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/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
46 #include "ext4_jbd2.h"
52 struct proc_dir_entry *ext4_proc_root;
53 static struct kset *ext4_kset;
55 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
56 unsigned long journal_devnum);
57 static int ext4_commit_super(struct super_block *sb,
58 struct ext4_super_block *es, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
186 * Wrappers for jbd2_journal_start/end.
188 * The only special thing we need to do here is to make sure that all
189 * journal_end calls result in the superblock being marked dirty, so
190 * that sync() will call the filesystem's write_super callback if
193 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
197 if (sb->s_flags & MS_RDONLY)
198 return ERR_PTR(-EROFS);
200 /* Special case here: if the journal has aborted behind our
201 * backs (eg. EIO in the commit thread), then we still need to
202 * take the FS itself readonly cleanly. */
203 journal = EXT4_SB(sb)->s_journal;
205 if (is_journal_aborted(journal)) {
206 ext4_abort(sb, __func__,
207 "Detected aborted journal");
208 return ERR_PTR(-EROFS);
210 return jbd2_journal_start(journal, nblocks);
213 * We're not journaling, return the appropriate indication.
215 current->journal_info = EXT4_NOJOURNAL_HANDLE;
216 return current->journal_info;
220 * The only special thing we need to do here is to make sure that all
221 * jbd2_journal_stop calls result in the superblock being marked dirty, so
222 * that sync() will call the filesystem's write_super callback if
225 int __ext4_journal_stop(const char *where, handle_t *handle)
227 struct super_block *sb;
231 if (!ext4_handle_valid(handle)) {
233 * Do this here since we don't call jbd2_journal_stop() in
236 current->journal_info = NULL;
239 sb = handle->h_transaction->t_journal->j_private;
241 rc = jbd2_journal_stop(handle);
246 __ext4_std_error(sb, where, err);
250 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
251 struct buffer_head *bh, handle_t *handle, int err)
254 const char *errstr = ext4_decode_error(NULL, err, nbuf);
256 BUG_ON(!ext4_handle_valid(handle));
259 BUFFER_TRACE(bh, "abort");
264 if (is_handle_aborted(handle))
267 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
268 caller, errstr, err_fn);
270 jbd2_journal_abort_handle(handle);
273 /* Deal with the reporting of failure conditions on a filesystem such as
274 * inconsistencies detected or read IO failures.
276 * On ext2, we can store the error state of the filesystem in the
277 * superblock. That is not possible on ext4, because we may have other
278 * write ordering constraints on the superblock which prevent us from
279 * writing it out straight away; and given that the journal is about to
280 * be aborted, we can't rely on the current, or future, transactions to
281 * write out the superblock safely.
283 * We'll just use the jbd2_journal_abort() error code to record an error in
284 * the journal instead. On recovery, the journal will compain about
285 * that error until we've noted it down and cleared it.
288 static void ext4_handle_error(struct super_block *sb)
290 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
292 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
293 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
295 if (sb->s_flags & MS_RDONLY)
298 if (!test_opt(sb, ERRORS_CONT)) {
299 journal_t *journal = EXT4_SB(sb)->s_journal;
301 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
303 jbd2_journal_abort(journal, -EIO);
305 if (test_opt(sb, ERRORS_RO)) {
306 printk(KERN_CRIT "Remounting filesystem read-only\n");
307 sb->s_flags |= MS_RDONLY;
309 ext4_commit_super(sb, es, 1);
310 if (test_opt(sb, ERRORS_PANIC))
311 panic("EXT4-fs (device %s): panic forced after error\n",
315 void ext4_error(struct super_block *sb, const char *function,
316 const char *fmt, ...)
321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
326 ext4_handle_error(sb);
329 static const char *ext4_decode_error(struct super_block *sb, int errno,
336 errstr = "IO failure";
339 errstr = "Out of memory";
342 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
343 errstr = "Journal has aborted";
345 errstr = "Readonly filesystem";
348 /* If the caller passed in an extra buffer for unknown
349 * errors, textualise them now. Else we just return
352 /* Check for truncated error codes... */
353 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
362 /* __ext4_std_error decodes expected errors from journaling functions
363 * automatically and invokes the appropriate error response. */
365 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
370 /* Special case: if the error is EROFS, and we're not already
371 * inside a transaction, then there's really no point in logging
373 if (errno == -EROFS && journal_current_handle() == NULL &&
374 (sb->s_flags & MS_RDONLY))
377 errstr = ext4_decode_error(sb, errno, nbuf);
378 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
379 sb->s_id, function, errstr);
381 ext4_handle_error(sb);
385 * ext4_abort is a much stronger failure handler than ext4_error. The
386 * abort function may be used to deal with unrecoverable failures such
387 * as journal IO errors or ENOMEM at a critical moment in log management.
389 * We unconditionally force the filesystem into an ABORT|READONLY state,
390 * unless the error response on the fs has been set to panic in which
391 * case we take the easy way out and panic immediately.
394 void ext4_abort(struct super_block *sb, const char *function,
395 const char *fmt, ...)
399 printk(KERN_CRIT "ext4_abort called.\n");
402 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
407 if (test_opt(sb, ERRORS_PANIC))
408 panic("EXT4-fs panic from previous error\n");
410 if (sb->s_flags & MS_RDONLY)
413 printk(KERN_CRIT "Remounting filesystem read-only\n");
414 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
415 sb->s_flags |= MS_RDONLY;
416 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
417 if (EXT4_SB(sb)->s_journal)
418 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
421 void ext4_warning(struct super_block *sb, const char *function,
422 const char *fmt, ...)
427 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
434 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
435 const char *function, const char *fmt, ...)
440 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
443 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
448 if (test_opt(sb, ERRORS_CONT)) {
449 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
450 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
451 ext4_commit_super(sb, es, 0);
454 ext4_unlock_group(sb, grp);
455 ext4_handle_error(sb);
457 * We only get here in the ERRORS_RO case; relocking the group
458 * may be dangerous, but nothing bad will happen since the
459 * filesystem will have already been marked read/only and the
460 * journal has been aborted. We return 1 as a hint to callers
461 * who might what to use the return value from
462 * ext4_grp_locked_error() to distinguish beween the
463 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
464 * aggressively from the ext4 function in question, with a
465 * more appropriate error code.
467 ext4_lock_group(sb, grp);
472 void ext4_update_dynamic_rev(struct super_block *sb)
474 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
476 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
479 ext4_warning(sb, __func__,
480 "updating to rev %d because of new feature flag, "
481 "running e2fsck is recommended",
484 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
485 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
486 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
487 /* leave es->s_feature_*compat flags alone */
488 /* es->s_uuid will be set by e2fsck if empty */
491 * The rest of the superblock fields should be zero, and if not it
492 * means they are likely already in use, so leave them alone. We
493 * can leave it up to e2fsck to clean up any inconsistencies there.
498 * Open the external journal device
500 static struct block_device *ext4_blkdev_get(dev_t dev)
502 struct block_device *bdev;
503 char b[BDEVNAME_SIZE];
505 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
511 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
512 __bdevname(dev, b), PTR_ERR(bdev));
517 * Release the journal device
519 static int ext4_blkdev_put(struct block_device *bdev)
522 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
525 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
527 struct block_device *bdev;
530 bdev = sbi->journal_bdev;
532 ret = ext4_blkdev_put(bdev);
533 sbi->journal_bdev = NULL;
538 static inline struct inode *orphan_list_entry(struct list_head *l)
540 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
543 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
547 printk(KERN_ERR "sb orphan head is %d\n",
548 le32_to_cpu(sbi->s_es->s_last_orphan));
550 printk(KERN_ERR "sb_info orphan list:\n");
551 list_for_each(l, &sbi->s_orphan) {
552 struct inode *inode = orphan_list_entry(l);
554 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
555 inode->i_sb->s_id, inode->i_ino, inode,
556 inode->i_mode, inode->i_nlink,
561 static void ext4_put_super(struct super_block *sb)
563 struct ext4_sb_info *sbi = EXT4_SB(sb);
564 struct ext4_super_block *es = sbi->s_es;
568 ext4_ext_release(sb);
569 ext4_xattr_put_super(sb);
570 if (sbi->s_journal) {
571 err = jbd2_journal_destroy(sbi->s_journal);
572 sbi->s_journal = NULL;
574 ext4_abort(sb, __func__,
575 "Couldn't clean up the journal");
577 if (!(sb->s_flags & MS_RDONLY)) {
578 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
579 es->s_state = cpu_to_le16(sbi->s_mount_state);
580 ext4_commit_super(sb, es, 1);
583 remove_proc_entry(sb->s_id, ext4_proc_root);
585 kobject_del(&sbi->s_kobj);
587 for (i = 0; i < sbi->s_gdb_count; i++)
588 brelse(sbi->s_group_desc[i]);
589 kfree(sbi->s_group_desc);
590 if (is_vmalloc_addr(sbi->s_flex_groups))
591 vfree(sbi->s_flex_groups);
593 kfree(sbi->s_flex_groups);
594 percpu_counter_destroy(&sbi->s_freeblocks_counter);
595 percpu_counter_destroy(&sbi->s_freeinodes_counter);
596 percpu_counter_destroy(&sbi->s_dirs_counter);
597 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
600 for (i = 0; i < MAXQUOTAS; i++)
601 kfree(sbi->s_qf_names[i]);
604 /* Debugging code just in case the in-memory inode orphan list
605 * isn't empty. The on-disk one can be non-empty if we've
606 * detected an error and taken the fs readonly, but the
607 * in-memory list had better be clean by this point. */
608 if (!list_empty(&sbi->s_orphan))
609 dump_orphan_list(sb, sbi);
610 J_ASSERT(list_empty(&sbi->s_orphan));
612 invalidate_bdev(sb->s_bdev);
613 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
615 * Invalidate the journal device's buffers. We don't want them
616 * floating about in memory - the physical journal device may
617 * hotswapped, and it breaks the `ro-after' testing code.
619 sync_blockdev(sbi->journal_bdev);
620 invalidate_bdev(sbi->journal_bdev);
621 ext4_blkdev_remove(sbi);
623 sb->s_fs_info = NULL;
625 * Now that we are completely done shutting down the
626 * superblock, we need to actually destroy the kobject.
