 * superblock write time. This avoids updating the superblock
3245 * write time when we are mounting the root file system
3246 * read/only but we need to replay the journal; at that point,
3247 * for people who are east of GMT and who make their clock
3248 * tick in localtime for Windows bug-for-bug compatibility,
3249 * the clock is set in the future, and this will cause e2fsck
3250 * to complain and force a full file system check.
3252 if (!(sb->s_flags & MS_RDONLY))
3253 es->s_wtime = cpu_to_le32(get_seconds());
3254 es->s_kbytes_written =
3255 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3256 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3257 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3258 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3259 &EXT4_SB(sb)->s_freeblocks_counter));
3260 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3261 &EXT4_SB(sb)->s_freeinodes_counter));
3263 BUFFER_TRACE(sbh, "marking dirty");
3264 mark_buffer_dirty(sbh);
3266 error = sync_dirty_buffer(sbh);
3270 error = buffer_write_io_error(sbh);
3272 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3274 clear_buffer_write_io_error(sbh);
3275 set_buffer_uptodate(sbh);
3282 * Have we just finished recovery? If so, and if we are mounting (or
3283 * remounting) the filesystem readonly, then we will end up with a
3284 * consistent fs on disk. Record that fact.
3286 static void ext4_mark_recovery_complete(struct super_block *sb,
3287 struct ext4_super_block *es)
3289 journal_t *journal = EXT4_SB(sb)->s_journal;
3291 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3292 BUG_ON(journal != NULL);
3295 jbd2_journal_lock_updates(journal);
3296 if (jbd2_journal_flush(journal) < 0)
3299 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3300 sb->s_flags & MS_RDONLY) {
3301 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3302 ext4_commit_super(sb, 1);
3306 jbd2_journal_unlock_updates(journal);
3310 * If we are mounting (or read-write remounting) a filesystem whose journal
3311 * has recorded an error from a previous lifetime, move that error to the
3312 * main filesystem now.
3314 static void ext4_clear_journal_err(struct super_block *sb,
3315 struct ext4_super_block *es)
3321 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3323 journal = EXT4_SB(sb)->s_journal;
3326 * Now check for any error status which may have been recorded in the
3327 * journal by a prior ext4_error() or ext4_abort()
3330 j_errno = jbd2_journal_errno(journal);
3334 errstr = ext4_decode_error(sb, j_errno, nbuf);
3335 ext4_warning(sb, __func__, "Filesystem error recorded "
3336 "from previous mount: %s", errstr);
3337 ext4_warning(sb, __func__, "Marking fs in need of "
3338 "filesystem check.");
3340 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3341 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3342 ext4_commit_super(sb, 1);
3344 jbd2_journal_clear_err(journal);
3349 * Force the running and committing transactions to commit,
3350 * and wait on the commit.
3352 int ext4_force_commit(struct super_block *sb)
3357 if (sb->s_flags & MS_RDONLY)
3360 journal = EXT4_SB(sb)->s_journal;
3362 ret = ext4_journal_force_commit(journal);
3367 static void ext4_write_super(struct super_block *sb)
3370 ext4_commit_super(sb, 1);
3374 static int ext4_sync_fs(struct super_block *sb, int wait)
3379 trace_ext4_sync_fs(sb, wait);
3380 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3382 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3388 * LVM calls this function before a (read-only) snapshot is created. This
3389 * gives us a chance to flush the journal completely and mark the fs clean.
3391 static int ext4_freeze(struct super_block *sb)
3396 if (sb->s_flags & MS_RDONLY)
3399 journal = EXT4_SB(sb)->s_journal;
3401 /* Now we set up the journal barrier. */
3402 jbd2_journal_lock_updates(journal);
3405 * Don't clear the needs_recovery flag if we failed to flush
3408 error = jbd2_journal_flush(journal);
3411 jbd2_journal_unlock_updates(journal);
3415 /* Journal blocked and flushed, clear needs_recovery flag. */
3416 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3417 error = ext4_commit_super(sb, 1);
3424 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3425 * flag here, even though the filesystem is not technically dirty yet.
3427 static int ext4_unfreeze(struct super_block *sb)
3429 if (sb->s_flags & MS_RDONLY)
3433 /* Reset the needs_recovery flag before the fs is unlocked. */
3434 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3435 ext4_commit_super(sb, 1);
3437 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3441 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3443 struct ext4_super_block *es;
3444 struct ext4_sb_info *sbi = EXT4_SB(sb);
3445 ext4_fsblk_t n_blocks_count = 0;
3446 unsigned long old_sb_flags;
3447 struct ext4_mount_options old_opts;
3449 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3457 /* Store the original options */
3459 old_sb_flags = sb->s_flags;
3460 old_opts.s_mount_opt = sbi->s_mount_opt;
3461 old_opts.s_resuid = sbi->s_resuid;
3462 old_opts.s_resgid = sbi->s_resgid;
3463 old_opts.s_commit_interval = sbi->s_commit_interval;
3464 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3465 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3467 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3468 for (i = 0; i < MAXQUOTAS; i++)
3469 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3471 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3472 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3475 * Allow the "check" option to be passed as a remount option.
