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;
586 ext4_commit_super(sb, 1);
588 ext4_release_system_zone(sb);
590 ext4_ext_release(sb);
591 ext4_xattr_put_super(sb);
592 if (sbi->s_journal) {
593 err = jbd2_journal_destroy(sbi->s_journal);
594 sbi->s_journal = NULL;
596 ext4_abort(sb, __func__,
597 "Couldn't clean up the journal");
599 if (!(sb->s_flags & MS_RDONLY)) {
600 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
601 es->s_state = cpu_to_le16(sbi->s_mount_state);
602 ext4_commit_super(sb, 1);
605 remove_proc_entry(sb->s_id, ext4_proc_root);
607 kobject_del(&sbi->s_kobj);
609 for (i = 0; i < sbi->s_gdb_count; i++)
610 brelse(sbi->s_group_desc[i]);
611 kfree(sbi->s_group_desc);
612 if (is_vmalloc_addr(sbi->s_flex_groups))
613 vfree(sbi->s_flex_groups);
615 kfree(sbi->s_flex_groups);
616 percpu_counter_destroy(&sbi->s_freeblocks_counter);
617 percpu_counter_destroy(&sbi->s_freeinodes_counter);
618 percpu_counter_destroy(&sbi->s_dirs_counter);
619 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
622 for (i = 0; i < MAXQUOTAS; i++)
623 kfree(sbi->s_qf_names[i]);
626 /* Debugging code just in case the in-memory inode orphan list
627 * isn't empty. The on-disk one can be non-empty if we've
628 * detected an error and taken the fs readonly, but the
629 * in-memory list had better be clean by this point. */
630 if (!list_empty(&sbi->s_orphan))
631 dump_orphan_list(sb, sbi);
632 J_ASSERT(list_empty(&sbi->s_orphan));
634 invalidate_bdev(sb->s_bdev);
635 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
637 * Invalidate the journal device's buffers. We don't want them
638 * floating about in memory - the physical journal device may
639 * hotswapped, and it breaks the `ro-after' testing code.
641 sync_blockdev(sbi->journal_bdev);
642 invalidate_bdev(sbi->journal_bdev);
643 ext4_blkdev_remove(sbi);
645 sb->s_fs_info = NULL;
647 * Now that we are completely done shutting down the
648 * superblock, we need to actually destroy the kobject.
652 kobject_put(&sbi->s_kobj);
653 wait_for_completion(&sbi->s_kobj_unregister);
654 kfree(sbi->s_blockgroup_lock);
658 static struct kmem_cache *ext4_inode_cachep;
661 * Called inside transaction, so use GFP_NOFS
663 static struct inode *ext4_alloc_inode(struct super_block *sb)
665 struct ext4_inode_info *ei;
667 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
671 ei->vfs_inode.i_version = 1;
672 ei->vfs_inode.i_data.writeback_index = 0;
673 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
674 INIT_LIST_HEAD(&ei->i_prealloc_list);
675 spin_lock_init(&ei->i_prealloc_lock);
677 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
678 * therefore it can be null here. Don't check it, just initialize
681 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
682 ei->i_reserved_data_blocks = 0;
683 ei->i_reserved_meta_blocks = 0;
684 ei->i_allocated_meta_blocks = 0;
685 ei->i_delalloc_reserved_flag = 0;
686 spin_lock_init(&(ei->i_block_reservation_lock));
688 return &ei->vfs_inode;
691 static void ext4_destroy_inode(struct inode *inode)
693 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
694 ext4_msg(inode->i_sb, KERN_ERR,
695 "Inode %lu (%p): orphan list check failed!",
696 inode->i_ino, EXT4_I(inode));
697 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
698 EXT4_I(inode), sizeof(struct ext4_inode_info),
702 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
705 static void init_once(void *foo)
707 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
709 INIT_LIST_HEAD(&ei->i_orphan);
710 #ifdef CONFIG_EXT4_FS_XATTR
711 init_rwsem(&ei->xattr_sem);
713 init_rwsem(&ei->i_data_sem);
714 inode_init_once(&ei->vfs_inode);
717 static int init_inodecache(void)
719 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
720 sizeof(struct ext4_inode_info),
721 0, (SLAB_RECLAIM_ACCOUNT|
724 if (ext4_inode_cachep == NULL)
729 static void destroy_inodecache(void)
731 kmem_cache_destroy(ext4_inode_cachep);
734 static void ext4_clear_inode(struct inode *inode)
736 ext4_discard_preallocations(inode);
737 if (EXT4_JOURNAL(inode))
738 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
739 &EXT4_I(inode)->jinode);
742 static inline void ext4_show_quota_options(struct seq_file *seq,
743 struct super_block *sb)
745 #if defined(CONFIG_QUOTA)
746 struct ext4_sb_info *sbi = EXT4_SB(sb);
748 if (sbi->s_jquota_fmt)
749 seq_printf(seq, ",jqfmt=%s",
750 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
752 if (sbi->s_qf_names[USRQUOTA])
753 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
755 if (sbi->s_qf_names[GRPQUOTA])
756 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
758 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
759 seq_puts(seq, ",usrquota");
761 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
762 seq_puts(seq, ",grpquota");
768 * - it's set to a non-default value OR
769 * - if the per-sb default is different from the global default
771 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
774 unsigned long def_mount_opts;
775 struct super_block *sb = vfs->mnt_sb;
776 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 struct ext4_super_block *es = sbi->s_es;
779 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
780 def_errors = le16_to_cpu(es->s_errors);
782 if (sbi->s_sb_block != 1)
783 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
784 if (test_opt(sb, MINIX_DF))
785 seq_puts(seq, ",minixdf");
786 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
787 seq_puts(seq, ",grpid");
788 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
789 seq_puts(seq, ",nogrpid");
790 if (sbi->s_resuid != EXT4_DEF_RESUID ||
791 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
792 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
794 if (sbi->s_resgid != EXT4_DEF_RESGID ||
795 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
796 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
798 if (test_opt(sb, ERRORS_RO)) {
799 if (def_errors == EXT4_ERRORS_PANIC ||
800 def_errors == EXT4_ERRORS_CONTINUE) {
801 seq_puts(seq, ",errors=remount-ro");
804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
805 seq_puts(seq, ",errors=continue");
806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
807 seq_puts(seq, ",errors=panic");
808 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
809 seq_puts(seq, ",nouid32");
810 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
811 seq_puts(seq, ",debug");
812 if (test_opt(sb, OLDALLOC))
813 seq_puts(seq, ",oldalloc");
814 #ifdef CONFIG_EXT4_FS_XATTR
815 if (test_opt(sb, XATTR_USER) &&
816 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
817 seq_puts(seq, ",user_xattr");
818 if (!test_opt(sb, XATTR_USER) &&
819 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
820 seq_puts(seq, ",nouser_xattr");
823 #ifdef CONFIG_EXT4_FS_POSIX_ACL
824 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
825 seq_puts(seq, ",acl");
826 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
827 seq_puts(seq, ",noacl");
829 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
830 seq_printf(seq, ",commit=%u",
831 (unsigned) (sbi->s_commit_interval / HZ));
833 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
834 seq_printf(seq, ",min_batch_time=%u",
835 (unsigned) sbi->s_min_batch_time);
837 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
838 seq_printf(seq, ",max_batch_time=%u",
839 (unsigned) sbi->s_min_batch_time);
843 * We're changing the default of barrier mount option, so
844 * let's always display its mount state so it's clear what its
847 seq_puts(seq, ",barrier=");
848 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
849 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
850 seq_puts(seq, ",journal_async_commit");
851 if (test_opt(sb, NOBH))
852 seq_puts(seq, ",nobh");
853 if (test_opt(sb, I_VERSION))
854 seq_puts(seq, ",i_version");
855 if (!test_opt(sb, DELALLOC))
856 seq_puts(seq, ",nodelalloc");
860 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
862 * journal mode get enabled in different ways
863 * So just print the value even if we didn't specify it
865 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
866 seq_puts(seq, ",data=journal");
867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
868 seq_puts(seq, ",data=ordered");
869 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
870 seq_puts(seq, ",data=writeback");
872 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
873 seq_printf(seq, ",inode_readahead_blks=%u",
874 sbi->s_inode_readahead_blks);
876 if (test_opt(sb, DATA_ERR_ABORT))
877 seq_puts(seq, ",data_err=abort");
879 if (test_opt(sb, NO_AUTO_DA_ALLOC))
880 seq_puts(seq, ",noauto_da_alloc");
882 ext4_show_quota_options(seq, sb);
887 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
888 u64 ino, u32 generation)
892 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
893 return ERR_PTR(-ESTALE);
894 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
895 return ERR_PTR(-ESTALE);
897 /* iget isn't really right if the inode is currently unallocated!!
899 * ext4_read_inode will return a bad_inode if the inode had been
900 * deleted, so we should be safe.
902 * Currently we don't know the generation for parent directory, so
903 * a generation of 0 means "accept any"
905 inode = ext4_iget(sb, ino);
907 return ERR_CAST(inode);
908 if (generation && inode->i_generation != generation) {
910 return ERR_PTR(-ESTALE);
916 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
917 int fh_len, int fh_type)
919 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
923 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
924 int fh_len, int fh_type)
926 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
931 * Try to release metadata pages (indirect blocks, directories) which are
932 * mapped via the block device. Since these pages could have journal heads
933 * which would prevent try_to_free_buffers() from freeing them, we must use
934 * jbd2 layer's try_to_free_buffers() function to release them.