630 kobject_put(&sbi->s_kobj);
631 wait_for_completion(&sbi->s_kobj_unregister);
634 kfree(sbi->s_blockgroup_lock);
639 static struct kmem_cache *ext4_inode_cachep;
642 * Called inside transaction, so use GFP_NOFS
644 static struct inode *ext4_alloc_inode(struct super_block *sb)
646 struct ext4_inode_info *ei;
648 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
651 #ifdef CONFIG_EXT4_FS_POSIX_ACL
652 ei->i_acl = EXT4_ACL_NOT_CACHED;
653 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
655 ei->vfs_inode.i_version = 1;
656 ei->vfs_inode.i_data.writeback_index = 0;
657 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
658 INIT_LIST_HEAD(&ei->i_prealloc_list);
659 spin_lock_init(&ei->i_prealloc_lock);
661 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
662 * therefore it can be null here. Don't check it, just initialize
665 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
666 ei->i_reserved_data_blocks = 0;
667 ei->i_reserved_meta_blocks = 0;
668 ei->i_allocated_meta_blocks = 0;
669 ei->i_delalloc_reserved_flag = 0;
670 spin_lock_init(&(ei->i_block_reservation_lock));
671 return &ei->vfs_inode;
674 static void ext4_destroy_inode(struct inode *inode)
676 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
677 printk("EXT4 Inode %p: orphan list check failed!\n",
679 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
680 EXT4_I(inode), sizeof(struct ext4_inode_info),
684 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
687 static void init_once(void *foo)
689 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
691 INIT_LIST_HEAD(&ei->i_orphan);
692 #ifdef CONFIG_EXT4_FS_XATTR
693 init_rwsem(&ei->xattr_sem);
695 init_rwsem(&ei->i_data_sem);
696 inode_init_once(&ei->vfs_inode);
699 static int init_inodecache(void)
701 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
702 sizeof(struct ext4_inode_info),
703 0, (SLAB_RECLAIM_ACCOUNT|
706 if (ext4_inode_cachep == NULL)
711 static void destroy_inodecache(void)
713 kmem_cache_destroy(ext4_inode_cachep);
716 static void ext4_clear_inode(struct inode *inode)
718 #ifdef CONFIG_EXT4_FS_POSIX_ACL
719 if (EXT4_I(inode)->i_acl &&
720 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
721 posix_acl_release(EXT4_I(inode)->i_acl);
722 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
724 if (EXT4_I(inode)->i_default_acl &&
725 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
726 posix_acl_release(EXT4_I(inode)->i_default_acl);
727 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
730 ext4_discard_preallocations(inode);
731 if (EXT4_JOURNAL(inode))
732 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
733 &EXT4_I(inode)->jinode);
736 static inline void ext4_show_quota_options(struct seq_file *seq,
737 struct super_block *sb)
739 #if defined(CONFIG_QUOTA)
740 struct ext4_sb_info *sbi = EXT4_SB(sb);
742 if (sbi->s_jquota_fmt)
743 seq_printf(seq, ",jqfmt=%s",
744 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
746 if (sbi->s_qf_names[USRQUOTA])
747 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
749 if (sbi->s_qf_names[GRPQUOTA])
750 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
752 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
753 seq_puts(seq, ",usrquota");
755 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
756 seq_puts(seq, ",grpquota");
762 * - it's set to a non-default value OR
763 * - if the per-sb default is different from the global default
765 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
768 unsigned long def_mount_opts;
769 struct super_block *sb = vfs->mnt_sb;
770 struct ext4_sb_info *sbi = EXT4_SB(sb);
771 struct ext4_super_block *es = sbi->s_es;
773 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
774 def_errors = le16_to_cpu(es->s_errors);
776 if (sbi->s_sb_block != 1)
777 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
778 if (test_opt(sb, MINIX_DF))
779 seq_puts(seq, ",minixdf");
780 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
781 seq_puts(seq, ",grpid");
782 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
783 seq_puts(seq, ",nogrpid");
784 if (sbi->s_resuid != EXT4_DEF_RESUID ||
785 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
786 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
788 if (sbi->s_resgid != EXT4_DEF_RESGID ||
789 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
790 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
792 if (test_opt(sb, ERRORS_RO)) {
793 if (def_errors == EXT4_ERRORS_PANIC ||
794 def_errors == EXT4_ERRORS_CONTINUE) {
795 seq_puts(seq, ",errors=remount-ro");
798 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
799 seq_puts(seq, ",errors=continue");
800 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
801 seq_puts(seq, ",errors=panic");
802 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
803 seq_puts(seq, ",nouid32");
804 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
805 seq_puts(seq, ",debug");
806 if (test_opt(sb, OLDALLOC))
807 seq_puts(seq, ",oldalloc");
808 #ifdef CONFIG_EXT4_FS_XATTR
809 if (test_opt(sb, XATTR_USER) &&
810 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
811 seq_puts(seq, ",user_xattr");
812 if (!test_opt(sb, XATTR_USER) &&
813 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
814 seq_puts(seq, ",nouser_xattr");
817 #ifdef CONFIG_EXT4_FS_POSIX_ACL
818 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
819 seq_puts(seq, ",acl");
820 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
821 seq_puts(seq, ",noacl");
823 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
824 seq_printf(seq, ",commit=%u",
825 (unsigned) (sbi->s_commit_interval / HZ));
827 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
828 seq_printf(seq, ",min_batch_time=%u",
829 (unsigned) sbi->s_min_batch_time);
831 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
832 seq_printf(seq, ",max_batch_time=%u",
833 (unsigned) sbi->s_min_batch_time);
837 * We're changing the default of barrier mount option, so
838 * let's always display its mount state so it's clear what its
841 seq_puts(seq, ",barrier=");
842 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
843 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
844 seq_puts(seq, ",journal_async_commit");
845 if (test_opt(sb, NOBH))
846 seq_puts(seq, ",nobh");
847 if (test_opt(sb, I_VERSION))
848 seq_puts(seq, ",i_version");
849 if (!test_opt(sb, DELALLOC))
850 seq_puts(seq, ",nodelalloc");
854 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
856 * journal mode get enabled in different ways
857 * So just print the value even if we didn't specify it
859 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
860 seq_puts(seq, ",data=journal");
861 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
862 seq_puts(seq, ",data=ordered");
863 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
864 seq_puts(seq, ",data=writeback");
866 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
867 seq_printf(seq, ",inode_readahead_blks=%u",
868 sbi->s_inode_readahead_blks);
870 if (test_opt(sb, DATA_ERR_ABORT))
871 seq_puts(seq, ",data_err=abort");
873 if (test_opt(sb, NO_AUTO_DA_ALLOC))
874 seq_puts(seq, ",noauto_da_alloc");
876 ext4_show_quota_options(seq, sb);
881 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
882 u64 ino, u32 generation)
886 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
887 return ERR_PTR(-ESTALE);
888 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
889 return ERR_PTR(-ESTALE);
891 /* iget isn't really right if the inode is currently unallocated!!
893 * ext4_read_inode will return a bad_inode if the inode had been
894 * deleted, so we should be safe.
896 * Currently we don't know the generation for parent directory, so
897 * a generation of 0 means "accept any"
899 inode = ext4_iget(sb, ino);
901 return ERR_CAST(inode);
902 if (generation && inode->i_generation != generation) {
904 return ERR_PTR(-ESTALE);
910 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
911 int fh_len, int fh_type)
913 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
917 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
918 int fh_len, int fh_type)
920 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
925 * Try to release metadata pages (indirect blocks, directories) which are
926 * mapped via the block device. Since these pages could have journal heads
927 * which would prevent try_to_free_buffers() from freeing them, we must use
928 * jbd2 layer's try_to_free_buffers() function to release them.
930 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
932 journal_t *journal = EXT4_SB(sb)->s_journal;
934 WARN_ON(PageChecked(page));
935 if (!page_has_buffers(page))
938 return jbd2_journal_try_to_free_buffers(journal, page,
940 return try_to_free_buffers(page);
944 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
945 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
947 static int ext4_write_dquot(struct dquot *dquot);
948 static int ext4_acquire_dquot(struct dquot *dquot);
949 static int ext4_release_dquot(struct dquot *dquot);
950 static int ext4_mark_dquot_dirty(struct dquot *dquot);
951 static int ext4_write_info(struct super_block *sb, int type);
952 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
953 char *path, int remount);
954 static int ext4_quota_on_mount(struct super_block *sb, int type);
955 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
956 size_t len, loff_t off);
957 static ssize_t ext4_quota_write(struct super_block *sb, int type,
958 const char *data, size_t len, loff_t off);
960 static struct dquot_operations ext4_quota_operations = {
961 .initialize = dquot_initialize,
963 .alloc_space = dquot_alloc_space,
964 .reserve_space = dquot_reserve_space,
965 .claim_space = dquot_claim_space,
966 .release_rsv = dquot_release_reserved_space,
967 .get_reserved_space = ext4_get_reserved_space,
968 .alloc_inode = dquot_alloc_inode,
969 .free_space = dquot_free_space,
970 .free_inode = dquot_free_inode,
971 .transfer = dquot_transfer,
972 .write_dquot = ext4_write_dquot,
973 .acquire_dquot = ext4_acquire_dquot,
974 .release_dquot = ext4_release_dquot,
975 .mark_dirty = ext4_mark_dquot_dirty,
976 .write_info = ext4_write_info,
977 .alloc_dquot = dquot_alloc,
978 .destroy_dquot = dquot_destroy,
981 static struct quotactl_ops ext4_qctl_operations = {
982 .quota_on = ext4_quota_on,
983 .quota_off = vfs_quota_off,
984 .quota_sync = vfs_quota_sync,
985 .get_info = vfs_get_dqinfo,
986 .set_info = vfs_set_dqinfo,
987 .get_dqblk = vfs_get_dqblk,
988 .set_dqblk = vfs_set_dqblk
992 static const struct super_operations ext4_sops = {
993 .alloc_inode = ext4_alloc_inode,
994 .destroy_inode = ext4_destroy_inode,
995 .write_inode = ext4_write_inode,
996 .dirty_inode = ext4_dirty_inode,
997 .delete_inode = ext4_delete_inode,
998 .put_super = ext4_put_super,
999 .write_super = ext4_write_super,
1000 .sync_fs = ext4_sync_fs,
1001 .freeze_fs = ext4_freeze,
1002 .unfreeze_fs = ext4_unfreeze,
1003 .statfs = ext4_statfs,
1004 .remount_fs = ext4_remount,
1005 .clear_inode = ext4_clear_inode,
1006 .show_options = ext4_show_options,
1008 .quota_read = ext4_quota_read,
1009 .quota_write = ext4_quota_write,
1011 .bdev_try_to_free_page = bdev_try_to_free_page,
1014 static const struct export_operations ext4_export_ops = {
1015 .fh_to_dentry = ext4_fh_to_dentry,
1016 .fh_to_parent = ext4_fh_to_parent,
1017 .get_parent = ext4_get_parent,
