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/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/marker.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
51 struct proc_dir_entry *ext4_proc_root;
52 static struct kset *ext4_kset;
54 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
55 unsigned long journal_devnum);
56 static int ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static int ext4_unfreeze(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static int ext4_freeze(struct super_block *sb);
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
96 __u32 ext4_free_blks_count(struct super_block *sb,
97 struct ext4_group_desc *bg)
99 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
104 __u32 ext4_free_inodes_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
107 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
112 __u32 ext4_used_dirs_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
115 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
120 __u32 ext4_itable_unused_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
123 return le16_to_cpu(bg->bg_itable_unused_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
128 void ext4_block_bitmap_set(struct super_block *sb,
129 struct ext4_group_desc *bg, ext4_fsblk_t blk)
131 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
132 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
133 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
136 void ext4_inode_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
144 void ext4_inode_table_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
152 void ext4_free_blks_set(struct super_block *sb,
153 struct ext4_group_desc *bg, __u32 count)
155 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
160 void ext4_free_inodes_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
163 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
168 void ext4_used_dirs_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
171 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
176 void ext4_itable_unused_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
179 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
185 * Wrappers for jbd2_journal_start/end.
187 * The only special thing we need to do here is to make sure that all
188 * journal_end calls result in the superblock being marked dirty, so
189 * that sync() will call the filesystem's write_super callback if
192 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
196 if (sb->s_flags & MS_RDONLY)
197 return ERR_PTR(-EROFS);
199 /* Special case here: if the journal has aborted behind our
200 * backs (eg. EIO in the commit thread), then we still need to
201 * take the FS itself readonly cleanly. */
202 journal = EXT4_SB(sb)->s_journal;
204 if (is_journal_aborted(journal)) {
205 ext4_abort(sb, __func__,
206 "Detected aborted journal");
207 return ERR_PTR(-EROFS);
209 return jbd2_journal_start(journal, nblocks);
212 * We're not journaling, return the appropriate indication.
214 current->journal_info = EXT4_NOJOURNAL_HANDLE;
215 return current->journal_info;
219 * The only special thing we need to do here is to make sure that all
220 * jbd2_journal_stop calls result in the superblock being marked dirty, so
221 * that sync() will call the filesystem's write_super callback if
224 int __ext4_journal_stop(const char *where, handle_t *handle)
226 struct super_block *sb;
230 if (!ext4_handle_valid(handle)) {
232 * Do this here since we don't call jbd2_journal_stop() in
235 current->journal_info = NULL;
238 sb = handle->h_transaction->t_journal->j_private;
240 rc = jbd2_journal_stop(handle);
245 __ext4_std_error(sb, where, err);
249 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
250 struct buffer_head *bh, handle_t *handle, int err)
253 const char *errstr = ext4_decode_error(NULL, err, nbuf);
255 BUG_ON(!ext4_handle_valid(handle));
258 BUFFER_TRACE(bh, "abort");
263 if (is_handle_aborted(handle))
266 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
267 caller, errstr, err_fn);
269 jbd2_journal_abort_handle(handle);
272 /* Deal with the reporting of failure conditions on a filesystem such as
273 * inconsistencies detected or read IO failures.
275 * On ext2, we can store the error state of the filesystem in the
276 * superblock. That is not possible on ext4, because we may have other
277 * write ordering constraints on the superblock which prevent us from
278 * writing it out straight away; and given that the journal is about to
279 * be aborted, we can't rely on the current, or future, transactions to
280 * write out the superblock safely.
282 * We'll just use the jbd2_journal_abort() error code to record an error in
283 * the journal instead. On recovery, the journal will compain about
284 * that error until we've noted it down and cleared it.
287 static void ext4_handle_error(struct super_block *sb)
289 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
291 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
294 if (sb->s_flags & MS_RDONLY)
297 if (!test_opt(sb, ERRORS_CONT)) {
298 journal_t *journal = EXT4_SB(sb)->s_journal;
300 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
302 jbd2_journal_abort(journal, -EIO);
304 if (test_opt(sb, ERRORS_RO)) {
305 printk(KERN_CRIT "Remounting filesystem read-only\n");
306 sb->s_flags |= MS_RDONLY;
308 ext4_commit_super(sb, es, 1);
309 if (test_opt(sb, ERRORS_PANIC))
310 panic("EXT4-fs (device %s): panic forced after error\n",
314 void ext4_error(struct super_block *sb, const char *function,
315 const char *fmt, ...)
320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 ext4_handle_error(sb);
328 static const char *ext4_decode_error(struct super_block *sb, int errno,
335 errstr = "IO failure";
338 errstr = "Out of memory";
341 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
342 errstr = "Journal has aborted";
344 errstr = "Readonly filesystem";
347 /* If the caller passed in an extra buffer for unknown
348 * errors, textualise them now. Else we just return
351 /* Check for truncated error codes... */
352 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
361 /* __ext4_std_error decodes expected errors from journaling functions
362 * automatically and invokes the appropriate error response. */
364 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
369 /* Special case: if the error is EROFS, and we're not already
370 * inside a transaction, then there's really no point in logging
372 if (errno == -EROFS && journal_current_handle() == NULL &&
373 (sb->s_flags & MS_RDONLY))
376 errstr = ext4_decode_error(sb, errno, nbuf);
377 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
378 sb->s_id, function, errstr);
380 ext4_handle_error(sb);
384 * ext4_abort is a much stronger failure handler than ext4_error. The
385 * abort function may be used to deal with unrecoverable failures such
386 * as journal IO errors or ENOMEM at a critical moment in log management.
388 * We unconditionally force the filesystem into an ABORT|READONLY state,
389 * unless the error response on the fs has been set to panic in which
390 * case we take the easy way out and panic immediately.
393 void ext4_abort(struct super_block *sb, const char *function,
394 const char *fmt, ...)
398 printk(KERN_CRIT "ext4_abort called.\n");
401 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
406 if (test_opt(sb, ERRORS_PANIC))
407 panic("EXT4-fs panic from previous error\n");
409 if (sb->s_flags & MS_RDONLY)
412 printk(KERN_CRIT "Remounting filesystem read-only\n");
413 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
414 sb->s_flags |= MS_RDONLY;
415 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
416 if (EXT4_SB(sb)->s_journal)
417 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
420 void ext4_warning(struct super_block *sb, const char *function,
421 const char *fmt, ...)
426 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
433 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
434 const char *function, const char *fmt, ...)
439 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
447 if (test_opt(sb, ERRORS_CONT)) {
448 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
449 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
450 ext4_commit_super(sb, es, 0);
453 ext4_unlock_group(sb, grp);
454 ext4_handle_error(sb);
456 * We only get here in the ERRORS_RO case; relocking the group
457 * may be dangerous, but nothing bad will happen since the
458 * filesystem will have already been marked read/only and the
459 * journal has been aborted. We return 1 as a hint to callers
460 * who might what to use the return value from
461 * ext4_grp_locked_error() to distinguish beween the
462 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
463 * aggressively from the ext4 function in question, with a
464 * more appropriate error code.
466 ext4_lock_group(sb, grp);
471 void ext4_update_dynamic_rev(struct super_block *sb)
473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
475 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
478 ext4_warning(sb, __func__,
479 "updating to rev %d because of new feature flag, "
480 "running e2fsck is recommended",
483 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
484 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
485 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
486 /* leave es->s_feature_*compat flags alone */
487 /* es->s_uuid will be set by e2fsck if empty */
490 * The rest of the superblock fields should be zero, and if not it
491 * means they are likely already in use, so leave them alone. We
492 * can leave it up to e2fsck to clean up any inconsistencies there.
497 * Open the external journal device
499 static struct block_device *ext4_blkdev_get(dev_t dev)
501 struct block_device *bdev;
502 char b[BDEVNAME_SIZE];
504 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
510 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
511 __bdevname(dev, b), PTR_ERR(bdev));
516 * Release the journal device
518 static int ext4_blkdev_put(struct block_device *bdev)
521 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
524 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
526 struct block_device *bdev;
529 bdev = sbi->journal_bdev;
531 ret = ext4_blkdev_put(bdev);
532 sbi->journal_bdev = NULL;
537 static inline struct inode *orphan_list_entry(struct list_head *l)
539 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
542 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
546 printk(KERN_ERR "sb orphan head is %d\n",
547 le32_to_cpu(sbi->s_es->s_last_orphan));
549 printk(KERN_ERR "sb_info orphan list:\n");
550 list_for_each(l, &sbi->s_orphan) {
551 struct inode *inode = orphan_list_entry(l);
553 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
554 inode->i_sb->s_id, inode->i_ino, inode,
555 inode->i_mode, inode->i_nlink,
560 static void ext4_put_super(struct super_block *sb)
562 struct ext4_sb_info *sbi = EXT4_SB(sb);
563 struct ext4_super_block *es = sbi->s_es;
567 ext4_ext_release(sb);
568 ext4_xattr_put_super(sb);
569 if (sbi->s_journal) {
570 err = jbd2_journal_destroy(sbi->s_journal);
571 sbi->s_journal = NULL;
573 ext4_abort(sb, __func__,
574 "Couldn't clean up the journal");
576 if (!(sb->s_flags & MS_RDONLY)) {
577 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
578 es->s_state = cpu_to_le16(sbi->s_mount_state);
579 ext4_commit_super(sb, es, 1);
582 remove_proc_entry(sb->s_id, ext4_proc_root);
584 kobject_del(&sbi->s_kobj);
586 for (i = 0; i < sbi->s_gdb_count; i++)
587 brelse(sbi->s_group_desc[i]);
588 kfree(sbi->s_group_desc);
589 kfree(sbi->s_flex_groups);
590 percpu_counter_destroy(&sbi->s_freeblocks_counter);
591 percpu_counter_destroy(&sbi->s_freeinodes_counter);
592 percpu_counter_destroy(&sbi->s_dirs_counter);
593 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
596 for (i = 0; i < MAXQUOTAS; i++)
597 kfree(sbi->s_qf_names[i]);
600 /* Debugging code just in case the in-memory inode orphan list
601 * isn't empty. The on-disk one can be non-empty if we've
602 * detected an error and taken the fs readonly, but the
603 * in-memory list had better be clean by this point. */
604 if (!list_empty(&sbi->s_orphan))
605 dump_orphan_list(sb, sbi);
606 J_ASSERT(list_empty(&sbi->s_orphan));
608 invalidate_bdev(sb->s_bdev);
609 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
611 * Invalidate the journal device's buffers. We don't want them
612 * floating about in memory - the physical journal device may
613 * hotswapped, and it breaks the `ro-after' testing code.
615 sync_blockdev(sbi->journal_bdev);
616 invalidate_bdev(sbi->journal_bdev);
617 ext4_blkdev_remove(sbi);
619 sb->s_fs_info = NULL;
621 * Now that we are completely done shutting down the
622 * superblock, we need to actually destroy the kobject.