3477 if (!parse_options(data, sb, NULL, &journal_ioprio,
3478 &n_blocks_count, 1)) {
3483 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3484 ext4_abort(sb, __func__, "Abort forced by user");
3486 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3487 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3491 if (sbi->s_journal) {
3492 ext4_init_journal_params(sb, sbi->s_journal);
3493 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3496 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3497 n_blocks_count > ext4_blocks_count(es)) {
3498 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3503 if (*flags & MS_RDONLY) {
3505 * First of all, the unconditional stuff we have to do
3506 * to disable replay of the journal when we next remount
3508 sb->s_flags |= MS_RDONLY;
3511 * OK, test if we are remounting a valid rw partition
3512 * readonly, and if so set the rdonly flag and then
3513 * mark the partition as valid again.
3515 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3516 (sbi->s_mount_state & EXT4_VALID_FS))
3517 es->s_state = cpu_to_le16(sbi->s_mount_state);
3520 ext4_mark_recovery_complete(sb, es);
3522 /* Make sure we can mount this feature set readwrite */
3523 if (!ext4_feature_set_ok(sb, 0)) {
3528 * Make sure the group descriptor checksums
3529 * are sane. If they aren't, refuse to remount r/w.
3531 for (g = 0; g < sbi->s_groups_count; g++) {
3532 struct ext4_group_desc *gdp =
3533 ext4_get_group_desc(sb, g, NULL);
3535 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3536 ext4_msg(sb, KERN_ERR,
3537 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3538 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3539 le16_to_cpu(gdp->bg_checksum));
3546 * If we have an unprocessed orphan list hanging
3547 * around from a previously readonly bdev mount,
3548 * require a full umount/remount for now.
3550 if (es->s_last_orphan) {
3551 ext4_msg(sb, KERN_WARNING, "Couldn't "
3552 "remount RDWR because of unprocessed "
3553 "orphan inode list. Please "
3554 "umount/remount instead");
3560 * Mounting a RDONLY partition read-write, so reread
3561 * and store the current valid flag. (It may have
3562 * been changed by e2fsck since we originally mounted
3566 ext4_clear_journal_err(sb, es);
3567 sbi->s_mount_state = le16_to_cpu(es->s_state);
3568 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3570 if (!ext4_setup_super(sb, es, 0))
3571 sb->s_flags &= ~MS_RDONLY;
3574 ext4_setup_system_zone(sb);
3575 if (sbi->s_journal == NULL)
3576 ext4_commit_super(sb, 1);
3579 /* Release old quota file names */
3580 for (i = 0; i < MAXQUOTAS; i++)
3581 if (old_opts.s_qf_names[i] &&
3582 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3583 kfree(old_opts.s_qf_names[i]);
3590 sb->s_flags = old_sb_flags;
3591 sbi->s_mount_opt = old_opts.s_mount_opt;
3592 sbi->s_resuid = old_opts.s_resuid;
3593 sbi->s_resgid = old_opts.s_resgid;
3594 sbi->s_commit_interval = old_opts.s_commit_interval;
3595 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3596 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3598 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3599 for (i = 0; i < MAXQUOTAS; i++) {
3600 if (sbi->s_qf_names[i] &&
3601 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3602 kfree(sbi->s_qf_names[i]);
3603 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3611 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3613 struct super_block *sb = dentry->d_sb;
3614 struct ext4_sb_info *sbi = EXT4_SB(sb);
3615 struct ext4_super_block *es = sbi->s_es;
3618 if (test_opt(sb, MINIX_DF)) {
3619 sbi->s_overhead_last = 0;
3620 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3621 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3622 ext4_fsblk_t overhead = 0;
3625 * Compute the overhead (FS structures). This is constant
3626 * for a given filesystem unless the number of block groups
3627 * changes so we cache the previous value until it does.
3631 * All of the blocks before first_data_block are
3634 overhead = le32_to_cpu(es->s_first_data_block);
3637 * Add the overhead attributed to the superblock and
3638 * block group descriptors. If the sparse superblocks
3639 * feature is turned on, then not all groups have this.