936 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
939 journal_t *journal = EXT4_SB(sb)->s_journal;
941 WARN_ON(PageChecked(page));
942 if (!page_has_buffers(page))
945 return jbd2_journal_try_to_free_buffers(journal, page,
947 return try_to_free_buffers(page);
951 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
952 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
954 static int ext4_write_dquot(struct dquot *dquot);
955 static int ext4_acquire_dquot(struct dquot *dquot);
956 static int ext4_release_dquot(struct dquot *dquot);
957 static int ext4_mark_dquot_dirty(struct dquot *dquot);
958 static int ext4_write_info(struct super_block *sb, int type);
959 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
960 char *path, int remount);
961 static int ext4_quota_on_mount(struct super_block *sb, int type);
962 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
963 size_t len, loff_t off);
964 static ssize_t ext4_quota_write(struct super_block *sb, int type,
965 const char *data, size_t len, loff_t off);
967 static const struct dquot_operations ext4_quota_operations = {
968 .initialize = dquot_initialize,
970 .alloc_space = dquot_alloc_space,
971 .reserve_space = dquot_reserve_space,
972 .claim_space = dquot_claim_space,
973 .release_rsv = dquot_release_reserved_space,
974 .get_reserved_space = ext4_get_reserved_space,
975 .alloc_inode = dquot_alloc_inode,
976 .free_space = dquot_free_space,
977 .free_inode = dquot_free_inode,
978 .transfer = dquot_transfer,
979 .write_dquot = ext4_write_dquot,
980 .acquire_dquot = ext4_acquire_dquot,
981 .release_dquot = ext4_release_dquot,
982 .mark_dirty = ext4_mark_dquot_dirty,
983 .write_info = ext4_write_info,
984 .alloc_dquot = dquot_alloc,
985 .destroy_dquot = dquot_destroy,
988 static const struct quotactl_ops ext4_qctl_operations = {
989 .quota_on = ext4_quota_on,
990 .quota_off = vfs_quota_off,
991 .quota_sync = vfs_quota_sync,
992 .get_info = vfs_get_dqinfo,
993 .set_info = vfs_set_dqinfo,
994 .get_dqblk = vfs_get_dqblk,
995 .set_dqblk = vfs_set_dqblk
999 static const struct super_operations ext4_sops = {
1000 .alloc_inode = ext4_alloc_inode,
1001 .destroy_inode = ext4_destroy_inode,
1002 .write_inode = ext4_write_inode,
1003 .dirty_inode = ext4_dirty_inode,
1004 .delete_inode = ext4_delete_inode,
1005 .put_super = ext4_put_super,
1006 .sync_fs = ext4_sync_fs,
1007 .freeze_fs = ext4_freeze,
1008 .unfreeze_fs = ext4_unfreeze,
1009 .statfs = ext4_statfs,
1010 .remount_fs = ext4_remount,
1011 .clear_inode = ext4_clear_inode,
1012 .show_options = ext4_show_options,
1014 .quota_read = ext4_quota_read,
1015 .quota_write = ext4_quota_write,
1017 .bdev_try_to_free_page = bdev_try_to_free_page,
1020 static const struct super_operations ext4_nojournal_sops = {
1021 .alloc_inode = ext4_alloc_inode,
1022 .destroy_inode = ext4_destroy_inode,
1023 .write_inode = ext4_write_inode,
1024 .dirty_inode = ext4_dirty_inode,
1025 .delete_inode = ext4_delete_inode,
1026 .write_super = ext4_write_super,
1027 .put_super = ext4_put_super,
1028 .statfs = ext4_statfs,
1029 .remount_fs = ext4_remount,
1030 .clear_inode = ext4_clear_inode,
1031 .show_options = ext4_show_options,
1033 .quota_read = ext4_quota_read,
1034 .quota_write = ext4_quota_write,
1036 .bdev_try_to_free_page = bdev_try_to_free_page,
1039 static const struct export_operations ext4_export_ops = {
1040 .fh_to_dentry = ext4_fh_to_dentry,
1041 .fh_to_parent = ext4_fh_to_parent,
1042 .get_parent = ext4_get_parent,
1046 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1047 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1048 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1049 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1050 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1051 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1052 Opt_journal_update, Opt_journal_dev,
1053 Opt_journal_checksum, Opt_journal_async_commit,
1054 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1055 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1056 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1057 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1058 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1059 Opt_usrquota, Opt_grpquota, Opt_i_version,
1060 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1061 Opt_block_validity, Opt_noblock_validity,
1062 Opt_inode_readahead_blks, Opt_journal_ioprio
1065 static const match_table_t tokens = {
1066 {Opt_bsd_df, "bsddf"},
1067 {Opt_minix_df, "minixdf"},
1068 {Opt_grpid, "grpid"},
1069 {Opt_grpid, "bsdgroups"},
1070 {Opt_nogrpid, "nogrpid"},
1071 {Opt_nogrpid, "sysvgroups"},
1072 {Opt_resgid, "resgid=%u"},
1073 {Opt_resuid, "resuid=%u"},
1075 {Opt_err_cont, "errors=continue"},
1076 {Opt_err_panic, "errors=panic"},
1077 {Opt_err_ro, "errors=remount-ro"},
1078 {Opt_nouid32, "nouid32"},
1079 {Opt_debug, "debug"},
1080 {Opt_oldalloc, "oldalloc"},
1081 {Opt_orlov, "orlov"},
1082 {Opt_user_xattr, "user_xattr"},
1083 {Opt_nouser_xattr, "nouser_xattr"},
1085 {Opt_noacl, "noacl"},
1086 {Opt_noload, "noload"},
1089 {Opt_commit, "commit=%u"},
1090 {Opt_min_batch_time, "min_batch_time=%u"},
1091 {Opt_max_batch_time, "max_batch_time=%u"},
1092 {Opt_journal_update, "journal=update"},
1093 {Opt_journal_dev, "journal_dev=%u"},
1094 {Opt_journal_checksum, "journal_checksum"},
1095 {Opt_journal_async_commit, "journal_async_commit"},
1096 {Opt_abort, "abort"},
1097 {Opt_data_journal, "data=journal"},
1098 {Opt_data_ordered, "data=ordered"},
1099 {Opt_data_writeback, "data=writeback"},
1100 {Opt_data_err_abort, "data_err=abort"},
1101 {Opt_data_err_ignore, "data_err=ignore"},
1102 {Opt_mb_history_length, "mb_history_length=%u"},
1103 {Opt_offusrjquota, "usrjquota="},
1104 {Opt_usrjquota, "usrjquota=%s"},
1105 {Opt_offgrpjquota, "grpjquota="},
1106 {Opt_grpjquota, "grpjquota=%s"},
1107 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1108 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1109 {Opt_grpquota, "grpquota"},
1110 {Opt_noquota, "noquota"},
1111 {Opt_quota, "quota"},
1112 {Opt_usrquota, "usrquota"},
1113 {Opt_barrier, "barrier=%u"},
1114 {Opt_barrier, "barrier"},
1115 {Opt_nobarrier, "nobarrier"},
1116 {Opt_i_version, "i_version"},
1117 {Opt_stripe, "stripe=%u"},
1118 {Opt_resize, "resize"},
1119 {Opt_delalloc, "delalloc"},
1120 {Opt_nodelalloc, "nodelalloc"},
1121 {Opt_block_validity, "block_validity"},
1122 {Opt_noblock_validity, "noblock_validity"},
1123 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1124 {Opt_journal_ioprio, "journal_ioprio=%u"},
1125 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1126 {Opt_auto_da_alloc, "auto_da_alloc"},
1127 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1131 static ext4_fsblk_t get_sb_block(void **data)
1133 ext4_fsblk_t sb_block;
1134 char *options = (char *) *data;
1136 if (!options || strncmp(options, "sb=", 3) != 0)
1137 return 1; /* Default location */
1140 /* TODO: use simple_strtoll with >32bit ext4 */
1141 sb_block = simple_strtoul(options, &options, 0);
1142 if (*options && *options != ',') {
1143 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1147 if (*options == ',')
1149 *data = (void *) options;
1154 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1156 static int parse_options(char *options, struct super_block *sb,
1157 unsigned long *journal_devnum,
1158 unsigned int *journal_ioprio,
1159 ext4_fsblk_t *n_blocks_count, int is_remount)
1161 struct ext4_sb_info *sbi = EXT4_SB(sb);
1163 substring_t args[MAX_OPT_ARGS];
1174 while ((p = strsep(&options, ",")) != NULL) {
1179 token = match_token(p, tokens, args);
1182 clear_opt(sbi->s_mount_opt, MINIX_DF);
1185 set_opt(sbi->s_mount_opt, MINIX_DF);
1188 set_opt(sbi->s_mount_opt, GRPID);
1191 clear_opt(sbi->s_mount_opt, GRPID);
1194 if (match_int(&args[0], &option))
1196 sbi->s_resuid = option;
1199 if (match_int(&args[0], &option))
1201 sbi->s_resgid = option;
1204 /* handled by get_sb_block() instead of here */
1205 /* *sb_block = match_int(&args[0]); */
1208 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1209 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1210 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1213 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1214 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1215 set_opt(sbi->s_mount_opt, ERRORS_RO);
1218 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1219 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1220 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1223 set_opt(sbi->s_mount_opt, NO_UID32);
1226 set_opt(sbi->s_mount_opt, DEBUG);
1229 set_opt(sbi->s_mount_opt, OLDALLOC);
1232 clear_opt(sbi->s_mount_opt, OLDALLOC);
1234 #ifdef CONFIG_EXT4_FS_XATTR
1235 case Opt_user_xattr:
1236 set_opt(sbi->s_mount_opt, XATTR_USER);
1238 case Opt_nouser_xattr:
1239 clear_opt(sbi->s_mount_opt, XATTR_USER);
1242 case Opt_user_xattr:
1243 case Opt_nouser_xattr:
1244 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1247 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1249 set_opt(sbi->s_mount_opt, POSIX_ACL);
1252 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1257 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1260 case Opt_journal_update:
1262 /* Eventually we will want to be able to create
1263 a journal file here. For now, only allow the
1264 user to specify an existing inode to be the
1267 ext4_msg(sb, KERN_ERR,
1268 "Cannot specify journal on remount");
1271 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1273 case Opt_journal_dev:
1275 ext4_msg(sb, KERN_ERR,
1276 "Cannot specify journal on remount");
1279 if (match_int(&args[0], &option))
1281 *journal_devnum = option;
1283 case Opt_journal_checksum:
1284 break; /* Kept for backwards compatibility */
1285 case Opt_journal_async_commit:
1286 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1289 set_opt(sbi->s_mount_opt, NOLOAD);
1292 if (match_int(&args[0], &option))
1297 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1298 sbi->s_commit_interval = HZ * option;
1300 case Opt_max_batch_time:
1301 if (match_int(&args[0], &option))
1306 option = EXT4_DEF_MAX_BATCH_TIME;
1307 sbi->s_max_batch_time = option;
1309 case Opt_min_batch_time:
1310 if (match_int(&args[0], &option))
1314 sbi->s_min_batch_time = option;
1316 case Opt_data_journal:
1317 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1319 case Opt_data_ordered:
1320 data_opt = EXT4_MOUNT_ORDERED_DATA;
1322 case Opt_data_writeback:
1323 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1326 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1328 ext4_msg(sb, KERN_ERR,
1329 "Cannot change data mode on remount");
1333 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1334 sbi->s_mount_opt |= data_opt;
1337 case Opt_data_err_abort:
1338 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1340 case Opt_data_err_ignore:
1341 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1343 case Opt_mb_history_length:
1344 if (match_int(&args[0], &option))
1348 sbi->s_mb_history_max = option;
1357 if (sb_any_quota_loaded(sb) &&
1358 !sbi->s_qf_names[qtype]) {
1359 ext4_msg(sb, KERN_ERR,
1360 "Cannot change journaled "
1361 "quota options when quota turned on");
1364 qname = match_strdup(&args[0]);
1366 ext4_msg(sb, KERN_ERR,
1367 "Not enough memory for "
1368 "storing quotafile name");