626 kobject_put(&sbi->s_kobj);
627 wait_for_completion(&sbi->s_kobj_unregister);
630 kfree(sbi->s_blockgroup_lock);
635 static struct kmem_cache *ext4_inode_cachep;
638 * Called inside transaction, so use GFP_NOFS
640 static struct inode *ext4_alloc_inode(struct super_block *sb)
642 struct ext4_inode_info *ei;
644 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
647 #ifdef CONFIG_EXT4_FS_POSIX_ACL
648 ei->i_acl = EXT4_ACL_NOT_CACHED;
649 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
651 ei->vfs_inode.i_version = 1;
652 ei->vfs_inode.i_data.writeback_index = 0;
653 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
654 INIT_LIST_HEAD(&ei->i_prealloc_list);
655 spin_lock_init(&ei->i_prealloc_lock);
657 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
658 * therefore it can be null here. Don't check it, just initialize
661 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
662 ei->i_reserved_data_blocks = 0;
663 ei->i_reserved_meta_blocks = 0;
664 ei->i_allocated_meta_blocks = 0;
665 ei->i_delalloc_reserved_flag = 0;
666 spin_lock_init(&(ei->i_block_reservation_lock));
667 return &ei->vfs_inode;
670 static void ext4_destroy_inode(struct inode *inode)
672 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
673 printk("EXT4 Inode %p: orphan list check failed!\n",
675 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
676 EXT4_I(inode), sizeof(struct ext4_inode_info),
680 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
683 static void init_once(void *foo)
685 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
687 INIT_LIST_HEAD(&ei->i_orphan);
688 #ifdef CONFIG_EXT4_FS_XATTR
689 init_rwsem(&ei->xattr_sem);
691 init_rwsem(&ei->i_data_sem);
692 inode_init_once(&ei->vfs_inode);
695 static int init_inodecache(void)
697 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
698 sizeof(struct ext4_inode_info),
699 0, (SLAB_RECLAIM_ACCOUNT|
702 if (ext4_inode_cachep == NULL)
707 static void destroy_inodecache(void)
709 kmem_cache_destroy(ext4_inode_cachep);
712 static void ext4_clear_inode(struct inode *inode)
714 #ifdef CONFIG_EXT4_FS_POSIX_ACL
715 if (EXT4_I(inode)->i_acl &&
716 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
717 posix_acl_release(EXT4_I(inode)->i_acl);
718 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
720 if (EXT4_I(inode)->i_default_acl &&
721 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
722 posix_acl_release(EXT4_I(inode)->i_default_acl);
723 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
726 ext4_discard_preallocations(inode);
727 if (EXT4_JOURNAL(inode))
728 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
729 &EXT4_I(inode)->jinode);
732 static inline void ext4_show_quota_options(struct seq_file *seq,
733 struct super_block *sb)
735 #if defined(CONFIG_QUOTA)
736 struct ext4_sb_info *sbi = EXT4_SB(sb);
738 if (sbi->s_jquota_fmt)
739 seq_printf(seq, ",jqfmt=%s",
740 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
742 if (sbi->s_qf_names[USRQUOTA])
743 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
745 if (sbi->s_qf_names[GRPQUOTA])
746 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
748 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
749 seq_puts(seq, ",usrquota");
751 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
752 seq_puts(seq, ",grpquota");
758 * - it's set to a non-default value OR
759 * - if the per-sb default is different from the global default
761 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
764 unsigned long def_mount_opts;
765 struct super_block *sb = vfs->mnt_sb;
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
769 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
770 def_errors = le16_to_cpu(es->s_errors);
772 if (sbi->s_sb_block != 1)
773 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
774 if (test_opt(sb, MINIX_DF))
775 seq_puts(seq, ",minixdf");
776 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
777 seq_puts(seq, ",grpid");
778 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
779 seq_puts(seq, ",nogrpid");
780 if (sbi->s_resuid != EXT4_DEF_RESUID ||
781 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
782 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
784 if (sbi->s_resgid != EXT4_DEF_RESGID ||
785 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
786 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
788 if (test_opt(sb, ERRORS_RO)) {
789 if (def_errors == EXT4_ERRORS_PANIC ||
790 def_errors == EXT4_ERRORS_CONTINUE) {
791 seq_puts(seq, ",errors=remount-ro");
794 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
795 seq_puts(seq, ",errors=continue");
796 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
797 seq_puts(seq, ",errors=panic");
798 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
799 seq_puts(seq, ",nouid32");
800 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
801 seq_puts(seq, ",debug");
802 if (test_opt(sb, OLDALLOC))
803 seq_puts(seq, ",oldalloc");
804 #ifdef CONFIG_EXT4_FS_XATTR
805 if (test_opt(sb, XATTR_USER) &&
806 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
807 seq_puts(seq, ",user_xattr");
808 if (!test_opt(sb, XATTR_USER) &&
809 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
810 seq_puts(seq, ",nouser_xattr");
813 #ifdef CONFIG_EXT4_FS_POSIX_ACL
814 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
815 seq_puts(seq, ",acl");
816 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
817 seq_puts(seq, ",noacl");
819 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
820 seq_printf(seq, ",commit=%u",
821 (unsigned) (sbi->s_commit_interval / HZ));
823 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
824 seq_printf(seq, ",min_batch_time=%u",
825 (unsigned) sbi->s_min_batch_time);
827 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
828 seq_printf(seq, ",max_batch_time=%u",
829 (unsigned) sbi->s_min_batch_time);
833 * We're changing the default of barrier mount option, so
834 * let's always display its mount state so it's clear what its
837 seq_puts(seq, ",barrier=");
838 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
839 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
840 seq_puts(seq, ",journal_async_commit");
841 if (test_opt(sb, NOBH))
842 seq_puts(seq, ",nobh");
843 if (test_opt(sb, I_VERSION))
844 seq_puts(seq, ",i_version");
845 if (!test_opt(sb, DELALLOC))
846 seq_puts(seq, ",nodelalloc");
850 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
852 * journal mode get enabled in different ways
853 * So just print the value even if we didn't specify it
855 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
856 seq_puts(seq, ",data=journal");
857 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
858 seq_puts(seq, ",data=ordered");
859 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
860 seq_puts(seq, ",data=writeback");
862 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
863 seq_printf(seq, ",inode_readahead_blks=%u",
864 sbi->s_inode_readahead_blks);
866 if (test_opt(sb, DATA_ERR_ABORT))
867 seq_puts(seq, ",data_err=abort");
869 if (test_opt(sb, NO_AUTO_DA_ALLOC))
870 seq_puts(seq, ",auto_da_alloc=0");
872 ext4_show_quota_options(seq, sb);
877 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
878 u64 ino, u32 generation)
882 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
883 return ERR_PTR(-ESTALE);
884 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
885 return ERR_PTR(-ESTALE);
887 /* iget isn't really right if the inode is currently unallocated!!
889 * ext4_read_inode will return a bad_inode if the inode had been
890 * deleted, so we should be safe.
892 * Currently we don't know the generation for parent directory, so
893 * a generation of 0 means "accept any"
895 inode = ext4_iget(sb, ino);
897 return ERR_CAST(inode);
898 if (generation && inode->i_generation != generation) {
900 return ERR_PTR(-ESTALE);
906 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
907 int fh_len, int fh_type)
909 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
913 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
914 int fh_len, int fh_type)
916 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
921 * Try to release metadata pages (indirect blocks, directories) which are
922 * mapped via the block device. Since these pages could have journal heads
923 * which would prevent try_to_free_buffers() from freeing them, we must use
924 * jbd2 layer's try_to_free_buffers() function to release them.
926 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
928 journal_t *journal = EXT4_SB(sb)->s_journal;
930 WARN_ON(PageChecked(page));
931 if (!page_has_buffers(page))
934 return jbd2_journal_try_to_free_buffers(journal, page,
936 return try_to_free_buffers(page);
940 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
941 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
943 static int ext4_write_dquot(struct dquot *dquot);
944 static int ext4_acquire_dquot(struct dquot *dquot);
945 static int ext4_release_dquot(struct dquot *dquot);
946 static int ext4_mark_dquot_dirty(struct dquot *dquot);
947 static int ext4_write_info(struct super_block *sb, int type);
948 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
949 char *path, int remount);
950 static int ext4_quota_on_mount(struct super_block *sb, int type);
951 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
952 size_t len, loff_t off);
953 static ssize_t ext4_quota_write(struct super_block *sb, int type,
954 const char *data, size_t len, loff_t off);
956 static struct dquot_operations ext4_quota_operations = {
957 .initialize = dquot_initialize,
959 .alloc_space = dquot_alloc_space,
960 .reserve_space = dquot_reserve_space,
961 .claim_space = dquot_claim_space,
962 .release_rsv = dquot_release_reserved_space,
963 .get_reserved_space = ext4_get_reserved_space,
964 .alloc_inode = dquot_alloc_inode,
965 .free_space = dquot_free_space,
966 .free_inode = dquot_free_inode,
967 .transfer = dquot_transfer,
968 .write_dquot = ext4_write_dquot,
969 .acquire_dquot = ext4_acquire_dquot,
970 .release_dquot = ext4_release_dquot,
971 .mark_dirty = ext4_mark_dquot_dirty,
972 .write_info = ext4_write_info,
973 .alloc_dquot = dquot_alloc,
974 .destroy_dquot = dquot_destroy,
977 static struct quotactl_ops ext4_qctl_operations = {
978 .quota_on = ext4_quota_on,
979 .quota_off = vfs_quota_off,
980 .quota_sync = vfs_quota_sync,
981 .get_info = vfs_get_dqinfo,
982 .set_info = vfs_set_dqinfo,
983 .get_dqblk = vfs_get_dqblk,
984 .set_dqblk = vfs_set_dqblk
988 static const struct super_operations ext4_sops = {
989 .alloc_inode = ext4_alloc_inode,
990 .destroy_inode = ext4_destroy_inode,
991 .write_inode = ext4_write_inode,
992 .dirty_inode = ext4_dirty_inode,
993 .delete_inode = ext4_delete_inode,
994 .put_super = ext4_put_super,
995 .write_super = ext4_write_super,
996 .sync_fs = ext4_sync_fs,
997 .freeze_fs = ext4_freeze,
998 .unfreeze_fs = ext4_unfreeze,
999 .statfs = ext4_statfs,
1000 .remount_fs = ext4_remount,
1001 .clear_inode = ext4_clear_inode,
1002 .show_options = ext4_show_options,
1004 .quota_read = ext4_quota_read,
1005 .quota_write = ext4_quota_write,
1007 .bdev_try_to_free_page = bdev_try_to_free_page,
