2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
46 #include "ext4_jbd2.h"
52 struct proc_dir_entry *ext4_proc_root;
53 static struct kset *ext4_kset;
55 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
56 unsigned long journal_devnum);
57 static int ext4_commit_super(struct super_block *sb, 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, 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, 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, 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 if (is_vmalloc_addr(sbi->s_flex_groups))
590 vfree(sbi->s_flex_groups);
592 kfree(sbi->s_flex_groups);
593 percpu_counter_destroy(&sbi->s_freeblocks_counter);
594 percpu_counter_destroy(&sbi->s_freeinodes_counter);
595 percpu_counter_destroy(&sbi->s_dirs_counter);
596 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
599 for (i = 0; i < MAXQUOTAS; i++)
600 kfree(sbi->s_qf_names[i]);
603 /* Debugging code just in case the in-memory inode orphan list
604 * isn't empty. The on-disk one can be non-empty if we've
605 * detected an error and taken the fs readonly, but the
606 * in-memory list had better be clean by this point. */
607 if (!list_empty(&sbi->s_orphan))
608 dump_orphan_list(sb, sbi);
609 J_ASSERT(list_empty(&sbi->s_orphan));
611 invalidate_bdev(sb->s_bdev);
612 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
614 * Invalidate the journal device's buffers. We don't want them
615 * floating about in memory - the physical journal device may
616 * hotswapped, and it breaks the `ro-after' testing code.
618 sync_blockdev(sbi->journal_bdev);
619 invalidate_bdev(sbi->journal_bdev);
620 ext4_blkdev_remove(sbi);
622 sb->s_fs_info = NULL;
624 * Now that we are completely done shutting down the
625 * superblock, we need to actually destroy the kobject.
629 kobject_put(&sbi->s_kobj);
630 wait_for_completion(&sbi->s_kobj_unregister);
633 kfree(sbi->s_blockgroup_lock);
638 static struct kmem_cache *ext4_inode_cachep;
641 * Called inside transaction, so use GFP_NOFS
643 static struct inode *ext4_alloc_inode(struct super_block *sb)
645 struct ext4_inode_info *ei;
647 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
650 #ifdef CONFIG_EXT4_FS_POSIX_ACL
651 ei->i_acl = EXT4_ACL_NOT_CACHED;
652 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
654 ei->vfs_inode.i_version = 1;
655 ei->vfs_inode.i_data.writeback_index = 0;
656 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
657 INIT_LIST_HEAD(&ei->i_prealloc_list);
658 spin_lock_init(&ei->i_prealloc_lock);
660 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
661 * therefore it can be null here. Don't check it, just initialize
664 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
665 ei->i_reserved_data_blocks = 0;
666 ei->i_reserved_meta_blocks = 0;
667 ei->i_allocated_meta_blocks = 0;
668 ei->i_delalloc_reserved_flag = 0;
669 spin_lock_init(&(ei->i_block_reservation_lock));
670 return &ei->vfs_inode;
673 static void ext4_destroy_inode(struct inode *inode)
675 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
676 printk("EXT4 Inode %p: orphan list check failed!\n",
678 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
679 EXT4_I(inode), sizeof(struct ext4_inode_info),
683 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
686 static void init_once(void *foo)
688 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
690 INIT_LIST_HEAD(&ei->i_orphan);
691 #ifdef CONFIG_EXT4_FS_XATTR
692 init_rwsem(&ei->xattr_sem);
694 init_rwsem(&ei->i_data_sem);
695 inode_init_once(&ei->vfs_inode);
698 static int init_inodecache(void)
700 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
701 sizeof(struct ext4_inode_info),
702 0, (SLAB_RECLAIM_ACCOUNT|
705 if (ext4_inode_cachep == NULL)
710 static void destroy_inodecache(void)
712 kmem_cache_destroy(ext4_inode_cachep);
715 static void ext4_clear_inode(struct inode *inode)
717 #ifdef CONFIG_EXT4_FS_POSIX_ACL
718 if (EXT4_I(inode)->i_acl &&
719 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
720 posix_acl_release(EXT4_I(inode)->i_acl);
721 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
723 if (EXT4_I(inode)->i_default_acl &&
724 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
725 posix_acl_release(EXT4_I(inode)->i_default_acl);
726 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
729 ext4_discard_preallocations(inode);
730 if (EXT4_JOURNAL(inode))
731 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
732 &EXT4_I(inode)->jinode);
735 static inline void ext4_show_quota_options(struct seq_file *seq,
736 struct super_block *sb)
738 #if defined(CONFIG_QUOTA)
739 struct ext4_sb_info *sbi = EXT4_SB(sb);
741 if (sbi->s_jquota_fmt)
742 seq_printf(seq, ",jqfmt=%s",
743 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
745 if (sbi->s_qf_names[USRQUOTA])
746 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
748 if (sbi->s_qf_names[GRPQUOTA])
749 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
751 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
752 seq_puts(seq, ",usrquota");
754 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
755 seq_puts(seq, ",grpquota");
761 * - it's set to a non-default value OR
762 * - if the per-sb default is different from the global default
764 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
767 unsigned long def_mount_opts;
768 struct super_block *sb = vfs->mnt_sb;
769 struct ext4_sb_info *sbi = EXT4_SB(sb);
770 struct ext4_super_block *es = sbi->s_es;
772 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
773 def_errors = le16_to_cpu(es->s_errors);
775 if (sbi->s_sb_block != 1)
776 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
777 if (test_opt(sb, MINIX_DF))
778 seq_puts(seq, ",minixdf");
779 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
780 seq_puts(seq, ",grpid");
781 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
782 seq_puts(seq, ",nogrpid");
783 if (sbi->s_resuid != EXT4_DEF_RESUID ||
784 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
785 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
787 if (sbi->s_resgid != EXT4_DEF_RESGID ||
788 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
789 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
791 if (test_opt(sb, ERRORS_RO)) {
792 if (def_errors == EXT4_ERRORS_PANIC ||
793 def_errors == EXT4_ERRORS_CONTINUE) {
794 seq_puts(seq, ",errors=remount-ro");
797 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
798 seq_puts(seq, ",errors=continue");
799 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
800 seq_puts(seq, ",errors=panic");
801 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
802 seq_puts(seq, ",nouid32");
803 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
804 seq_puts(seq, ",debug");
805 if (test_opt(sb, OLDALLOC))
806 seq_puts(seq, ",oldalloc");
807 #ifdef CONFIG_EXT4_FS_XATTR
808 if (test_opt(sb, XATTR_USER) &&
809 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
810 seq_puts(seq, ",user_xattr");
811 if (!test_opt(sb, XATTR_USER) &&
812 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
813 seq_puts(seq, ",nouser_xattr");
816 #ifdef CONFIG_EXT4_FS_POSIX_ACL
817 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
818 seq_puts(seq, ",acl");
819 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
820 seq_puts(seq, ",noacl");
822 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
823 seq_printf(seq, ",commit=%u",
824 (unsigned) (sbi->s_commit_interval / HZ));
826 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
827 seq_printf(seq, ",min_batch_time=%u",
828 (unsigned) sbi->s_min_batch_time);
830 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
831 seq_printf(seq, ",max_batch_time=%u",
832 (unsigned) sbi->s_min_batch_time);
836 * We're changing the default of barrier mount option, so
837 * let's always display its mount state so it's clear what its
840 seq_puts(seq, ",barrier=");
841 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
842 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
843 seq_puts(seq, ",journal_async_commit");
844 if (test_opt(sb, NOBH))
845 seq_puts(seq, ",nobh");
846 if (test_opt(sb, I_VERSION))
847 seq_puts(seq, ",i_version");
848 if (!test_opt(sb, DELALLOC))
849 seq_puts(seq, ",nodelalloc");
853 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
855 * journal mode get enabled in different ways
856 * So just print the value even if we didn't specify it
858 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
859 seq_puts(seq, ",data=journal");
860 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
861 seq_puts(seq, ",data=ordered");
862 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
863 seq_puts(seq, ",data=writeback");
865 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
866 seq_printf(seq, ",inode_readahead_blks=%u",
867 sbi->s_inode_readahead_blks);
869 if (test_opt(sb, DATA_ERR_ABORT))
870 seq_puts(seq, ",data_err=abort");
872 if (test_opt(sb, NO_AUTO_DA_ALLOC))
873 seq_puts(seq, ",noauto_da_alloc");
875 ext4_show_quota_options(seq, sb);
880 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
881 u64 ino, u32 generation)
885 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
886 return ERR_PTR(-ESTALE);
887 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
888 return ERR_PTR(-ESTALE);
890 /* iget isn't really right if the inode is currently unallocated!!
892 * ext4_read_inode will return a bad_inode if the inode had been
893 * deleted, so we should be safe.
895 * Currently we don't know the generation for parent directory, so
896 * a generation of 0 means "accept any"
898 inode = ext4_iget(sb, ino);
900 return ERR_CAST(inode);
901 if (generation && inode->i_generation != generation) {
903 return ERR_PTR(-ESTALE);
909 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
910 int fh_len, int fh_type)
912 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
916 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
917 int fh_len, int fh_type)
919 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
924 * Try to release metadata pages (indirect blocks, directories) which are
925 * mapped via the block device. Since these pages could have journal heads
926 * which would prevent try_to_free_buffers() from freeing them, we must use
927 * jbd2 layer's try_to_free_buffers() function to release them.