1371 if (sbi->s_qf_names[qtype] &&
1372 strcmp(sbi->s_qf_names[qtype], qname)) {
1373 ext4_msg(sb, KERN_ERR,
1374 "%s quota file already "
1375 "specified", QTYPE2NAME(qtype));
1379 sbi->s_qf_names[qtype] = qname;
1380 if (strchr(sbi->s_qf_names[qtype], '/')) {
1381 ext4_msg(sb, KERN_ERR,
1382 "quotafile must be on "
1384 kfree(sbi->s_qf_names[qtype]);
1385 sbi->s_qf_names[qtype] = NULL;
1388 set_opt(sbi->s_mount_opt, QUOTA);
1390 case Opt_offusrjquota:
1393 case Opt_offgrpjquota:
1396 if (sb_any_quota_loaded(sb) &&
1397 sbi->s_qf_names[qtype]) {
1398 ext4_msg(sb, KERN_ERR, "Cannot change "
1399 "journaled quota options when "
1404 * The space will be released later when all options
1405 * are confirmed to be correct
1407 sbi->s_qf_names[qtype] = NULL;
1409 case Opt_jqfmt_vfsold:
1410 qfmt = QFMT_VFS_OLD;
1412 case Opt_jqfmt_vfsv0:
1415 if (sb_any_quota_loaded(sb) &&
1416 sbi->s_jquota_fmt != qfmt) {
1417 ext4_msg(sb, KERN_ERR, "Cannot change "
1418 "journaled quota options when "
1422 sbi->s_jquota_fmt = qfmt;
1426 set_opt(sbi->s_mount_opt, QUOTA);
1427 set_opt(sbi->s_mount_opt, USRQUOTA);
1430 set_opt(sbi->s_mount_opt, QUOTA);
1431 set_opt(sbi->s_mount_opt, GRPQUOTA);
1434 if (sb_any_quota_loaded(sb)) {
1435 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1436 "options when quota turned on");
1439 clear_opt(sbi->s_mount_opt, QUOTA);
1440 clear_opt(sbi->s_mount_opt, USRQUOTA);
1441 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1447 ext4_msg(sb, KERN_ERR,
1448 "quota options not supported");
1452 case Opt_offusrjquota:
1453 case Opt_offgrpjquota:
1454 case Opt_jqfmt_vfsold:
1455 case Opt_jqfmt_vfsv0:
1456 ext4_msg(sb, KERN_ERR,
1457 "journaled quota options not supported");
1463 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1466 clear_opt(sbi->s_mount_opt, BARRIER);
1469 if (match_int(&args[0], &option)) {
1470 set_opt(sbi->s_mount_opt, BARRIER);
1474 set_opt(sbi->s_mount_opt, BARRIER);
1476 clear_opt(sbi->s_mount_opt, BARRIER);
1482 ext4_msg(sb, KERN_ERR,
1483 "resize option only available "
1487 if (match_int(&args[0], &option) != 0)
1489 *n_blocks_count = option;
1492 set_opt(sbi->s_mount_opt, NOBH);
1495 clear_opt(sbi->s_mount_opt, NOBH);
1498 set_opt(sbi->s_mount_opt, I_VERSION);
1499 sb->s_flags |= MS_I_VERSION;
1501 case Opt_nodelalloc:
1502 clear_opt(sbi->s_mount_opt, DELALLOC);
1505 if (match_int(&args[0], &option))
1509 sbi->s_stripe = option;
1512 set_opt(sbi->s_mount_opt, DELALLOC);
1514 case Opt_block_validity:
1515 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1517 case Opt_noblock_validity:
1518 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1520 case Opt_inode_readahead_blks:
1521 if (match_int(&args[0], &option))
1523 if (option < 0 || option > (1 << 30))
1525 if (!is_power_of_2(option)) {
1526 ext4_msg(sb, KERN_ERR,
1527 "EXT4-fs: inode_readahead_blks"
1528 " must be a power of 2");
1531 sbi->s_inode_readahead_blks = option;
1533 case Opt_journal_ioprio:
1534 if (match_int(&args[0], &option))
1536 if (option < 0 || option > 7)
1538 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1541 case Opt_noauto_da_alloc:
1542 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1544 case Opt_auto_da_alloc:
1545 if (match_int(&args[0], &option)) {
1546 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1550 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1552 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1555 ext4_msg(sb, KERN_ERR,
1556 "Unrecognized mount option \"%s\" "
1557 "or missing value", p);
1562 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1563 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1564 sbi->s_qf_names[USRQUOTA])
1565 clear_opt(sbi->s_mount_opt, USRQUOTA);
1567 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1568 sbi->s_qf_names[GRPQUOTA])
1569 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1571 if ((sbi->s_qf_names[USRQUOTA] &&
1572 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1573 (sbi->s_qf_names[GRPQUOTA] &&
1574 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1575 ext4_msg(sb, KERN_ERR, "old and new quota "
1580 if (!sbi->s_jquota_fmt) {
1581 ext4_msg(sb, KERN_ERR, "journaled quota format "
1586 if (sbi->s_jquota_fmt) {
1587 ext4_msg(sb, KERN_ERR, "journaled quota format "
1588 "specified with no journaling "
1597 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
1603 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1604 ext4_msg(sb, KERN_ERR, "revision level too high, "
1605 "forcing read-only mode");
1610 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1611 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1612 "running e2fsck is recommended");
1613 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1614 ext4_msg(sb, KERN_WARNING,
1615 "warning: mounting fs with errors, "
1616 "running e2fsck is recommended");
1617 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1618 le16_to_cpu(es->s_mnt_count) >=
1619 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1620 ext4_msg(sb, KERN_WARNING,
1621 "warning: maximal mount count reached, "
1622 "running e2fsck is recommended");
1623 else if (le32_to_cpu(es->s_checkinterval) &&
1624 (le32_to_cpu(es->s_lastcheck) +
1625 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1626 ext4_msg(sb, KERN_WARNING,
1627 "warning: checktime reached, "
1628 "running e2fsck is recommended");
1629 if (!sbi->s_journal)
1630 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1631 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1632 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1633 le16_add_cpu(&es->s_mnt_count, 1);
1634 es->s_mtime = cpu_to_le32(get_seconds());
1635 ext4_update_dynamic_rev(sb);
1637 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1639 ext4_commit_super(sb, 1);
1640 if (test_opt(sb, DEBUG))
1641 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1642 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1644 sbi->s_groups_count,
1645 EXT4_BLOCKS_PER_GROUP(sb),
1646 EXT4_INODES_PER_GROUP(sb),
1649 if (EXT4_SB(sb)->s_journal) {
1650 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1651 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1652 "external", EXT4_SB(sb)->s_journal->j_devname);
1654 ext4_msg(sb, KERN_INFO, "no journal");
1659 static int ext4_fill_flex_info(struct super_block *sb)
1661 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 struct ext4_group_desc *gdp = NULL;
1663 ext4_group_t flex_group_count;
1664 ext4_group_t flex_group;
1665 int groups_per_flex = 0;
1669 if (!sbi->s_es->s_log_groups_per_flex) {
1670 sbi->s_log_groups_per_flex = 0;
1674 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1675 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1677 /* We allocate both existing and potentially added groups */
1678 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1679 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1680 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1681 size = flex_group_count * sizeof(struct flex_groups);
1682 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1683 if (sbi->s_flex_groups == NULL) {
1684 sbi->s_flex_groups = vmalloc(size);
1685 if (sbi->s_flex_groups)
1686 memset(sbi->s_flex_groups, 0, size);
1688 if (sbi->s_flex_groups == NULL) {
1689 ext4_msg(sb, KERN_ERR, "not enough memory for "
1690 "%u flex groups", flex_group_count);
1694 for (i = 0; i < sbi->s_groups_count; i++) {
1695 gdp = ext4_get_group_desc(sb, i, NULL);
1697 flex_group = ext4_flex_group(sbi, i);
1698 atomic_add(ext4_free_inodes_count(sb, gdp),
1699 &sbi->s_flex_groups[flex_group].free_inodes);
1700 atomic_add(ext4_free_blks_count(sb, gdp),
1701 &sbi->s_flex_groups[flex_group].free_blocks);
1702 atomic_add(ext4_used_dirs_count(sb, gdp),
1703 &sbi->s_flex_groups[flex_group].used_dirs);
1711 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1712 struct ext4_group_desc *gdp)
1716 if (sbi->s_es->s_feature_ro_compat &
1717 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1718 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1719 __le32 le_group = cpu_to_le32(block_group);
1721 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1722 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1723 crc = crc16(crc, (__u8 *)gdp, offset);
1724 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1725 /* for checksum of struct ext4_group_desc do the rest...*/
1726 if ((sbi->s_es->s_feature_incompat &
1727 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1728 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1729 crc = crc16(crc, (__u8 *)gdp + offset,
1730 le16_to_cpu(sbi->s_es->s_desc_size) -
1734 return cpu_to_le16(crc);
1737 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1738 struct ext4_group_desc *gdp)
1740 if ((sbi->s_es->s_feature_ro_compat &
1741 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1742 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1748 /* Called at mount-time, super-block is locked */
1749 static int ext4_check_descriptors(struct super_block *sb)
1751 struct ext4_sb_info *sbi = EXT4_SB(sb);
1752 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1753 ext4_fsblk_t last_block;
1754 ext4_fsblk_t block_bitmap;
1755 ext4_fsblk_t inode_bitmap;
1756 ext4_fsblk_t inode_table;
1757 int flexbg_flag = 0;
1760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1763 ext4_debug("Checking group descriptors");
1765 for (i = 0; i < sbi->s_groups_count; i++) {
1766 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1768 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1769 last_block = ext4_blocks_count(sbi->s_es) - 1;
1771 last_block = first_block +
1772 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1774 block_bitmap = ext4_block_bitmap(sb, gdp);
1775 if (block_bitmap < first_block || block_bitmap > last_block) {
1776 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1777 "Block bitmap for group %u not in group "
1778 "(block %llu)!", i, block_bitmap);
1781 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1782 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1783 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1784 "Inode bitmap for group %u not in group "
1785 "(block %llu)!", i, inode_bitmap);
1788 inode_table = ext4_inode_table(sb, gdp);
1789 if (inode_table < first_block ||
1790 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1791 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1792 "Inode table for group %u not in group "
1793 "(block %llu)!", i, inode_table);
1796 ext4_lock_group(sb, i);
1797 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1798 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1799 "Checksum for group %u failed (%u!=%u)",
1800 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1801 gdp)), le16_to_cpu(gdp->bg_checksum));
1802 if (!(sb->s_flags & MS_RDONLY)) {
1803 ext4_unlock_group(sb, i);
1807 ext4_unlock_group(sb, i);
1809 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1812 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1813 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1817 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1818 * the superblock) which were deleted from all directories, but held open by
1819 * a process at the time of a crash. We walk the list and try to delete these
1820 * inodes at recovery time (only with a read-write filesystem).