1010 static const struct export_operations ext4_export_ops = {
1011 .fh_to_dentry = ext4_fh_to_dentry,
1012 .fh_to_parent = ext4_fh_to_parent,
1013 .get_parent = ext4_get_parent,
1017 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1018 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1019 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1020 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1021 Opt_auto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1022 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1023 Opt_journal_update, Opt_journal_dev,
1024 Opt_journal_checksum, Opt_journal_async_commit,
1025 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1026 Opt_data_err_abort, Opt_data_err_ignore,
1027 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1028 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1029 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1030 Opt_grpquota, Opt_i_version,
1031 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1032 Opt_inode_readahead_blks, Opt_journal_ioprio
1035 static const match_table_t tokens = {
1036 {Opt_bsd_df, "bsddf"},
1037 {Opt_minix_df, "minixdf"},
1038 {Opt_grpid, "grpid"},
1039 {Opt_grpid, "bsdgroups"},
1040 {Opt_nogrpid, "nogrpid"},
1041 {Opt_nogrpid, "sysvgroups"},
1042 {Opt_resgid, "resgid=%u"},
1043 {Opt_resuid, "resuid=%u"},
1045 {Opt_err_cont, "errors=continue"},
1046 {Opt_err_panic, "errors=panic"},
1047 {Opt_err_ro, "errors=remount-ro"},
1048 {Opt_nouid32, "nouid32"},
1049 {Opt_debug, "debug"},
1050 {Opt_oldalloc, "oldalloc"},
1051 {Opt_orlov, "orlov"},
1052 {Opt_user_xattr, "user_xattr"},
1053 {Opt_nouser_xattr, "nouser_xattr"},
1055 {Opt_noacl, "noacl"},
1056 {Opt_noload, "noload"},
1059 {Opt_commit, "commit=%u"},
1060 {Opt_min_batch_time, "min_batch_time=%u"},
1061 {Opt_max_batch_time, "max_batch_time=%u"},
1062 {Opt_journal_update, "journal=update"},
1063 {Opt_journal_dev, "journal_dev=%u"},
1064 {Opt_journal_checksum, "journal_checksum"},
1065 {Opt_journal_async_commit, "journal_async_commit"},
1066 {Opt_abort, "abort"},
1067 {Opt_data_journal, "data=journal"},
1068 {Opt_data_ordered, "data=ordered"},
1069 {Opt_data_writeback, "data=writeback"},
1070 {Opt_data_err_abort, "data_err=abort"},
1071 {Opt_data_err_ignore, "data_err=ignore"},
1072 {Opt_offusrjquota, "usrjquota="},
1073 {Opt_usrjquota, "usrjquota=%s"},
1074 {Opt_offgrpjquota, "grpjquota="},
1075 {Opt_grpjquota, "grpjquota=%s"},
1076 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1077 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1078 {Opt_grpquota, "grpquota"},
1079 {Opt_noquota, "noquota"},
1080 {Opt_quota, "quota"},
1081 {Opt_usrquota, "usrquota"},
1082 {Opt_barrier, "barrier=%u"},
1083 {Opt_i_version, "i_version"},
1084 {Opt_stripe, "stripe=%u"},
1085 {Opt_resize, "resize"},
1086 {Opt_delalloc, "delalloc"},
1087 {Opt_nodelalloc, "nodelalloc"},
1088 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1089 {Opt_journal_ioprio, "journal_ioprio=%u"},
1090 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1094 static ext4_fsblk_t get_sb_block(void **data)
1096 ext4_fsblk_t sb_block;
1097 char *options = (char *) *data;
1099 if (!options || strncmp(options, "sb=", 3) != 0)
1100 return 1; /* Default location */
1102 /*todo: use simple_strtoll with >32bit ext4 */
1103 sb_block = simple_strtoul(options, &options, 0);
1104 if (*options && *options != ',') {
1105 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1109 if (*options == ',')
1111 *data = (void *) options;
1115 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1117 static int parse_options(char *options, struct super_block *sb,
1118 unsigned long *journal_devnum,
1119 unsigned int *journal_ioprio,
1120 ext4_fsblk_t *n_blocks_count, int is_remount)
1122 struct ext4_sb_info *sbi = EXT4_SB(sb);
1124 substring_t args[MAX_OPT_ARGS];
1135 while ((p = strsep(&options, ",")) != NULL) {
1140 token = match_token(p, tokens, args);
1143 clear_opt(sbi->s_mount_opt, MINIX_DF);
1146 set_opt(sbi->s_mount_opt, MINIX_DF);
1149 set_opt(sbi->s_mount_opt, GRPID);
1152 clear_opt(sbi->s_mount_opt, GRPID);
1155 if (match_int(&args[0], &option))
1157 sbi->s_resuid = option;
1160 if (match_int(&args[0], &option))
1162 sbi->s_resgid = option;
1165 /* handled by get_sb_block() instead of here */
1166 /* *sb_block = match_int(&args[0]); */
1169 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1170 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1171 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1174 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1175 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1176 set_opt(sbi->s_mount_opt, ERRORS_RO);
1179 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1180 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1181 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1184 set_opt(sbi->s_mount_opt, NO_UID32);
1187 set_opt(sbi->s_mount_opt, DEBUG);
1190 set_opt(sbi->s_mount_opt, OLDALLOC);
1193 clear_opt(sbi->s_mount_opt, OLDALLOC);
1195 #ifdef CONFIG_EXT4_FS_XATTR
1196 case Opt_user_xattr:
1197 set_opt(sbi->s_mount_opt, XATTR_USER);
1199 case Opt_nouser_xattr:
1200 clear_opt(sbi->s_mount_opt, XATTR_USER);
1203 case Opt_user_xattr:
1204 case Opt_nouser_xattr:
1205 printk(KERN_ERR "EXT4 (no)user_xattr options "
1209 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1211 set_opt(sbi->s_mount_opt, POSIX_ACL);
1214 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1219 printk(KERN_ERR "EXT4 (no)acl options "
1223 case Opt_journal_update:
1225 /* Eventually we will want to be able to create
1226 a journal file here. For now, only allow the
1227 user to specify an existing inode to be the
1230 printk(KERN_ERR "EXT4-fs: cannot specify "
1231 "journal on remount\n");
1234 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1236 case Opt_journal_dev:
1238 printk(KERN_ERR "EXT4-fs: cannot specify "
1239 "journal on remount\n");
1242 if (match_int(&args[0], &option))
1244 *journal_devnum = option;
1246 case Opt_journal_checksum:
1247 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1249 case Opt_journal_async_commit:
1250 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1251 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1254 set_opt(sbi->s_mount_opt, NOLOAD);
1257 if (match_int(&args[0], &option))
1262 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1263 sbi->s_commit_interval = HZ * option;
1265 case Opt_max_batch_time:
1266 if (match_int(&args[0], &option))
1271 option = EXT4_DEF_MAX_BATCH_TIME;
1272 sbi->s_max_batch_time = option;
1274 case Opt_min_batch_time:
1275 if (match_int(&args[0], &option))
1279 sbi->s_min_batch_time = option;
1281 case Opt_data_journal:
1282 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1284 case Opt_data_ordered:
1285 data_opt = EXT4_MOUNT_ORDERED_DATA;
1287 case Opt_data_writeback:
1288 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1291 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1294 "EXT4-fs: cannot change data "
1295 "mode on remount\n");
1299 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1300 sbi->s_mount_opt |= data_opt;
1303 case Opt_data_err_abort:
1304 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1306 case Opt_data_err_ignore:
1307 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1316 if (sb_any_quota_loaded(sb) &&
1317 !sbi->s_qf_names[qtype]) {
1319 "EXT4-fs: Cannot change journaled "
1320 "quota options when quota turned on.\n");
1323 qname = match_strdup(&args[0]);
1326 "EXT4-fs: not enough memory for "
1327 "storing quotafile name.\n");
1330 if (sbi->s_qf_names[qtype] &&
1331 strcmp(sbi->s_qf_names[qtype], qname)) {
1333 "EXT4-fs: %s quota file already "
1334 "specified.\n", QTYPE2NAME(qtype));
1338 sbi->s_qf_names[qtype] = qname;
1339 if (strchr(sbi->s_qf_names[qtype], '/')) {
1341 "EXT4-fs: quotafile must be on "
1342 "filesystem root.\n");
1343 kfree(sbi->s_qf_names[qtype]);
1344 sbi->s_qf_names[qtype] = NULL;
1347 set_opt(sbi->s_mount_opt, QUOTA);
1349 case Opt_offusrjquota:
1352 case Opt_offgrpjquota:
1355 if (sb_any_quota_loaded(sb) &&
1356 sbi->s_qf_names[qtype]) {
1357 printk(KERN_ERR "EXT4-fs: Cannot change "
1358 "journaled quota options when "
1359 "quota turned on.\n");
1363 * The space will be released later when all options
1364 * are confirmed to be correct
1366 sbi->s_qf_names[qtype] = NULL;
1368 case Opt_jqfmt_vfsold:
1369 qfmt = QFMT_VFS_OLD;
1371 case Opt_jqfmt_vfsv0:
1374 if (sb_any_quota_loaded(sb) &&
1375 sbi->s_jquota_fmt != qfmt) {
1376 printk(KERN_ERR "EXT4-fs: Cannot change "
1377 "journaled quota options when "
1378 "quota turned on.\n");
1381 sbi->s_jquota_fmt = qfmt;
1385 set_opt(sbi->s_mount_opt, QUOTA);
1386 set_opt(sbi->s_mount_opt, USRQUOTA);
1389 set_opt(sbi->s_mount_opt, QUOTA);
1390 set_opt(sbi->s_mount_opt, GRPQUOTA);
1393 if (sb_any_quota_loaded(sb)) {
1394 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1395 "options when quota turned on.\n");
1398 clear_opt(sbi->s_mount_opt, QUOTA);
1399 clear_opt(sbi->s_mount_opt, USRQUOTA);
1400 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1407 "EXT4-fs: quota options not supported.\n");
1411 case Opt_offusrjquota:
1412 case Opt_offgrpjquota:
1413 case Opt_jqfmt_vfsold:
1414 case Opt_jqfmt_vfsv0:
1416 "EXT4-fs: journaled quota options not "
1423 set_opt(sbi->s_mount_opt, ABORT);
1426 if (match_int(&args[0], &option))
1429 set_opt(sbi->s_mount_opt, BARRIER);
1431 clear_opt(sbi->s_mount_opt, BARRIER);
1437 printk("EXT4-fs: resize option only available "
1441 if (match_int(&args[0], &option) != 0)
1443 *n_blocks_count = option;
1446 set_opt(sbi->s_mount_opt, NOBH);
1449 clear_opt(sbi->s_mount_opt, NOBH);
1452 set_opt(sbi->s_mount_opt, I_VERSION);
1453 sb->s_flags |= MS_I_VERSION;
1455 case Opt_nodelalloc:
1456 clear_opt(sbi->s_mount_opt, DELALLOC);
1459 if (match_int(&args[0], &option))
1463 sbi->s_stripe = option;
1466 set_opt(sbi->s_mount_opt, DELALLOC);
1468 case Opt_inode_readahead_blks:
1469 if (match_int(&args[0], &option))
1471 if (option < 0 || option > (1 << 30))
1473 if (option & (option - 1)) {
1474 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1475 " must be a power of 2\n");
1478 sbi->s_inode_readahead_blks = option;
1480 case Opt_journal_ioprio:
1481 if (match_int(&args[0], &option))
1483 if (option < 0 || option > 7)
1485 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1488 case Opt_auto_da_alloc:
1489 if (match_int(&args[0], &option))
1492 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1494 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1498 "EXT4-fs: Unrecognized mount option \"%s\" "
1499 "or missing value\n", p);
1504 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1505 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1506 sbi->s_qf_names[USRQUOTA])
1507 clear_opt(sbi->s_mount_opt, USRQUOTA);
1509 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1510 sbi->s_qf_names[GRPQUOTA])
1511 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1513 if ((sbi->s_qf_names[USRQUOTA] &&
1514 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1515 (sbi->s_qf_names[GRPQUOTA] &&
1516 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1517 printk(KERN_ERR "EXT4-fs: old and new quota "
1518 "format mixing.\n");