929 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
931 journal_t *journal = EXT4_SB(sb)->s_journal;
933 WARN_ON(PageChecked(page));
934 if (!page_has_buffers(page))
937 return jbd2_journal_try_to_free_buffers(journal, page,
939 return try_to_free_buffers(page);
943 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
944 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
946 static int ext4_write_dquot(struct dquot *dquot);
947 static int ext4_acquire_dquot(struct dquot *dquot);
948 static int ext4_release_dquot(struct dquot *dquot);
949 static int ext4_mark_dquot_dirty(struct dquot *dquot);
950 static int ext4_write_info(struct super_block *sb, int type);
951 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
952 char *path, int remount);
953 static int ext4_quota_on_mount(struct super_block *sb, int type);
954 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
955 size_t len, loff_t off);
956 static ssize_t ext4_quota_write(struct super_block *sb, int type,
957 const char *data, size_t len, loff_t off);
959 static struct dquot_operations ext4_quota_operations = {
960 .initialize = dquot_initialize,
962 .alloc_space = dquot_alloc_space,
963 .reserve_space = dquot_reserve_space,
964 .claim_space = dquot_claim_space,
965 .release_rsv = dquot_release_reserved_space,
966 .get_reserved_space = ext4_get_reserved_space,
967 .alloc_inode = dquot_alloc_inode,
968 .free_space = dquot_free_space,
969 .free_inode = dquot_free_inode,
970 .transfer = dquot_transfer,
971 .write_dquot = ext4_write_dquot,
972 .acquire_dquot = ext4_acquire_dquot,
973 .release_dquot = ext4_release_dquot,
974 .mark_dirty = ext4_mark_dquot_dirty,
975 .write_info = ext4_write_info,
976 .alloc_dquot = dquot_alloc,
977 .destroy_dquot = dquot_destroy,
980 static struct quotactl_ops ext4_qctl_operations = {
981 .quota_on = ext4_quota_on,
982 .quota_off = vfs_quota_off,
983 .quota_sync = vfs_quota_sync,
984 .get_info = vfs_get_dqinfo,
985 .set_info = vfs_set_dqinfo,
986 .get_dqblk = vfs_get_dqblk,
987 .set_dqblk = vfs_set_dqblk
991 static const struct super_operations ext4_sops = {
992 .alloc_inode = ext4_alloc_inode,
993 .destroy_inode = ext4_destroy_inode,
994 .write_inode = ext4_write_inode,
995 .dirty_inode = ext4_dirty_inode,
996 .delete_inode = ext4_delete_inode,
997 .put_super = ext4_put_super,
998 .sync_fs = ext4_sync_fs,
999 .freeze_fs = ext4_freeze,
1000 .unfreeze_fs = ext4_unfreeze,
1001 .statfs = ext4_statfs,
1002 .remount_fs = ext4_remount,
1003 .clear_inode = ext4_clear_inode,
1004 .show_options = ext4_show_options,
1006 .quota_read = ext4_quota_read,
1007 .quota_write = ext4_quota_write,
1009 .bdev_try_to_free_page = bdev_try_to_free_page,
1012 static const struct super_operations ext4_nojournal_sops = {
1013 .alloc_inode = ext4_alloc_inode,
1014 .destroy_inode = ext4_destroy_inode,
1015 .write_inode = ext4_write_inode,
1016 .dirty_inode = ext4_dirty_inode,
1017 .delete_inode = ext4_delete_inode,
1018 .write_super = ext4_write_super,
1019 .put_super = ext4_put_super,
1020 .statfs = ext4_statfs,
1021 .remount_fs = ext4_remount,
1022 .clear_inode = ext4_clear_inode,
1023 .show_options = ext4_show_options,
1025 .quota_read = ext4_quota_read,
1026 .quota_write = ext4_quota_write,
1028 .bdev_try_to_free_page = bdev_try_to_free_page,
1031 static const struct export_operations ext4_export_ops = {
1032 .fh_to_dentry = ext4_fh_to_dentry,
1033 .fh_to_parent = ext4_fh_to_parent,
1034 .get_parent = ext4_get_parent,
1038 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1039 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1040 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1041 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1042 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1043 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1044 Opt_journal_update, Opt_journal_dev,
1045 Opt_journal_checksum, Opt_journal_async_commit,
1046 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1047 Opt_data_err_abort, Opt_data_err_ignore,
1048 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1049 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1050 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1051 Opt_usrquota, Opt_grpquota, Opt_i_version,
1052 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1053 Opt_inode_readahead_blks, Opt_journal_ioprio
1056 static const match_table_t tokens = {
1057 {Opt_bsd_df, "bsddf"},
1058 {Opt_minix_df, "minixdf"},
1059 {Opt_grpid, "grpid"},
1060 {Opt_grpid, "bsdgroups"},
1061 {Opt_nogrpid, "nogrpid"},
1062 {Opt_nogrpid, "sysvgroups"},
1063 {Opt_resgid, "resgid=%u"},
1064 {Opt_resuid, "resuid=%u"},
1066 {Opt_err_cont, "errors=continue"},
1067 {Opt_err_panic, "errors=panic"},
1068 {Opt_err_ro, "errors=remount-ro"},
1069 {Opt_nouid32, "nouid32"},
1070 {Opt_debug, "debug"},
1071 {Opt_oldalloc, "oldalloc"},
1072 {Opt_orlov, "orlov"},
1073 {Opt_user_xattr, "user_xattr"},
1074 {Opt_nouser_xattr, "nouser_xattr"},
1076 {Opt_noacl, "noacl"},
1077 {Opt_noload, "noload"},
1080 {Opt_commit, "commit=%u"},
1081 {Opt_min_batch_time, "min_batch_time=%u"},
1082 {Opt_max_batch_time, "max_batch_time=%u"},
1083 {Opt_journal_update, "journal=update"},
1084 {Opt_journal_dev, "journal_dev=%u"},
1085 {Opt_journal_checksum, "journal_checksum"},
1086 {Opt_journal_async_commit, "journal_async_commit"},
1087 {Opt_abort, "abort"},
1088 {Opt_data_journal, "data=journal"},
1089 {Opt_data_ordered, "data=ordered"},
1090 {Opt_data_writeback, "data=writeback"},
1091 {Opt_data_err_abort, "data_err=abort"},
1092 {Opt_data_err_ignore, "data_err=ignore"},
1093 {Opt_offusrjquota, "usrjquota="},
1094 {Opt_usrjquota, "usrjquota=%s"},
1095 {Opt_offgrpjquota, "grpjquota="},
1096 {Opt_grpjquota, "grpjquota=%s"},
1097 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1098 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1099 {Opt_grpquota, "grpquota"},
1100 {Opt_noquota, "noquota"},
1101 {Opt_quota, "quota"},
1102 {Opt_usrquota, "usrquota"},
1103 {Opt_barrier, "barrier=%u"},
1104 {Opt_barrier, "barrier"},
1105 {Opt_nobarrier, "nobarrier"},
1106 {Opt_i_version, "i_version"},
1107 {Opt_stripe, "stripe=%u"},
1108 {Opt_resize, "resize"},
1109 {Opt_delalloc, "delalloc"},
1110 {Opt_nodelalloc, "nodelalloc"},
1111 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1112 {Opt_journal_ioprio, "journal_ioprio=%u"},
1113 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1114 {Opt_auto_da_alloc, "auto_da_alloc"},
1115 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1119 static ext4_fsblk_t get_sb_block(void **data)
1121 ext4_fsblk_t sb_block;
1122 char *options = (char *) *data;
1124 if (!options || strncmp(options, "sb=", 3) != 0)
1125 return 1; /* Default location */
1127 /*todo: use simple_strtoll with >32bit ext4 */
1128 sb_block = simple_strtoul(options, &options, 0);
1129 if (*options && *options != ',') {
1130 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1134 if (*options == ',')
1136 *data = (void *) options;
1140 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1142 static int parse_options(char *options, struct super_block *sb,
1143 unsigned long *journal_devnum,
1144 unsigned int *journal_ioprio,
1145 ext4_fsblk_t *n_blocks_count, int is_remount)
1147 struct ext4_sb_info *sbi = EXT4_SB(sb);
1149 substring_t args[MAX_OPT_ARGS];
1160 while ((p = strsep(&options, ",")) != NULL) {
1165 token = match_token(p, tokens, args);
1168 clear_opt(sbi->s_mount_opt, MINIX_DF);
1171 set_opt(sbi->s_mount_opt, MINIX_DF);
1174 set_opt(sbi->s_mount_opt, GRPID);
1177 clear_opt(sbi->s_mount_opt, GRPID);
1180 if (match_int(&args[0], &option))
1182 sbi->s_resuid = option;
1185 if (match_int(&args[0], &option))
1187 sbi->s_resgid = option;
1190 /* handled by get_sb_block() instead of here */
1191 /* *sb_block = match_int(&args[0]); */
1194 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1195 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1196 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1199 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1200 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1201 set_opt(sbi->s_mount_opt, ERRORS_RO);
1204 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1205 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1206 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1209 set_opt(sbi->s_mount_opt, NO_UID32);
1212 set_opt(sbi->s_mount_opt, DEBUG);
1215 set_opt(sbi->s_mount_opt, OLDALLOC);
1218 clear_opt(sbi->s_mount_opt, OLDALLOC);
1220 #ifdef CONFIG_EXT4_FS_XATTR
1221 case Opt_user_xattr:
1222 set_opt(sbi->s_mount_opt, XATTR_USER);
1224 case Opt_nouser_xattr:
1225 clear_opt(sbi->s_mount_opt, XATTR_USER);
1228 case Opt_user_xattr:
1229 case Opt_nouser_xattr:
1230 printk(KERN_ERR "EXT4 (no)user_xattr options "
1234 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1236 set_opt(sbi->s_mount_opt, POSIX_ACL);
1239 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1244 printk(KERN_ERR "EXT4 (no)acl options "
1248 case Opt_journal_update:
1250 /* Eventually we will want to be able to create
1251 a journal file here. For now, only allow the
1252 user to specify an existing inode to be the
1255 printk(KERN_ERR "EXT4-fs: cannot specify "
1256 "journal on remount\n");
1259 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1261 case Opt_journal_dev:
1263 printk(KERN_ERR "EXT4-fs: cannot specify "
1264 "journal on remount\n");
1267 if (match_int(&args[0], &option))
1269 *journal_devnum = option;
1271 case Opt_journal_checksum:
1272 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1274 case Opt_journal_async_commit:
1275 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1276 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1279 set_opt(sbi->s_mount_opt, NOLOAD);
1282 if (match_int(&args[0], &option))
1287 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1288 sbi->s_commit_interval = HZ * option;
1290 case Opt_max_batch_time:
1291 if (match_int(&args[0], &option))
1296 option = EXT4_DEF_MAX_BATCH_TIME;
1297 sbi->s_max_batch_time = option;
1299 case Opt_min_batch_time:
1300 if (match_int(&args[0], &option))
1304 sbi->s_min_batch_time = option;
1306 case Opt_data_journal:
1307 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1309 case Opt_data_ordered:
1310 data_opt = EXT4_MOUNT_ORDERED_DATA;
1312 case Opt_data_writeback:
1313 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1316 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1319 "EXT4-fs: cannot change data "
1320 "mode on remount\n");
1324 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1325 sbi->s_mount_opt |= data_opt;
1328 case Opt_data_err_abort:
1329 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1331 case Opt_data_err_ignore:
1332 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1341 if (sb_any_quota_loaded(sb) &&
1342 !sbi->s_qf_names[qtype]) {
1344 "EXT4-fs: Cannot change journaled "
1345 "quota options when quota turned on.\n");
1348 qname = match_strdup(&args[0]);
1351 "EXT4-fs: not enough memory for "
1352 "storing quotafile name.\n");
1355 if (sbi->s_qf_names[qtype] &&
1356 strcmp(sbi->s_qf_names[qtype], qname)) {
1358 "EXT4-fs: %s quota file already "