1822 * In order to keep the orphan inode chain consistent during traversal (in
1823 * case of crash during recovery), we link each inode into the superblock
1824 * orphan list_head and handle it the same way as an inode deletion during
1825 * normal operation (which journals the operations for us).
1827 * We only do an iget() and an iput() on each inode, which is very safe if we
1828 * accidentally point at an in-use or already deleted inode. The worst that
1829 * can happen in this case is that we get a "bit already cleared" message from
1830 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1831 * e2fsck was run on this filesystem, and it must have already done the orphan
1832 * inode cleanup for us, so we can safely abort without any further action.
1834 static void ext4_orphan_cleanup(struct super_block *sb,
1835 struct ext4_super_block *es)
1837 unsigned int s_flags = sb->s_flags;
1838 int nr_orphans = 0, nr_truncates = 0;
1842 if (!es->s_last_orphan) {
1843 jbd_debug(4, "no orphan inodes to clean up\n");
1847 if (bdev_read_only(sb->s_bdev)) {
1848 ext4_msg(sb, KERN_ERR, "write access "
1849 "unavailable, skipping orphan cleanup");
1853 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1854 if (es->s_last_orphan)
1855 jbd_debug(1, "Errors on filesystem, "
1856 "clearing orphan list.\n");
1857 es->s_last_orphan = 0;
1858 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1862 if (s_flags & MS_RDONLY) {
1863 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1864 sb->s_flags &= ~MS_RDONLY;
1867 /* Needed for iput() to work correctly and not trash data */
1868 sb->s_flags |= MS_ACTIVE;
1869 /* Turn on quotas so that they are updated correctly */
1870 for (i = 0; i < MAXQUOTAS; i++) {
1871 if (EXT4_SB(sb)->s_qf_names[i]) {
1872 int ret = ext4_quota_on_mount(sb, i);
1874 ext4_msg(sb, KERN_ERR,
1875 "Cannot turn on journaled "
1876 "quota: error %d", ret);
1881 while (es->s_last_orphan) {
1882 struct inode *inode;
1884 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1885 if (IS_ERR(inode)) {
1886 es->s_last_orphan = 0;
1890 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1892 if (inode->i_nlink) {
1893 ext4_msg(sb, KERN_DEBUG,
1894 "%s: truncating inode %lu to %lld bytes",
1895 __func__, inode->i_ino, inode->i_size);
1896 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1897 inode->i_ino, inode->i_size);
1898 ext4_truncate(inode);
1901 ext4_msg(sb, KERN_DEBUG,
1902 "%s: deleting unreferenced inode %lu",
1903 __func__, inode->i_ino);
1904 jbd_debug(2, "deleting unreferenced inode %lu\n",
1908 iput(inode); /* The delete magic happens here! */
1911 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1914 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1915 PLURAL(nr_orphans));
1917 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1918 PLURAL(nr_truncates));
1920 /* Turn quotas off */
1921 for (i = 0; i < MAXQUOTAS; i++) {
1922 if (sb_dqopt(sb)->files[i])
1923 vfs_quota_off(sb, i, 0);
1926 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1930 * Maximal extent format file size.
1931 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1932 * extent format containers, within a sector_t, and within i_blocks
1933 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1934 * so that won't be a limiting factor.
1936 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1938 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1941 loff_t upper_limit = MAX_LFS_FILESIZE;
1943 /* small i_blocks in vfs inode? */
1944 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1946 * CONFIG_LBDAF is not enabled implies the inode
1947 * i_block represent total blocks in 512 bytes
1948 * 32 == size of vfs inode i_blocks * 8
1950 upper_limit = (1LL << 32) - 1;
1952 /* total blocks in file system block size */
1953 upper_limit >>= (blkbits - 9);
1954 upper_limit <<= blkbits;
1957 /* 32-bit extent-start container, ee_block */
1962 /* Sanity check against vm- & vfs- imposed limits */
1963 if (res > upper_limit)
1970 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1971 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1972 * We need to be 1 filesystem block less than the 2^48 sector limit.
1974 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1976 loff_t res = EXT4_NDIR_BLOCKS;
1979 /* This is calculated to be the largest file size for a dense, block
1980 * mapped file such that the file's total number of 512-byte sectors,
1981 * including data and all indirect blocks, does not exceed (2^48 - 1).
1983 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1984 * number of 512-byte sectors of the file.
1987 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1989 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1990 * the inode i_block field represents total file blocks in
1991 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1993 upper_limit = (1LL << 32) - 1;
1995 /* total blocks in file system block size */
1996 upper_limit >>= (bits - 9);
2000 * We use 48 bit ext4_inode i_blocks
2001 * With EXT4_HUGE_FILE_FL set the i_blocks
2002 * represent total number of blocks in
2003 * file system block size
2005 upper_limit = (1LL << 48) - 1;
2009 /* indirect blocks */
2011 /* double indirect blocks */
2012 meta_blocks += 1 + (1LL << (bits-2));
2013 /* tripple indirect blocks */
2014 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2016 upper_limit -= meta_blocks;
2017 upper_limit <<= bits;
2019 res += 1LL << (bits-2);
2020 res += 1LL << (2*(bits-2));
2021 res += 1LL << (3*(bits-2));
2023 if (res > upper_limit)
2026 if (res > MAX_LFS_FILESIZE)
2027 res = MAX_LFS_FILESIZE;
2032 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2033 ext4_fsblk_t logical_sb_block, int nr)
2035 struct ext4_sb_info *sbi = EXT4_SB(sb);
2036 ext4_group_t bg, first_meta_bg;
2039 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2041 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2043 return logical_sb_block + nr + 1;
2044 bg = sbi->s_desc_per_block * nr;
2045 if (ext4_bg_has_super(sb, bg))
2048 return (has_super + ext4_group_first_block_no(sb, bg));
2052 * ext4_get_stripe_size: Get the stripe size.
2053 * @sbi: In memory super block info
2055 * If we have specified it via mount option, then
2056 * use the mount option value. If the value specified at mount time is
2057 * greater than the blocks per group use the super block value.
2058 * If the super block value is greater than blocks per group return 0.
2059 * Allocator needs it be less than blocks per group.
2062 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2064 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2065 unsigned long stripe_width =
2066 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2068 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2069 return sbi->s_stripe;
2071 if (stripe_width <= sbi->s_blocks_per_group)
2072 return stripe_width;
2074 if (stride <= sbi->s_blocks_per_group)
2083 struct attribute attr;
2084 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2085 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2086 const char *, size_t);
2090 static int parse_strtoul(const char *buf,
2091 unsigned long max, unsigned long *value)
2095 while (*buf && isspace(*buf))
2097 *value = simple_strtoul(buf, &endp, 0);
2098 while (*endp && isspace(*endp))
2100 if (*endp || *value > max)
2106 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2107 struct ext4_sb_info *sbi,
2110 return snprintf(buf, PAGE_SIZE, "%llu\n",
2111 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2114 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2115 struct ext4_sb_info *sbi, char *buf)
2117 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2119 return snprintf(buf, PAGE_SIZE, "%lu\n",
2120 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2121 sbi->s_sectors_written_start) >> 1);
2124 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2125 struct ext4_sb_info *sbi, char *buf)
2127 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2129 return snprintf(buf, PAGE_SIZE, "%llu\n",
2130 sbi->s_kbytes_written +
2131 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2135 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2136 struct ext4_sb_info *sbi,
2137 const char *buf, size_t count)
2141 if (parse_strtoul(buf, 0x40000000, &t))
2144 if (!is_power_of_2(t))
2147 sbi->s_inode_readahead_blks = t;
2151 static ssize_t sbi_ui_show(struct ext4_attr *a,
2152 struct ext4_sb_info *sbi, char *buf)
2154 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2156 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2159 static ssize_t sbi_ui_store(struct ext4_attr *a,
2160 struct ext4_sb_info *sbi,
2161 const char *buf, size_t count)
2163 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2166 if (parse_strtoul(buf, 0xffffffff, &t))
2172 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2173 static struct ext4_attr ext4_attr_##_name = { \
2174 .attr = {.name = __stringify(_name), .mode = _mode }, \
2177 .offset = offsetof(struct ext4_sb_info, _elname), \
2179 #define EXT4_ATTR(name, mode, show, store) \
2180 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2182 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2183 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2184 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2185 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2186 #define ATTR_LIST(name) &ext4_attr_##name.attr
2188 EXT4_RO_ATTR(delayed_allocation_blocks);
2189 EXT4_RO_ATTR(session_write_kbytes);
2190 EXT4_RO_ATTR(lifetime_write_kbytes);
2191 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2192 inode_readahead_blks_store, s_inode_readahead_blks);
2193 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2194 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2195 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2196 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2197 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2198 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2199 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2200 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2202 static struct attribute *ext4_attrs[] = {
2203 ATTR_LIST(delayed_allocation_blocks),
2204 ATTR_LIST(session_write_kbytes),
2205 ATTR_LIST(lifetime_write_kbytes),
2206 ATTR_LIST(inode_readahead_blks),
2207 ATTR_LIST(inode_goal),
2208 ATTR_LIST(mb_stats),
2209 ATTR_LIST(mb_max_to_scan),
2210 ATTR_LIST(mb_min_to_scan),
2211 ATTR_LIST(mb_order2_req),
2212 ATTR_LIST(mb_stream_req),
2213 ATTR_LIST(mb_group_prealloc),
2214 ATTR_LIST(max_writeback_mb_bump),
2218 static ssize_t ext4_attr_show(struct kobject *kobj,
2219 struct attribute *attr, char *buf)
2221 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2223 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2225 return a->show ? a->show(a, sbi, buf) : 0;
2228 static ssize_t ext4_attr_store(struct kobject *kobj,
2229 struct attribute *attr,
2230 const char *buf, size_t len)
2232 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2234 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2236 return a->store ? a->store(a, sbi, buf, len) : 0;
2239 static void ext4_sb_release(struct kobject *kobj)
2241 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2243 complete(&sbi->s_kobj_unregister);
2247 static struct sysfs_ops ext4_attr_ops = {
2248 .show = ext4_attr_show,
2249 .store = ext4_attr_store,
2252 static struct kobj_type ext4_ktype = {
2253 .default_attrs = ext4_attrs,
2254 .sysfs_ops = &ext4_attr_ops,
2255 .release = ext4_sb_release,
2259 * Check whether this filesystem can be mounted based on
2260 * the features present and the RDONLY/RDWR mount requested.