1522 if (!sbi->s_jquota_fmt) {
1523 printk(KERN_ERR "EXT4-fs: journaled quota format "
1524 "not specified.\n");
1528 if (sbi->s_jquota_fmt) {
1529 printk(KERN_ERR "EXT4-fs: journaled quota format "
1530 "specified with no journaling "
1539 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1542 struct ext4_sb_info *sbi = EXT4_SB(sb);
1545 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1546 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1547 "forcing read-only mode\n");
1552 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1553 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1554 "running e2fsck is recommended\n");
1555 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1557 "EXT4-fs warning: mounting fs with errors, "
1558 "running e2fsck is recommended\n");
1559 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1560 le16_to_cpu(es->s_mnt_count) >=
1561 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1563 "EXT4-fs warning: maximal mount count reached, "
1564 "running e2fsck is recommended\n");
1565 else if (le32_to_cpu(es->s_checkinterval) &&
1566 (le32_to_cpu(es->s_lastcheck) +
1567 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1569 "EXT4-fs warning: checktime reached, "
1570 "running e2fsck is recommended\n");
1571 if (!sbi->s_journal)
1572 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1573 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1574 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1575 le16_add_cpu(&es->s_mnt_count, 1);
1576 es->s_mtime = cpu_to_le32(get_seconds());
1577 ext4_update_dynamic_rev(sb);
1579 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1581 ext4_commit_super(sb, es, 1);
1582 if (test_opt(sb, DEBUG))
1583 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1584 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1586 sbi->s_groups_count,
1587 EXT4_BLOCKS_PER_GROUP(sb),
1588 EXT4_INODES_PER_GROUP(sb),
1591 if (EXT4_SB(sb)->s_journal) {
1592 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1593 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1594 "external", EXT4_SB(sb)->s_journal->j_devname);
1596 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1601 static int ext4_fill_flex_info(struct super_block *sb)
1603 struct ext4_sb_info *sbi = EXT4_SB(sb);
1604 struct ext4_group_desc *gdp = NULL;
1605 struct buffer_head *bh;
1606 ext4_group_t flex_group_count;
1607 ext4_group_t flex_group;
1608 int groups_per_flex = 0;
1611 if (!sbi->s_es->s_log_groups_per_flex) {
1612 sbi->s_log_groups_per_flex = 0;
1616 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1617 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1619 /* We allocate both existing and potentially added groups */
1620 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1621 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1622 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1623 sbi->s_flex_groups = kzalloc(flex_group_count *
1624 sizeof(struct flex_groups), GFP_KERNEL);
1625 if (sbi->s_flex_groups == NULL) {
1626 printk(KERN_ERR "EXT4-fs: not enough memory for "
1627 "%u flex groups\n", flex_group_count);
1631 for (i = 0; i < sbi->s_groups_count; i++) {
1632 gdp = ext4_get_group_desc(sb, i, &bh);
1634 flex_group = ext4_flex_group(sbi, i);
1635 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1636 ext4_free_inodes_count(sb, gdp));
1637 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1638 ext4_free_blks_count(sb, gdp));
1639 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1640 ext4_used_dirs_count(sb, gdp));
1648 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1649 struct ext4_group_desc *gdp)
1653 if (sbi->s_es->s_feature_ro_compat &
1654 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1655 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1656 __le32 le_group = cpu_to_le32(block_group);
1658 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1659 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1660 crc = crc16(crc, (__u8 *)gdp, offset);
1661 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1662 /* for checksum of struct ext4_group_desc do the rest...*/
1663 if ((sbi->s_es->s_feature_incompat &
1664 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1665 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1666 crc = crc16(crc, (__u8 *)gdp + offset,
1667 le16_to_cpu(sbi->s_es->s_desc_size) -
1671 return cpu_to_le16(crc);
1674 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1675 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 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1685 /* Called at mount-time, super-block is locked */
1686 static int ext4_check_descriptors(struct super_block *sb)
1688 struct ext4_sb_info *sbi = EXT4_SB(sb);
1689 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1690 ext4_fsblk_t last_block;
1691 ext4_fsblk_t block_bitmap;
1692 ext4_fsblk_t inode_bitmap;
1693 ext4_fsblk_t inode_table;
1694 int flexbg_flag = 0;
1697 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1700 ext4_debug("Checking group descriptors");
1702 for (i = 0; i < sbi->s_groups_count; i++) {
1703 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1705 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1706 last_block = ext4_blocks_count(sbi->s_es) - 1;
1708 last_block = first_block +
1709 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1711 block_bitmap = ext4_block_bitmap(sb, gdp);
1712 if (block_bitmap < first_block || block_bitmap > last_block) {
1713 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1714 "Block bitmap for group %u not in group "
1715 "(block %llu)!\n", i, block_bitmap);
1718 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1719 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1720 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1721 "Inode bitmap for group %u not in group "
1722 "(block %llu)!\n", i, inode_bitmap);
1725 inode_table = ext4_inode_table(sb, gdp);
1726 if (inode_table < first_block ||
1727 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1728 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1729 "Inode table for group %u not in group "
1730 "(block %llu)!\n", i, inode_table);
1733 spin_lock(sb_bgl_lock(sbi, i));
1734 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1735 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1736 "Checksum for group %u failed (%u!=%u)\n",
1737 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1738 gdp)), le16_to_cpu(gdp->bg_checksum));
1739 if (!(sb->s_flags & MS_RDONLY)) {
1740 spin_unlock(sb_bgl_lock(sbi, i));
1744 spin_unlock(sb_bgl_lock(sbi, i));
1746 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1749 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1750 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1754 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1755 * the superblock) which were deleted from all directories, but held open by
1756 * a process at the time of a crash. We walk the list and try to delete these
1757 * inodes at recovery time (only with a read-write filesystem).
1759 * In order to keep the orphan inode chain consistent during traversal (in
1760 * case of crash during recovery), we link each inode into the superblock
1761 * orphan list_head and handle it the same way as an inode deletion during
1762 * normal operation (which journals the operations for us).
1764 * We only do an iget() and an iput() on each inode, which is very safe if we
1765 * accidentally point at an in-use or already deleted inode. The worst that
1766 * can happen in this case is that we get a "bit already cleared" message from
1767 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1768 * e2fsck was run on this filesystem, and it must have already done the orphan
1769 * inode cleanup for us, so we can safely abort without any further action.
1771 static void ext4_orphan_cleanup(struct super_block *sb,
1772 struct ext4_super_block *es)
1774 unsigned int s_flags = sb->s_flags;
1775 int nr_orphans = 0, nr_truncates = 0;
1779 if (!es->s_last_orphan) {
1780 jbd_debug(4, "no orphan inodes to clean up\n");
1784 if (bdev_read_only(sb->s_bdev)) {
1785 printk(KERN_ERR "EXT4-fs: write access "
1786 "unavailable, skipping orphan cleanup.\n");
1790 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1791 if (es->s_last_orphan)
1792 jbd_debug(1, "Errors on filesystem, "
1793 "clearing orphan list.\n");
1794 es->s_last_orphan = 0;
1795 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1799 if (s_flags & MS_RDONLY) {
1800 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1802 sb->s_flags &= ~MS_RDONLY;
1805 /* Needed for iput() to work correctly and not trash data */
1806 sb->s_flags |= MS_ACTIVE;
1807 /* Turn on quotas so that they are updated correctly */
1808 for (i = 0; i < MAXQUOTAS; i++) {
1809 if (EXT4_SB(sb)->s_qf_names[i]) {
1810 int ret = ext4_quota_on_mount(sb, i);
1813 "EXT4-fs: Cannot turn on journaled "
1814 "quota: error %d\n", ret);
1819 while (es->s_last_orphan) {
1820 struct inode *inode;
1822 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1823 if (IS_ERR(inode)) {
1824 es->s_last_orphan = 0;
1828 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1830 if (inode->i_nlink) {
1832 "%s: truncating inode %lu to %lld bytes\n",
1833 __func__, inode->i_ino, inode->i_size);
1834 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1835 inode->i_ino, inode->i_size);
1836 ext4_truncate(inode);
1840 "%s: deleting unreferenced inode %lu\n",
1841 __func__, inode->i_ino);
1842 jbd_debug(2, "deleting unreferenced inode %lu\n",
1846 iput(inode); /* The delete magic happens here! */
1849 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1852 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1853 sb->s_id, PLURAL(nr_orphans));
1855 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1856 sb->s_id, PLURAL(nr_truncates));
1858 /* Turn quotas off */
1859 for (i = 0; i < MAXQUOTAS; i++) {
1860 if (sb_dqopt(sb)->files[i])
1861 vfs_quota_off(sb, i, 0);
1864 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1867 * Maximal extent format file size.
1868 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1869 * extent format containers, within a sector_t, and within i_blocks
1870 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1871 * so that won't be a limiting factor.
1873 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1875 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1878 loff_t upper_limit = MAX_LFS_FILESIZE;
1880 /* small i_blocks in vfs inode? */
1881 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1883 * CONFIG_LBD is not enabled implies the inode
1884 * i_block represent total blocks in 512 bytes
1885 * 32 == size of vfs inode i_blocks * 8
1887 upper_limit = (1LL << 32) - 1;
1889 /* total blocks in file system block size */
1890 upper_limit >>= (blkbits - 9);
1891 upper_limit <<= blkbits;
1894 /* 32-bit extent-start container, ee_block */
1899 /* Sanity check against vm- & vfs- imposed limits */
1900 if (res > upper_limit)
1907 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1908 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1909 * We need to be 1 filesystem block less than the 2^48 sector limit.