3641 for (i = 0; i < ngroups; i++) {
3642 overhead += ext4_bg_has_super(sb, i) +
3643 ext4_bg_num_gdb(sb, i);
3648 * Every block group has an inode bitmap, a block
3649 * bitmap, and an inode table.
3651 overhead += ngroups * (2 + sbi->s_itb_per_group);
3652 sbi->s_overhead_last = overhead;
3654 sbi->s_blocks_last = ext4_blocks_count(es);
3657 buf->f_type = EXT4_SUPER_MAGIC;
3658 buf->f_bsize = sb->s_blocksize;
3659 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3660 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3661 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3662 ext4_free_blocks_count_set(es, buf->f_bfree);
3663 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3664 if (buf->f_bfree < ext4_r_blocks_count(es))
3666 buf->f_files = le32_to_cpu(es->s_inodes_count);
3667 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3668 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3669 buf->f_namelen = EXT4_NAME_LEN;
3670 fsid = le64_to_cpup((void *)es->s_uuid) ^
3671 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3672 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3673 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3678 /* Helper function for writing quotas on sync - we need to start transaction
3679 * before quota file is locked for write. Otherwise the are possible deadlocks:
3680 * Process 1 Process 2
3681 * ext4_create() quota_sync()
3682 * jbd2_journal_start() write_dquot()
3683 * vfs_dq_init() down(dqio_mutex)
3684 * down(dqio_mutex) jbd2_journal_start()
3690 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3692 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3695 static int ext4_write_dquot(struct dquot *dquot)
3699 struct inode *inode;
3701 inode = dquot_to_inode(dquot);
3702 handle = ext4_journal_start(inode,
3703 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3705 return PTR_ERR(handle);
3706 ret = dquot_commit(dquot);
3707 err = ext4_journal_stop(handle);
3713 static int ext4_acquire_dquot(struct dquot *dquot)
3718 handle = ext4_journal_start(dquot_to_inode(dquot),
3719 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3721 return PTR_ERR(handle);
3722 ret = dquot_acquire(dquot);
3723 err = ext4_journal_stop(handle);
3729 static int ext4_release_dquot(struct dquot *dquot)
3734 handle = ext4_journal_start(dquot_to_inode(dquot),
3735 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3736 if (IS_ERR(handle)) {
3737 /* Release dquot anyway to avoid endless cycle in dqput() */
3738 dquot_release(dquot);
3739 return PTR_ERR(handle);
3741 ret = dquot_release(dquot);
3742 err = ext4_journal_stop(handle);
3748 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3750 /* Are we journaling quotas? */
3751 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3752 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3753 dquot_mark_dquot_dirty(dquot);
3754 return ext4_write_dquot(dquot);
3756 return dquot_mark_dquot_dirty(dquot);
3760 static int ext4_write_info(struct super_block *sb, int type)
3765 /* Data block + inode block */
3766 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3768 return PTR_ERR(handle);
3769 ret = dquot_commit_info(sb, type);
3770 err = ext4_journal_stop(handle);
3777 * Turn on quotas during mount time - we need to find
3778 * the quota file and such...
3780 static int ext4_quota_on_mount(struct super_block *sb, int type)
3782 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3783 EXT4_SB(sb)->s_jquota_fmt, type);
3787 * Standard function to be called on quota_on
3789 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3790 char *name, int remount)
3795 if (!test_opt(sb, QUOTA))
3797 /* When remounting, no checks are needed and in fact, name is NULL */
3799 return vfs_quota_on(sb, type, format_id, name, remount);
3801 err = kern_path(name, LOOKUP_FOLLOW, &path);
3805 /* Quotafile not on the same filesystem? */
3806 if (path.mnt->mnt_sb != sb) {
3810 /* Journaling quota? */
3811 if (EXT4_SB(sb)->s_qf_names[type]) {
3812 /* Quotafile not in fs root? */
3813 if (path.dentry->d_parent != sb->s_root)
3814 ext4_msg(sb, KERN_WARNING,
3815 "Quota file not on filesystem root. "
3816 "Journaled quota will not work");
3820 * When we journal data on quota file, we have to flush journal to see
3821 * all updates to the file when we bypass pagecache...
3823 if (EXT4_SB(sb)->s_journal &&
3824 ext4_should_journal_data(path.dentry->d_inode)) {
3826 * We don't need to lock updates but journal_flush() could
3827 * otherwise be livelocked...