1021 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1022 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1023 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1024 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1025 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1026 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1027 Opt_journal_update, Opt_journal_dev,
1028 Opt_journal_checksum, Opt_journal_async_commit,
1029 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1030 Opt_data_err_abort, Opt_data_err_ignore,
1031 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1032 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1033 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1034 Opt_usrquota, Opt_grpquota, Opt_i_version,
1035 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1036 Opt_inode_readahead_blks, Opt_journal_ioprio
1039 static const match_table_t tokens = {
1040 {Opt_bsd_df, "bsddf"},
1041 {Opt_minix_df, "minixdf"},
1042 {Opt_grpid, "grpid"},
1043 {Opt_grpid, "bsdgroups"},
1044 {Opt_nogrpid, "nogrpid"},
1045 {Opt_nogrpid, "sysvgroups"},
1046 {Opt_resgid, "resgid=%u"},
1047 {Opt_resuid, "resuid=%u"},
1049 {Opt_err_cont, "errors=continue"},
1050 {Opt_err_panic, "errors=panic"},
1051 {Opt_err_ro, "errors=remount-ro"},
1052 {Opt_nouid32, "nouid32"},
1053 {Opt_debug, "debug"},
1054 {Opt_oldalloc, "oldalloc"},
1055 {Opt_orlov, "orlov"},
1056 {Opt_user_xattr, "user_xattr"},
1057 {Opt_nouser_xattr, "nouser_xattr"},
1059 {Opt_noacl, "noacl"},
1060 {Opt_noload, "noload"},
1063 {Opt_commit, "commit=%u"},
1064 {Opt_min_batch_time, "min_batch_time=%u"},
1065 {Opt_max_batch_time, "max_batch_time=%u"},
1066 {Opt_journal_update, "journal=update"},
1067 {Opt_journal_dev, "journal_dev=%u"},
1068 {Opt_journal_checksum, "journal_checksum"},
1069 {Opt_journal_async_commit, "journal_async_commit"},
1070 {Opt_abort, "abort"},
1071 {Opt_data_journal, "data=journal"},
1072 {Opt_data_ordered, "data=ordered"},
1073 {Opt_data_writeback, "data=writeback"},
1074 {Opt_data_err_abort, "data_err=abort"},
1075 {Opt_data_err_ignore, "data_err=ignore"},
1076 {Opt_offusrjquota, "usrjquota="},
1077 {Opt_usrjquota, "usrjquota=%s"},
1078 {Opt_offgrpjquota, "grpjquota="},
1079 {Opt_grpjquota, "grpjquota=%s"},
1080 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1081 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1082 {Opt_grpquota, "grpquota"},
1083 {Opt_noquota, "noquota"},
1084 {Opt_quota, "quota"},
1085 {Opt_usrquota, "usrquota"},
1086 {Opt_barrier, "barrier=%u"},
1087 {Opt_barrier, "barrier"},
1088 {Opt_nobarrier, "nobarrier"},
1089 {Opt_i_version, "i_version"},
1090 {Opt_stripe, "stripe=%u"},
1091 {Opt_resize, "resize"},
1092 {Opt_delalloc, "delalloc"},
1093 {Opt_nodelalloc, "nodelalloc"},
1094 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1095 {Opt_journal_ioprio, "journal_ioprio=%u"},
1096 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1097 {Opt_auto_da_alloc, "auto_da_alloc"},
1098 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1102 static ext4_fsblk_t get_sb_block(void **data)
1104 ext4_fsblk_t sb_block;
1105 char *options = (char *) *data;
1107 if (!options || strncmp(options, "sb=", 3) != 0)
1108 return 1; /* Default location */
1110 /*todo: use simple_strtoll with >32bit ext4 */
1111 sb_block = simple_strtoul(options, &options, 0);
1112 if (*options && *options != ',') {
1113 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1117 if (*options == ',')
1119 *data = (void *) options;
1123 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1125 static int parse_options(char *options, struct super_block *sb,
1126 unsigned long *journal_devnum,
1127 unsigned int *journal_ioprio,
1128 ext4_fsblk_t *n_blocks_count, int is_remount)
1130 struct ext4_sb_info *sbi = EXT4_SB(sb);
1132 substring_t args[MAX_OPT_ARGS];
1143 while ((p = strsep(&options, ",")) != NULL) {
1148 token = match_token(p, tokens, args);
1151 clear_opt(sbi->s_mount_opt, MINIX_DF);
1154 set_opt(sbi->s_mount_opt, MINIX_DF);
1157 set_opt(sbi->s_mount_opt, GRPID);
1160 clear_opt(sbi->s_mount_opt, GRPID);
1163 if (match_int(&args[0], &option))
1165 sbi->s_resuid = option;
1168 if (match_int(&args[0], &option))
1170 sbi->s_resgid = option;
1173 /* handled by get_sb_block() instead of here */
1174 /* *sb_block = match_int(&args[0]); */
1177 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1178 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1179 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1182 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1183 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1184 set_opt(sbi->s_mount_opt, ERRORS_RO);
1187 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1188 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1189 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1192 set_opt(sbi->s_mount_opt, NO_UID32);
1195 set_opt(sbi->s_mount_opt, DEBUG);
1198 set_opt(sbi->s_mount_opt, OLDALLOC);
1201 clear_opt(sbi->s_mount_opt, OLDALLOC);
1203 #ifdef CONFIG_EXT4_FS_XATTR
1204 case Opt_user_xattr:
1205 set_opt(sbi->s_mount_opt, XATTR_USER);
1207 case Opt_nouser_xattr:
1208 clear_opt(sbi->s_mount_opt, XATTR_USER);
1211 case Opt_user_xattr:
1212 case Opt_nouser_xattr:
1213 printk(KERN_ERR "EXT4 (no)user_xattr options "
1217 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1219 set_opt(sbi->s_mount_opt, POSIX_ACL);
1222 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1227 printk(KERN_ERR "EXT4 (no)acl options "
1231 case Opt_journal_update:
1233 /* Eventually we will want to be able to create
1234 a journal file here. For now, only allow the
1235 user to specify an existing inode to be the
1238 printk(KERN_ERR "EXT4-fs: cannot specify "
1239 "journal on remount\n");
1242 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1244 case Opt_journal_dev:
1246 printk(KERN_ERR "EXT4-fs: cannot specify "
1247 "journal on remount\n");
1250 if (match_int(&args[0], &option))
1252 *journal_devnum = option;
1254 case Opt_journal_checksum:
1255 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1257 case Opt_journal_async_commit:
1258 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1259 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1262 set_opt(sbi->s_mount_opt, NOLOAD);
1265 if (match_int(&args[0], &option))
1270 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1271 sbi->s_commit_interval = HZ * option;
1273 case Opt_max_batch_time:
1274 if (match_int(&args[0], &option))
1279 option = EXT4_DEF_MAX_BATCH_TIME;
1280 sbi->s_max_batch_time = option;
1282 case Opt_min_batch_time:
1283 if (match_int(&args[0], &option))
1287 sbi->s_min_batch_time = option;
1289 case Opt_data_journal:
1290 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1292 case Opt_data_ordered:
1293 data_opt = EXT4_MOUNT_ORDERED_DATA;
1295 case Opt_data_writeback:
1296 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1299 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1302 "EXT4-fs: cannot change data "
1303 "mode on remount\n");
1307 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1308 sbi->s_mount_opt |= data_opt;
1311 case Opt_data_err_abort:
1312 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1314 case Opt_data_err_ignore:
1315 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1324 if (sb_any_quota_loaded(sb) &&
1325 !sbi->s_qf_names[qtype]) {
1327 "EXT4-fs: Cannot change journaled "
1328 "quota options when quota turned on.\n");
1331 qname = match_strdup(&args[0]);
1334 "EXT4-fs: not enough memory for "
1335 "storing quotafile name.\n");
1338 if (sbi->s_qf_names[qtype] &&
1339 strcmp(sbi->s_qf_names[qtype], qname)) {
1341 "EXT4-fs: %s quota file already "
1342 "specified.\n", QTYPE2NAME(qtype));
1346 sbi->s_qf_names[qtype] = qname;
1347 if (strchr(sbi->s_qf_names[qtype], '/')) {
1349 "EXT4-fs: quotafile must be on "
1350 "filesystem root.\n");
1351 kfree(sbi->s_qf_names[qtype]);
1352 sbi->s_qf_names[qtype] = NULL;
1355 set_opt(sbi->s_mount_opt, QUOTA);
1357 case Opt_offusrjquota:
1360 case Opt_offgrpjquota:
1363 if (sb_any_quota_loaded(sb) &&
1364 sbi->s_qf_names[qtype]) {
1365 printk(KERN_ERR "EXT4-fs: Cannot change "
1366 "journaled quota options when "
1367 "quota turned on.\n");
1371 * The space will be released later when all options
1372 * are confirmed to be correct
1374 sbi->s_qf_names[qtype] = NULL;
1376 case Opt_jqfmt_vfsold:
1377 qfmt = QFMT_VFS_OLD;
1379 case Opt_jqfmt_vfsv0:
1382 if (sb_any_quota_loaded(sb) &&
1383 sbi->s_jquota_fmt != qfmt) {
1384 printk(KERN_ERR "EXT4-fs: Cannot change "
1385 "journaled quota options when "
1386 "quota turned on.\n");
1389 sbi->s_jquota_fmt = qfmt;
1393 set_opt(sbi->s_mount_opt, QUOTA);
1394 set_opt(sbi->s_mount_opt, USRQUOTA);
1397 set_opt(sbi->s_mount_opt, QUOTA);
1398 set_opt(sbi->s_mount_opt, GRPQUOTA);
1401 if (sb_any_quota_loaded(sb)) {
1402 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1403 "options when quota turned on.\n");
1406 clear_opt(sbi->s_mount_opt, QUOTA);
1407 clear_opt(sbi->s_mount_opt, USRQUOTA);
1408 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1415 "EXT4-fs: quota options not supported.\n");
1419 case Opt_offusrjquota:
1420 case Opt_offgrpjquota:
1421 case Opt_jqfmt_vfsold:
1422 case Opt_jqfmt_vfsv0:
1424 "EXT4-fs: journaled quota options not "
1431 set_opt(sbi->s_mount_opt, ABORT);
1434 clear_opt(sbi->s_mount_opt, BARRIER);
1437 if (match_int(&args[0], &option)) {
1438 set_opt(sbi->s_mount_opt, BARRIER);
1442 set_opt(sbi->s_mount_opt, BARRIER);
1444 clear_opt(sbi->s_mount_opt, BARRIER);
1450 printk("EXT4-fs: resize option only available "
1454 if (match_int(&args[0], &option) != 0)
1456 *n_blocks_count = option;
1459 set_opt(sbi->s_mount_opt, NOBH);
1462 clear_opt(sbi->s_mount_opt, NOBH);
1465 set_opt(sbi->s_mount_opt, I_VERSION);
1466 sb->s_flags |= MS_I_VERSION;
1468 case Opt_nodelalloc:
1469 clear_opt(sbi->s_mount_opt, DELALLOC);
1472 if (match_int(&args[0], &option))
1476 sbi->s_stripe = option;
1479 set_opt(sbi->s_mount_opt, DELALLOC);
1481 case Opt_inode_readahead_blks:
1482 if (match_int(&args[0], &option))
1484 if (option < 0 || option > (1 << 30))
1486 if (!is_power_of_2(option)) {
1487 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1488 " must be a power of 2\n");
1491 sbi->s_inode_readahead_blks = option;
1493 case Opt_journal_ioprio:
1494 if (match_int(&args[0], &option))
1496 if (option < 0 || option > 7)
1498 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1501 case Opt_noauto_da_alloc:
1502 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1504 case Opt_auto_da_alloc:
1505 if (match_int(&args[0], &option)) {
1506 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1510 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1512 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1516 "EXT4-fs: Unrecognized mount option \"%s\" "
1517 "or missing value\n", p);
1522 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1523 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1524 sbi->s_qf_names[USRQUOTA])
1525 clear_opt(sbi->s_mount_opt, USRQUOTA);
1527 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1528 sbi->s_qf_names[GRPQUOTA])
1529 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1531 if ((sbi->s_qf_names[USRQUOTA] &&
1532 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1533 (sbi->s_qf_names[GRPQUOTA] &&
1534 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1535 printk(KERN_ERR "EXT4-fs: old and new quota "
1536 "format mixing.\n");