1911 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1913 loff_t res = EXT4_NDIR_BLOCKS;
1916 /* This is calculated to be the largest file size for a
1917 * dense, bitmapped file such that the total number of
1918 * sectors in the file, including data and all indirect blocks,
1919 * does not exceed 2^48 -1
1920 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1921 * total number of 512 bytes blocks of the file
1924 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1926 * !has_huge_files or CONFIG_LBD is not enabled
1927 * implies the inode i_block represent total blocks in
1928 * 512 bytes 32 == size of vfs inode i_blocks * 8
1930 upper_limit = (1LL << 32) - 1;
1932 /* total blocks in file system block size */
1933 upper_limit >>= (bits - 9);
1937 * We use 48 bit ext4_inode i_blocks
1938 * With EXT4_HUGE_FILE_FL set the i_blocks
1939 * represent total number of blocks in
1940 * file system block size
1942 upper_limit = (1LL << 48) - 1;
1946 /* indirect blocks */
1948 /* double indirect blocks */
1949 meta_blocks += 1 + (1LL << (bits-2));
1950 /* tripple indirect blocks */
1951 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1953 upper_limit -= meta_blocks;
1954 upper_limit <<= bits;
1956 res += 1LL << (bits-2);
1957 res += 1LL << (2*(bits-2));
1958 res += 1LL << (3*(bits-2));
1960 if (res > upper_limit)
1963 if (res > MAX_LFS_FILESIZE)
1964 res = MAX_LFS_FILESIZE;
1969 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1970 ext4_fsblk_t logical_sb_block, int nr)
1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973 ext4_group_t bg, first_meta_bg;
1976 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1978 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1980 return logical_sb_block + nr + 1;
1981 bg = sbi->s_desc_per_block * nr;
1982 if (ext4_bg_has_super(sb, bg))
1984 return (has_super + ext4_group_first_block_no(sb, bg));
1988 * ext4_get_stripe_size: Get the stripe size.
1989 * @sbi: In memory super block info
1991 * If we have specified it via mount option, then
1992 * use the mount option value. If the value specified at mount time is
1993 * greater than the blocks per group use the super block value.
1994 * If the super block value is greater than blocks per group return 0.
1995 * Allocator needs it be less than blocks per group.
1998 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2000 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2001 unsigned long stripe_width =
2002 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2004 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2005 return sbi->s_stripe;
2007 if (stripe_width <= sbi->s_blocks_per_group)
2008 return stripe_width;
2010 if (stride <= sbi->s_blocks_per_group)
2019 struct attribute attr;
2020 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2021 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2022 const char *, size_t);
2026 static int parse_strtoul(const char *buf,
2027 unsigned long max, unsigned long *value)
2031 while (*buf && isspace(*buf))
2033 *value = simple_strtoul(buf, &endp, 0);
2034 while (*endp && isspace(*endp))
2036 if (*endp || *value > max)
2042 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2043 struct ext4_sb_info *sbi,
2046 return snprintf(buf, PAGE_SIZE, "%llu\n",
2047 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2050 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2051 struct ext4_sb_info *sbi, char *buf)
2053 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2055 return snprintf(buf, PAGE_SIZE, "%lu\n",
2056 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2057 sbi->s_sectors_written_start) >> 1);
2060 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2061 struct ext4_sb_info *sbi, char *buf)
2063 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2065 return snprintf(buf, PAGE_SIZE, "%llu\n",
2066 sbi->s_kbytes_written +
2067 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2068 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2071 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2072 struct ext4_sb_info *sbi,
2073 const char *buf, size_t count)
2077 if (parse_strtoul(buf, 0x40000000, &t))
2080 /* inode_readahead_blks must be a power of 2 */
2084 sbi->s_inode_readahead_blks = t;
2088 static ssize_t sbi_ui_show(struct ext4_attr *a,
2089 struct ext4_sb_info *sbi, char *buf)
2091 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2093 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2096 static ssize_t sbi_ui_store(struct ext4_attr *a,
2097 struct ext4_sb_info *sbi,
2098 const char *buf, size_t count)
2100 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2103 if (parse_strtoul(buf, 0xffffffff, &t))
2109 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2110 static struct ext4_attr ext4_attr_##_name = { \
2111 .attr = {.name = __stringify(_name), .mode = _mode }, \
2114 .offset = offsetof(struct ext4_sb_info, _elname), \
2116 #define EXT4_ATTR(name, mode, show, store) \
2117 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2119 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2120 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2121 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2122 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2123 #define ATTR_LIST(name) &ext4_attr_##name.attr
2125 EXT4_RO_ATTR(delayed_allocation_blocks);
2126 EXT4_RO_ATTR(session_write_kbytes);
2127 EXT4_RO_ATTR(lifetime_write_kbytes);
2128 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2129 inode_readahead_blks_store, s_inode_readahead_blks);
2130 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2131 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2132 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2133 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2134 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2135 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2137 static struct attribute *ext4_attrs[] = {
2138 ATTR_LIST(delayed_allocation_blocks),
2139 ATTR_LIST(session_write_kbytes),
2140 ATTR_LIST(lifetime_write_kbytes),
2141 ATTR_LIST(inode_readahead_blks),
2142 ATTR_LIST(mb_stats),
2143 ATTR_LIST(mb_max_to_scan),
2144 ATTR_LIST(mb_min_to_scan),
2145 ATTR_LIST(mb_order2_req),
2146 ATTR_LIST(mb_stream_req),
2147 ATTR_LIST(mb_group_prealloc),
2151 static ssize_t ext4_attr_show(struct kobject *kobj,
2152 struct attribute *attr, char *buf)
2154 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2156 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2158 return a->show ? a->show(a, sbi, buf) : 0;
2161 static ssize_t ext4_attr_store(struct kobject *kobj,
2162 struct attribute *attr,
2163 const char *buf, size_t len)
2165 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2167 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2169 return a->store ? a->store(a, sbi, buf, len) : 0;
2172 static void ext4_sb_release(struct kobject *kobj)
2174 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2176 complete(&sbi->s_kobj_unregister);
2180 static struct sysfs_ops ext4_attr_ops = {
2181 .show = ext4_attr_show,
2182 .store = ext4_attr_store,
2185 static struct kobj_type ext4_ktype = {
2186 .default_attrs = ext4_attrs,
2187 .sysfs_ops = &ext4_attr_ops,
2188 .release = ext4_sb_release,
2191 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2192 __releases(kernel_lock)
2193 __acquires(kernel_lock)
2196 struct buffer_head *bh;
2197 struct ext4_super_block *es = NULL;
2198 struct ext4_sb_info *sbi;
2200 ext4_fsblk_t sb_block = get_sb_block(&data);
2201 ext4_fsblk_t logical_sb_block;
2202 unsigned long offset = 0;
2203 unsigned long journal_devnum = 0;
2204 unsigned long def_mount_opts;
2210 unsigned int db_count;
2212 int needs_recovery, has_huge_files;
2216 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2218 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2222 sbi->s_blockgroup_lock =
2223 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2224 if (!sbi->s_blockgroup_lock) {
2228 sb->s_fs_info = sbi;
2229 sbi->s_mount_opt = 0;
2230 sbi->s_resuid = EXT4_DEF_RESUID;
2231 sbi->s_resgid = EXT4_DEF_RESGID;
2232 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2233 sbi->s_sb_block = sb_block;
2234 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2239 /* Cleanup superblock name */
2240 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2243 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2245 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2250 * The ext4 superblock will not be buffer aligned for other than 1kB
2251 * block sizes. We need to calculate the offset from buffer start.
2253 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2254 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2255 offset = do_div(logical_sb_block, blocksize);
2257 logical_sb_block = sb_block;
2260 if (!(bh = sb_bread(sb, logical_sb_block))) {
2261 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2265 * Note: s_es must be initialized as soon as possible because
2266 * some ext4 macro-instructions depend on its value
2268 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2270 sb->s_magic = le16_to_cpu(es->s_magic);
2271 if (sb->s_magic != EXT4_SUPER_MAGIC)
2273 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2275 /* Set defaults before we parse the mount options */
2276 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2277 if (def_mount_opts & EXT4_DEFM_DEBUG)
2278 set_opt(sbi->s_mount_opt, DEBUG);
2279 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2280 set_opt(sbi->s_mount_opt, GRPID);
2281 if (def_mount_opts & EXT4_DEFM_UID16)
2282 set_opt(sbi->s_mount_opt, NO_UID32);
2283 #ifdef CONFIG_EXT4_FS_XATTR
2284 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2285 set_opt(sbi->s_mount_opt, XATTR_USER);
2287 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2288 if (def_mount_opts & EXT4_DEFM_ACL)
2289 set_opt(sbi->s_mount_opt, POSIX_ACL);
2291 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2292 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2293 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2294 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2295 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2296 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2298 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2299 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2300 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2301 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2303 set_opt(sbi->s_mount_opt, ERRORS_RO);
2305 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2306 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2307 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2308 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2309 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2311 set_opt(sbi->s_mount_opt, BARRIER);
2314 * enable delayed allocation by default
2315 * Use -o nodelalloc to turn it off
2317 set_opt(sbi->s_mount_opt, DELALLOC);
2320 if (!parse_options((char *) data, sb, &journal_devnum,
2321 &journal_ioprio, NULL, 0))
2324 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2325 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2327 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2328 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2329 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2330 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2332 "EXT4-fs warning: feature flags set on rev 0 fs, "
2333 "running e2fsck is recommended\n");
2336 * Check feature flags regardless of the revision level, since we
2337 * previously didn't change the revision level when setting the flags,
2338 * so there is a chance incompat flags are set on a rev 0 filesystem.
2340 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2342 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2343 "unsupported optional features (%x).\n", sb->s_id,
2344 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2345 ~EXT4_FEATURE_INCOMPAT_SUPP));
2348 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2349 if (!(sb->s_flags & MS_RDONLY) && features) {
2350 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2351 "unsupported optional features (%x).\n", sb->s_id,
2352 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2353 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2356 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2357 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2358 if (has_huge_files) {
2360 * Large file size enabled file system can only be
2361 * mount if kernel is build with CONFIG_LBD
2363 if (sizeof(root->i_blocks) < sizeof(u64) &&
2364 !(sb->s_flags & MS_RDONLY)) {
2365 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2366 "files cannot be mounted read-write "
2367 "without CONFIG_LBD.\n", sb->s_id);
2371 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2373 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2374 blocksize > EXT4_MAX_BLOCK_SIZE) {
2376 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2377 blocksize, sb->s_id);
2381 if (sb->s_blocksize != blocksize) {
2383 /* Validate the filesystem blocksize */
2384 if (!sb_set_blocksize(sb, blocksize)) {
2385 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2391 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2392 offset = do_div(logical_sb_block, blocksize);
2393 bh = sb_bread(sb, logical_sb_block);
2396 "EXT4-fs: Can't read superblock on 2nd try.\n");
2399 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2401 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2403 "EXT4-fs: Magic mismatch, very weird !\n");
2408 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2410 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2412 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2413 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2414 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2416 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2417 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2418 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2419 (!is_power_of_2(sbi->s_inode_size)) ||
2420 (sbi->s_inode_size > blocksize)) {
2422 "EXT4-fs: unsupported inode size: %d\n",
2426 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2427 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2429 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2430 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2431 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2432 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2433 !is_power_of_2(sbi->s_desc_size)) {
2435 "EXT4-fs: unsupported descriptor size %lu\n",
2440 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2441 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2442 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2443 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2445 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2446 if (sbi->s_inodes_per_block == 0)
2448 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2449 sbi->s_inodes_per_block;
2450 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2452 sbi->s_mount_state = le16_to_cpu(es->s_state);
2453 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2454 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2455 for (i = 0; i < 4; i++)
2456 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2457 sbi->s_def_hash_version = es->s_def_hash_version;
2458 i = le32_to_cpu(es->s_flags);
2459 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2460 sbi->s_hash_unsigned = 3;
2461 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2462 #ifdef __CHAR_UNSIGNED__
2463 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2464 sbi->s_hash_unsigned = 3;
2466 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2471 if (sbi->s_blocks_per_group > blocksize * 8) {
2473 "EXT4-fs: #blocks per group too big: %lu\n",
2474 sbi->s_blocks_per_group);
2477 if (sbi->s_inodes_per_group > blocksize * 8) {
2479 "EXT4-fs: #inodes per group too big: %lu\n",
2480 sbi->s_inodes_per_group);
2484 if (ext4_blocks_count(es) >
2485 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2486 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2487 " too large to mount safely\n", sb->s_id);
2488 if (sizeof(sector_t) < 8)
2489 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2494 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2498 * It makes no sense for the first data block to be beyond the end
2499 * of the filesystem.