2261 * Returns 1 if this filesystem can be mounted as requested,
2262 * 0 if it cannot be.
2264 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2266 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2267 ext4_msg(sb, KERN_ERR,
2268 "Couldn't mount because of "
2269 "unsupported optional features (%x)",
2270 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2271 ~EXT4_FEATURE_INCOMPAT_SUPP));
2278 /* Check that feature set is OK for a read-write mount */
2279 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2280 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2281 "unsupported optional features (%x)",
2282 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2283 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2287 * Large file size enabled file system can only be mounted
2288 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2291 if (sizeof(blkcnt_t) < sizeof(u64)) {
2292 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2293 "cannot be mounted RDWR without "
2301 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2302 __releases(kernel_lock)
2303 __acquires(kernel_lock)
2305 struct buffer_head *bh;
2306 struct ext4_super_block *es = NULL;
2307 struct ext4_sb_info *sbi;
2309 ext4_fsblk_t sb_block = get_sb_block(&data);
2310 ext4_fsblk_t logical_sb_block;
2311 unsigned long offset = 0;
2312 unsigned long journal_devnum = 0;
2313 unsigned long def_mount_opts;
2319 unsigned int db_count;
2321 int needs_recovery, has_huge_files;
2324 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2326 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2330 sbi->s_blockgroup_lock =
2331 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2332 if (!sbi->s_blockgroup_lock) {
2336 sb->s_fs_info = sbi;
2337 sbi->s_mount_opt = 0;
2338 sbi->s_resuid = EXT4_DEF_RESUID;
2339 sbi->s_resgid = EXT4_DEF_RESGID;
2340 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2341 sbi->s_sb_block = sb_block;
2342 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2347 /* Cleanup superblock name */
2348 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2351 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2353 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2358 * The ext4 superblock will not be buffer aligned for other than 1kB
2359 * block sizes. We need to calculate the offset from buffer start.
2361 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2362 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2363 offset = do_div(logical_sb_block, blocksize);
2365 logical_sb_block = sb_block;
2368 if (!(bh = sb_bread(sb, logical_sb_block))) {
2369 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2373 * Note: s_es must be initialized as soon as possible because
2374 * some ext4 macro-instructions depend on its value
2376 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2378 sb->s_magic = le16_to_cpu(es->s_magic);
2379 if (sb->s_magic != EXT4_SUPER_MAGIC)
2381 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2383 /* Set defaults before we parse the mount options */
2384 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2385 if (def_mount_opts & EXT4_DEFM_DEBUG)
2386 set_opt(sbi->s_mount_opt, DEBUG);
2387 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2388 set_opt(sbi->s_mount_opt, GRPID);
2389 if (def_mount_opts & EXT4_DEFM_UID16)
2390 set_opt(sbi->s_mount_opt, NO_UID32);
2391 #ifdef CONFIG_EXT4_FS_XATTR
2392 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2393 set_opt(sbi->s_mount_opt, XATTR_USER);
2395 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2396 if (def_mount_opts & EXT4_DEFM_ACL)
2397 set_opt(sbi->s_mount_opt, POSIX_ACL);
2399 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2400 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2401 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2402 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2403 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2404 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2406 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2407 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2408 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2409 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2411 set_opt(sbi->s_mount_opt, ERRORS_RO);
2413 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2414 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2415 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2416 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2417 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2418 sbi->s_mb_history_max = default_mb_history_length;
2420 set_opt(sbi->s_mount_opt, BARRIER);
2423 * enable delayed allocation by default
2424 * Use -o nodelalloc to turn it off
2426 set_opt(sbi->s_mount_opt, DELALLOC);
2428 if (!parse_options((char *) data, sb, &journal_devnum,
2429 &journal_ioprio, NULL, 0))
2432 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2433 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2435 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2436 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2437 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2438 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2439 ext4_msg(sb, KERN_WARNING,
2440 "feature flags set on rev 0 fs, "
2441 "running e2fsck is recommended");
2444 * Check feature flags regardless of the revision level, since we
2445 * previously didn't change the revision level when setting the flags,
2446 * so there is a chance incompat flags are set on a rev 0 filesystem.
2448 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2451 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2453 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2454 blocksize > EXT4_MAX_BLOCK_SIZE) {
2455 ext4_msg(sb, KERN_ERR,
2456 "Unsupported filesystem blocksize %d", blocksize);
2460 if (sb->s_blocksize != blocksize) {
2461 /* Validate the filesystem blocksize */
2462 if (!sb_set_blocksize(sb, blocksize)) {
2463 ext4_msg(sb, KERN_ERR, "bad block size %d",
2469 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2470 offset = do_div(logical_sb_block, blocksize);
2471 bh = sb_bread(sb, logical_sb_block);
2473 ext4_msg(sb, KERN_ERR,
2474 "Can't read superblock on 2nd try");
2477 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2479 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2480 ext4_msg(sb, KERN_ERR,
2481 "Magic mismatch, very weird!");
2486 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2487 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2488 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2490 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2492 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2493 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2494 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2496 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2497 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2498 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2499 (!is_power_of_2(sbi->s_inode_size)) ||
2500 (sbi->s_inode_size > blocksize)) {
2501 ext4_msg(sb, KERN_ERR,
2502 "unsupported inode size: %d",
2506 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2507 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2510 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2511 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2512 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2513 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2514 !is_power_of_2(sbi->s_desc_size)) {
2515 ext4_msg(sb, KERN_ERR,
2516 "unsupported descriptor size %lu",
2521 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2523 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2524 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2525 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2528 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2529 if (sbi->s_inodes_per_block == 0)
2531 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2532 sbi->s_inodes_per_block;
2533 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2535 sbi->s_mount_state = le16_to_cpu(es->s_state);
2536 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2537 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2539 for (i = 0; i < 4; i++)
2540 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2541 sbi->s_def_hash_version = es->s_def_hash_version;
2542 i = le32_to_cpu(es->s_flags);
2543 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2544 sbi->s_hash_unsigned = 3;
2545 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2546 #ifdef __CHAR_UNSIGNED__
2547 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2548 sbi->s_hash_unsigned = 3;
2550 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2555 if (sbi->s_blocks_per_group > blocksize * 8) {
2556 ext4_msg(sb, KERN_ERR,
2557 "#blocks per group too big: %lu",
2558 sbi->s_blocks_per_group);
2561 if (sbi->s_inodes_per_group > blocksize * 8) {
2562 ext4_msg(sb, KERN_ERR,
2563 "#inodes per group too big: %lu",
2564 sbi->s_inodes_per_group);
2569 * Test whether we have more sectors than will fit in sector_t,
2570 * and whether the max offset is addressable by the page cache.
2572 if ((ext4_blocks_count(es) >
2573 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2574 (ext4_blocks_count(es) >
2575 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2576 ext4_msg(sb, KERN_ERR, "filesystem"
2577 " too large to mount safely on this system");
2578 if (sizeof(sector_t) < 8)
2579 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2584 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2587 /* check blocks count against device size */
2588 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2589 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2590 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2591 "exceeds size of device (%llu blocks)",
2592 ext4_blocks_count(es), blocks_count);
2597 * It makes no sense for the first data block to be beyond the end
2598 * of the filesystem.
2600 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2601 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2602 "block %u is beyond end of filesystem (%llu)",
2603 le32_to_cpu(es->s_first_data_block),
2604 ext4_blocks_count(es));
2607 blocks_count = (ext4_blocks_count(es) -
2608 le32_to_cpu(es->s_first_data_block) +
2609 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2610 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2611 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2612 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2613 "(block count %llu, first data block %u, "
2614 "blocks per group %lu)", sbi->s_groups_count,
2615 ext4_blocks_count(es),
2616 le32_to_cpu(es->s_first_data_block),
2617 EXT4_BLOCKS_PER_GROUP(sb));
2620 sbi->s_groups_count = blocks_count;
2621 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2622 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2623 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2624 EXT4_DESC_PER_BLOCK(sb);
2625 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2627 if (sbi->s_group_desc == NULL) {
2628 ext4_msg(sb, KERN_ERR, "not enough memory");
2632 #ifdef CONFIG_PROC_FS
2634 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2637 bgl_lock_init(sbi->s_blockgroup_lock);
2639 for (i = 0; i < db_count; i++) {
2640 block = descriptor_loc(sb, logical_sb_block, i);
2641 sbi->s_group_desc[i] = sb_bread(sb, block);
2642 if (!sbi->s_group_desc[i]) {
2643 ext4_msg(sb, KERN_ERR,
2644 "can't read group descriptor %d", i);
2649 if (!ext4_check_descriptors(sb)) {
2650 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2654 if (!ext4_fill_flex_info(sb)) {
2655 ext4_msg(sb, KERN_ERR,
2656 "unable to initialize "
2657 "flex_bg meta info!");
2661 sbi->s_gdb_count = db_count;
2662 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2663 spin_lock_init(&sbi->s_next_gen_lock);
2665 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2666 ext4_count_free_blocks(sb));
2668 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2669 ext4_count_free_inodes(sb));
2672 err = percpu_counter_init(&sbi->s_dirs_counter,
2673 ext4_count_dirs(sb));
2676 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2679 ext4_msg(sb, KERN_ERR, "insufficient memory");
2683 sbi->s_stripe = ext4_get_stripe_size(sbi);
2684 sbi->s_max_writeback_mb_bump = 128;
2687 * set up enough so that it can read an inode
2689 if (!test_opt(sb, NOLOAD) &&
2690 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2691 sb->s_op = &ext4_sops;
2693 sb->s_op = &ext4_nojournal_sops;
2694 sb->s_export_op = &ext4_export_ops;
2695 sb->s_xattr = ext4_xattr_handlers;
2697 sb->s_qcop = &ext4_qctl_operations;
2698 sb->dq_op = &ext4_quota_operations;
2700 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2701 mutex_init(&sbi->s_orphan_lock);
2702 mutex_init(&sbi->s_resize_lock);
2706 needs_recovery = (es->s_last_orphan != 0 ||
2707 EXT4_HAS_INCOMPAT_FEATURE(sb,
2708 EXT4_FEATURE_INCOMPAT_RECOVER));
2711 * The first inode we look at is the journal inode. Don't try
2712 * root first: it may be modified in the journal!