1540 if (!sbi->s_jquota_fmt) {
1541 printk(KERN_ERR "EXT4-fs: journaled quota format "
1542 "not specified.\n");
1546 if (sbi->s_jquota_fmt) {
1547 printk(KERN_ERR "EXT4-fs: journaled quota format "
1548 "specified with no journaling "
1557 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1560 struct ext4_sb_info *sbi = EXT4_SB(sb);
1563 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1564 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1565 "forcing read-only mode\n");
1570 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1571 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1572 "running e2fsck is recommended\n");
1573 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1575 "EXT4-fs warning: mounting fs with errors, "
1576 "running e2fsck is recommended\n");
1577 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1578 le16_to_cpu(es->s_mnt_count) >=
1579 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1581 "EXT4-fs warning: maximal mount count reached, "
1582 "running e2fsck is recommended\n");
1583 else if (le32_to_cpu(es->s_checkinterval) &&
1584 (le32_to_cpu(es->s_lastcheck) +
1585 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1587 "EXT4-fs warning: checktime reached, "
1588 "running e2fsck is recommended\n");
1589 if (!sbi->s_journal)
1590 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1591 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1592 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1593 le16_add_cpu(&es->s_mnt_count, 1);
1594 es->s_mtime = cpu_to_le32(get_seconds());
1595 ext4_update_dynamic_rev(sb);
1597 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1599 ext4_commit_super(sb, es, 1);
1600 if (test_opt(sb, DEBUG))
1601 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1602 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1604 sbi->s_groups_count,
1605 EXT4_BLOCKS_PER_GROUP(sb),
1606 EXT4_INODES_PER_GROUP(sb),
1609 if (EXT4_SB(sb)->s_journal) {
1610 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1611 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1612 "external", EXT4_SB(sb)->s_journal->j_devname);
1614 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1619 static int ext4_fill_flex_info(struct super_block *sb)
1621 struct ext4_sb_info *sbi = EXT4_SB(sb);
1622 struct ext4_group_desc *gdp = NULL;
1623 struct buffer_head *bh;
1624 ext4_group_t flex_group_count;
1625 ext4_group_t flex_group;
1626 int groups_per_flex = 0;
1630 if (!sbi->s_es->s_log_groups_per_flex) {
1631 sbi->s_log_groups_per_flex = 0;
1635 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1636 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1638 /* We allocate both existing and potentially added groups */
1639 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1640 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1641 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1642 size = flex_group_count * sizeof(struct flex_groups);
1643 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1644 if (sbi->s_flex_groups == NULL) {
1645 sbi->s_flex_groups = vmalloc(size);
1646 if (sbi->s_flex_groups)
1647 memset(sbi->s_flex_groups, 0, size);
1649 if (sbi->s_flex_groups == NULL) {
1650 printk(KERN_ERR "EXT4-fs: not enough memory for "
1651 "%u flex groups\n", flex_group_count);
1655 for (i = 0; i < sbi->s_groups_count; i++) {
1656 gdp = ext4_get_group_desc(sb, i, &bh);
1658 flex_group = ext4_flex_group(sbi, i);
1659 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1660 ext4_free_inodes_count(sb, gdp));
1661 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1662 ext4_free_blks_count(sb, gdp));
1663 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1664 ext4_used_dirs_count(sb, gdp));
1672 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1673 struct ext4_group_desc *gdp)
1677 if (sbi->s_es->s_feature_ro_compat &
1678 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1679 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1680 __le32 le_group = cpu_to_le32(block_group);
1682 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1683 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1684 crc = crc16(crc, (__u8 *)gdp, offset);
1685 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1686 /* for checksum of struct ext4_group_desc do the rest...*/
1687 if ((sbi->s_es->s_feature_incompat &
1688 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1689 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1690 crc = crc16(crc, (__u8 *)gdp + offset,
1691 le16_to_cpu(sbi->s_es->s_desc_size) -
1695 return cpu_to_le16(crc);
1698 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1699 struct ext4_group_desc *gdp)
1701 if ((sbi->s_es->s_feature_ro_compat &
1702 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1703 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1709 /* Called at mount-time, super-block is locked */
1710 static int ext4_check_descriptors(struct super_block *sb)
1712 struct ext4_sb_info *sbi = EXT4_SB(sb);
1713 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1714 ext4_fsblk_t last_block;
1715 ext4_fsblk_t block_bitmap;
1716 ext4_fsblk_t inode_bitmap;
1717 ext4_fsblk_t inode_table;
1718 int flexbg_flag = 0;
1721 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1724 ext4_debug("Checking group descriptors");
1726 for (i = 0; i < sbi->s_groups_count; i++) {
1727 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1729 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1730 last_block = ext4_blocks_count(sbi->s_es) - 1;
1732 last_block = first_block +
1733 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1735 block_bitmap = ext4_block_bitmap(sb, gdp);
1736 if (block_bitmap < first_block || block_bitmap > last_block) {
1737 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1738 "Block bitmap for group %u not in group "
1739 "(block %llu)!\n", i, block_bitmap);
1742 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1743 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1744 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1745 "Inode bitmap for group %u not in group "
1746 "(block %llu)!\n", i, inode_bitmap);
1749 inode_table = ext4_inode_table(sb, gdp);
1750 if (inode_table < first_block ||
1751 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1752 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1753 "Inode table for group %u not in group "
1754 "(block %llu)!\n", i, inode_table);
1757 spin_lock(sb_bgl_lock(sbi, i));
1758 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1759 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1760 "Checksum for group %u failed (%u!=%u)\n",
1761 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1762 gdp)), le16_to_cpu(gdp->bg_checksum));
1763 if (!(sb->s_flags & MS_RDONLY)) {
1764 spin_unlock(sb_bgl_lock(sbi, i));
1768 spin_unlock(sb_bgl_lock(sbi, i));
1770 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1773 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1774 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1778 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1779 * the superblock) which were deleted from all directories, but held open by
1780 * a process at the time of a crash. We walk the list and try to delete these
1781 * inodes at recovery time (only with a read-write filesystem).
1783 * In order to keep the orphan inode chain consistent during traversal (in
1784 * case of crash during recovery), we link each inode into the superblock
1785 * orphan list_head and handle it the same way as an inode deletion during
1786 * normal operation (which journals the operations for us).
1788 * We only do an iget() and an iput() on each inode, which is very safe if we
1789 * accidentally point at an in-use or already deleted inode. The worst that
1790 * can happen in this case is that we get a "bit already cleared" message from
1791 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1792 * e2fsck was run on this filesystem, and it must have already done the orphan
1793 * inode cleanup for us, so we can safely abort without any further action.
1795 static void ext4_orphan_cleanup(struct super_block *sb,
1796 struct ext4_super_block *es)
1798 unsigned int s_flags = sb->s_flags;
1799 int nr_orphans = 0, nr_truncates = 0;
1803 if (!es->s_last_orphan) {
1804 jbd_debug(4, "no orphan inodes to clean up\n");
1808 if (bdev_read_only(sb->s_bdev)) {
1809 printk(KERN_ERR "EXT4-fs: write access "
1810 "unavailable, skipping orphan cleanup.\n");
1814 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1815 if (es->s_last_orphan)
1816 jbd_debug(1, "Errors on filesystem, "
1817 "clearing orphan list.\n");
1818 es->s_last_orphan = 0;
1819 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1823 if (s_flags & MS_RDONLY) {
1824 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1826 sb->s_flags &= ~MS_RDONLY;
1829 /* Needed for iput() to work correctly and not trash data */
1830 sb->s_flags |= MS_ACTIVE;
1831 /* Turn on quotas so that they are updated correctly */
1832 for (i = 0; i < MAXQUOTAS; i++) {
1833 if (EXT4_SB(sb)->s_qf_names[i]) {
1834 int ret = ext4_quota_on_mount(sb, i);
1837 "EXT4-fs: Cannot turn on journaled "
1838 "quota: error %d\n", ret);
1843 while (es->s_last_orphan) {
1844 struct inode *inode;
1846 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1847 if (IS_ERR(inode)) {
1848 es->s_last_orphan = 0;
1852 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1854 if (inode->i_nlink) {
1856 "%s: truncating inode %lu to %lld bytes\n",
1857 __func__, inode->i_ino, inode->i_size);
1858 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1859 inode->i_ino, inode->i_size);
1860 ext4_truncate(inode);
1864 "%s: deleting unreferenced inode %lu\n",
1865 __func__, inode->i_ino);
1866 jbd_debug(2, "deleting unreferenced inode %lu\n",
1870 iput(inode); /* The delete magic happens here! */
1873 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1876 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1877 sb->s_id, PLURAL(nr_orphans));
1879 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1880 sb->s_id, PLURAL(nr_truncates));
1882 /* Turn quotas off */
1883 for (i = 0; i < MAXQUOTAS; i++) {
1884 if (sb_dqopt(sb)->files[i])
1885 vfs_quota_off(sb, i, 0);
1888 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1891 * Maximal extent format file size.
1892 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1893 * extent format containers, within a sector_t, and within i_blocks
1894 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1895 * so that won't be a limiting factor.
1897 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1899 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1902 loff_t upper_limit = MAX_LFS_FILESIZE;
1904 /* small i_blocks in vfs inode? */
1905 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1907 * CONFIG_LBD is not enabled implies the inode
1908 * i_block represent total blocks in 512 bytes
1909 * 32 == size of vfs inode i_blocks * 8
1911 upper_limit = (1LL << 32) - 1;
1913 /* total blocks in file system block size */
1914 upper_limit >>= (blkbits - 9);
1915 upper_limit <<= blkbits;
1918 /* 32-bit extent-start container, ee_block */
1923 /* Sanity check against vm- & vfs- imposed limits */
1924 if (res > upper_limit)
1931 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1932 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1933 * We need to be 1 filesystem block less than the 2^48 sector limit.