2501 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2502 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2503 "block %u is beyond end of filesystem (%llu)\n",
2504 le32_to_cpu(es->s_first_data_block),
2505 ext4_blocks_count(es));
2508 blocks_count = (ext4_blocks_count(es) -
2509 le32_to_cpu(es->s_first_data_block) +
2510 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2511 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2512 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2513 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2514 "(block count %llu, first data block %u, "
2515 "blocks per group %lu)\n", sbi->s_groups_count,
2516 ext4_blocks_count(es),
2517 le32_to_cpu(es->s_first_data_block),
2518 EXT4_BLOCKS_PER_GROUP(sb));
2521 sbi->s_groups_count = blocks_count;
2522 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2523 EXT4_DESC_PER_BLOCK(sb);
2524 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2526 if (sbi->s_group_desc == NULL) {
2527 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2531 #ifdef CONFIG_PROC_FS
2533 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2536 bgl_lock_init(sbi->s_blockgroup_lock);
2538 for (i = 0; i < db_count; i++) {
2539 block = descriptor_loc(sb, logical_sb_block, i);
2540 sbi->s_group_desc[i] = sb_bread(sb, block);
2541 if (!sbi->s_group_desc[i]) {
2542 printk(KERN_ERR "EXT4-fs: "
2543 "can't read group descriptor %d\n", i);
2548 if (!ext4_check_descriptors(sb)) {
2549 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2552 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2553 if (!ext4_fill_flex_info(sb)) {
2555 "EXT4-fs: unable to initialize "
2556 "flex_bg meta info!\n");
2560 sbi->s_gdb_count = db_count;
2561 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2562 spin_lock_init(&sbi->s_next_gen_lock);
2564 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2565 ext4_count_free_blocks(sb));
2567 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2568 ext4_count_free_inodes(sb));
2571 err = percpu_counter_init(&sbi->s_dirs_counter,
2572 ext4_count_dirs(sb));
2575 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2578 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2582 sbi->s_stripe = ext4_get_stripe_size(sbi);
2585 * set up enough so that it can read an inode
2587 sb->s_op = &ext4_sops;
2588 sb->s_export_op = &ext4_export_ops;
2589 sb->s_xattr = ext4_xattr_handlers;
2591 sb->s_qcop = &ext4_qctl_operations;
2592 sb->dq_op = &ext4_quota_operations;
2594 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2598 needs_recovery = (es->s_last_orphan != 0 ||
2599 EXT4_HAS_INCOMPAT_FEATURE(sb,
2600 EXT4_FEATURE_INCOMPAT_RECOVER));
2603 * The first inode we look at is the journal inode. Don't try
2604 * root first: it may be modified in the journal!
2606 if (!test_opt(sb, NOLOAD) &&
2607 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2608 if (ext4_load_journal(sb, es, journal_devnum))
2610 if (!(sb->s_flags & MS_RDONLY) &&
2611 EXT4_SB(sb)->s_journal->j_failed_commit) {
2612 printk(KERN_CRIT "EXT4-fs error (device %s): "
2613 "ext4_fill_super: Journal transaction "
2614 "%u is corrupt\n", sb->s_id,
2615 EXT4_SB(sb)->s_journal->j_failed_commit);
2616 if (test_opt(sb, ERRORS_RO)) {
2618 "Mounting filesystem read-only\n");
2619 sb->s_flags |= MS_RDONLY;
2620 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2621 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2623 if (test_opt(sb, ERRORS_PANIC)) {
2624 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2625 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2626 ext4_commit_super(sb, es, 1);
2630 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2631 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2632 printk(KERN_ERR "EXT4-fs: required journal recovery "
2633 "suppressed and not mounted read-only\n");
2636 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2637 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2638 sbi->s_journal = NULL;
2643 if (ext4_blocks_count(es) > 0xffffffffULL &&
2644 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2645 JBD2_FEATURE_INCOMPAT_64BIT)) {
2646 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2650 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2651 jbd2_journal_set_features(sbi->s_journal,
2652 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2653 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2654 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2655 jbd2_journal_set_features(sbi->s_journal,
2656 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2657 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2658 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2660 jbd2_journal_clear_features(sbi->s_journal,
2661 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2662 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2665 /* We have now updated the journal if required, so we can
2666 * validate the data journaling mode. */
2667 switch (test_opt(sb, DATA_FLAGS)) {
2669 /* No mode set, assume a default based on the journal
2670 * capabilities: ORDERED_DATA if the journal can
2671 * cope, else JOURNAL_DATA
2673 if (jbd2_journal_check_available_features
2674 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2675 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2677 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2680 case EXT4_MOUNT_ORDERED_DATA:
2681 case EXT4_MOUNT_WRITEBACK_DATA:
2682 if (!jbd2_journal_check_available_features
2683 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2684 printk(KERN_ERR "EXT4-fs: Journal does not support "
2685 "requested data journaling mode\n");
2691 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2695 if (test_opt(sb, NOBH)) {
2696 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2697 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2698 "its supported only with writeback mode\n");
2699 clear_opt(sbi->s_mount_opt, NOBH);
2703 * The jbd2_journal_load will have done any necessary log recovery,
2704 * so we can safely mount the rest of the filesystem now.
2707 root = ext4_iget(sb, EXT4_ROOT_INO);
2709 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2710 ret = PTR_ERR(root);
2713 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2715 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2718 sb->s_root = d_alloc_root(root);
2720 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2726 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2728 /* determine the minimum size of new large inodes, if present */
2729 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2730 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2731 EXT4_GOOD_OLD_INODE_SIZE;
2732 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2733 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2734 if (sbi->s_want_extra_isize <
2735 le16_to_cpu(es->s_want_extra_isize))
2736 sbi->s_want_extra_isize =
2737 le16_to_cpu(es->s_want_extra_isize);
2738 if (sbi->s_want_extra_isize <
2739 le16_to_cpu(es->s_min_extra_isize))
2740 sbi->s_want_extra_isize =
2741 le16_to_cpu(es->s_min_extra_isize);
2744 /* Check if enough inode space is available */
2745 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2746 sbi->s_inode_size) {
2747 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2748 EXT4_GOOD_OLD_INODE_SIZE;
2749 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2753 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2754 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2755 "requested data journaling mode\n");
2756 clear_opt(sbi->s_mount_opt, DELALLOC);
2757 } else if (test_opt(sb, DELALLOC))
2758 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2761 err = ext4_mb_init(sb, needs_recovery);
2763 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2768 sbi->s_kobj.kset = ext4_kset;
2769 init_completion(&sbi->s_kobj_unregister);
2770 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2773 ext4_mb_release(sb);
2774 ext4_ext_release(sb);
2779 * akpm: core read_super() calls in here with the superblock locked.
2780 * That deadlocks, because orphan cleanup needs to lock the superblock
2781 * in numerous places. Here we just pop the lock - it's relatively
2782 * harmless, because we are now ready to accept write_super() requests,
2783 * and aviro says that's the only reason for hanging onto the
2786 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2787 ext4_orphan_cleanup(sb, es);
2788 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2789 if (needs_recovery) {
2790 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2791 ext4_mark_recovery_complete(sb, es);
2793 if (EXT4_SB(sb)->s_journal) {
2794 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2795 descr = " journalled data mode";
2796 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2797 descr = " ordered data mode";
2799 descr = " writeback data mode";
2801 descr = "out journal";
2803 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2811 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2816 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2817 if (sbi->s_journal) {
2818 jbd2_journal_destroy(sbi->s_journal);
2819 sbi->s_journal = NULL;
2822 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2823 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2824 percpu_counter_destroy(&sbi->s_dirs_counter);
2825 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2827 for (i = 0; i < db_count; i++)
2828 brelse(sbi->s_group_desc[i]);
2829 kfree(sbi->s_group_desc);
2832 remove_proc_entry(sb->s_id, ext4_proc_root);
2835 for (i = 0; i < MAXQUOTAS; i++)
2836 kfree(sbi->s_qf_names[i]);
2838 ext4_blkdev_remove(sbi);
2841 sb->s_fs_info = NULL;
2848 * Setup any per-fs journal parameters now. We'll do this both on
2849 * initial mount, once the journal has been initialised but before we've
2850 * done any recovery; and again on any subsequent remount.