2714 if (!test_opt(sb, NOLOAD) &&
2715 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2716 if (ext4_load_journal(sb, es, journal_devnum))
2718 if (!(sb->s_flags & MS_RDONLY) &&
2719 EXT4_SB(sb)->s_journal->j_failed_commit) {
2720 ext4_msg(sb, KERN_CRIT, "error: "
2721 "ext4_fill_super: Journal transaction "
2723 EXT4_SB(sb)->s_journal->j_failed_commit);
2724 if (test_opt(sb, ERRORS_RO)) {
2725 ext4_msg(sb, KERN_CRIT,
2726 "Mounting filesystem read-only");
2727 sb->s_flags |= MS_RDONLY;
2728 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2729 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2731 if (test_opt(sb, ERRORS_PANIC)) {
2732 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2733 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2734 ext4_commit_super(sb, 1);
2738 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2739 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2740 ext4_msg(sb, KERN_ERR, "required journal recovery "
2741 "suppressed and not mounted read-only");
2744 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2745 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2746 sbi->s_journal = NULL;
2751 if (ext4_blocks_count(es) > 0xffffffffULL &&
2752 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2753 JBD2_FEATURE_INCOMPAT_64BIT)) {
2754 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2758 jbd2_journal_set_features(sbi->s_journal,
2759 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2760 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2761 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2762 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2764 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2765 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2767 /* We have now updated the journal if required, so we can
2768 * validate the data journaling mode. */
2769 switch (test_opt(sb, DATA_FLAGS)) {
2771 /* No mode set, assume a default based on the journal
2772 * capabilities: ORDERED_DATA if the journal can
2773 * cope, else JOURNAL_DATA
2775 if (jbd2_journal_check_available_features
2776 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2777 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2779 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2782 case EXT4_MOUNT_ORDERED_DATA:
2783 case EXT4_MOUNT_WRITEBACK_DATA:
2784 if (!jbd2_journal_check_available_features
2785 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2786 ext4_msg(sb, KERN_ERR, "Journal does not support "
2787 "requested data journaling mode");
2793 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2797 if (test_opt(sb, NOBH)) {
2798 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2799 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2800 "its supported only with writeback mode");
2801 clear_opt(sbi->s_mount_opt, NOBH);
2805 * The jbd2_journal_load will have done any necessary log recovery,
2806 * so we can safely mount the rest of the filesystem now.
2809 root = ext4_iget(sb, EXT4_ROOT_INO);
2811 ext4_msg(sb, KERN_ERR, "get root inode failed");
2812 ret = PTR_ERR(root);
2815 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2817 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2820 sb->s_root = d_alloc_root(root);
2822 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2828 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2830 /* determine the minimum size of new large inodes, if present */
2831 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2832 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2833 EXT4_GOOD_OLD_INODE_SIZE;
2834 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2835 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2836 if (sbi->s_want_extra_isize <
2837 le16_to_cpu(es->s_want_extra_isize))
2838 sbi->s_want_extra_isize =
2839 le16_to_cpu(es->s_want_extra_isize);
2840 if (sbi->s_want_extra_isize <
2841 le16_to_cpu(es->s_min_extra_isize))
2842 sbi->s_want_extra_isize =
2843 le16_to_cpu(es->s_min_extra_isize);
2846 /* Check if enough inode space is available */
2847 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2848 sbi->s_inode_size) {
2849 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2850 EXT4_GOOD_OLD_INODE_SIZE;
2851 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2855 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2856 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2857 "requested data journaling mode");
2858 clear_opt(sbi->s_mount_opt, DELALLOC);
2859 } else if (test_opt(sb, DELALLOC))
2860 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2862 err = ext4_setup_system_zone(sb);
2864 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2865 "zone (%d)\n", err);
2870 err = ext4_mb_init(sb, needs_recovery);
2872 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2877 sbi->s_kobj.kset = ext4_kset;
2878 init_completion(&sbi->s_kobj_unregister);
2879 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2882 ext4_mb_release(sb);
2883 ext4_ext_release(sb);
2887 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2888 ext4_orphan_cleanup(sb, es);
2889 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2890 if (needs_recovery) {
2891 ext4_msg(sb, KERN_INFO, "recovery complete");
2892 ext4_mark_recovery_complete(sb, es);
2894 if (EXT4_SB(sb)->s_journal) {
2895 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2896 descr = " journalled data mode";
2897 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2898 descr = " ordered data mode";
2900 descr = " writeback data mode";
2902 descr = "out journal";
2904 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2911 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2915 ext4_msg(sb, KERN_ERR, "mount failed");
2916 ext4_release_system_zone(sb);
2917 if (sbi->s_journal) {
2918 jbd2_journal_destroy(sbi->s_journal);
2919 sbi->s_journal = NULL;
2922 if (sbi->s_flex_groups) {
2923 if (is_vmalloc_addr(sbi->s_flex_groups))
2924 vfree(sbi->s_flex_groups);
2926 kfree(sbi->s_flex_groups);
2928 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2929 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2930 percpu_counter_destroy(&sbi->s_dirs_counter);
2931 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2933 for (i = 0; i < db_count; i++)
2934 brelse(sbi->s_group_desc[i]);
2935 kfree(sbi->s_group_desc);
2938 remove_proc_entry(sb->s_id, ext4_proc_root);
2941 for (i = 0; i < MAXQUOTAS; i++)
2942 kfree(sbi->s_qf_names[i]);
2944 ext4_blkdev_remove(sbi);
2947 sb->s_fs_info = NULL;
2948 kfree(sbi->s_blockgroup_lock);
2955 * Setup any per-fs journal parameters now. We'll do this both on
2956 * initial mount, once the journal has been initialised but before we've
2957 * done any recovery; and again on any subsequent remount.
2959 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2961 struct ext4_sb_info *sbi = EXT4_SB(sb);
2963 journal->j_commit_interval = sbi->s_commit_interval;
2964 journal->j_min_batch_time = sbi->s_min_batch_time;
2965 journal->j_max_batch_time = sbi->s_max_batch_time;
2967 spin_lock(&journal->j_state_lock);
2968 if (test_opt(sb, BARRIER))
2969 journal->j_flags |= JBD2_BARRIER;
2971 journal->j_flags &= ~JBD2_BARRIER;
2972 if (test_opt(sb, DATA_ERR_ABORT))
2973 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2975 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2976 spin_unlock(&journal->j_state_lock);
2979 static journal_t *ext4_get_journal(struct super_block *sb,
2980 unsigned int journal_inum)
2982 struct inode *journal_inode;
2985 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2987 /* First, test for the existence of a valid inode on disk. Bad
2988 * things happen if we iget() an unused inode, as the subsequent
2989 * iput() will try to delete it. */
2991 journal_inode = ext4_iget(sb, journal_inum);
2992 if (IS_ERR(journal_inode)) {
2993 ext4_msg(sb, KERN_ERR, "no journal found");
2996 if (!journal_inode->i_nlink) {
2997 make_bad_inode(journal_inode);
2998 iput(journal_inode);
2999 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3003 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3004 journal_inode, journal_inode->i_size);
3005 if (!S_ISREG(journal_inode->i_mode)) {
3006 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3007 iput(journal_inode);
3011 journal = jbd2_journal_init_inode(journal_inode);
3013 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3014 iput(journal_inode);
3017 journal->j_private = sb;
3018 ext4_init_journal_params(sb, journal);
3022 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3025 struct buffer_head *bh;
3029 int hblock, blocksize;
3030 ext4_fsblk_t sb_block;
3031 unsigned long offset;
3032 struct ext4_super_block *es;
3033 struct block_device *bdev;
3035 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3037 bdev = ext4_blkdev_get(j_dev, sb);
3041 if (bd_claim(bdev, sb)) {
3042 ext4_msg(sb, KERN_ERR,
3043 "failed to claim external journal device");
3044 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3048 blocksize = sb->s_blocksize;
3049 hblock = bdev_logical_block_size(bdev);
3050 if (blocksize < hblock) {
3051 ext4_msg(sb, KERN_ERR,
3052 "blocksize too small for journal device");
3056 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3057 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3058 set_blocksize(bdev, blocksize);
3059 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3060 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3061 "external journal");
3065 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3066 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3067 !(le32_to_cpu(es->s_feature_incompat) &
3068 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3069 ext4_msg(sb, KERN_ERR, "external journal has "
3075 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3076 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3081 len = ext4_blocks_count(es);
3082 start = sb_block + 1;
3083 brelse(bh); /* we're done with the superblock */
3085 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3086 start, len, blocksize);
3088 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3091 journal->j_private = sb;
3092 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3093 wait_on_buffer(journal->j_sb_buffer);
3094 if (!buffer_uptodate(journal->j_sb_buffer)) {
3095 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3098 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3099 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3100 "user (unsupported) - %d",
3101 be32_to_cpu(journal->j_superblock->s_nr_users));
3104 EXT4_SB(sb)->journal_bdev = bdev;
3105 ext4_init_journal_params(sb, journal);
3109 jbd2_journal_destroy(journal);
3111 ext4_blkdev_put(bdev);
3115 static int ext4_load_journal(struct super_block *sb,
3116 struct ext4_super_block *es,
3117 unsigned long journal_devnum)
3120 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3123 int really_read_only;
3125 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3127 if (journal_devnum &&
3128 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3129 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3130 "numbers have changed");
3131 journal_dev = new_decode_dev(journal_devnum);
3133 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3135 really_read_only = bdev_read_only(sb->s_bdev);
3138 * Are we loading a blank journal or performing recovery after a
3139 * crash? For recovery, we need to check in advance whether we
3140 * can get read-write access to the device.