1935 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1937 loff_t res = EXT4_NDIR_BLOCKS;
1940 /* This is calculated to be the largest file size for a
1941 * dense, bitmapped file such that the total number of
1942 * sectors in the file, including data and all indirect blocks,
1943 * does not exceed 2^48 -1
1944 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1945 * total number of 512 bytes blocks of the file
1948 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1950 * !has_huge_files or CONFIG_LBD is not enabled
1951 * implies the inode i_block represent total blocks in
1952 * 512 bytes 32 == size of vfs inode i_blocks * 8
1954 upper_limit = (1LL << 32) - 1;
1956 /* total blocks in file system block size */
1957 upper_limit >>= (bits - 9);
1961 * We use 48 bit ext4_inode i_blocks
1962 * With EXT4_HUGE_FILE_FL set the i_blocks
1963 * represent total number of blocks in
1964 * file system block size
1966 upper_limit = (1LL << 48) - 1;
1970 /* indirect blocks */
1972 /* double indirect blocks */
1973 meta_blocks += 1 + (1LL << (bits-2));
1974 /* tripple indirect blocks */
1975 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1977 upper_limit -= meta_blocks;
1978 upper_limit <<= bits;
1980 res += 1LL << (bits-2);
1981 res += 1LL << (2*(bits-2));
1982 res += 1LL << (3*(bits-2));
1984 if (res > upper_limit)
1987 if (res > MAX_LFS_FILESIZE)
1988 res = MAX_LFS_FILESIZE;
1993 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1994 ext4_fsblk_t logical_sb_block, int nr)
1996 struct ext4_sb_info *sbi = EXT4_SB(sb);
1997 ext4_group_t bg, first_meta_bg;
2000 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2002 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2004 return logical_sb_block + nr + 1;
2005 bg = sbi->s_desc_per_block * nr;
2006 if (ext4_bg_has_super(sb, bg))
2008 return (has_super + ext4_group_first_block_no(sb, bg));
2012 * ext4_get_stripe_size: Get the stripe size.
2013 * @sbi: In memory super block info
2015 * If we have specified it via mount option, then
2016 * use the mount option value. If the value specified at mount time is
2017 * greater than the blocks per group use the super block value.
2018 * If the super block value is greater than blocks per group return 0.
2019 * Allocator needs it be less than blocks per group.
2022 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2024 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2025 unsigned long stripe_width =
2026 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2028 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2029 return sbi->s_stripe;
2031 if (stripe_width <= sbi->s_blocks_per_group)
2032 return stripe_width;
2034 if (stride <= sbi->s_blocks_per_group)
2043 struct attribute attr;
2044 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2045 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2046 const char *, size_t);
2050 static int parse_strtoul(const char *buf,
2051 unsigned long max, unsigned long *value)
2055 while (*buf && isspace(*buf))
2057 *value = simple_strtoul(buf, &endp, 0);
2058 while (*endp && isspace(*endp))
2060 if (*endp || *value > max)
2066 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2067 struct ext4_sb_info *sbi,
2070 return snprintf(buf, PAGE_SIZE, "%llu\n",
2071 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2074 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2075 struct ext4_sb_info *sbi, char *buf)
2077 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2079 return snprintf(buf, PAGE_SIZE, "%lu\n",
2080 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2081 sbi->s_sectors_written_start) >> 1);
2084 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2085 struct ext4_sb_info *sbi, char *buf)
2087 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2089 return snprintf(buf, PAGE_SIZE, "%llu\n",
2090 sbi->s_kbytes_written +
2091 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2092 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2095 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2096 struct ext4_sb_info *sbi,
2097 const char *buf, size_t count)
2101 if (parse_strtoul(buf, 0x40000000, &t))
2104 if (!is_power_of_2(t))
2107 sbi->s_inode_readahead_blks = t;
2111 static ssize_t sbi_ui_show(struct ext4_attr *a,
2112 struct ext4_sb_info *sbi, char *buf)
2114 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2116 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2119 static ssize_t sbi_ui_store(struct ext4_attr *a,
2120 struct ext4_sb_info *sbi,
2121 const char *buf, size_t count)
2123 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2126 if (parse_strtoul(buf, 0xffffffff, &t))
2132 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2133 static struct ext4_attr ext4_attr_##_name = { \
2134 .attr = {.name = __stringify(_name), .mode = _mode }, \
2137 .offset = offsetof(struct ext4_sb_info, _elname), \
2139 #define EXT4_ATTR(name, mode, show, store) \
2140 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2142 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2143 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2144 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2145 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2146 #define ATTR_LIST(name) &ext4_attr_##name.attr
2148 EXT4_RO_ATTR(delayed_allocation_blocks);
2149 EXT4_RO_ATTR(session_write_kbytes);
2150 EXT4_RO_ATTR(lifetime_write_kbytes);
2151 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2152 inode_readahead_blks_store, s_inode_readahead_blks);
2153 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2154 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2155 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2156 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2157 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2158 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2160 static struct attribute *ext4_attrs[] = {
2161 ATTR_LIST(delayed_allocation_blocks),
2162 ATTR_LIST(session_write_kbytes),
2163 ATTR_LIST(lifetime_write_kbytes),
2164 ATTR_LIST(inode_readahead_blks),
2165 ATTR_LIST(mb_stats),
2166 ATTR_LIST(mb_max_to_scan),
2167 ATTR_LIST(mb_min_to_scan),
2168 ATTR_LIST(mb_order2_req),
2169 ATTR_LIST(mb_stream_req),
2170 ATTR_LIST(mb_group_prealloc),
2174 static ssize_t ext4_attr_show(struct kobject *kobj,
2175 struct attribute *attr, char *buf)
2177 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2179 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2181 return a->show ? a->show(a, sbi, buf) : 0;
2184 static ssize_t ext4_attr_store(struct kobject *kobj,
2185 struct attribute *attr,
2186 const char *buf, size_t len)
2188 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2190 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2192 return a->store ? a->store(a, sbi, buf, len) : 0;
2195 static void ext4_sb_release(struct kobject *kobj)
2197 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2199 complete(&sbi->s_kobj_unregister);
2203 static struct sysfs_ops ext4_attr_ops = {
2204 .show = ext4_attr_show,
2205 .store = ext4_attr_store,
2208 static struct kobj_type ext4_ktype = {
2209 .default_attrs = ext4_attrs,
2210 .sysfs_ops = &ext4_attr_ops,
2211 .release = ext4_sb_release,
2214 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2215 __releases(kernel_lock)
2216 __acquires(kernel_lock)
2219 struct buffer_head *bh;
2220 struct ext4_super_block *es = NULL;
2221 struct ext4_sb_info *sbi;
2223 ext4_fsblk_t sb_block = get_sb_block(&data);
2224 ext4_fsblk_t logical_sb_block;
2225 unsigned long offset = 0;
2226 unsigned long journal_devnum = 0;
2227 unsigned long def_mount_opts;
2233 unsigned int db_count;
2235 int needs_recovery, has_huge_files;
2239 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2241 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2245 sbi->s_blockgroup_lock =
2246 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2247 if (!sbi->s_blockgroup_lock) {
2251 sb->s_fs_info = sbi;
2252 sbi->s_mount_opt = 0;
2253 sbi->s_resuid = EXT4_DEF_RESUID;
2254 sbi->s_resgid = EXT4_DEF_RESGID;
2255 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2256 sbi->s_sb_block = sb_block;
2257 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2262 /* Cleanup superblock name */
2263 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2266 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2268 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2273 * The ext4 superblock will not be buffer aligned for other than 1kB
2274 * block sizes. We need to calculate the offset from buffer start.
2276 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2277 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2278 offset = do_div(logical_sb_block, blocksize);
2280 logical_sb_block = sb_block;
2283 if (!(bh = sb_bread(sb, logical_sb_block))) {
2284 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2288 * Note: s_es must be initialized as soon as possible because
2289 * some ext4 macro-instructions depend on its value
2291 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2293 sb->s_magic = le16_to_cpu(es->s_magic);
2294 if (sb->s_magic != EXT4_SUPER_MAGIC)
2296 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2298 /* Set defaults before we parse the mount options */
2299 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2300 if (def_mount_opts & EXT4_DEFM_DEBUG)
2301 set_opt(sbi->s_mount_opt, DEBUG);
2302 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2303 set_opt(sbi->s_mount_opt, GRPID);
2304 if (def_mount_opts & EXT4_DEFM_UID16)
2305 set_opt(sbi->s_mount_opt, NO_UID32);
2306 #ifdef CONFIG_EXT4_FS_XATTR
2307 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2308 set_opt(sbi->s_mount_opt, XATTR_USER);
2310 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2311 if (def_mount_opts & EXT4_DEFM_ACL)
2312 set_opt(sbi->s_mount_opt, POSIX_ACL);
2314 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2315 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2316 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2317 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2318 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2319 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2321 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2322 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2323 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2324 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2326 set_opt(sbi->s_mount_opt, ERRORS_RO);
2328 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2329 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2330 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2331 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2332 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2334 set_opt(sbi->s_mount_opt, BARRIER);
2337 * enable delayed allocation by default
2338 * Use -o nodelalloc to turn it off
2340 set_opt(sbi->s_mount_opt, DELALLOC);
2343 if (!parse_options((char *) data, sb, &journal_devnum,
2344 &journal_ioprio, NULL, 0))
2347 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2348 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2350 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2351 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2352 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2353 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2355 "EXT4-fs warning: feature flags set on rev 0 fs, "
2356 "running e2fsck is recommended\n");
2359 * Check feature flags regardless of the revision level, since we
2360 * previously didn't change the revision level when setting the flags,
2361 * so there is a chance incompat flags are set on a rev 0 filesystem.
2363 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2365 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2366 "unsupported optional features (%x).\n", sb->s_id,
2367 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2368 ~EXT4_FEATURE_INCOMPAT_SUPP));
2371 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2372 if (!(sb->s_flags & MS_RDONLY) && features) {
2373 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2374 "unsupported optional features (%x).\n", sb->s_id,
2375 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2376 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2379 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2380 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2381 if (has_huge_files) {
2383 * Large file size enabled file system can only be
2384 * mount if kernel is build with CONFIG_LBD
2386 if (sizeof(root->i_blocks) < sizeof(u64) &&
2387 !(sb->s_flags & MS_RDONLY)) {
2388 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2389 "files cannot be mounted read-write "
2390 "without CONFIG_LBD.\n", sb->s_id);
2394 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2396 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2397 blocksize > EXT4_MAX_BLOCK_SIZE) {
2399 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2400 blocksize, sb->s_id);
2404 if (sb->s_blocksize != blocksize) {
2406 /* Validate the filesystem blocksize */
2407 if (!sb_set_blocksize(sb, blocksize)) {
2408 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2414 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2415 offset = do_div(logical_sb_block, blocksize);
2416 bh = sb_bread(sb, logical_sb_block);
2419 "EXT4-fs: Can't read superblock on 2nd try.\n");
2422 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2424 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2426 "EXT4-fs: Magic mismatch, very weird !\n");
2431 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2433 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2435 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2436 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2437 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2439 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2440 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2441 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2442 (!is_power_of_2(sbi->s_inode_size)) ||
2443 (sbi->s_inode_size > blocksize)) {
2445 "EXT4-fs: unsupported inode size: %d\n",
2449 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2450 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2452 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2453 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2454 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2455 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2456 !is_power_of_2(sbi->s_desc_size)) {
2458 "EXT4-fs: unsupported descriptor size %lu\n",
2463 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2464 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2465 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2466 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2468 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2469 if (sbi->s_inodes_per_block == 0)
2471 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2472 sbi->s_inodes_per_block;
2473 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2475 sbi->s_mount_state = le16_to_cpu(es->s_state);
2476 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2477 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2478 for (i = 0; i < 4; i++)
2479 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2480 sbi->s_def_hash_version = es->s_def_hash_version;
2481 i = le32_to_cpu(es->s_flags);
2482 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2483 sbi->s_hash_unsigned = 3;
2484 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2485 #ifdef __CHAR_UNSIGNED__
2486 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2487 sbi->s_hash_unsigned = 3;
2489 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2494 if (sbi->s_blocks_per_group > blocksize * 8) {
2496 "EXT4-fs: #blocks per group too big: %lu\n",
2497 sbi->s_blocks_per_group);
2500 if (sbi->s_inodes_per_group > blocksize * 8) {
2502 "EXT4-fs: #inodes per group too big: %lu\n",
2503 sbi->s_inodes_per_group);
2507 if (ext4_blocks_count(es) >
2508 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2509 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2510 " too large to mount safely\n", sb->s_id);
2511 if (sizeof(sector_t) < 8)
2512 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2517 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2520 /* check blocks count against device size */
2521 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2522 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2523 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2524 "exceeds size of device (%llu blocks)\n",
2525 ext4_blocks_count(es), blocks_count);
2530 * It makes no sense for the first data block to be beyond the end
2531 * of the filesystem.