1359 "specified.\n", QTYPE2NAME(qtype));
1363 sbi->s_qf_names[qtype] = qname;
1364 if (strchr(sbi->s_qf_names[qtype], '/')) {
1366 "EXT4-fs: quotafile must be on "
1367 "filesystem root.\n");
1368 kfree(sbi->s_qf_names[qtype]);
1369 sbi->s_qf_names[qtype] = NULL;
1372 set_opt(sbi->s_mount_opt, QUOTA);
1374 case Opt_offusrjquota:
1377 case Opt_offgrpjquota:
1380 if (sb_any_quota_loaded(sb) &&
1381 sbi->s_qf_names[qtype]) {
1382 printk(KERN_ERR "EXT4-fs: Cannot change "
1383 "journaled quota options when "
1384 "quota turned on.\n");
1388 * The space will be released later when all options
1389 * are confirmed to be correct
1391 sbi->s_qf_names[qtype] = NULL;
1393 case Opt_jqfmt_vfsold:
1394 qfmt = QFMT_VFS_OLD;
1396 case Opt_jqfmt_vfsv0:
1399 if (sb_any_quota_loaded(sb) &&
1400 sbi->s_jquota_fmt != qfmt) {
1401 printk(KERN_ERR "EXT4-fs: Cannot change "
1402 "journaled quota options when "
1403 "quota turned on.\n");
1406 sbi->s_jquota_fmt = qfmt;
1410 set_opt(sbi->s_mount_opt, QUOTA);
1411 set_opt(sbi->s_mount_opt, USRQUOTA);
1414 set_opt(sbi->s_mount_opt, QUOTA);
1415 set_opt(sbi->s_mount_opt, GRPQUOTA);
1418 if (sb_any_quota_loaded(sb)) {
1419 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1420 "options when quota turned on.\n");
1423 clear_opt(sbi->s_mount_opt, QUOTA);
1424 clear_opt(sbi->s_mount_opt, USRQUOTA);
1425 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1432 "EXT4-fs: quota options not supported.\n");
1436 case Opt_offusrjquota:
1437 case Opt_offgrpjquota:
1438 case Opt_jqfmt_vfsold:
1439 case Opt_jqfmt_vfsv0:
1441 "EXT4-fs: journaled quota options not "
1448 set_opt(sbi->s_mount_opt, ABORT);
1451 clear_opt(sbi->s_mount_opt, BARRIER);
1454 if (match_int(&args[0], &option)) {
1455 set_opt(sbi->s_mount_opt, BARRIER);
1459 set_opt(sbi->s_mount_opt, BARRIER);
1461 clear_opt(sbi->s_mount_opt, BARRIER);
1467 printk("EXT4-fs: resize option only available "
1471 if (match_int(&args[0], &option) != 0)
1473 *n_blocks_count = option;
1476 set_opt(sbi->s_mount_opt, NOBH);
1479 clear_opt(sbi->s_mount_opt, NOBH);
1482 set_opt(sbi->s_mount_opt, I_VERSION);
1483 sb->s_flags |= MS_I_VERSION;
1485 case Opt_nodelalloc:
1486 clear_opt(sbi->s_mount_opt, DELALLOC);
1489 if (match_int(&args[0], &option))
1493 sbi->s_stripe = option;
1496 set_opt(sbi->s_mount_opt, DELALLOC);
1498 case Opt_inode_readahead_blks:
1499 if (match_int(&args[0], &option))
1501 if (option < 0 || option > (1 << 30))
1503 if (!is_power_of_2(option)) {
1504 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1505 " must be a power of 2\n");
1508 sbi->s_inode_readahead_blks = option;
1510 case Opt_journal_ioprio:
1511 if (match_int(&args[0], &option))
1513 if (option < 0 || option > 7)
1515 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1518 case Opt_noauto_da_alloc:
1519 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1521 case Opt_auto_da_alloc:
1522 if (match_int(&args[0], &option)) {
1523 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1527 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1529 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1533 "EXT4-fs: Unrecognized mount option \"%s\" "
1534 "or missing value\n", p);
1539 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1540 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1541 sbi->s_qf_names[USRQUOTA])
1542 clear_opt(sbi->s_mount_opt, USRQUOTA);
1544 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1545 sbi->s_qf_names[GRPQUOTA])
1546 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1548 if ((sbi->s_qf_names[USRQUOTA] &&
1549 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1550 (sbi->s_qf_names[GRPQUOTA] &&
1551 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1552 printk(KERN_ERR "EXT4-fs: old and new quota "
1553 "format mixing.\n");
1557 if (!sbi->s_jquota_fmt) {
1558 printk(KERN_ERR "EXT4-fs: journaled quota format "
1559 "not specified.\n");
1563 if (sbi->s_jquota_fmt) {
1564 printk(KERN_ERR "EXT4-fs: journaled quota format "
1565 "specified with no journaling "
1574 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1577 struct ext4_sb_info *sbi = EXT4_SB(sb);
1580 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1581 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1582 "forcing read-only mode\n");
1587 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1588 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1589 "running e2fsck is recommended\n");
1590 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1592 "EXT4-fs warning: mounting fs with errors, "
1593 "running e2fsck is recommended\n");
1594 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1595 le16_to_cpu(es->s_mnt_count) >=
1596 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1598 "EXT4-fs warning: maximal mount count reached, "
1599 "running e2fsck is recommended\n");
1600 else if (le32_to_cpu(es->s_checkinterval) &&
1601 (le32_to_cpu(es->s_lastcheck) +
1602 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1604 "EXT4-fs warning: checktime reached, "
1605 "running e2fsck is recommended\n");
1606 if (!sbi->s_journal)
1607 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1608 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1609 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1610 le16_add_cpu(&es->s_mnt_count, 1);
1611 es->s_mtime = cpu_to_le32(get_seconds());
1612 ext4_update_dynamic_rev(sb);
1614 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1616 ext4_commit_super(sb, 1);
1617 if (test_opt(sb, DEBUG))
1618 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1619 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1621 sbi->s_groups_count,
1622 EXT4_BLOCKS_PER_GROUP(sb),
1623 EXT4_INODES_PER_GROUP(sb),
1626 if (EXT4_SB(sb)->s_journal) {
1627 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1628 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1629 "external", EXT4_SB(sb)->s_journal->j_devname);
1631 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1636 static int ext4_fill_flex_info(struct super_block *sb)
1638 struct ext4_sb_info *sbi = EXT4_SB(sb);
1639 struct ext4_group_desc *gdp = NULL;
1640 struct buffer_head *bh;
1641 ext4_group_t flex_group_count;
1642 ext4_group_t flex_group;
1643 int groups_per_flex = 0;
1647 if (!sbi->s_es->s_log_groups_per_flex) {
1648 sbi->s_log_groups_per_flex = 0;
1652 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1653 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1655 /* We allocate both existing and potentially added groups */
1656 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1657 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1658 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1659 size = flex_group_count * sizeof(struct flex_groups);
1660 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1661 if (sbi->s_flex_groups == NULL) {
1662 sbi->s_flex_groups = vmalloc(size);
1663 if (sbi->s_flex_groups)
1664 memset(sbi->s_flex_groups, 0, size);
1666 if (sbi->s_flex_groups == NULL) {
1667 printk(KERN_ERR "EXT4-fs: not enough memory for "
1668 "%u flex groups\n", flex_group_count);
1672 for (i = 0; i < sbi->s_groups_count; i++) {
1673 gdp = ext4_get_group_desc(sb, i, &bh);
1675 flex_group = ext4_flex_group(sbi, i);
1676 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1677 ext4_free_inodes_count(sb, gdp));
1678 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1679 ext4_free_blks_count(sb, gdp));
1680 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1681 ext4_used_dirs_count(sb, gdp));
1689 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1690 struct ext4_group_desc *gdp)
1694 if (sbi->s_es->s_feature_ro_compat &
1695 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1696 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1697 __le32 le_group = cpu_to_le32(block_group);
1699 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1700 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1701 crc = crc16(crc, (__u8 *)gdp, offset);
1702 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1703 /* for checksum of struct ext4_group_desc do the rest...*/
1704 if ((sbi->s_es->s_feature_incompat &
1705 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1706 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1707 crc = crc16(crc, (__u8 *)gdp + offset,
1708 le16_to_cpu(sbi->s_es->s_desc_size) -
1712 return cpu_to_le16(crc);
1715 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1716 struct ext4_group_desc *gdp)
1718 if ((sbi->s_es->s_feature_ro_compat &
1719 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1720 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1726 /* Called at mount-time, super-block is locked */
1727 static int ext4_check_descriptors(struct super_block *sb)
1729 struct ext4_sb_info *sbi = EXT4_SB(sb);
1730 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1731 ext4_fsblk_t last_block;
1732 ext4_fsblk_t block_bitmap;
1733 ext4_fsblk_t inode_bitmap;
1734 ext4_fsblk_t inode_table;
1735 int flexbg_flag = 0;
1738 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1741 ext4_debug("Checking group descriptors");
1743 for (i = 0; i < sbi->s_groups_count; i++) {
1744 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1746 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1747 last_block = ext4_blocks_count(sbi->s_es) - 1;
1749 last_block = first_block +
1750 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1752 block_bitmap = ext4_block_bitmap(sb, gdp);
1753 if (block_bitmap < first_block || block_bitmap > last_block) {
1754 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1755 "Block bitmap for group %u not in group "
1756 "(block %llu)!\n", i, block_bitmap);
1759 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1760 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1761 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1762 "Inode bitmap for group %u not in group "
1763 "(block %llu)!\n", i, inode_bitmap);
1766 inode_table = ext4_inode_table(sb, gdp);
1767 if (inode_table < first_block ||
1768 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1769 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1770 "Inode table for group %u not in group "
1771 "(block %llu)!\n", i, inode_table);
1774 spin_lock(sb_bgl_lock(sbi, i));
1775 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1776 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1777 "Checksum for group %u failed (%u!=%u)\n",
1778 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1779 gdp)), le16_to_cpu(gdp->bg_checksum));
1780 if (!(sb->s_flags & MS_RDONLY)) {
1781 spin_unlock(sb_bgl_lock(sbi, i));
1785 spin_unlock(sb_bgl_lock(sbi, i));
1787 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1790 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1791 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1795 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1796 * the superblock) which were deleted from all directories, but held open by
1797 * a process at the time of a crash. We walk the list and try to delete these
1798 * inodes at recovery time (only with a read-write filesystem).