3829 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3830 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3831 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3838 err = vfs_quota_on_path(sb, type, format_id, &path);
3843 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3844 * acquiring the locks... As quota files are never truncated and quota code
3845 * itself serializes the operations (and noone else should touch the files)
3846 * we don't have to be afraid of races */
3847 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3848 size_t len, loff_t off)
3850 struct inode *inode = sb_dqopt(sb)->files[type];
3851 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3853 int offset = off & (sb->s_blocksize - 1);
3856 struct buffer_head *bh;
3857 loff_t i_size = i_size_read(inode);
3861 if (off+len > i_size)
3864 while (toread > 0) {
3865 tocopy = sb->s_blocksize - offset < toread ?
3866 sb->s_blocksize - offset : toread;
3867 bh = ext4_bread(NULL, inode, blk, 0, &err);
3870 if (!bh) /* A hole? */
3871 memset(data, 0, tocopy);
3873 memcpy(data, bh->b_data+offset, tocopy);
3883 /* Write to quotafile (we know the transaction is already started and has
3884 * enough credits) */
3885 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3886 const char *data, size_t len, loff_t off)
3888 struct inode *inode = sb_dqopt(sb)->files[type];
3889 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3891 int offset = off & (sb->s_blocksize - 1);
3893 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3894 size_t towrite = len;
3895 struct buffer_head *bh;
3896 handle_t *handle = journal_current_handle();
3898 if (EXT4_SB(sb)->s_journal && !handle) {
3899 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3900 " cancelled because transaction is not started",
3901 (unsigned long long)off, (unsigned long long)len);
3904 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3905 while (towrite > 0) {
3906 tocopy = sb->s_blocksize - offset < towrite ?
3907 sb->s_blocksize - offset : towrite;
3908 bh = ext4_bread(handle, inode, blk, 1, &err);
3911 if (journal_quota) {
3912 err = ext4_journal_get_write_access(handle, bh);
3919 memcpy(bh->b_data+offset, data, tocopy);
3920 flush_dcache_page(bh->b_page);
3923 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3925 /* Always do at least ordered writes for quotas */
3926 err = ext4_jbd2_file_inode(handle, inode);
3927 mark_buffer_dirty(bh);
3938 if (len == towrite) {
3939 mutex_unlock(&inode->i_mutex);
3942 if (inode->i_size < off+len-towrite) {
3943 i_size_write(inode, off+len-towrite);
3944 EXT4_I(inode)->i_disksize = inode->i_size;
3946 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3947 ext4_mark_inode_dirty(handle, inode);
3948 mutex_unlock(&inode->i_mutex);
3949 return len - towrite;
3954 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3955 const char *dev_name, void *data, struct vfsmount *mnt)
3957 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3960 static struct file_system_type ext4_fs_type = {
3961 .owner = THIS_MODULE,
3963 .get_sb = ext4_get_sb,
3964 .kill_sb = kill_block_super,
3965 .fs_flags = FS_REQUIRES_DEV,
3968 #ifdef CONFIG_EXT4DEV_COMPAT
3969 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3970 const char *dev_name, void *data,struct vfsmount *mnt)
3972 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3973 "to mount using ext4\n", dev_name);
3974 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3975 "will go away by 2.6.31\n", dev_name);
3976 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3979 static struct file_system_type ext4dev_fs_type = {
3980 .owner = THIS_MODULE,
3982 .get_sb = ext4dev_get_sb,
3983 .kill_sb = kill_block_super,
3984 .fs_flags = FS_REQUIRES_DEV,
3986 MODULE_ALIAS("ext4dev");
3989 static int __init init_ext4_fs(void)
3993 err = init_ext4_system_zone();
3996 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3999 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4000 err = init_ext4_mballoc();
4004 err = init_ext4_xattr();
4007 err = init_inodecache();
4010 err = register_filesystem(&ext4_fs_type);
4013 #ifdef CONFIG_EXT4DEV_COMPAT
4014 err = register_filesystem(&ext4dev_fs_type);
4016 unregister_filesystem(&ext4_fs_type);
4022 destroy_inodecache();
4026 exit_ext4_mballoc();
4028 remove_proc_entry("fs/ext4", NULL);
4029 kset_unregister(ext4_kset);
4031 exit_ext4_system_zone();
4035 static void __exit exit_ext4_fs(void)
4037 unregister_filesystem(&ext4_fs_type);
4038 #ifdef CONFIG_EXT4DEV_COMPAT
4039 unregister_filesystem(&ext4dev_fs_type);
4041 destroy_inodecache();
4043 exit_ext4_mballoc();
4044 remove_proc_entry("fs/ext4", NULL);
4045 kset_unregister(ext4_kset);
4046 exit_ext4_system_zone();
4049 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4050 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4051 MODULE_LICENSE("GPL");
4052 module_init(init_ext4_fs)
4053 module_exit(exit_ext4_fs)