2533 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2534 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2535 "block %u is beyond end of filesystem (%llu)\n",
2536 le32_to_cpu(es->s_first_data_block),
2537 ext4_blocks_count(es));
2540 blocks_count = (ext4_blocks_count(es) -
2541 le32_to_cpu(es->s_first_data_block) +
2542 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2543 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2544 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2545 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2546 "(block count %llu, first data block %u, "
2547 "blocks per group %lu)\n", sbi->s_groups_count,
2548 ext4_blocks_count(es),
2549 le32_to_cpu(es->s_first_data_block),
2550 EXT4_BLOCKS_PER_GROUP(sb));
2553 sbi->s_groups_count = blocks_count;
2554 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2555 EXT4_DESC_PER_BLOCK(sb);
2556 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2558 if (sbi->s_group_desc == NULL) {
2559 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2563 #ifdef CONFIG_PROC_FS
2565 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2568 bgl_lock_init(sbi->s_blockgroup_lock);
2570 for (i = 0; i < db_count; i++) {
2571 block = descriptor_loc(sb, logical_sb_block, i);
2572 sbi->s_group_desc[i] = sb_bread(sb, block);
2573 if (!sbi->s_group_desc[i]) {
2574 printk(KERN_ERR "EXT4-fs: "
2575 "can't read group descriptor %d\n", i);
2580 if (!ext4_check_descriptors(sb)) {
2581 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2584 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2585 if (!ext4_fill_flex_info(sb)) {
2587 "EXT4-fs: unable to initialize "
2588 "flex_bg meta info!\n");
2592 sbi->s_gdb_count = db_count;
2593 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2594 spin_lock_init(&sbi->s_next_gen_lock);
2596 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2597 ext4_count_free_blocks(sb));
2599 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2600 ext4_count_free_inodes(sb));
2603 err = percpu_counter_init(&sbi->s_dirs_counter,
2604 ext4_count_dirs(sb));
2607 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2610 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2614 sbi->s_stripe = ext4_get_stripe_size(sbi);
2617 * set up enough so that it can read an inode
2619 sb->s_op = &ext4_sops;
2620 sb->s_export_op = &ext4_export_ops;
2621 sb->s_xattr = ext4_xattr_handlers;
2623 sb->s_qcop = &ext4_qctl_operations;
2624 sb->dq_op = &ext4_quota_operations;
2626 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2630 needs_recovery = (es->s_last_orphan != 0 ||
2631 EXT4_HAS_INCOMPAT_FEATURE(sb,
2632 EXT4_FEATURE_INCOMPAT_RECOVER));
2635 * The first inode we look at is the journal inode. Don't try
2636 * root first: it may be modified in the journal!
2638 if (!test_opt(sb, NOLOAD) &&
2639 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2640 if (ext4_load_journal(sb, es, journal_devnum))
2642 if (!(sb->s_flags & MS_RDONLY) &&
2643 EXT4_SB(sb)->s_journal->j_failed_commit) {
2644 printk(KERN_CRIT "EXT4-fs error (device %s): "
2645 "ext4_fill_super: Journal transaction "
2646 "%u is corrupt\n", sb->s_id,
2647 EXT4_SB(sb)->s_journal->j_failed_commit);
2648 if (test_opt(sb, ERRORS_RO)) {
2650 "Mounting filesystem read-only\n");
2651 sb->s_flags |= MS_RDONLY;
2652 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2653 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2655 if (test_opt(sb, ERRORS_PANIC)) {
2656 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2657 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2658 ext4_commit_super(sb, es, 1);
2662 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2663 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2664 printk(KERN_ERR "EXT4-fs: required journal recovery "
2665 "suppressed and not mounted read-only\n");
2668 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2669 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2670 sbi->s_journal = NULL;
2675 if (ext4_blocks_count(es) > 0xffffffffULL &&
2676 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2677 JBD2_FEATURE_INCOMPAT_64BIT)) {
2678 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2682 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2683 jbd2_journal_set_features(sbi->s_journal,
2684 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2685 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2686 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2687 jbd2_journal_set_features(sbi->s_journal,
2688 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2689 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2690 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2692 jbd2_journal_clear_features(sbi->s_journal,
2693 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2694 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2697 /* We have now updated the journal if required, so we can
2698 * validate the data journaling mode. */
2699 switch (test_opt(sb, DATA_FLAGS)) {
2701 /* No mode set, assume a default based on the journal
2702 * capabilities: ORDERED_DATA if the journal can
2703 * cope, else JOURNAL_DATA
2705 if (jbd2_journal_check_available_features
2706 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2707 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2709 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2712 case EXT4_MOUNT_ORDERED_DATA:
2713 case EXT4_MOUNT_WRITEBACK_DATA:
2714 if (!jbd2_journal_check_available_features
2715 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2716 printk(KERN_ERR "EXT4-fs: Journal does not support "
2717 "requested data journaling mode\n");
2723 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2727 if (test_opt(sb, NOBH)) {
2728 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2729 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2730 "its supported only with writeback mode\n");
2731 clear_opt(sbi->s_mount_opt, NOBH);
2735 * The jbd2_journal_load will have done any necessary log recovery,
2736 * so we can safely mount the rest of the filesystem now.
2739 root = ext4_iget(sb, EXT4_ROOT_INO);
2741 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2742 ret = PTR_ERR(root);
2745 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2747 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2750 sb->s_root = d_alloc_root(root);
2752 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2758 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2760 /* determine the minimum size of new large inodes, if present */
2761 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2762 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2763 EXT4_GOOD_OLD_INODE_SIZE;
2764 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2765 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2766 if (sbi->s_want_extra_isize <
2767 le16_to_cpu(es->s_want_extra_isize))
2768 sbi->s_want_extra_isize =
2769 le16_to_cpu(es->s_want_extra_isize);
2770 if (sbi->s_want_extra_isize <
2771 le16_to_cpu(es->s_min_extra_isize))
2772 sbi->s_want_extra_isize =
2773 le16_to_cpu(es->s_min_extra_isize);
2776 /* Check if enough inode space is available */
2777 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2778 sbi->s_inode_size) {
2779 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2780 EXT4_GOOD_OLD_INODE_SIZE;
2781 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2785 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2786 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2787 "requested data journaling mode\n");
2788 clear_opt(sbi->s_mount_opt, DELALLOC);
2789 } else if (test_opt(sb, DELALLOC))
2790 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2793 err = ext4_mb_init(sb, needs_recovery);
2795 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2800 sbi->s_kobj.kset = ext4_kset;
2801 init_completion(&sbi->s_kobj_unregister);
2802 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2805 ext4_mb_release(sb);
2806 ext4_ext_release(sb);
2811 * akpm: core read_super() calls in here with the superblock locked.
2812 * That deadlocks, because orphan cleanup needs to lock the superblock
2813 * in numerous places. Here we just pop the lock - it's relatively
2814 * harmless, because we are now ready to accept write_super() requests,
2815 * and aviro says that's the only reason for hanging onto the
2818 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2819 ext4_orphan_cleanup(sb, es);
2820 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2821 if (needs_recovery) {
2822 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2823 ext4_mark_recovery_complete(sb, es);
2825 if (EXT4_SB(sb)->s_journal) {
2826 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2827 descr = " journalled data mode";
2828 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2829 descr = " ordered data mode";
2831 descr = " writeback data mode";
2833 descr = "out journal";
2835 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2843 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2848 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2849 if (sbi->s_journal) {
2850 jbd2_journal_destroy(sbi->s_journal);
2851 sbi->s_journal = NULL;
2854 if (sbi->s_flex_groups) {
2855 if (is_vmalloc_addr(sbi->s_flex_groups))
2856 vfree(sbi->s_flex_groups);
2858 kfree(sbi->s_flex_groups);
2860 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2861 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2862 percpu_counter_destroy(&sbi->s_dirs_counter);
2863 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2865 for (i = 0; i < db_count; i++)
2866 brelse(sbi->s_group_desc[i]);
2867 kfree(sbi->s_group_desc);
2870 remove_proc_entry(sb->s_id, ext4_proc_root);
2873 for (i = 0; i < MAXQUOTAS; i++)
2874 kfree(sbi->s_qf_names[i]);
2876 ext4_blkdev_remove(sbi);
2879 sb->s_fs_info = NULL;
2886 * Setup any per-fs journal parameters now. We'll do this both on
2887 * initial mount, once the journal has been initialised but before we've
2888 * done any recovery; and again on any subsequent remount.