3142 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3143 if (sb->s_flags & MS_RDONLY) {
3144 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3145 "required on readonly filesystem");
3146 if (really_read_only) {
3147 ext4_msg(sb, KERN_ERR, "write access "
3148 "unavailable, cannot proceed");
3151 ext4_msg(sb, KERN_INFO, "write access will "
3152 "be enabled during recovery");
3156 if (journal_inum && journal_dev) {
3157 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3158 "and inode journals!");
3163 if (!(journal = ext4_get_journal(sb, journal_inum)))
3166 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3170 if (journal->j_flags & JBD2_BARRIER)
3171 ext4_msg(sb, KERN_INFO, "barriers enabled");
3173 ext4_msg(sb, KERN_INFO, "barriers disabled");
3175 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3176 err = jbd2_journal_update_format(journal);
3178 ext4_msg(sb, KERN_ERR, "error updating journal");
3179 jbd2_journal_destroy(journal);
3184 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3185 err = jbd2_journal_wipe(journal, !really_read_only);
3187 err = jbd2_journal_load(journal);
3190 ext4_msg(sb, KERN_ERR, "error loading journal");
3191 jbd2_journal_destroy(journal);
3195 EXT4_SB(sb)->s_journal = journal;
3196 ext4_clear_journal_err(sb, es);
3198 if (journal_devnum &&
3199 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3200 es->s_journal_dev = cpu_to_le32(journal_devnum);
3202 /* Make sure we flush the recovery flag to disk. */
3203 ext4_commit_super(sb, 1);
3209 static int ext4_commit_super(struct super_block *sb, int sync)
3211 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3212 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3217 if (buffer_write_io_error(sbh)) {
3219 * Oh, dear. A previous attempt to write the
3220 * superblock failed. This could happen because the
3221 * USB device was yanked out. Or it could happen to
3222 * be a transient write error and maybe the block will
3223 * be remapped. Nothing we can do but to retry the
3224 * write and hope for the best.
3226 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3227 "superblock detected");
3228 clear_buffer_write_io_error(sbh);
3229 set_buffer_uptodate(sbh);
3232 * If the file system is mounted read-only, don't update the
3233 * superblock write time. This avoids updating the superblock
3234 * write time when we are mounting the root file system
3235 * read/only but we need to replay the journal; at that point,
3236 * for people who are east of GMT and who make their clock
3237 * tick in localtime for Windows bug-for-bug compatibility,
3238 * the clock is set in the future, and this will cause e2fsck
3239 * to complain and force a full file system check.
3241 if (!(sb->s_flags & MS_RDONLY))
3242 es->s_wtime = cpu_to_le32(get_seconds());
3243 es->s_kbytes_written =
3244 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3245 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3246 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3247 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3248 &EXT4_SB(sb)->s_freeblocks_counter));
3249 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3250 &EXT4_SB(sb)->s_freeinodes_counter));
3252 BUFFER_TRACE(sbh, "marking dirty");
3253 mark_buffer_dirty(sbh);
3255 error = sync_dirty_buffer(sbh);
3259 error = buffer_write_io_error(sbh);
3261 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3263 clear_buffer_write_io_error(sbh);
3264 set_buffer_uptodate(sbh);
3271 * Have we just finished recovery? If so, and if we are mounting (or
3272 * remounting) the filesystem readonly, then we will end up with a
3273 * consistent fs on disk. Record that fact.
3275 static void ext4_mark_recovery_complete(struct super_block *sb,
3276 struct ext4_super_block *es)
3278 journal_t *journal = EXT4_SB(sb)->s_journal;
3280 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3281 BUG_ON(journal != NULL);
3284 jbd2_journal_lock_updates(journal);
3285 if (jbd2_journal_flush(journal) < 0)
3288 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3289 sb->s_flags & MS_RDONLY) {
3290 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3291 ext4_commit_super(sb, 1);
3295 jbd2_journal_unlock_updates(journal);
3299 * If we are mounting (or read-write remounting) a filesystem whose journal
3300 * has recorded an error from a previous lifetime, move that error to the
3301 * main filesystem now.
3303 static void ext4_clear_journal_err(struct super_block *sb,
3304 struct ext4_super_block *es)
3310 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3312 journal = EXT4_SB(sb)->s_journal;
3315 * Now check for any error status which may have been recorded in the
3316 * journal by a prior ext4_error() or ext4_abort()
3319 j_errno = jbd2_journal_errno(journal);
3323 errstr = ext4_decode_error(sb, j_errno, nbuf);
3324 ext4_warning(sb, __func__, "Filesystem error recorded "
3325 "from previous mount: %s", errstr);
3326 ext4_warning(sb, __func__, "Marking fs in need of "
3327 "filesystem check.");
3329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3330 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3331 ext4_commit_super(sb, 1);
3333 jbd2_journal_clear_err(journal);
3338 * Force the running and committing transactions to commit,
3339 * and wait on the commit.
3341 int ext4_force_commit(struct super_block *sb)
3346 if (sb->s_flags & MS_RDONLY)
3349 journal = EXT4_SB(sb)->s_journal;
3351 ret = ext4_journal_force_commit(journal);
3356 static void ext4_write_super(struct super_block *sb)
3359 ext4_commit_super(sb, 1);
3363 static int ext4_sync_fs(struct super_block *sb, int wait)
3368 trace_ext4_sync_fs(sb, wait);
3369 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3371 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3377 * LVM calls this function before a (read-only) snapshot is created. This
3378 * gives us a chance to flush the journal completely and mark the fs clean.
3380 static int ext4_freeze(struct super_block *sb)
3385 if (sb->s_flags & MS_RDONLY)
3388 journal = EXT4_SB(sb)->s_journal;
3390 /* Now we set up the journal barrier. */
3391 jbd2_journal_lock_updates(journal);
3394 * Don't clear the needs_recovery flag if we failed to flush
3397 error = jbd2_journal_flush(journal);
3400 jbd2_journal_unlock_updates(journal);
3404 /* Journal blocked and flushed, clear needs_recovery flag. */
3405 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3406 error = ext4_commit_super(sb, 1);
3413 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3414 * flag here, even though the filesystem is not technically dirty yet.
3416 static int ext4_unfreeze(struct super_block *sb)
3418 if (sb->s_flags & MS_RDONLY)
3422 /* Reset the needs_recovery flag before the fs is unlocked. */
3423 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3424 ext4_commit_super(sb, 1);
3426 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3430 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3432 struct ext4_super_block *es;
3433 struct ext4_sb_info *sbi = EXT4_SB(sb);
3434 ext4_fsblk_t n_blocks_count = 0;
3435 unsigned long old_sb_flags;
3436 struct ext4_mount_options old_opts;
3438 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3446 /* Store the original options */
3448 old_sb_flags = sb->s_flags;
3449 old_opts.s_mount_opt = sbi->s_mount_opt;
3450 old_opts.s_resuid = sbi->s_resuid;
3451 old_opts.s_resgid = sbi->s_resgid;
3452 old_opts.s_commit_interval = sbi->s_commit_interval;
3453 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3454 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3456 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3457 for (i = 0; i < MAXQUOTAS; i++)
3458 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3460 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3461 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3464 * Allow the "check" option to be passed as a remount option.
3466 if (!parse_options(data, sb, NULL, &journal_ioprio,
3467 &n_blocks_count, 1)) {
3472 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3473 ext4_abort(sb, __func__, "Abort forced by user");
3475 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3476 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3480 if (sbi->s_journal) {
3481 ext4_init_journal_params(sb, sbi->s_journal);
3482 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3485 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3486 n_blocks_count > ext4_blocks_count(es)) {
3487 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3492 if (*flags & MS_RDONLY) {
3494 * First of all, the unconditional stuff we have to do
3495 * to disable replay of the journal when we next remount
3497 sb->s_flags |= MS_RDONLY;
3500 * OK, test if we are remounting a valid rw partition
3501 * readonly, and if so set the rdonly flag and then
3502 * mark the partition as valid again.
3504 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3505 (sbi->s_mount_state & EXT4_VALID_FS))
3506 es->s_state = cpu_to_le16(sbi->s_mount_state);
3509 ext4_mark_recovery_complete(sb, es);
3511 /* Make sure we can mount this feature set readwrite */
3512 if (!ext4_feature_set_ok(sb, 0)) {
3517 * Make sure the group descriptor checksums
3518 * are sane. If they aren't, refuse to remount r/w.
3520 for (g = 0; g < sbi->s_groups_count; g++) {
3521 struct ext4_group_desc *gdp =
3522 ext4_get_group_desc(sb, g, NULL);
3524 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3525 ext4_msg(sb, KERN_ERR,
3526 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3527 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3528 le16_to_cpu(gdp->bg_checksum));
3535 * If we have an unprocessed orphan list hanging
3536 * around from a previously readonly bdev mount,
3537 * require a full umount/remount for now.
3539 if (es->s_last_orphan) {
3540 ext4_msg(sb, KERN_WARNING, "Couldn't "
3541 "remount RDWR because of unprocessed "
3542 "orphan inode list. Please "
3543 "umount/remount instead");
3549 * Mounting a RDONLY partition read-write, so reread
3550 * and store the current valid flag. (It may have
3551 * been changed by e2fsck since we originally mounted
3555 ext4_clear_journal_err(sb, es);
3556 sbi->s_mount_state = le16_to_cpu(es->s_state);
3557 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3559 if (!ext4_setup_super(sb, es, 0))
3560 sb->s_flags &= ~MS_RDONLY;
3563 ext4_setup_system_zone(sb);
3564 if (sbi->s_journal == NULL)
3565 ext4_commit_super(sb, 1);
3568 /* Release old quota file names */
3569 for (i = 0; i < MAXQUOTAS; i++)
3570 if (old_opts.s_qf_names[i] &&
3571 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3572 kfree(old_opts.s_qf_names[i]);
3579 sb->s_flags = old_sb_flags;
3580 sbi->s_mount_opt = old_opts.s_mount_opt;
3581 sbi->s_resuid = old_opts.s_resuid;
3582 sbi->s_resgid = old_opts.s_resgid;
3583 sbi->s_commit_interval = old_opts.s_commit_interval;
3584 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3585 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3587 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3588 for (i = 0; i < MAXQUOTAS; i++) {
3589 if (sbi->s_qf_names[i] &&
3590 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3591 kfree(sbi->s_qf_names[i]);
3592 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3600 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3602 struct super_block *sb = dentry->d_sb;
3603 struct ext4_sb_info *sbi = EXT4_SB(sb);
3604 struct ext4_super_block *es = sbi->s_es;
3607 if (test_opt(sb, MINIX_DF)) {
3608 sbi->s_overhead_last = 0;
3609 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3610 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3611 ext4_fsblk_t overhead = 0;
3614 * Compute the overhead (FS structures). This is constant
3615 * for a given filesystem unless the number of block groups
3616 * changes so we cache the previous value until it does.