2852 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2854 struct ext4_sb_info *sbi = EXT4_SB(sb);
2856 journal->j_commit_interval = sbi->s_commit_interval;
2857 journal->j_min_batch_time = sbi->s_min_batch_time;
2858 journal->j_max_batch_time = sbi->s_max_batch_time;
2860 spin_lock(&journal->j_state_lock);
2861 if (test_opt(sb, BARRIER))
2862 journal->j_flags |= JBD2_BARRIER;
2864 journal->j_flags &= ~JBD2_BARRIER;
2865 if (test_opt(sb, DATA_ERR_ABORT))
2866 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2868 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2869 spin_unlock(&journal->j_state_lock);
2872 static journal_t *ext4_get_journal(struct super_block *sb,
2873 unsigned int journal_inum)
2875 struct inode *journal_inode;
2878 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2880 /* First, test for the existence of a valid inode on disk. Bad
2881 * things happen if we iget() an unused inode, as the subsequent
2882 * iput() will try to delete it. */
2884 journal_inode = ext4_iget(sb, journal_inum);
2885 if (IS_ERR(journal_inode)) {
2886 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2889 if (!journal_inode->i_nlink) {
2890 make_bad_inode(journal_inode);
2891 iput(journal_inode);
2892 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2896 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2897 journal_inode, journal_inode->i_size);
2898 if (!S_ISREG(journal_inode->i_mode)) {
2899 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2900 iput(journal_inode);
2904 journal = jbd2_journal_init_inode(journal_inode);
2906 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2907 iput(journal_inode);
2910 journal->j_private = sb;
2911 ext4_init_journal_params(sb, journal);
2915 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2918 struct buffer_head *bh;
2922 int hblock, blocksize;
2923 ext4_fsblk_t sb_block;
2924 unsigned long offset;
2925 struct ext4_super_block *es;
2926 struct block_device *bdev;
2928 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2930 bdev = ext4_blkdev_get(j_dev);
2934 if (bd_claim(bdev, sb)) {
2936 "EXT4-fs: failed to claim external journal device.\n");
2937 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2941 blocksize = sb->s_blocksize;
2942 hblock = bdev_hardsect_size(bdev);
2943 if (blocksize < hblock) {
2945 "EXT4-fs: blocksize too small for journal device.\n");
2949 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2950 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2951 set_blocksize(bdev, blocksize);
2952 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2953 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2954 "external journal\n");
2958 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2959 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2960 !(le32_to_cpu(es->s_feature_incompat) &
2961 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2962 printk(KERN_ERR "EXT4-fs: external journal has "
2963 "bad superblock\n");
2968 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2969 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2974 len = ext4_blocks_count(es);
2975 start = sb_block + 1;
2976 brelse(bh); /* we're done with the superblock */
2978 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2979 start, len, blocksize);
2981 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2984 journal->j_private = sb;
2985 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2986 wait_on_buffer(journal->j_sb_buffer);
2987 if (!buffer_uptodate(journal->j_sb_buffer)) {
2988 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2991 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2992 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2993 "user (unsupported) - %d\n",
2994 be32_to_cpu(journal->j_superblock->s_nr_users));
2997 EXT4_SB(sb)->journal_bdev = bdev;
2998 ext4_init_journal_params(sb, journal);
3001 jbd2_journal_destroy(journal);
3003 ext4_blkdev_put(bdev);
3007 static int ext4_load_journal(struct super_block *sb,
3008 struct ext4_super_block *es,
3009 unsigned long journal_devnum)
3012 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3015 int really_read_only;
3017 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3019 if (journal_devnum &&
3020 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3021 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3022 "numbers have changed\n");
3023 journal_dev = new_decode_dev(journal_devnum);
3025 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3027 really_read_only = bdev_read_only(sb->s_bdev);
3030 * Are we loading a blank journal or performing recovery after a
3031 * crash? For recovery, we need to check in advance whether we
3032 * can get read-write access to the device.
3035 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3036 if (sb->s_flags & MS_RDONLY) {
3037 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3038 "required on readonly filesystem.\n");
3039 if (really_read_only) {
3040 printk(KERN_ERR "EXT4-fs: write access "
3041 "unavailable, cannot proceed.\n");
3044 printk(KERN_INFO "EXT4-fs: write access will "
3045 "be enabled during recovery.\n");
3049 if (journal_inum && journal_dev) {
3050 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3051 "and inode journals!\n");
3056 if (!(journal = ext4_get_journal(sb, journal_inum)))
3059 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3063 if (journal->j_flags & JBD2_BARRIER)
3064 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3066 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3068 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3069 err = jbd2_journal_update_format(journal);
3071 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3072 jbd2_journal_destroy(journal);
3077 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3078 err = jbd2_journal_wipe(journal, !really_read_only);
3080 err = jbd2_journal_load(journal);
3083 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3084 jbd2_journal_destroy(journal);
3088 EXT4_SB(sb)->s_journal = journal;
3089 ext4_clear_journal_err(sb, es);
3091 if (journal_devnum &&
3092 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3093 es->s_journal_dev = cpu_to_le32(journal_devnum);
3096 /* Make sure we flush the recovery flag to disk. */
3097 ext4_commit_super(sb, es, 1);
3103 static int ext4_commit_super(struct super_block *sb,
3104 struct ext4_super_block *es, int sync)
3106 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3111 if (buffer_write_io_error(sbh)) {
3113 * Oh, dear. A previous attempt to write the
3114 * superblock failed. This could happen because the
3115 * USB device was yanked out. Or it could happen to
3116 * be a transient write error and maybe the block will
3117 * be remapped. Nothing we can do but to retry the
3118 * write and hope for the best.
3120 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3121 "superblock detected for %s.\n", sb->s_id);
3122 clear_buffer_write_io_error(sbh);
3123 set_buffer_uptodate(sbh);
3125 es->s_wtime = cpu_to_le32(get_seconds());
3126 es->s_kbytes_written =
3127 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3128 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3129 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3130 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3131 &EXT4_SB(sb)->s_freeblocks_counter));
3132 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3133 &EXT4_SB(sb)->s_freeinodes_counter));
3135 BUFFER_TRACE(sbh, "marking dirty");
3136 mark_buffer_dirty(sbh);
3138 error = sync_dirty_buffer(sbh);
3142 error = buffer_write_io_error(sbh);
3144 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3145 "superblock for %s.\n", sb->s_id);
3146 clear_buffer_write_io_error(sbh);
3147 set_buffer_uptodate(sbh);
3155 * Have we just finished recovery? If so, and if we are mounting (or
3156 * remounting) the filesystem readonly, then we will end up with a
3157 * consistent fs on disk. Record that fact.
3159 static void ext4_mark_recovery_complete(struct super_block *sb,
3160 struct ext4_super_block *es)
3162 journal_t *journal = EXT4_SB(sb)->s_journal;
3164 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3165 BUG_ON(journal != NULL);
3168 jbd2_journal_lock_updates(journal);
3169 if (jbd2_journal_flush(journal) < 0)
3173 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3174 sb->s_flags & MS_RDONLY) {
3175 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3177 ext4_commit_super(sb, es, 1);
3182 jbd2_journal_unlock_updates(journal);
3186 * If we are mounting (or read-write remounting) a filesystem whose journal
3187 * has recorded an error from a previous lifetime, move that error to the
3188 * main filesystem now.
3190 static void ext4_clear_journal_err(struct super_block *sb,
3191 struct ext4_super_block *es)
3197 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3199 journal = EXT4_SB(sb)->s_journal;
3202 * Now check for any error status which may have been recorded in the
3203 * journal by a prior ext4_error() or ext4_abort()
3206 j_errno = jbd2_journal_errno(journal);
3210 errstr = ext4_decode_error(sb, j_errno, nbuf);
3211 ext4_warning(sb, __func__, "Filesystem error recorded "
3212 "from previous mount: %s", errstr);
3213 ext4_warning(sb, __func__, "Marking fs in need of "
3214 "filesystem check.");
3216 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3217 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3218 ext4_commit_super(sb, es, 1);
3220 jbd2_journal_clear_err(journal);
3225 * Force the running and committing transactions to commit,
3226 * and wait on the commit.
3228 int ext4_force_commit(struct super_block *sb)
3233 if (sb->s_flags & MS_RDONLY)
3236 journal = EXT4_SB(sb)->s_journal;
3239 ret = ext4_journal_force_commit(journal);
3246 * Ext4 always journals updates to the superblock itself, so we don't
3247 * have to propagate any other updates to the superblock on disk at this
3248 * point. (We can probably nuke this function altogether, and remove
3249 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3251 static void ext4_write_super(struct super_block *sb)
3253 if (EXT4_SB(sb)->s_journal) {
3254 if (mutex_trylock(&sb->s_lock) != 0)
3258 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3262 static int ext4_sync_fs(struct super_block *sb, int wait)
3267 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3269 if (EXT4_SB(sb)->s_journal) {
3270 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3273 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3277 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3283 * LVM calls this function before a (read-only) snapshot is created. This
3284 * gives us a chance to flush the journal completely and mark the fs clean.
3286 static int ext4_freeze(struct super_block *sb)
3292 if (!(sb->s_flags & MS_RDONLY)) {
3293 journal = EXT4_SB(sb)->s_journal;
3296 /* Now we set up the journal barrier. */
3297 jbd2_journal_lock_updates(journal);
3300 * We don't want to clear needs_recovery flag when we
3301 * failed to flush the journal.
3303 error = jbd2_journal_flush(journal);
3308 /* Journal blocked and flushed, clear needs_recovery flag. */
3309 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3310 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3316 jbd2_journal_unlock_updates(journal);
3321 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3322 * flag here, even though the filesystem is not technically dirty yet.
3324 static int ext4_unfreeze(struct super_block *sb)
3326 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3328 /* Reser the needs_recovery flag before the fs is unlocked. */
3329 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3330 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3332 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3337 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3339 struct ext4_super_block *es;
3340 struct ext4_sb_info *sbi = EXT4_SB(sb);
3341 ext4_fsblk_t n_blocks_count = 0;
3342 unsigned long old_sb_flags;
3343 struct ext4_mount_options old_opts;
3345 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3351 /* Store the original options */
3352 old_sb_flags = sb->s_flags;
3353 old_opts.s_mount_opt = sbi->s_mount_opt;
3354 old_opts.s_resuid = sbi->s_resuid;
3355 old_opts.s_resgid = sbi->s_resgid;
3356 old_opts.s_commit_interval = sbi->s_commit_interval;
3357 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3358 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3360 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3361 for (i = 0; i < MAXQUOTAS; i++)
3362 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3364 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3365 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3368 * Allow the "check" option to be passed as a remount option.
3370 if (!parse_options(data, sb, NULL, &journal_ioprio,
3371 &n_blocks_count, 1)) {
3376 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3377 ext4_abort(sb, __func__, "Abort forced by user");
3379 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3380 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3384 if (sbi->s_journal) {
3385 ext4_init_journal_params(sb, sbi->s_journal);
3386 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3389 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3390 n_blocks_count > ext4_blocks_count(es)) {
3391 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3396 if (*flags & MS_RDONLY) {
3398 * First of all, the unconditional stuff we have to do
3399 * to disable replay of the journal when we next remount
3401 sb->s_flags |= MS_RDONLY;
3404 * OK, test if we are remounting a valid rw partition
3405 * readonly, and if so set the rdonly flag and then
3406 * mark the partition as valid again.
3408 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3409 (sbi->s_mount_state & EXT4_VALID_FS))
3410 es->s_state = cpu_to_le16(sbi->s_mount_state);
3413 * We have to unlock super so that we can wait for
3416 if (sbi->s_journal) {
3418 ext4_mark_recovery_complete(sb, es);
3423 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3424 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3425 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3426 "remount RDWR because of unsupported "
3427 "optional features (%x).\n", sb->s_id,
3428 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3429 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3435 * Make sure the group descriptor checksums
3436 * are sane. If they aren't, refuse to
3439 for (g = 0; g < sbi->s_groups_count; g++) {
3440 struct ext4_group_desc *gdp =
3441 ext4_get_group_desc(sb, g, NULL);
3443 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3445 "EXT4-fs: ext4_remount: "
3446 "Checksum for group %u failed (%u!=%u)\n",
3447 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3448 le16_to_cpu(gdp->bg_checksum));
3455 * If we have an unprocessed orphan list hanging
3456 * around from a previously readonly bdev mount,
3457 * require a full umount/remount for now.