1800 * In order to keep the orphan inode chain consistent during traversal (in
1801 * case of crash during recovery), we link each inode into the superblock
1802 * orphan list_head and handle it the same way as an inode deletion during
1803 * normal operation (which journals the operations for us).
1805 * We only do an iget() and an iput() on each inode, which is very safe if we
1806 * accidentally point at an in-use or already deleted inode. The worst that
1807 * can happen in this case is that we get a "bit already cleared" message from
1808 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1809 * e2fsck was run on this filesystem, and it must have already done the orphan
1810 * inode cleanup for us, so we can safely abort without any further action.
1812 static void ext4_orphan_cleanup(struct super_block *sb,
1813 struct ext4_super_block *es)
1815 unsigned int s_flags = sb->s_flags;
1816 int nr_orphans = 0, nr_truncates = 0;
1820 if (!es->s_last_orphan) {
1821 jbd_debug(4, "no orphan inodes to clean up\n");
1825 if (bdev_read_only(sb->s_bdev)) {
1826 printk(KERN_ERR "EXT4-fs: write access "
1827 "unavailable, skipping orphan cleanup.\n");
1831 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1832 if (es->s_last_orphan)
1833 jbd_debug(1, "Errors on filesystem, "
1834 "clearing orphan list.\n");
1835 es->s_last_orphan = 0;
1836 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1840 if (s_flags & MS_RDONLY) {
1841 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1843 sb->s_flags &= ~MS_RDONLY;
1846 /* Needed for iput() to work correctly and not trash data */
1847 sb->s_flags |= MS_ACTIVE;
1848 /* Turn on quotas so that they are updated correctly */
1849 for (i = 0; i < MAXQUOTAS; i++) {
1850 if (EXT4_SB(sb)->s_qf_names[i]) {
1851 int ret = ext4_quota_on_mount(sb, i);
1854 "EXT4-fs: Cannot turn on journaled "
1855 "quota: error %d\n", ret);
1860 while (es->s_last_orphan) {
1861 struct inode *inode;
1863 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1864 if (IS_ERR(inode)) {
1865 es->s_last_orphan = 0;
1869 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1871 if (inode->i_nlink) {
1873 "%s: truncating inode %lu to %lld bytes\n",
1874 __func__, inode->i_ino, inode->i_size);
1875 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1876 inode->i_ino, inode->i_size);
1877 ext4_truncate(inode);
1881 "%s: deleting unreferenced inode %lu\n",
1882 __func__, inode->i_ino);
1883 jbd_debug(2, "deleting unreferenced inode %lu\n",
1887 iput(inode); /* The delete magic happens here! */
1890 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1893 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1894 sb->s_id, PLURAL(nr_orphans));
1896 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1897 sb->s_id, PLURAL(nr_truncates));
1899 /* Turn quotas off */
1900 for (i = 0; i < MAXQUOTAS; i++) {
1901 if (sb_dqopt(sb)->files[i])
1902 vfs_quota_off(sb, i, 0);
1905 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1908 * Maximal extent format file size.
1909 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1910 * extent format containers, within a sector_t, and within i_blocks
1911 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1912 * so that won't be a limiting factor.
1914 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1916 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1919 loff_t upper_limit = MAX_LFS_FILESIZE;
1921 /* small i_blocks in vfs inode? */
1922 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1924 * CONFIG_LBD is not enabled implies the inode
1925 * i_block represent total blocks in 512 bytes
1926 * 32 == size of vfs inode i_blocks * 8
1928 upper_limit = (1LL << 32) - 1;
1930 /* total blocks in file system block size */
1931 upper_limit >>= (blkbits - 9);
1932 upper_limit <<= blkbits;
1935 /* 32-bit extent-start container, ee_block */
1940 /* Sanity check against vm- & vfs- imposed limits */
1941 if (res > upper_limit)
1948 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1949 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1950 * We need to be 1 filesystem block less than the 2^48 sector limit.
1952 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1954 loff_t res = EXT4_NDIR_BLOCKS;
1957 /* This is calculated to be the largest file size for a
1958 * dense, bitmapped file such that the total number of
1959 * sectors in the file, including data and all indirect blocks,
1960 * does not exceed 2^48 -1
1961 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1962 * total number of 512 bytes blocks of the file
1965 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1967 * !has_huge_files or CONFIG_LBD is not enabled
1968 * implies the inode i_block represent total blocks in
1969 * 512 bytes 32 == size of vfs inode i_blocks * 8
1971 upper_limit = (1LL << 32) - 1;
1973 /* total blocks in file system block size */
1974 upper_limit >>= (bits - 9);
1978 * We use 48 bit ext4_inode i_blocks
1979 * With EXT4_HUGE_FILE_FL set the i_blocks
1980 * represent total number of blocks in
1981 * file system block size
1983 upper_limit = (1LL << 48) - 1;
1987 /* indirect blocks */
1989 /* double indirect blocks */
1990 meta_blocks += 1 + (1LL << (bits-2));
1991 /* tripple indirect blocks */
1992 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1994 upper_limit -= meta_blocks;
1995 upper_limit <<= bits;
1997 res += 1LL << (bits-2);
1998 res += 1LL << (2*(bits-2));
1999 res += 1LL << (3*(bits-2));
2001 if (res > upper_limit)
2004 if (res > MAX_LFS_FILESIZE)
2005 res = MAX_LFS_FILESIZE;
2010 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2011 ext4_fsblk_t logical_sb_block, int nr)
2013 struct ext4_sb_info *sbi = EXT4_SB(sb);
2014 ext4_group_t bg, first_meta_bg;
2017 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2019 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2021 return logical_sb_block + nr + 1;
2022 bg = sbi->s_desc_per_block * nr;
2023 if (ext4_bg_has_super(sb, bg))
2025 return (has_super + ext4_group_first_block_no(sb, bg));
2029 * ext4_get_stripe_size: Get the stripe size.
2030 * @sbi: In memory super block info
2032 * If we have specified it via mount option, then
2033 * use the mount option value. If the value specified at mount time is
2034 * greater than the blocks per group use the super block value.
2035 * If the super block value is greater than blocks per group return 0.
2036 * Allocator needs it be less than blocks per group.