3620 * All of the blocks before first_data_block are
3623 overhead = le32_to_cpu(es->s_first_data_block);
3626 * Add the overhead attributed to the superblock and
3627 * block group descriptors. If the sparse superblocks
3628 * feature is turned on, then not all groups have this.
3630 for (i = 0; i < ngroups; i++) {
3631 overhead += ext4_bg_has_super(sb, i) +
3632 ext4_bg_num_gdb(sb, i);
3637 * Every block group has an inode bitmap, a block
3638 * bitmap, and an inode table.
3640 overhead += ngroups * (2 + sbi->s_itb_per_group);
3641 sbi->s_overhead_last = overhead;
3643 sbi->s_blocks_last = ext4_blocks_count(es);
3646 buf->f_type = EXT4_SUPER_MAGIC;
3647 buf->f_bsize = sb->s_blocksize;
3648 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3649 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3650 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3651 ext4_free_blocks_count_set(es, buf->f_bfree);
3652 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3653 if (buf->f_bfree < ext4_r_blocks_count(es))
3655 buf->f_files = le32_to_cpu(es->s_inodes_count);
3656 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3657 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3658 buf->f_namelen = EXT4_NAME_LEN;
3659 fsid = le64_to_cpup((void *)es->s_uuid) ^
3660 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3661 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3662 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3667 /* Helper function for writing quotas on sync - we need to start transaction
3668 * before quota file is locked for write. Otherwise the are possible deadlocks:
3669 * Process 1 Process 2
3670 * ext4_create() quota_sync()
3671 * jbd2_journal_start() write_dquot()
3672 * vfs_dq_init() down(dqio_mutex)
3673 * down(dqio_mutex) jbd2_journal_start()
3679 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3681 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3684 static int ext4_write_dquot(struct dquot *dquot)
3688 struct inode *inode;
3690 inode = dquot_to_inode(dquot);
3691 handle = ext4_journal_start(inode,
3692 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3694 return PTR_ERR(handle);
3695 ret = dquot_commit(dquot);
3696 err = ext4_journal_stop(handle);
3702 static int ext4_acquire_dquot(struct dquot *dquot)
3707 handle = ext4_journal_start(dquot_to_inode(dquot),
3708 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3710 return PTR_ERR(handle);
3711 ret = dquot_acquire(dquot);
3712 err = ext4_journal_stop(handle);
3718 static int ext4_release_dquot(struct dquot *dquot)
3723 handle = ext4_journal_start(dquot_to_inode(dquot),
3724 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3725 if (IS_ERR(handle)) {
3726 /* Release dquot anyway to avoid endless cycle in dqput() */
3727 dquot_release(dquot);
3728 return PTR_ERR(handle);
3730 ret = dquot_release(dquot);
3731 err = ext4_journal_stop(handle);
3737 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3739 /* Are we journaling quotas? */
3740 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3741 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3742 dquot_mark_dquot_dirty(dquot);
3743 return ext4_write_dquot(dquot);
3745 return dquot_mark_dquot_dirty(dquot);
3749 static int ext4_write_info(struct super_block *sb, int type)
3754 /* Data block + inode block */
3755 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3757 return PTR_ERR(handle);
3758 ret = dquot_commit_info(sb, type);
3759 err = ext4_journal_stop(handle);
3766 * Turn on quotas during mount time - we need to find
3767 * the quota file and such...
3769 static int ext4_quota_on_mount(struct super_block *sb, int type)
3771 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3772 EXT4_SB(sb)->s_jquota_fmt, type);
3776 * Standard function to be called on quota_on
3778 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3779 char *name, int remount)
3784 if (!test_opt(sb, QUOTA))
3786 /* When remounting, no checks are needed and in fact, name is NULL */
3788 return vfs_quota_on(sb, type, format_id, name, remount);
3790 err = kern_path(name, LOOKUP_FOLLOW, &path);
3794 /* Quotafile not on the same filesystem? */
3795 if (path.mnt->mnt_sb != sb) {
3799 /* Journaling quota? */
3800 if (EXT4_SB(sb)->s_qf_names[type]) {
3801 /* Quotafile not in fs root? */
3802 if (path.dentry->d_parent != sb->s_root)
3803 ext4_msg(sb, KERN_WARNING,
3804 "Quota file not on filesystem root. "
3805 "Journaled quota will not work");
3809 * When we journal data on quota file, we have to flush journal to see
3810 * all updates to the file when we bypass pagecache...
3812 if (EXT4_SB(sb)->s_journal &&
3813 ext4_should_journal_data(path.dentry->d_inode)) {
3815 * We don't need to lock updates but journal_flush() could
3816 * otherwise be livelocked...
3818 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3819 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3820 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3827 err = vfs_quota_on_path(sb, type, format_id, &path);
3832 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3833 * acquiring the locks... As quota files are never truncated and quota code
3834 * itself serializes the operations (and noone else should touch the files)
3835 * we don't have to be afraid of races */
3836 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3837 size_t len, loff_t off)
3839 struct inode *inode = sb_dqopt(sb)->files[type];
3840 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3842 int offset = off & (sb->s_blocksize - 1);
3845 struct buffer_head *bh;
3846 loff_t i_size = i_size_read(inode);
3850 if (off+len > i_size)
3853 while (toread > 0) {
3854 tocopy = sb->s_blocksize - offset < toread ?
3855 sb->s_blocksize - offset : toread;
3856 bh = ext4_bread(NULL, inode, blk, 0, &err);
3859 if (!bh) /* A hole? */
3860 memset(data, 0, tocopy);
3862 memcpy(data, bh->b_data+offset, tocopy);
3872 /* Write to quotafile (we know the transaction is already started and has
3873 * enough credits) */
3874 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3875 const char *data, size_t len, loff_t off)
3877 struct inode *inode = sb_dqopt(sb)->files[type];
3878 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3880 int offset = off & (sb->s_blocksize - 1);
3882 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3883 size_t towrite = len;
3884 struct buffer_head *bh;
3885 handle_t *handle = journal_current_handle();
3887 if (EXT4_SB(sb)->s_journal && !handle) {
3888 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3889 " cancelled because transaction is not started",
3890 (unsigned long long)off, (unsigned long long)len);
3893 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3894 while (towrite > 0) {
3895 tocopy = sb->s_blocksize - offset < towrite ?
3896 sb->s_blocksize - offset : towrite;
3897 bh = ext4_bread(handle, inode, blk, 1, &err);
3900 if (journal_quota) {
3901 err = ext4_journal_get_write_access(handle, bh);
3908 memcpy(bh->b_data+offset, data, tocopy);
3909 flush_dcache_page(bh->b_page);
3912 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3914 /* Always do at least ordered writes for quotas */
3915 err = ext4_jbd2_file_inode(handle, inode);
3916 mark_buffer_dirty(bh);
3927 if (len == towrite) {
3928 mutex_unlock(&inode->i_mutex);
3931 if (inode->i_size < off+len-towrite) {
3932 i_size_write(inode, off+len-towrite);
3933 EXT4_I(inode)->i_disksize = inode->i_size;
3935 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3936 ext4_mark_inode_dirty(handle, inode);
3937 mutex_unlock(&inode->i_mutex);
3938 return len - towrite;
3943 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3944 const char *dev_name, void *data, struct vfsmount *mnt)
3946 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3949 static struct file_system_type ext4_fs_type = {
3950 .owner = THIS_MODULE,
3952 .get_sb = ext4_get_sb,
3953 .kill_sb = kill_block_super,
3954 .fs_flags = FS_REQUIRES_DEV,
3957 #ifdef CONFIG_EXT4DEV_COMPAT
3958 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3959 const char *dev_name, void *data,struct vfsmount *mnt)
3961 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3962 "to mount using ext4\n", dev_name);
3963 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3964 "will go away by 2.6.31\n", dev_name);
3965 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3968 static struct file_system_type ext4dev_fs_type = {
3969 .owner = THIS_MODULE,
3971 .get_sb = ext4dev_get_sb,
3972 .kill_sb = kill_block_super,
3973 .fs_flags = FS_REQUIRES_DEV,
3975 MODULE_ALIAS("ext4dev");
3978 static int __init init_ext4_fs(void)
3982 err = init_ext4_system_zone();
3985 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3988 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3989 err = init_ext4_mballoc();
3993 err = init_ext4_xattr();
3996 err = init_inodecache();
3999 err = register_filesystem(&ext4_fs_type);
4002 #ifdef CONFIG_EXT4DEV_COMPAT
4003 err = register_filesystem(&ext4dev_fs_type);
4005 unregister_filesystem(&ext4_fs_type);
4011 destroy_inodecache();
4015 exit_ext4_mballoc();
4017 remove_proc_entry("fs/ext4", NULL);
4018 kset_unregister(ext4_kset);
4020 exit_ext4_system_zone();
4024 static void __exit exit_ext4_fs(void)
4026 unregister_filesystem(&ext4_fs_type);
4027 #ifdef CONFIG_EXT4DEV_COMPAT
4028 unregister_filesystem(&ext4dev_fs_type);
4030 destroy_inodecache();
4032 exit_ext4_mballoc();
4033 remove_proc_entry("fs/ext4", NULL);
4034 kset_unregister(ext4_kset);
4035 exit_ext4_system_zone();
4038 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4039 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4040 MODULE_LICENSE("GPL");
4041 module_init(init_ext4_fs)
4042 module_exit(exit_ext4_fs)