3459 if (es->s_last_orphan) {
3460 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3461 "remount RDWR because of unprocessed "
3462 "orphan inode list. Please "
3463 "umount/remount instead.\n",
3470 * Mounting a RDONLY partition read-write, so reread
3471 * and store the current valid flag. (It may have
3472 * been changed by e2fsck since we originally mounted
3476 ext4_clear_journal_err(sb, es);
3477 sbi->s_mount_state = le16_to_cpu(es->s_state);
3478 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3480 if (!ext4_setup_super(sb, es, 0))
3481 sb->s_flags &= ~MS_RDONLY;
3484 if (sbi->s_journal == NULL)
3485 ext4_commit_super(sb, es, 1);
3488 /* Release old quota file names */
3489 for (i = 0; i < MAXQUOTAS; i++)
3490 if (old_opts.s_qf_names[i] &&
3491 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3492 kfree(old_opts.s_qf_names[i]);
3496 sb->s_flags = old_sb_flags;
3497 sbi->s_mount_opt = old_opts.s_mount_opt;
3498 sbi->s_resuid = old_opts.s_resuid;
3499 sbi->s_resgid = old_opts.s_resgid;
3500 sbi->s_commit_interval = old_opts.s_commit_interval;
3501 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3502 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3504 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3505 for (i = 0; i < MAXQUOTAS; i++) {
3506 if (sbi->s_qf_names[i] &&
3507 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3508 kfree(sbi->s_qf_names[i]);
3509 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3515 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3517 struct super_block *sb = dentry->d_sb;
3518 struct ext4_sb_info *sbi = EXT4_SB(sb);
3519 struct ext4_super_block *es = sbi->s_es;
3522 if (test_opt(sb, MINIX_DF)) {
3523 sbi->s_overhead_last = 0;
3524 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3525 ext4_group_t ngroups = sbi->s_groups_count, i;
3526 ext4_fsblk_t overhead = 0;
3530 * Compute the overhead (FS structures). This is constant
3531 * for a given filesystem unless the number of block groups
3532 * changes so we cache the previous value until it does.
3536 * All of the blocks before first_data_block are
3539 overhead = le32_to_cpu(es->s_first_data_block);
3542 * Add the overhead attributed to the superblock and
3543 * block group descriptors. If the sparse superblocks
3544 * feature is turned on, then not all groups have this.
3546 for (i = 0; i < ngroups; i++) {
3547 overhead += ext4_bg_has_super(sb, i) +
3548 ext4_bg_num_gdb(sb, i);
3553 * Every block group has an inode bitmap, a block
3554 * bitmap, and an inode table.
3556 overhead += ngroups * (2 + sbi->s_itb_per_group);
3557 sbi->s_overhead_last = overhead;
3559 sbi->s_blocks_last = ext4_blocks_count(es);
3562 buf->f_type = EXT4_SUPER_MAGIC;
3563 buf->f_bsize = sb->s_blocksize;
3564 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3565 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3566 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3567 ext4_free_blocks_count_set(es, buf->f_bfree);
3568 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3569 if (buf->f_bfree < ext4_r_blocks_count(es))
3571 buf->f_files = le32_to_cpu(es->s_inodes_count);
3572 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3573 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3574 buf->f_namelen = EXT4_NAME_LEN;
3575 fsid = le64_to_cpup((void *)es->s_uuid) ^
3576 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3577 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3578 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3582 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3583 * is locked for write. Otherwise the are possible deadlocks:
3584 * Process 1 Process 2
3585 * ext4_create() quota_sync()
3586 * jbd2_journal_start() write_dquot()
3587 * vfs_dq_init() down(dqio_mutex)
3588 * down(dqio_mutex) jbd2_journal_start()
3594 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3596 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3599 static int ext4_write_dquot(struct dquot *dquot)
3603 struct inode *inode;
3605 inode = dquot_to_inode(dquot);
3606 handle = ext4_journal_start(inode,
3607 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3609 return PTR_ERR(handle);
3610 ret = dquot_commit(dquot);
3611 err = ext4_journal_stop(handle);
3617 static int ext4_acquire_dquot(struct dquot *dquot)
3622 handle = ext4_journal_start(dquot_to_inode(dquot),
3623 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3625 return PTR_ERR(handle);
3626 ret = dquot_acquire(dquot);
3627 err = ext4_journal_stop(handle);
3633 static int ext4_release_dquot(struct dquot *dquot)
3638 handle = ext4_journal_start(dquot_to_inode(dquot),
3639 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3640 if (IS_ERR(handle)) {
3641 /* Release dquot anyway to avoid endless cycle in dqput() */
3642 dquot_release(dquot);
3643 return PTR_ERR(handle);
3645 ret = dquot_release(dquot);
3646 err = ext4_journal_stop(handle);
3652 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3654 /* Are we journaling quotas? */
3655 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3656 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3657 dquot_mark_dquot_dirty(dquot);
3658 return ext4_write_dquot(dquot);
3660 return dquot_mark_dquot_dirty(dquot);
3664 static int ext4_write_info(struct super_block *sb, int type)
3669 /* Data block + inode block */
3670 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3672 return PTR_ERR(handle);
3673 ret = dquot_commit_info(sb, type);
3674 err = ext4_journal_stop(handle);
3681 * Turn on quotas during mount time - we need to find
3682 * the quota file and such...
3684 static int ext4_quota_on_mount(struct super_block *sb, int type)
3686 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3687 EXT4_SB(sb)->s_jquota_fmt, type);
3691 * Standard function to be called on quota_on
3693 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3694 char *name, int remount)
3699 if (!test_opt(sb, QUOTA))
3701 /* When remounting, no checks are needed and in fact, name is NULL */
3703 return vfs_quota_on(sb, type, format_id, name, remount);
3705 err = kern_path(name, LOOKUP_FOLLOW, &path);
3709 /* Quotafile not on the same filesystem? */
3710 if (path.mnt->mnt_sb != sb) {
3714 /* Journaling quota? */
3715 if (EXT4_SB(sb)->s_qf_names[type]) {
3716 /* Quotafile not in fs root? */
3717 if (path.dentry->d_parent != sb->s_root)
3719 "EXT4-fs: Quota file not on filesystem root. "
3720 "Journaled quota will not work.\n");
3724 * When we journal data on quota file, we have to flush journal to see
3725 * all updates to the file when we bypass pagecache...
3727 if (EXT4_SB(sb)->s_journal &&
3728 ext4_should_journal_data(path.dentry->d_inode)) {
3730 * We don't need to lock updates but journal_flush() could
3731 * otherwise be livelocked...
3733 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3734 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3735 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3742 err = vfs_quota_on_path(sb, type, format_id, &path);
3747 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3748 * acquiring the locks... As quota files are never truncated and quota code
3749 * itself serializes the operations (and noone else should touch the files)
3750 * we don't have to be afraid of races */
3751 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3752 size_t len, loff_t off)
3754 struct inode *inode = sb_dqopt(sb)->files[type];
3755 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3757 int offset = off & (sb->s_blocksize - 1);
3760 struct buffer_head *bh;
3761 loff_t i_size = i_size_read(inode);
3765 if (off+len > i_size)
3768 while (toread > 0) {
3769 tocopy = sb->s_blocksize - offset < toread ?
3770 sb->s_blocksize - offset : toread;
3771 bh = ext4_bread(NULL, inode, blk, 0, &err);
3774 if (!bh) /* A hole? */
3775 memset(data, 0, tocopy);
3777 memcpy(data, bh->b_data+offset, tocopy);
3787 /* Write to quotafile (we know the transaction is already started and has
3788 * enough credits) */
3789 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3790 const char *data, 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);
3797 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3798 size_t towrite = len;
3799 struct buffer_head *bh;
3800 handle_t *handle = journal_current_handle();
3802 if (EXT4_SB(sb)->s_journal && !handle) {
3803 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3804 " cancelled because transaction is not started.\n",
3805 (unsigned long long)off, (unsigned long long)len);
3808 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3809 while (towrite > 0) {
3810 tocopy = sb->s_blocksize - offset < towrite ?
3811 sb->s_blocksize - offset : towrite;
3812 bh = ext4_bread(handle, inode, blk, 1, &err);
3815 if (journal_quota) {
3816 err = ext4_journal_get_write_access(handle, bh);
3823 memcpy(bh->b_data+offset, data, tocopy);
3824 flush_dcache_page(bh->b_page);
3827 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3829 /* Always do at least ordered writes for quotas */
3830 err = ext4_jbd2_file_inode(handle, inode);
3831 mark_buffer_dirty(bh);
3842 if (len == towrite) {
3843 mutex_unlock(&inode->i_mutex);
3846 if (inode->i_size < off+len-towrite) {
3847 i_size_write(inode, off+len-towrite);
3848 EXT4_I(inode)->i_disksize = inode->i_size;
3850 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3851 ext4_mark_inode_dirty(handle, inode);
3852 mutex_unlock(&inode->i_mutex);
3853 return len - towrite;
3858 static int ext4_get_sb(struct file_system_type *fs_type,
3859 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3861 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3864 static struct file_system_type ext4_fs_type = {
3865 .owner = THIS_MODULE,
3867 .get_sb = ext4_get_sb,
3868 .kill_sb = kill_block_super,
3869 .fs_flags = FS_REQUIRES_DEV,
3872 #ifdef CONFIG_EXT4DEV_COMPAT
3873 static int ext4dev_get_sb(struct file_system_type *fs_type,
3874 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3876 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3877 "to mount using ext4\n");
3878 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3879 "will go away by 2.6.31\n");
3880 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3883 static struct file_system_type ext4dev_fs_type = {
3884 .owner = THIS_MODULE,
3886 .get_sb = ext4dev_get_sb,
3887 .kill_sb = kill_block_super,
3888 .fs_flags = FS_REQUIRES_DEV,
3890 MODULE_ALIAS("ext4dev");
3893 static int __init init_ext4_fs(void)
3897 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3900 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3901 err = init_ext4_mballoc();
3905 err = init_ext4_xattr();
3908 err = init_inodecache();
3911 err = register_filesystem(&ext4_fs_type);
3914 #ifdef CONFIG_EXT4DEV_COMPAT
3915 err = register_filesystem(&ext4dev_fs_type);
3917 unregister_filesystem(&ext4_fs_type);
3923 destroy_inodecache();
3927 exit_ext4_mballoc();
3931 static void __exit exit_ext4_fs(void)
3933 unregister_filesystem(&ext4_fs_type);
3934 #ifdef CONFIG_EXT4DEV_COMPAT
3935 unregister_filesystem(&ext4dev_fs_type);
3937 destroy_inodecache();
3939 exit_ext4_mballoc();
3940 remove_proc_entry("fs/ext4", NULL);
3941 kset_unregister(ext4_kset);
3944 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3945 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3946 MODULE_LICENSE("GPL");
3947 module_init(init_ext4_fs)
3948 module_exit(exit_ext4_fs)