2039 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2041 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2042 unsigned long stripe_width =
2043 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2045 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2046 return sbi->s_stripe;
2048 if (stripe_width <= sbi->s_blocks_per_group)
2049 return stripe_width;
2051 if (stride <= sbi->s_blocks_per_group)
2060 struct attribute attr;
2061 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2062 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2063 const char *, size_t);
2067 static int parse_strtoul(const char *buf,
2068 unsigned long max, unsigned long *value)
2072 while (*buf && isspace(*buf))
2074 *value = simple_strtoul(buf, &endp, 0);
2075 while (*endp && isspace(*endp))
2077 if (*endp || *value > max)
2083 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2084 struct ext4_sb_info *sbi,
2087 return snprintf(buf, PAGE_SIZE, "%llu\n",
2088 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2091 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2092 struct ext4_sb_info *sbi, char *buf)
2094 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2096 return snprintf(buf, PAGE_SIZE, "%lu\n",
2097 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2098 sbi->s_sectors_written_start) >> 1);
2101 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2102 struct ext4_sb_info *sbi, char *buf)
2104 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2106 return snprintf(buf, PAGE_SIZE, "%llu\n",
2107 sbi->s_kbytes_written +
2108 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2109 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2112 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2113 struct ext4_sb_info *sbi,
2114 const char *buf, size_t count)
2118 if (parse_strtoul(buf, 0x40000000, &t))
2121 if (!is_power_of_2(t))
2124 sbi->s_inode_readahead_blks = t;
2128 static ssize_t sbi_ui_show(struct ext4_attr *a,
2129 struct ext4_sb_info *sbi, char *buf)
2131 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2133 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2136 static ssize_t sbi_ui_store(struct ext4_attr *a,
2137 struct ext4_sb_info *sbi,
2138 const char *buf, size_t count)
2140 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2143 if (parse_strtoul(buf, 0xffffffff, &t))
2149 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2150 static struct ext4_attr ext4_attr_##_name = { \
2151 .attr = {.name = __stringify(_name), .mode = _mode }, \
2154 .offset = offsetof(struct ext4_sb_info, _elname), \
2156 #define EXT4_ATTR(name, mode, show, store) \
2157 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2159 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2160 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2161 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2162 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2163 #define ATTR_LIST(name) &ext4_attr_##name.attr
2165 EXT4_RO_ATTR(delayed_allocation_blocks);
2166 EXT4_RO_ATTR(session_write_kbytes);
2167 EXT4_RO_ATTR(lifetime_write_kbytes);
2168 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2169 inode_readahead_blks_store, s_inode_readahead_blks);
2170 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2171 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2172 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2173 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2174 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2175 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2177 static struct attribute *ext4_attrs[] = {
2178 ATTR_LIST(delayed_allocation_blocks),
2179 ATTR_LIST(session_write_kbytes),
2180 ATTR_LIST(lifetime_write_kbytes),
2181 ATTR_LIST(inode_readahead_blks),
2182 ATTR_LIST(mb_stats),
2183 ATTR_LIST(mb_max_to_scan),
2184 ATTR_LIST(mb_min_to_scan),
2185 ATTR_LIST(mb_order2_req),
2186 ATTR_LIST(mb_stream_req),
2187 ATTR_LIST(mb_group_prealloc),
2191 static ssize_t ext4_attr_show(struct kobject *kobj,
2192 struct attribute *attr, char *buf)
2194 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2196 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2198 return a->show ? a->show(a, sbi, buf) : 0;
2201 static ssize_t ext4_attr_store(struct kobject *kobj,
2202 struct attribute *attr,
2203 const char *buf, size_t len)
2205 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2207 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2209 return a->store ? a->store(a, sbi, buf, len) : 0;
2212 static void ext4_sb_release(struct kobject *kobj)
2214 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2216 complete(&sbi->s_kobj_unregister);
2220 static struct sysfs_ops ext4_attr_ops = {
2221 .show = ext4_attr_show,
2222 .store = ext4_attr_store,
2225 static struct kobj_type ext4_ktype = {
2226 .default_attrs = ext4_attrs,
2227 .sysfs_ops = &ext4_attr_ops,
2228 .release = ext4_sb_release,
2231 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2232 __releases(kernel_lock)
2233 __acquires(kernel_lock)
2236 struct buffer_head *bh;
2237 struct ext4_super_block *es = NULL;
2238 struct ext4_sb_info *sbi;
2240 ext4_fsblk_t sb_block = get_sb_block(&data);
2241 ext4_fsblk_t logical_sb_block;
2242 unsigned long offset = 0;
2243 unsigned long journal_devnum = 0;
2244 unsigned long def_mount_opts;
2250 unsigned int db_count;
2252 int needs_recovery, has_huge_files;
2256 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2258 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2262 sbi->s_blockgroup_lock =
2263 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2264 if (!sbi->s_blockgroup_lock) {
2268 sb->s_fs_info = sbi;
2269 sbi->s_mount_opt = 0;
2270 sbi->s_resuid = EXT4_DEF_RESUID;
2271 sbi->s_resgid = EXT4_DEF_RESGID;
2272 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2273 sbi->s_sb_block = sb_block;
2274 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2279 /* Cleanup superblock name */
2280 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2283 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2285 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2290 * The ext4 superblock will not be buffer aligned for other than 1kB
2291 * block sizes. We need to calculate the offset from buffer start.
2293 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2294 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2295 offset = do_div(logical_sb_block, blocksize);
2297 logical_sb_block = sb_block;
2300 if (!(bh = sb_bread(sb, logical_sb_block))) {
2301 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2305 * Note: s_es must be initialized as soon as possible because
2306 * some ext4 macro-instructions depend on its value
2308 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2310 sb->s_magic = le16_to_cpu(es->s_magic);
2311 if (sb->s_magic != EXT4_SUPER_MAGIC)
2313 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2315 /* Set defaults before we parse the mount options */
2316 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2317 if (def_mount_opts & EXT4_DEFM_DEBUG)
2318 set_opt(sbi->s_mount_opt, DEBUG);
2319 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2320 set_opt(sbi->s_mount_opt, GRPID);
2321 if (def_mount_opts & EXT4_DEFM_UID16)
2322 set_opt(sbi->s_mount_opt, NO_UID32);
2323 #ifdef CONFIG_EXT4_FS_XATTR
2324 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2325 set_opt(sbi->s_mount_opt, XATTR_USER);
2327 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2328 if (def_mount_opts & EXT4_DEFM_ACL)
2329 set_opt(sbi->s_mount_opt, POSIX_ACL);
2331 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2332 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2333 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2334 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2335 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2336 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2338 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2339 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2340 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2341 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2343 set_opt(sbi->s_mount_opt, ERRORS_RO);
2345 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2346 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2347 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2348 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2349 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2351 set_opt(sbi->s_mount_opt, BARRIER);
2354 * enable delayed allocation by default
2355 * Use -o nodelalloc to turn it off
2357 set_opt(sbi->s_mount_opt, DELALLOC);
2360 if (!parse_options((char *) data, sb, &journal_devnum,
2361 &journal_ioprio, NULL, 0))
2364 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2365 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2367 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2368 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2369 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2370 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2372 "EXT4-fs warning: feature flags set on rev 0 fs, "
2373 "running e2fsck is recommended\n");
2376 * Check feature flags regardless of the revision level, since we
2377 * previously didn't change the revision level when setting the flags,
2378 * so there is a chance incompat flags are set on a rev 0 filesystem.
2380 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2382 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2383 "unsupported optional features (%x).\n", sb->s_id,
2384 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2385 ~EXT4_FEATURE_INCOMPAT_SUPP));
2388 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2389 if (!(sb->s_flags & MS_RDONLY) && features) {
2390 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2391 "unsupported optional features (%x).\n", sb->s_id,
2392 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2393 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2396 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2397 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2398 if (has_huge_files) {
2400 * Large file size enabled file system can only be
2401 * mount if kernel is build with CONFIG_LBD
2403 if (sizeof(root->i_blocks) < sizeof(u64) &&
2404 !(sb->s_flags & MS_RDONLY)) {
2405 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2406 "files cannot be mounted read-write "
2407 "without CONFIG_LBD.\n", sb->s_id);
2411 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2413 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2414 blocksize > EXT4_MAX_BLOCK_SIZE) {
2416 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2417 blocksize, sb->s_id);
2421 if (sb->s_blocksize != blocksize) {
2423 /* Validate the filesystem blocksize */
2424 if (!sb_set_blocksize(sb, blocksize)) {
2425 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2431 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2432 offset = do_div(logical_sb_block, blocksize);
2433 bh = sb_bread(sb, logical_sb_block);
2436 "EXT4-fs: Can't read superblock on 2nd try.\n");
2439 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2441 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2443 "EXT4-fs: Magic mismatch, very weird !\n");
2448 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2450 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2452 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2453 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2454 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2456 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2457 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2458 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2459 (!is_power_of_2(sbi->s_inode_size)) ||
2460 (sbi->s_inode_size > blocksize)) {
2462 "EXT4-fs: unsupported inode size: %d\n",
2466 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2467 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2469 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2470 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2471 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2472 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2473 !is_power_of_2(sbi->s_desc_size)) {
2475 "EXT4-fs: unsupported descriptor size %lu\n",
2480 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2481 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2482 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2483 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2485 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2486 if (sbi->s_inodes_per_block == 0)
2488 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2489 sbi->s_inodes_per_block;
2490 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2492 sbi->s_mount_state = le16_to_cpu(es->s_state);
2493 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2494 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2495 for (i = 0; i < 4; i++)
2496 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2497 sbi->s_def_hash_version = es->s_def_hash_version;
2498 i = le32_to_cpu(es->s_flags);
2499 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2500 sbi->s_hash_unsigned = 3;
2501 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2502 #ifdef __CHAR_UNSIGNED__
2503 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2504 sbi->s_hash_unsigned = 3;
2506 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2511 if (sbi->s_blocks_per_group > blocksize * 8) {
2513 "EXT4-fs: #blocks per group too big: %lu\n",
2514 sbi->s_blocks_per_group);
2517 if (sbi->s_inodes_per_group > blocksize * 8) {
2519 "EXT4-fs: #inodes per group too big: %lu\n",
2520 sbi->s_inodes_per_group);
2524 if (ext4_blocks_count(es) >
2525 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2526 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2527 " too large to mount safely\n", sb->s_id);
2528 if (sizeof(sector_t) < 8)
2529 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2534 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2537 /* check blocks count against device size */
2538 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2539 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2540 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2541 "exceeds size of device (%llu blocks)\n",
2542 ext4_blocks_count(es), blocks_count);
2547 * It makes no sense for the first data block to be beyond the end
2548 * of the filesystem.