2890 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2892 struct ext4_sb_info *sbi = EXT4_SB(sb);
2894 journal->j_commit_interval = sbi->s_commit_interval;
2895 journal->j_min_batch_time = sbi->s_min_batch_time;
2896 journal->j_max_batch_time = sbi->s_max_batch_time;
2898 spin_lock(&journal->j_state_lock);
2899 if (test_opt(sb, BARRIER))
2900 journal->j_flags |= JBD2_BARRIER;
2902 journal->j_flags &= ~JBD2_BARRIER;
2903 if (test_opt(sb, DATA_ERR_ABORT))
2904 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2906 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2907 spin_unlock(&journal->j_state_lock);
2910 static journal_t *ext4_get_journal(struct super_block *sb,
2911 unsigned int journal_inum)
2913 struct inode *journal_inode;
2916 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2918 /* First, test for the existence of a valid inode on disk. Bad
2919 * things happen if we iget() an unused inode, as the subsequent
2920 * iput() will try to delete it. */
2922 journal_inode = ext4_iget(sb, journal_inum);
2923 if (IS_ERR(journal_inode)) {
2924 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2927 if (!journal_inode->i_nlink) {
2928 make_bad_inode(journal_inode);
2929 iput(journal_inode);
2930 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2934 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2935 journal_inode, journal_inode->i_size);
2936 if (!S_ISREG(journal_inode->i_mode)) {
2937 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2938 iput(journal_inode);
2942 journal = jbd2_journal_init_inode(journal_inode);
2944 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2945 iput(journal_inode);
2948 journal->j_private = sb;
2949 ext4_init_journal_params(sb, journal);
2953 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2956 struct buffer_head *bh;
2960 int hblock, blocksize;
2961 ext4_fsblk_t sb_block;
2962 unsigned long offset;
2963 struct ext4_super_block *es;
2964 struct block_device *bdev;
2966 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2968 bdev = ext4_blkdev_get(j_dev);
2972 if (bd_claim(bdev, sb)) {
2974 "EXT4-fs: failed to claim external journal device.\n");
2975 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2979 blocksize = sb->s_blocksize;
2980 hblock = bdev_hardsect_size(bdev);
2981 if (blocksize < hblock) {
2983 "EXT4-fs: blocksize too small for journal device.\n");
2987 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2988 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2989 set_blocksize(bdev, blocksize);
2990 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2991 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2992 "external journal\n");
2996 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2997 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2998 !(le32_to_cpu(es->s_feature_incompat) &
2999 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3000 printk(KERN_ERR "EXT4-fs: external journal has "
3001 "bad superblock\n");
3006 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3007 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3012 len = ext4_blocks_count(es);
3013 start = sb_block + 1;
3014 brelse(bh); /* we're done with the superblock */
3016 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3017 start, len, blocksize);
3019 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3022 journal->j_private = sb;
3023 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3024 wait_on_buffer(journal->j_sb_buffer);
3025 if (!buffer_uptodate(journal->j_sb_buffer)) {
3026 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3029 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3030 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3031 "user (unsupported) - %d\n",
3032 be32_to_cpu(journal->j_superblock->s_nr_users));
3035 EXT4_SB(sb)->journal_bdev = bdev;
3036 ext4_init_journal_params(sb, journal);
3039 jbd2_journal_destroy(journal);
3041 ext4_blkdev_put(bdev);
3045 static int ext4_load_journal(struct super_block *sb,
3046 struct ext4_super_block *es,
3047 unsigned long journal_devnum)
3050 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3053 int really_read_only;
3055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3057 if (journal_devnum &&
3058 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3059 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3060 "numbers have changed\n");
3061 journal_dev = new_decode_dev(journal_devnum);
3063 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3065 really_read_only = bdev_read_only(sb->s_bdev);
3068 * Are we loading a blank journal or performing recovery after a
3069 * crash? For recovery, we need to check in advance whether we
3070 * can get read-write access to the device.
3073 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3074 if (sb->s_flags & MS_RDONLY) {
3075 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3076 "required on readonly filesystem.\n");
3077 if (really_read_only) {
3078 printk(KERN_ERR "EXT4-fs: write access "
3079 "unavailable, cannot proceed.\n");
3082 printk(KERN_INFO "EXT4-fs: write access will "
3083 "be enabled during recovery.\n");
3087 if (journal_inum && journal_dev) {
3088 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3089 "and inode journals!\n");
3094 if (!(journal = ext4_get_journal(sb, journal_inum)))
3097 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3101 if (journal->j_flags & JBD2_BARRIER)
3102 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3104 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3106 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3107 err = jbd2_journal_update_format(journal);
3109 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3110 jbd2_journal_destroy(journal);
3115 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3116 err = jbd2_journal_wipe(journal, !really_read_only);
3118 err = jbd2_journal_load(journal);
3121 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3122 jbd2_journal_destroy(journal);
3126 EXT4_SB(sb)->s_journal = journal;
3127 ext4_clear_journal_err(sb, es);
3129 if (journal_devnum &&
3130 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3131 es->s_journal_dev = cpu_to_le32(journal_devnum);
3134 /* Make sure we flush the recovery flag to disk. */
3135 ext4_commit_super(sb, es, 1);
3141 static int ext4_commit_super(struct super_block *sb,
3142 struct ext4_super_block *es, int sync)
3144 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3149 if (buffer_write_io_error(sbh)) {
3151 * Oh, dear. A previous attempt to write the
3152 * superblock failed. This could happen because the
3153 * USB device was yanked out. Or it could happen to
3154 * be a transient write error and maybe the block will
3155 * be remapped. Nothing we can do but to retry the
3156 * write and hope for the best.
3158 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3159 "superblock detected for %s.\n", sb->s_id);
3160 clear_buffer_write_io_error(sbh);
3161 set_buffer_uptodate(sbh);
3163 es->s_wtime = cpu_to_le32(get_seconds());
3164 es->s_kbytes_written =
3165 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3166 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3167 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3168 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3169 &EXT4_SB(sb)->s_freeblocks_counter));
3170 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3171 &EXT4_SB(sb)->s_freeinodes_counter));
3173 BUFFER_TRACE(sbh, "marking dirty");
3174 mark_buffer_dirty(sbh);
3176 error = sync_dirty_buffer(sbh);
3180 error = buffer_write_io_error(sbh);
3182 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3183 "superblock for %s.\n", sb->s_id);
3184 clear_buffer_write_io_error(sbh);
3185 set_buffer_uptodate(sbh);
3193 * Have we just finished recovery? If so, and if we are mounting (or
3194 * remounting) the filesystem readonly, then we will end up with a
3195 * consistent fs on disk. Record that fact.
3197 static void ext4_mark_recovery_complete(struct super_block *sb,
3198 struct ext4_super_block *es)
3200 journal_t *journal = EXT4_SB(sb)->s_journal;
3202 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3203 BUG_ON(journal != NULL);
3206 jbd2_journal_lock_updates(journal);
3207 if (jbd2_journal_flush(journal) < 0)
3211 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3212 sb->s_flags & MS_RDONLY) {
3213 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3215 ext4_commit_super(sb, es, 1);
3220 jbd2_journal_unlock_updates(journal);
3224 * If we are mounting (or read-write remounting) a filesystem whose journal
3225 * has recorded an error from a previous lifetime, move that error to the
3226 * main filesystem now.
3228 static void ext4_clear_journal_err(struct super_block *sb,
3229 struct ext4_super_block *es)
3235 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3237 journal = EXT4_SB(sb)->s_journal;
3240 * Now check for any error status which may have been recorded in the
3241 * journal by a prior ext4_error() or ext4_abort()
3244 j_errno = jbd2_journal_errno(journal);
3248 errstr = ext4_decode_error(sb, j_errno, nbuf);
3249 ext4_warning(sb, __func__, "Filesystem error recorded "
3250 "from previous mount: %s", errstr);
3251 ext4_warning(sb, __func__, "Marking fs in need of "
3252 "filesystem check.");
3254 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3255 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3256 ext4_commit_super(sb, es, 1);
3258 jbd2_journal_clear_err(journal);
3263 * Force the running and committing transactions to commit,
3264 * and wait on the commit.
3266 int ext4_force_commit(struct super_block *sb)
3271 if (sb->s_flags & MS_RDONLY)
3274 journal = EXT4_SB(sb)->s_journal;
3277 ret = ext4_journal_force_commit(journal);
3284 * Ext4 always journals updates to the superblock itself, so we don't
3285 * have to propagate any other updates to the superblock on disk at this
3286 * point. (We can probably nuke this function altogether, and remove
3287 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3289 static void ext4_write_super(struct super_block *sb)
3291 if (EXT4_SB(sb)->s_journal) {
3292 if (mutex_trylock(&sb->s_lock) != 0)
3296 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3300 static int ext4_sync_fs(struct super_block *sb, int wait)
3305 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3307 if (EXT4_SB(sb)->s_journal) {
3308 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3311 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3315 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3321 * LVM calls this function before a (read-only) snapshot is created. This
3322 * gives us a chance to flush the journal completely and mark the fs clean.
3324 static int ext4_freeze(struct super_block *sb)
3330 if (!(sb->s_flags & MS_RDONLY)) {
3331 journal = EXT4_SB(sb)->s_journal;
3334 /* Now we set up the journal barrier. */
3335 jbd2_journal_lock_updates(journal);
3338 * We don't want to clear needs_recovery flag when we
3339 * failed to flush the journal.
3341 error = jbd2_journal_flush(journal);
3346 /* Journal blocked and flushed, clear needs_recovery flag. */
3347 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3348 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3354 jbd2_journal_unlock_updates(journal);
3359 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3360 * flag here, even though the filesystem is not technically dirty yet.
3362 static int ext4_unfreeze(struct super_block *sb)
3364 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3366 /* Reser the needs_recovery flag before the fs is unlocked. */
3367 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3368 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3370 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3375 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3377 struct ext4_super_block *es;
3378 struct ext4_sb_info *sbi = EXT4_SB(sb);
3379 ext4_fsblk_t n_blocks_count = 0;
3380 unsigned long old_sb_flags;
3381 struct ext4_mount_options old_opts;
3383 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3389 /* Store the original options */
3390 old_sb_flags = sb->s_flags;
3391 old_opts.s_mount_opt = sbi->s_mount_opt;
3392 old_opts.s_resuid = sbi->s_resuid;
3393 old_opts.s_resgid = sbi->s_resgid;
3394 old_opts.s_commit_interval = sbi->s_commit_interval;
3395 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3396 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3398 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3399 for (i = 0; i < MAXQUOTAS; i++)
3400 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3402 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3403 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3406 * Allow the "check" option to be passed as a remount option.
3408 if (!parse_options(data, sb, NULL, &journal_ioprio,
3409 &n_blocks_count, 1)) {
3414 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3415 ext4_abort(sb, __func__, "Abort forced by user");
3417 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3418 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3422 if (sbi->s_journal) {
3423 ext4_init_journal_params(sb, sbi->s_journal);
3424 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3427 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3428 n_blocks_count > ext4_blocks_count(es)) {
3429 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3434 if (*flags & MS_RDONLY) {
3436 * First of all, the unconditional stuff we have to do
3437 * to disable replay of the journal when we next remount
3439 sb->s_flags |= MS_RDONLY;
3442 * OK, test if we are remounting a valid rw partition
3443 * readonly, and if so set the rdonly flag and then
3444 * mark the partition as valid again.
3446 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3447 (sbi->s_mount_state & EXT4_VALID_FS))
3448 es->s_state = cpu_to_le16(sbi->s_mount_state);
3451 * We have to unlock super so that we can wait for
3454 if (sbi->s_journal) {
3456 ext4_mark_recovery_complete(sb, es);
3461 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3462 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3463 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3464 "remount RDWR because of unsupported "
3465 "optional features (%x).\n", sb->s_id,
3466 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3467 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3473 * Make sure the group descriptor checksums
3474 * are sane. If they aren't, refuse to
3477 for (g = 0; g < sbi->s_groups_count; g++) {
3478 struct ext4_group_desc *gdp =
3479 ext4_get_group_desc(sb, g, NULL);
3481 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3483 "EXT4-fs: ext4_remount: "
3484 "Checksum for group %u failed (%u!=%u)\n",
3485 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3486 le16_to_cpu(gdp->bg_checksum));
3493 * If we have an unprocessed orphan list hanging
3494 * around from a previously readonly bdev mount,
3495 * require a full umount/remount for now.