2550 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2551 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2552 "block %u is beyond end of filesystem (%llu)\n",
2553 le32_to_cpu(es->s_first_data_block),
2554 ext4_blocks_count(es));
2557 blocks_count = (ext4_blocks_count(es) -
2558 le32_to_cpu(es->s_first_data_block) +
2559 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2560 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2561 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2562 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2563 "(block count %llu, first data block %u, "
2564 "blocks per group %lu)\n", sbi->s_groups_count,
2565 ext4_blocks_count(es),
2566 le32_to_cpu(es->s_first_data_block),
2567 EXT4_BLOCKS_PER_GROUP(sb));
2570 sbi->s_groups_count = blocks_count;
2571 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2572 EXT4_DESC_PER_BLOCK(sb);
2573 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2575 if (sbi->s_group_desc == NULL) {
2576 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2580 #ifdef CONFIG_PROC_FS
2582 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2585 bgl_lock_init(sbi->s_blockgroup_lock);
2587 for (i = 0; i < db_count; i++) {
2588 block = descriptor_loc(sb, logical_sb_block, i);
2589 sbi->s_group_desc[i] = sb_bread(sb, block);
2590 if (!sbi->s_group_desc[i]) {
2591 printk(KERN_ERR "EXT4-fs: "
2592 "can't read group descriptor %d\n", i);
2597 if (!ext4_check_descriptors(sb)) {
2598 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2601 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2602 if (!ext4_fill_flex_info(sb)) {
2604 "EXT4-fs: unable to initialize "
2605 "flex_bg meta info!\n");
2609 sbi->s_gdb_count = db_count;
2610 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2611 spin_lock_init(&sbi->s_next_gen_lock);
2613 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2614 ext4_count_free_blocks(sb));
2616 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2617 ext4_count_free_inodes(sb));
2620 err = percpu_counter_init(&sbi->s_dirs_counter,
2621 ext4_count_dirs(sb));
2624 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2627 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2631 sbi->s_stripe = ext4_get_stripe_size(sbi);
2634 * set up enough so that it can read an inode
2636 if (!test_opt(sb, NOLOAD) &&
2637 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2638 sb->s_op = &ext4_sops;
2640 sb->s_op = &ext4_nojournal_sops;
2641 sb->s_export_op = &ext4_export_ops;
2642 sb->s_xattr = ext4_xattr_handlers;
2644 sb->s_qcop = &ext4_qctl_operations;
2645 sb->dq_op = &ext4_quota_operations;
2647 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2651 needs_recovery = (es->s_last_orphan != 0 ||
2652 EXT4_HAS_INCOMPAT_FEATURE(sb,
2653 EXT4_FEATURE_INCOMPAT_RECOVER));
2656 * The first inode we look at is the journal inode. Don't try
2657 * root first: it may be modified in the journal!
2659 if (!test_opt(sb, NOLOAD) &&
2660 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2661 if (ext4_load_journal(sb, es, journal_devnum))
2663 if (!(sb->s_flags & MS_RDONLY) &&
2664 EXT4_SB(sb)->s_journal->j_failed_commit) {
2665 printk(KERN_CRIT "EXT4-fs error (device %s): "
2666 "ext4_fill_super: Journal transaction "
2667 "%u is corrupt\n", sb->s_id,
2668 EXT4_SB(sb)->s_journal->j_failed_commit);
2669 if (test_opt(sb, ERRORS_RO)) {
2671 "Mounting filesystem read-only\n");
2672 sb->s_flags |= MS_RDONLY;
2673 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2674 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2676 if (test_opt(sb, ERRORS_PANIC)) {
2677 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2678 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2679 ext4_commit_super(sb, 1);
2683 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2684 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2685 printk(KERN_ERR "EXT4-fs: required journal recovery "
2686 "suppressed and not mounted read-only\n");
2689 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2690 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2691 sbi->s_journal = NULL;
2696 if (ext4_blocks_count(es) > 0xffffffffULL &&
2697 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2698 JBD2_FEATURE_INCOMPAT_64BIT)) {
2699 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2703 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2704 jbd2_journal_set_features(sbi->s_journal,
2705 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2706 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2707 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2708 jbd2_journal_set_features(sbi->s_journal,
2709 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2710 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2711 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2713 jbd2_journal_clear_features(sbi->s_journal,
2714 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2715 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2718 /* We have now updated the journal if required, so we can
2719 * validate the data journaling mode. */
2720 switch (test_opt(sb, DATA_FLAGS)) {
2722 /* No mode set, assume a default based on the journal
2723 * capabilities: ORDERED_DATA if the journal can
2724 * cope, else JOURNAL_DATA
2726 if (jbd2_journal_check_available_features
2727 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2728 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2730 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2733 case EXT4_MOUNT_ORDERED_DATA:
2734 case EXT4_MOUNT_WRITEBACK_DATA:
2735 if (!jbd2_journal_check_available_features
2736 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2737 printk(KERN_ERR "EXT4-fs: Journal does not support "
2738 "requested data journaling mode\n");
2744 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2748 if (test_opt(sb, NOBH)) {
2749 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2750 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2751 "its supported only with writeback mode\n");
2752 clear_opt(sbi->s_mount_opt, NOBH);
2756 * The jbd2_journal_load will have done any necessary log recovery,
2757 * so we can safely mount the rest of the filesystem now.
2760 root = ext4_iget(sb, EXT4_ROOT_INO);
2762 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2763 ret = PTR_ERR(root);
2766 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2768 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2771 sb->s_root = d_alloc_root(root);
2773 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2779 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2781 /* determine the minimum size of new large inodes, if present */
2782 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2783 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2784 EXT4_GOOD_OLD_INODE_SIZE;
2785 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2786 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2787 if (sbi->s_want_extra_isize <
2788 le16_to_cpu(es->s_want_extra_isize))
2789 sbi->s_want_extra_isize =
2790 le16_to_cpu(es->s_want_extra_isize);
2791 if (sbi->s_want_extra_isize <
2792 le16_to_cpu(es->s_min_extra_isize))
2793 sbi->s_want_extra_isize =
2794 le16_to_cpu(es->s_min_extra_isize);
2797 /* Check if enough inode space is available */
2798 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2799 sbi->s_inode_size) {
2800 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2801 EXT4_GOOD_OLD_INODE_SIZE;
2802 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2806 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2807 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2808 "requested data journaling mode\n");
2809 clear_opt(sbi->s_mount_opt, DELALLOC);
2810 } else if (test_opt(sb, DELALLOC))
2811 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2814 err = ext4_mb_init(sb, needs_recovery);
2816 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2821 sbi->s_kobj.kset = ext4_kset;
2822 init_completion(&sbi->s_kobj_unregister);
2823 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2826 ext4_mb_release(sb);
2827 ext4_ext_release(sb);
2831 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2832 ext4_orphan_cleanup(sb, es);
2833 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2834 if (needs_recovery) {
2835 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2836 ext4_mark_recovery_complete(sb, es);
2838 if (EXT4_SB(sb)->s_journal) {
2839 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2840 descr = " journalled data mode";
2841 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2842 descr = " ordered data mode";
2844 descr = " writeback data mode";
2846 descr = "out journal";
2848 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2856 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2861 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2862 if (sbi->s_journal) {
2863 jbd2_journal_destroy(sbi->s_journal);
2864 sbi->s_journal = NULL;
2867 if (sbi->s_flex_groups) {
2868 if (is_vmalloc_addr(sbi->s_flex_groups))
2869 vfree(sbi->s_flex_groups);
2871 kfree(sbi->s_flex_groups);
2873 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2874 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2875 percpu_counter_destroy(&sbi->s_dirs_counter);
2876 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2878 for (i = 0; i < db_count; i++)
2879 brelse(sbi->s_group_desc[i]);
2880 kfree(sbi->s_group_desc);
2883 remove_proc_entry(sb->s_id, ext4_proc_root);
2886 for (i = 0; i < MAXQUOTAS; i++)
2887 kfree(sbi->s_qf_names[i]);
2889 ext4_blkdev_remove(sbi);
2892 sb->s_fs_info = NULL;
2899 * Setup any per-fs journal parameters now. We'll do this both on
2900 * initial mount, once the journal has been initialised but before we've
2901 * done any recovery; and again on any subsequent remount.
2903 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2905 struct ext4_sb_info *sbi = EXT4_SB(sb);
2907 journal->j_commit_interval = sbi->s_commit_interval;
2908 journal->j_min_batch_time = sbi->s_min_batch_time;
2909 journal->j_max_batch_time = sbi->s_max_batch_time;
2911 spin_lock(&journal->j_state_lock);
2912 if (test_opt(sb, BARRIER))
2913 journal->j_flags |= JBD2_BARRIER;
2915 journal->j_flags &= ~JBD2_BARRIER;
2916 if (test_opt(sb, DATA_ERR_ABORT))
2917 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2919 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2920 spin_unlock(&journal->j_state_lock);
2923 static journal_t *ext4_get_journal(struct super_block *sb,
2924 unsigned int journal_inum)
2926 struct inode *journal_inode;
2929 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2931 /* First, test for the existence of a valid inode on disk. Bad
2932 * things happen if we iget() an unused inode, as the subsequent
2933 * iput() will try to delete it. */
2935 journal_inode = ext4_iget(sb, journal_inum);
2936 if (IS_ERR(journal_inode)) {
2937 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2940 if (!journal_inode->i_nlink) {
2941 make_bad_inode(journal_inode);
2942 iput(journal_inode);
2943 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2947 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2948 journal_inode, journal_inode->i_size);
2949 if (!S_ISREG(journal_inode->i_mode)) {
2950 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2951 iput(journal_inode);
2955 journal = jbd2_journal_init_inode(journal_inode);
2957 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2958 iput(journal_inode);
2961 journal->j_private = sb;
2962 ext4_init_journal_params(sb, journal);
2966 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2969 struct buffer_head *bh;
2973 int hblock, blocksize;
2974 ext4_fsblk_t sb_block;
2975 unsigned long offset;
2976 struct ext4_super_block *es;
2977 struct block_device *bdev;
2979 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2981 bdev = ext4_blkdev_get(j_dev);
2985 if (bd_claim(bdev, sb)) {
2987 "EXT4-fs: failed to claim external journal device.\n");
2988 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2992 blocksize = sb->s_blocksize;
2993 hblock = bdev_hardsect_size(bdev);
2994 if (blocksize < hblock) {
2996 "EXT4-fs: blocksize too small for journal device.\n");
3000 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3001 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3002 set_blocksize(bdev, blocksize);
3003 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3004 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
3005 "external journal\n");
3009 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3010 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3011 !(le32_to_cpu(es->s_feature_incompat) &
3012 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3013 printk(KERN_ERR "EXT4-fs: external journal has "
3014 "bad superblock\n");
3019 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3020 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3025 len = ext4_blocks_count(es);
3026 start = sb_block + 1;
3027 brelse(bh); /* we're done with the superblock */
3029 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3030 start, len, blocksize);
3032 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3035 journal->j_private = sb;
3036 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3037 wait_on_buffer(journal->j_sb_buffer);
3038 if (!buffer_uptodate(journal->j_sb_buffer)) {
3039 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3042 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3043 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3044 "user (unsupported) - %d\n",
3045 be32_to_cpu(journal->j_superblock->s_nr_users));
3048 EXT4_SB(sb)->journal_bdev = bdev;
3049 ext4_init_journal_params(sb, journal);
3052 jbd2_journal_destroy(journal);
3054 ext4_blkdev_put(bdev);
3058 static int ext4_load_journal(struct super_block *sb,
3059 struct ext4_super_block *es,
3060 unsigned long journal_devnum)
3063 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3066 int really_read_only;
3068 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3070 if (journal_devnum &&
3071 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3072 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3073 "numbers have changed\n");
3074 journal_dev = new_decode_dev(journal_devnum);
3076 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3078 really_read_only = bdev_read_only(sb->s_bdev);
3081 * Are we loading a blank journal or performing recovery after a
3082 * crash? For recovery, we need to check in advance whether we
3083 * can get read-write access to the device.