3497 if (es->s_last_orphan) {
3498 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3499 "remount RDWR because of unprocessed "
3500 "orphan inode list. Please "
3501 "umount/remount instead.\n",
3508 * Mounting a RDONLY partition read-write, so reread
3509 * and store the current valid flag. (It may have
3510 * been changed by e2fsck since we originally mounted
3514 ext4_clear_journal_err(sb, es);
3515 sbi->s_mount_state = le16_to_cpu(es->s_state);
3516 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3518 if (!ext4_setup_super(sb, es, 0))
3519 sb->s_flags &= ~MS_RDONLY;
3522 if (sbi->s_journal == NULL)
3523 ext4_commit_super(sb, es, 1);
3526 /* Release old quota file names */
3527 for (i = 0; i < MAXQUOTAS; i++)
3528 if (old_opts.s_qf_names[i] &&
3529 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3530 kfree(old_opts.s_qf_names[i]);
3534 sb->s_flags = old_sb_flags;
3535 sbi->s_mount_opt = old_opts.s_mount_opt;
3536 sbi->s_resuid = old_opts.s_resuid;
3537 sbi->s_resgid = old_opts.s_resgid;
3538 sbi->s_commit_interval = old_opts.s_commit_interval;
3539 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3540 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3542 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3543 for (i = 0; i < MAXQUOTAS; i++) {
3544 if (sbi->s_qf_names[i] &&
3545 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3546 kfree(sbi->s_qf_names[i]);
3547 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3553 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3555 struct super_block *sb = dentry->d_sb;
3556 struct ext4_sb_info *sbi = EXT4_SB(sb);
3557 struct ext4_super_block *es = sbi->s_es;
3560 if (test_opt(sb, MINIX_DF)) {
3561 sbi->s_overhead_last = 0;
3562 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3563 ext4_group_t ngroups = sbi->s_groups_count, i;
3564 ext4_fsblk_t overhead = 0;
3568 * Compute the overhead (FS structures). This is constant
3569 * for a given filesystem unless the number of block groups
3570 * changes so we cache the previous value until it does.
3574 * All of the blocks before first_data_block are
3577 overhead = le32_to_cpu(es->s_first_data_block);
3580 * Add the overhead attributed to the superblock and
3581 * block group descriptors. If the sparse superblocks
3582 * feature is turned on, then not all groups have this.
3584 for (i = 0; i < ngroups; i++) {
3585 overhead += ext4_bg_has_super(sb, i) +
3586 ext4_bg_num_gdb(sb, i);
3591 * Every block group has an inode bitmap, a block
3592 * bitmap, and an inode table.
3594 overhead += ngroups * (2 + sbi->s_itb_per_group);
3595 sbi->s_overhead_last = overhead;
3597 sbi->s_blocks_last = ext4_blocks_count(es);
3600 buf->f_type = EXT4_SUPER_MAGIC;
3601 buf->f_bsize = sb->s_blocksize;
3602 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3603 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3604 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3605 ext4_free_blocks_count_set(es, buf->f_bfree);
3606 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3607 if (buf->f_bfree < ext4_r_blocks_count(es))
3609 buf->f_files = le32_to_cpu(es->s_inodes_count);
3610 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3611 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3612 buf->f_namelen = EXT4_NAME_LEN;
3613 fsid = le64_to_cpup((void *)es->s_uuid) ^
3614 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3615 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3616 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3620 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3621 * is locked for write. Otherwise the are possible deadlocks:
3622 * Process 1 Process 2
3623 * ext4_create() quota_sync()
3624 * jbd2_journal_start() write_dquot()
3625 * vfs_dq_init() down(dqio_mutex)
3626 * down(dqio_mutex) jbd2_journal_start()
3632 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3634 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3637 static int ext4_write_dquot(struct dquot *dquot)
3641 struct inode *inode;
3643 inode = dquot_to_inode(dquot);
3644 handle = ext4_journal_start(inode,
3645 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3647 return PTR_ERR(handle);
3648 ret = dquot_commit(dquot);
3649 err = ext4_journal_stop(handle);
3655 static int ext4_acquire_dquot(struct dquot *dquot)
3660 handle = ext4_journal_start(dquot_to_inode(dquot),
3661 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3663 return PTR_ERR(handle);
3664 ret = dquot_acquire(dquot);
3665 err = ext4_journal_stop(handle);
3671 static int ext4_release_dquot(struct dquot *dquot)
3676 handle = ext4_journal_start(dquot_to_inode(dquot),
3677 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3678 if (IS_ERR(handle)) {
3679 /* Release dquot anyway to avoid endless cycle in dqput() */
3680 dquot_release(dquot);
3681 return PTR_ERR(handle);
3683 ret = dquot_release(dquot);
3684 err = ext4_journal_stop(handle);
3690 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3692 /* Are we journaling quotas? */
3693 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3694 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3695 dquot_mark_dquot_dirty(dquot);
3696 return ext4_write_dquot(dquot);
3698 return dquot_mark_dquot_dirty(dquot);
3702 static int ext4_write_info(struct super_block *sb, int type)
3707 /* Data block + inode block */
3708 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3710 return PTR_ERR(handle);
3711 ret = dquot_commit_info(sb, type);
3712 err = ext4_journal_stop(handle);
3719 * Turn on quotas during mount time - we need to find
3720 * the quota file and such...
3722 static int ext4_quota_on_mount(struct super_block *sb, int type)
3724 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3725 EXT4_SB(sb)->s_jquota_fmt, type);
3729 * Standard function to be called on quota_on
3731 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3732 char *name, int remount)
3737 if (!test_opt(sb, QUOTA))
3739 /* When remounting, no checks are needed and in fact, name is NULL */
3741 return vfs_quota_on(sb, type, format_id, name, remount);
3743 err = kern_path(name, LOOKUP_FOLLOW, &path);
3747 /* Quotafile not on the same filesystem? */
3748 if (path.mnt->mnt_sb != sb) {
3752 /* Journaling quota? */
3753 if (EXT4_SB(sb)->s_qf_names[type]) {
3754 /* Quotafile not in fs root? */
3755 if (path.dentry->d_parent != sb->s_root)
3757 "EXT4-fs: Quota file not on filesystem root. "
3758 "Journaled quota will not work.\n");
3762 * When we journal data on quota file, we have to flush journal to see
3763 * all updates to the file when we bypass pagecache...
3765 if (EXT4_SB(sb)->s_journal &&
3766 ext4_should_journal_data(path.dentry->d_inode)) {
3768 * We don't need to lock updates but journal_flush() could
3769 * otherwise be livelocked...
3771 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3772 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3773 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3780 err = vfs_quota_on_path(sb, type, format_id, &path);
3785 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3786 * acquiring the locks... As quota files are never truncated and quota code
3787 * itself serializes the operations (and noone else should touch the files)
3788 * we don't have to be afraid of races */
3789 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3790 size_t len, loff_t off)
3792 struct inode *inode = sb_dqopt(sb)->files[type];
3793 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3795 int offset = off & (sb->s_blocksize - 1);
3798 struct buffer_head *bh;
3799 loff_t i_size = i_size_read(inode);
3803 if (off+len > i_size)
3806 while (toread > 0) {
3807 tocopy = sb->s_blocksize - offset < toread ?
3808 sb->s_blocksize - offset : toread;
3809 bh = ext4_bread(NULL, inode, blk, 0, &err);
3812 if (!bh) /* A hole? */
3813 memset(data, 0, tocopy);
3815 memcpy(data, bh->b_data+offset, tocopy);
3825 /* Write to quotafile (we know the transaction is already started and has
3826 * enough credits) */
3827 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3828 const char *data, size_t len, loff_t off)
3830 struct inode *inode = sb_dqopt(sb)->files[type];
3831 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3833 int offset = off & (sb->s_blocksize - 1);
3835 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3836 size_t towrite = len;
3837 struct buffer_head *bh;
3838 handle_t *handle = journal_current_handle();
3840 if (EXT4_SB(sb)->s_journal && !handle) {
3841 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3842 " cancelled because transaction is not started.\n",
3843 (unsigned long long)off, (unsigned long long)len);
3846 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3847 while (towrite > 0) {
3848 tocopy = sb->s_blocksize - offset < towrite ?
3849 sb->s_blocksize - offset : towrite;
3850 bh = ext4_bread(handle, inode, blk, 1, &err);
3853 if (journal_quota) {
3854 err = ext4_journal_get_write_access(handle, bh);
3861 memcpy(bh->b_data+offset, data, tocopy);
3862 flush_dcache_page(bh->b_page);
3865 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3867 /* Always do at least ordered writes for quotas */
3868 err = ext4_jbd2_file_inode(handle, inode);
3869 mark_buffer_dirty(bh);
3880 if (len == towrite) {
3881 mutex_unlock(&inode->i_mutex);
3884 if (inode->i_size < off+len-towrite) {
3885 i_size_write(inode, off+len-towrite);
3886 EXT4_I(inode)->i_disksize = inode->i_size;
3888 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3889 ext4_mark_inode_dirty(handle, inode);
3890 mutex_unlock(&inode->i_mutex);
3891 return len - towrite;
3896 static int ext4_get_sb(struct file_system_type *fs_type,
3897 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3899 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3902 static struct file_system_type ext4_fs_type = {
3903 .owner = THIS_MODULE,
3905 .get_sb = ext4_get_sb,
3906 .kill_sb = kill_block_super,
3907 .fs_flags = FS_REQUIRES_DEV,
3910 #ifdef CONFIG_EXT4DEV_COMPAT
3911 static int ext4dev_get_sb(struct file_system_type *fs_type,
3912 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3914 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3915 "to mount using ext4\n");
3916 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3917 "will go away by 2.6.31\n");
3918 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3921 static struct file_system_type ext4dev_fs_type = {
3922 .owner = THIS_MODULE,
3924 .get_sb = ext4dev_get_sb,
3925 .kill_sb = kill_block_super,
3926 .fs_flags = FS_REQUIRES_DEV,
3928 MODULE_ALIAS("ext4dev");
3931 static int __init init_ext4_fs(void)
3935 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3938 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3939 err = init_ext4_mballoc();
3943 err = init_ext4_xattr();
3946 err = init_inodecache();
3949 err = register_filesystem(&ext4_fs_type);
3952 #ifdef CONFIG_EXT4DEV_COMPAT
3953 err = register_filesystem(&ext4dev_fs_type);
3955 unregister_filesystem(&ext4_fs_type);
3961 destroy_inodecache();
3965 exit_ext4_mballoc();
3969 static void __exit exit_ext4_fs(void)
3971 unregister_filesystem(&ext4_fs_type);
3972 #ifdef CONFIG_EXT4DEV_COMPAT
3973 unregister_filesystem(&ext4dev_fs_type);
3975 destroy_inodecache();
3977 exit_ext4_mballoc();
3978 remove_proc_entry("fs/ext4", NULL);
3979 kset_unregister(ext4_kset);
3982 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3983 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3984 MODULE_LICENSE("GPL");
3985 module_init(init_ext4_fs)
3986 module_exit(exit_ext4_fs)