3086 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3087 if (sb->s_flags & MS_RDONLY) {
3088 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3089 "required on readonly filesystem.\n");
3090 if (really_read_only) {
3091 printk(KERN_ERR "EXT4-fs: write access "
3092 "unavailable, cannot proceed.\n");
3095 printk(KERN_INFO "EXT4-fs: write access will "
3096 "be enabled during recovery.\n");
3100 if (journal_inum && journal_dev) {
3101 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3102 "and inode journals!\n");
3107 if (!(journal = ext4_get_journal(sb, journal_inum)))
3110 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3114 if (journal->j_flags & JBD2_BARRIER)
3115 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3117 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3119 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3120 err = jbd2_journal_update_format(journal);
3122 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3123 jbd2_journal_destroy(journal);
3128 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3129 err = jbd2_journal_wipe(journal, !really_read_only);
3131 err = jbd2_journal_load(journal);
3134 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3135 jbd2_journal_destroy(journal);
3139 EXT4_SB(sb)->s_journal = journal;
3140 ext4_clear_journal_err(sb, es);
3142 if (journal_devnum &&
3143 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3144 es->s_journal_dev = cpu_to_le32(journal_devnum);
3146 /* Make sure we flush the recovery flag to disk. */
3147 ext4_commit_super(sb, 1);
3153 static int ext4_commit_super(struct super_block *sb, int sync)
3155 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3156 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3161 if (buffer_write_io_error(sbh)) {
3163 * Oh, dear. A previous attempt to write the
3164 * superblock failed. This could happen because the
3165 * USB device was yanked out. Or it could happen to
3166 * be a transient write error and maybe the block will
3167 * be remapped. Nothing we can do but to retry the
3168 * write and hope for the best.
3170 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3171 "superblock detected for %s.\n", sb->s_id);
3172 clear_buffer_write_io_error(sbh);
3173 set_buffer_uptodate(sbh);
3175 es->s_wtime = cpu_to_le32(get_seconds());
3176 es->s_kbytes_written =
3177 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3178 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3179 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3180 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3181 &EXT4_SB(sb)->s_freeblocks_counter));
3182 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3183 &EXT4_SB(sb)->s_freeinodes_counter));
3185 BUFFER_TRACE(sbh, "marking dirty");
3186 mark_buffer_dirty(sbh);
3188 error = sync_dirty_buffer(sbh);
3192 error = buffer_write_io_error(sbh);
3194 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3195 "superblock for %s.\n", sb->s_id);
3196 clear_buffer_write_io_error(sbh);
3197 set_buffer_uptodate(sbh);
3205 * Have we just finished recovery? If so, and if we are mounting (or
3206 * remounting) the filesystem readonly, then we will end up with a
3207 * consistent fs on disk. Record that fact.
3209 static void ext4_mark_recovery_complete(struct super_block *sb,
3210 struct ext4_super_block *es)
3212 journal_t *journal = EXT4_SB(sb)->s_journal;
3214 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3215 BUG_ON(journal != NULL);
3218 jbd2_journal_lock_updates(journal);
3219 if (jbd2_journal_flush(journal) < 0)
3222 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3223 sb->s_flags & MS_RDONLY) {
3224 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3225 ext4_commit_super(sb, 1);
3229 jbd2_journal_unlock_updates(journal);
3233 * If we are mounting (or read-write remounting) a filesystem whose journal
3234 * has recorded an error from a previous lifetime, move that error to the
3235 * main filesystem now.
3237 static void ext4_clear_journal_err(struct super_block *sb,
3238 struct ext4_super_block *es)
3244 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3246 journal = EXT4_SB(sb)->s_journal;
3249 * Now check for any error status which may have been recorded in the
3250 * journal by a prior ext4_error() or ext4_abort()
3253 j_errno = jbd2_journal_errno(journal);
3257 errstr = ext4_decode_error(sb, j_errno, nbuf);
3258 ext4_warning(sb, __func__, "Filesystem error recorded "
3259 "from previous mount: %s", errstr);
3260 ext4_warning(sb, __func__, "Marking fs in need of "
3261 "filesystem check.");
3263 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3264 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3265 ext4_commit_super(sb, 1);
3267 jbd2_journal_clear_err(journal);
3272 * Force the running and committing transactions to commit,
3273 * and wait on the commit.
3275 int ext4_force_commit(struct super_block *sb)
3280 if (sb->s_flags & MS_RDONLY)
3283 journal = EXT4_SB(sb)->s_journal;
3285 ret = ext4_journal_force_commit(journal);
3290 static void ext4_write_super(struct super_block *sb)
3292 ext4_commit_super(sb, 1);
3295 static int ext4_sync_fs(struct super_block *sb, int wait)
3300 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3301 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3303 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3309 * LVM calls this function before a (read-only) snapshot is created. This
3310 * gives us a chance to flush the journal completely and mark the fs clean.
3312 static int ext4_freeze(struct super_block *sb)
3317 if (sb->s_flags & MS_RDONLY)
3320 journal = EXT4_SB(sb)->s_journal;
3322 /* Now we set up the journal barrier. */
3323 jbd2_journal_lock_updates(journal);
3326 * Don't clear the needs_recovery flag if we failed to flush
3329 error = jbd2_journal_flush(journal);
3332 jbd2_journal_unlock_updates(journal);
3336 /* Journal blocked and flushed, clear needs_recovery flag. */
3337 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3338 error = ext4_commit_super(sb, 1);
3345 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3346 * flag here, even though the filesystem is not technically dirty yet.
3348 static int ext4_unfreeze(struct super_block *sb)
3350 if (sb->s_flags & MS_RDONLY)
3354 /* Reset the needs_recovery flag before the fs is unlocked. */
3355 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3356 ext4_commit_super(sb, 1);
3358 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3362 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3364 struct ext4_super_block *es;
3365 struct ext4_sb_info *sbi = EXT4_SB(sb);
3366 ext4_fsblk_t n_blocks_count = 0;
3367 unsigned long old_sb_flags;
3368 struct ext4_mount_options old_opts;
3370 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3376 /* Store the original options */
3377 old_sb_flags = sb->s_flags;
3378 old_opts.s_mount_opt = sbi->s_mount_opt;
3379 old_opts.s_resuid = sbi->s_resuid;
3380 old_opts.s_resgid = sbi->s_resgid;
3381 old_opts.s_commit_interval = sbi->s_commit_interval;
3382 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3383 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3385 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3386 for (i = 0; i < MAXQUOTAS; i++)
3387 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3389 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3390 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3393 * Allow the "check" option to be passed as a remount option.
3395 if (!parse_options(data, sb, NULL, &journal_ioprio,
3396 &n_blocks_count, 1)) {
3401 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3402 ext4_abort(sb, __func__, "Abort forced by user");
3404 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3405 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3409 if (sbi->s_journal) {
3410 ext4_init_journal_params(sb, sbi->s_journal);
3411 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3414 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3415 n_blocks_count > ext4_blocks_count(es)) {
3416 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3421 if (*flags & MS_RDONLY) {
3423 * First of all, the unconditional stuff we have to do
3424 * to disable replay of the journal when we next remount
3426 sb->s_flags |= MS_RDONLY;
3429 * OK, test if we are remounting a valid rw partition
3430 * readonly, and if so set the rdonly flag and then
3431 * mark the partition as valid again.
3433 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3434 (sbi->s_mount_state & EXT4_VALID_FS))
3435 es->s_state = cpu_to_le16(sbi->s_mount_state);
3438 ext4_mark_recovery_complete(sb, es);
3441 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3442 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3443 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3444 "remount RDWR because of unsupported "
3445 "optional features (%x).\n", sb->s_id,
3446 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3447 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3453 * Make sure the group descriptor checksums
3454 * are sane. If they aren't, refuse to
3457 for (g = 0; g < sbi->s_groups_count; g++) {
3458 struct ext4_group_desc *gdp =
3459 ext4_get_group_desc(sb, g, NULL);
3461 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3463 "EXT4-fs: ext4_remount: "
3464 "Checksum for group %u failed (%u!=%u)\n",
3465 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3466 le16_to_cpu(gdp->bg_checksum));
3473 * If we have an unprocessed orphan list hanging
3474 * around from a previously readonly bdev mount,
3475 * require a full umount/remount for now.
3477 if (es->s_last_orphan) {
3478 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3479 "remount RDWR because of unprocessed "
3480 "orphan inode list. Please "
3481 "umount/remount instead.\n",
3488 * Mounting a RDONLY partition read-write, so reread
3489 * and store the current valid flag. (It may have
3490 * been changed by e2fsck since we originally mounted
3494 ext4_clear_journal_err(sb, es);
3495 sbi->s_mount_state = le16_to_cpu(es->s_state);
3496 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3498 if (!ext4_setup_super(sb, es, 0))
3499 sb->s_flags &= ~MS_RDONLY;
3502 if (sbi->s_journal == NULL)
3503 ext4_commit_super(sb, 1);
3506 /* Release old quota file names */
3507 for (i = 0; i < MAXQUOTAS; i++)
3508 if (old_opts.s_qf_names[i] &&
3509 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3510 kfree(old_opts.s_qf_names[i]);
3514 sb->s_flags = old_sb_flags;
3515 sbi->s_mount_opt = old_opts.s_mount_opt;
3516 sbi->s_resuid = old_opts.s_resuid;
3517 sbi->s_resgid = old_opts.s_resgid;
3518 sbi->s_commit_interval = old_opts.s_commit_interval;
3519 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3520 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3522 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3523 for (i = 0; i < MAXQUOTAS; i++) {
3524 if (sbi->s_qf_names[i] &&
3525 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3526 kfree(sbi->s_qf_names[i]);
3527 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3533 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3535 struct super_block *sb = dentry->d_sb;
3536 struct ext4_sb_info *sbi = EXT4_SB(sb);
3537 struct ext4_super_block *es = sbi->s_es;
3540 if (test_opt(sb, MINIX_DF)) {
3541 sbi->s_overhead_last = 0;
3542 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3543 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3544 ext4_fsblk_t overhead = 0;
3547 * Compute the overhead (FS structures). This is constant
3548 * for a given filesystem unless the number of block groups
3549 * changes so we cache the previous value until it does.
3553 * All of the blocks before first_data_block are
3556 overhead = le32_to_cpu(es->s_first_data_block);
3559 * Add the overhead attributed to the superblock and
3560 * block group descriptors. If the sparse superblocks
3561 * feature is turned on, then not all groups have this.
3563 for (i = 0; i < ngroups; i++) {
3564 overhead += ext4_bg_has_super(sb, i) +
3565 ext4_bg_num_gdb(sb, i);
3570 * Every block group has an inode bitmap, a block
3571 * bitmap, and an inode table.
3573 overhead += ngroups * (2 + sbi->s_itb_per_group);
3574 sbi->s_overhead_last = overhead;
3576 sbi->s_blocks_last = ext4_blocks_count(es);
3579 buf->f_type = EXT4_SUPER_MAGIC;
3580 buf->f_bsize = sb->s_blocksize;
3581 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3582 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3583 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3584 ext4_free_blocks_count_set(es, buf->f_bfree);
3585 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3586 if (buf->f_bfree < ext4_r_blocks_count(es))
3588 buf->f_files = le32_to_cpu(es->s_inodes_count);
3589 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3590 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3591 buf->f_namelen = EXT4_NAME_LEN;
3592 fsid = le64_to_cpup((void *)es->s_uuid) ^
3593 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3594 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3595 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3599 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3600 * is locked for write. Otherwise the are possible deadlocks:
3601 * Process 1 Process 2
3602 * ext4_create() quota_sync()
3603 * jbd2_journal_start() write_dquot()
3604 * vfs_dq_init() down(dqio_mutex)
3605 * down(dqio_mutex) jbd2_journal_start()
3611 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3613 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3616 static int ext4_write_dquot(struct dquot *dquot)
3620 struct inode *inode;
3622 inode = dquot_to_inode(dquot);
3623 handle = ext4_journal_start(inode,
3624 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3626 return PTR_ERR(handle);
3627 ret = dquot_commit(dquot);
3628 err = ext4_journal_stop(handle);
3634 static int ext4_acquire_dquot(struct dquot *dquot)
3639 handle = ext4_journal_start(dquot_to_inode(dquot),
3640 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3642 return PTR_ERR(handle);
3643 ret = dquot_acquire(dquot);
3644 err = ext4_journal_stop(handle);
3650 static int ext4_release_dquot(struct dquot *dquot)
3655 handle = ext4_journal_start(dquot_to_inode(dquot),
3656 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3657 if (IS_ERR(handle)) {
3658 /* Release dquot anyway to avoid endless cycle in dqput() */
3659 dquot_release(dquot);
3660 return PTR_ERR(handle);
3662 ret = dquot_release(dquot);
3663 err = ext4_journal_stop(handle);
3669 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3671 /* Are we journaling quotas? */
3672 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3673 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3674 dquot_mark_dquot_dirty(dquot);
3675 return ext4_write_dquot(dquot);
3677 return dquot_mark_dquot_dirty(dquot);
3681 static int ext4_write_info(struct super_block *sb, int type)
3686 /* Data block + inode block */
3687 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3689 return PTR_ERR(handle);
3690 ret = dquot_commit_info(sb, type);
3691 err = ext4_journal_stop(handle);
3698 * Turn on quotas during mount time - we need to find
3699 * the quota file and such...
3701 static int ext4_quota_on_mount(struct super_block *sb, int type)
3703 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3704 EXT4_SB(sb)->s_jquota_fmt, type);
3708 * Standard function to be called on quota_on
3710 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3711 char *name, int remount)
3716 if (!test_opt(sb, QUOTA))
3718 /* When remounting, no checks are needed and in fact, name is NULL */
3720 return vfs_quota_on(sb, type, format_id, name, remount);
3722 err = kern_path(name, LOOKUP_FOLLOW, &path);
3726 /* Quotafile not on the same filesystem? */
3727 if (path.mnt->mnt_sb != sb) {
3731 /* Journaling quota? */
3732 if (EXT4_SB(sb)->s_qf_names[type]) {
3733 /* Quotafile not in fs root? */
3734 if (path.dentry->d_parent != sb->s_root)
3736 "EXT4-fs: Quota file not on filesystem root. "
3737 "Journaled quota will not work.\n");
3741 * When we journal data on quota file, we have to flush journal to see
3742 * all updates to the file when we bypass pagecache...
3744 if (EXT4_SB(sb)->s_journal &&
3745 ext4_should_journal_data(path.dentry->d_inode)) {
3747 * We don't need to lock updates but journal_flush() could
3748 * otherwise be livelocked...
3750 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3751 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3752 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3759 err = vfs_quota_on_path(sb, type, format_id, &path);
3764 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3765 * acquiring the locks... As quota files are never truncated and quota code
3766 * itself serializes the operations (and noone else should touch the files)
3767 * we don't have to be afraid of races */
3768 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3769 size_t len, loff_t off)
3771 struct inode *inode = sb_dqopt(sb)->files[type];
3772 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3774 int offset = off & (sb->s_blocksize - 1);
3777 struct buffer_head *bh;
3778 loff_t i_size = i_size_read(inode);
3782 if (off+len > i_size)
3785 while (toread > 0) {
3786 tocopy = sb->s_blocksize - offset < toread ?
3787 sb->s_blocksize - offset : toread;
3788 bh = ext4_bread(NULL, inode, blk, 0, &err);
3791 if (!bh) /* A hole? */
3792 memset(data, 0, tocopy);
3794 memcpy(data, bh->b_data+offset, tocopy);
3804 /* Write to quotafile (we know the transaction is already started and has
3805 * enough credits) */
3806 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3807 const char *data, size_t len, loff_t off)
3809 struct inode *inode = sb_dqopt(sb)->files[type];
3810 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3812 int offset = off & (sb->s_blocksize - 1);
3814 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3815 size_t towrite = len;
3816 struct buffer_head *bh;
3817 handle_t *handle = journal_current_handle();
3819 if (EXT4_SB(sb)->s_journal && !handle) {
3820 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3821 " cancelled because transaction is not started.\n",
3822 (unsigned long long)off, (unsigned long long)len);
3825 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3826 while (towrite > 0) {
3827 tocopy = sb->s_blocksize - offset < towrite ?
3828 sb->s_blocksize - offset : towrite;
3829 bh = ext4_bread(handle, inode, blk, 1, &err);
3832 if (journal_quota) {
3833 err = ext4_journal_get_write_access(handle, bh);
3840 memcpy(bh->b_data+offset, data, tocopy);
3841 flush_dcache_page(bh->b_page);
3844 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3846 /* Always do at least ordered writes for quotas */
3847 err = ext4_jbd2_file_inode(handle, inode);
3848 mark_buffer_dirty(bh);
3859 if (len == towrite) {
3860 mutex_unlock(&inode->i_mutex);
3863 if (inode->i_size < off+len-towrite) {
3864 i_size_write(inode, off+len-towrite);
3865 EXT4_I(inode)->i_disksize = inode->i_size;
3867 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3868 ext4_mark_inode_dirty(handle, inode);
3869 mutex_unlock(&inode->i_mutex);
3870 return len - towrite;
3875 static int ext4_get_sb(struct file_system_type *fs_type,
3876 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3878 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3881 static struct file_system_type ext4_fs_type = {
3882 .owner = THIS_MODULE,
3884 .get_sb = ext4_get_sb,
3885 .kill_sb = kill_block_super,
3886 .fs_flags = FS_REQUIRES_DEV,
3889 #ifdef CONFIG_EXT4DEV_COMPAT
3890 static int ext4dev_get_sb(struct file_system_type *fs_type,
3891 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3893 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3894 "to mount using ext4\n");
3895 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3896 "will go away by 2.6.31\n");
3897 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3900 static struct file_system_type ext4dev_fs_type = {
3901 .owner = THIS_MODULE,
3903 .get_sb = ext4dev_get_sb,
3904 .kill_sb = kill_block_super,
3905 .fs_flags = FS_REQUIRES_DEV,
3907 MODULE_ALIAS("ext4dev");
3910 static int __init init_ext4_fs(void)
3914 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3917 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3918 err = init_ext4_mballoc();
3922 err = init_ext4_xattr();
3925 err = init_inodecache();
3928 err = register_filesystem(&ext4_fs_type);
3931 #ifdef CONFIG_EXT4DEV_COMPAT
3932 err = register_filesystem(&ext4dev_fs_type);
3934 unregister_filesystem(&ext4_fs_type);
3940 destroy_inodecache();
3944 exit_ext4_mballoc();
3948 static void __exit exit_ext4_fs(void)
3950 unregister_filesystem(&ext4_fs_type);
3951 #ifdef CONFIG_EXT4DEV_COMPAT
3952 unregister_filesystem(&ext4dev_fs_type);
3954 destroy_inodecache();
3956 exit_ext4_mballoc();
3957 remove_proc_entry("fs/ext4", NULL);
3958 kset_unregister(ext4_kset);
3961 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3962 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3963 MODULE_LICENSE("GPL");
3964 module_init(init_ext4_fs)
3965 module_exit(exit_ext4_fs)