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 .write_super = ext4_write_super,
999 .sync_fs = ext4_sync_fs,
1000 .freeze_fs = ext4_freeze,
1001 .unfreeze_fs = ext4_unfreeze,
1002 .statfs = ext4_statfs,
1003 .remount_fs = ext4_remount,
1004 .clear_inode = ext4_clear_inode,
1005 .show_options = ext4_show_options,
1007 .quota_read = ext4_quota_read,
1008 .quota_write = ext4_quota_write,
1010 .bdev_try_to_free_page = bdev_try_to_free_page,
1013 static const struct export_operations ext4_export_ops = {
1014 .fh_to_dentry = ext4_fh_to_dentry,
1015 .fh_to_parent = ext4_fh_to_parent,
1016 .get_parent = ext4_get_parent,
1020 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1021 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1022 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1023 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1024 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1025 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1026 Opt_journal_update, Opt_journal_dev,
1027 Opt_journal_checksum, Opt_journal_async_commit,
1028 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1029 Opt_data_err_abort, Opt_data_err_ignore,
1030 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1031 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1032 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1033 Opt_usrquota, Opt_grpquota, Opt_i_version,
1034 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1035 Opt_inode_readahead_blks, Opt_journal_ioprio
1038 static const match_table_t tokens = {
1039 {Opt_bsd_df, "bsddf"},
1040 {Opt_minix_df, "minixdf"},
1041 {Opt_grpid, "grpid"},
1042 {Opt_grpid, "bsdgroups"},
1043 {Opt_nogrpid, "nogrpid"},
1044 {Opt_nogrpid, "sysvgroups"},
1045 {Opt_resgid, "resgid=%u"},
1046 {Opt_resuid, "resuid=%u"},
1048 {Opt_err_cont, "errors=continue"},
1049 {Opt_err_panic, "errors=panic"},
1050 {Opt_err_ro, "errors=remount-ro"},
1051 {Opt_nouid32, "nouid32"},
1052 {Opt_debug, "debug"},
1053 {Opt_oldalloc, "oldalloc"},
1054 {Opt_orlov, "orlov"},
1055 {Opt_user_xattr, "user_xattr"},
1056 {Opt_nouser_xattr, "nouser_xattr"},
1058 {Opt_noacl, "noacl"},
1059 {Opt_noload, "noload"},
1062 {Opt_commit, "commit=%u"},
1063 {Opt_min_batch_time, "min_batch_time=%u"},
1064 {Opt_max_batch_time, "max_batch_time=%u"},
1065 {Opt_journal_update, "journal=update"},
1066 {Opt_journal_dev, "journal_dev=%u"},
1067 {Opt_journal_checksum, "journal_checksum"},
1068 {Opt_journal_async_commit, "journal_async_commit"},
1069 {Opt_abort, "abort"},
1070 {Opt_data_journal, "data=journal"},
1071 {Opt_data_ordered, "data=ordered"},
1072 {Opt_data_writeback, "data=writeback"},
1073 {Opt_data_err_abort, "data_err=abort"},
1074 {Opt_data_err_ignore, "data_err=ignore"},
1075 {Opt_offusrjquota, "usrjquota="},
1076 {Opt_usrjquota, "usrjquota=%s"},
1077 {Opt_offgrpjquota, "grpjquota="},
1078 {Opt_grpjquota, "grpjquota=%s"},
1079 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1080 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1081 {Opt_grpquota, "grpquota"},
1082 {Opt_noquota, "noquota"},
1083 {Opt_quota, "quota"},
1084 {Opt_usrquota, "usrquota"},
1085 {Opt_barrier, "barrier=%u"},
1086 {Opt_barrier, "barrier"},
1087 {Opt_nobarrier, "nobarrier"},
1088 {Opt_i_version, "i_version"},
1089 {Opt_stripe, "stripe=%u"},
1090 {Opt_resize, "resize"},
1091 {Opt_delalloc, "delalloc"},
1092 {Opt_nodelalloc, "nodelalloc"},
1093 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1094 {Opt_journal_ioprio, "journal_ioprio=%u"},
1095 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1096 {Opt_auto_da_alloc, "auto_da_alloc"},
1097 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1101 static ext4_fsblk_t get_sb_block(void **data)
1103 ext4_fsblk_t sb_block;
1104 char *options = (char *) *data;
1106 if (!options || strncmp(options, "sb=", 3) != 0)
1107 return 1; /* Default location */
1109 /*todo: use simple_strtoll with >32bit ext4 */
1110 sb_block = simple_strtoul(options, &options, 0);
1111 if (*options && *options != ',') {
1112 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1116 if (*options == ',')
1118 *data = (void *) options;
1122 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1124 static int parse_options(char *options, struct super_block *sb,
1125 unsigned long *journal_devnum,
1126 unsigned int *journal_ioprio,
1127 ext4_fsblk_t *n_blocks_count, int is_remount)
1129 struct ext4_sb_info *sbi = EXT4_SB(sb);
1131 substring_t args[MAX_OPT_ARGS];
1142 while ((p = strsep(&options, ",")) != NULL) {
1147 token = match_token(p, tokens, args);
1150 clear_opt(sbi->s_mount_opt, MINIX_DF);
1153 set_opt(sbi->s_mount_opt, MINIX_DF);
1156 set_opt(sbi->s_mount_opt, GRPID);
1159 clear_opt(sbi->s_mount_opt, GRPID);
1162 if (match_int(&args[0], &option))
1164 sbi->s_resuid = option;
1167 if (match_int(&args[0], &option))
1169 sbi->s_resgid = option;
1172 /* handled by get_sb_block() instead of here */
1173 /* *sb_block = match_int(&args[0]); */
1176 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1177 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1178 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1181 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1182 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1183 set_opt(sbi->s_mount_opt, ERRORS_RO);
1186 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1187 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1188 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1191 set_opt(sbi->s_mount_opt, NO_UID32);
1194 set_opt(sbi->s_mount_opt, DEBUG);
1197 set_opt(sbi->s_mount_opt, OLDALLOC);
1200 clear_opt(sbi->s_mount_opt, OLDALLOC);
1202 #ifdef CONFIG_EXT4_FS_XATTR
1203 case Opt_user_xattr:
1204 set_opt(sbi->s_mount_opt, XATTR_USER);
1206 case Opt_nouser_xattr:
1207 clear_opt(sbi->s_mount_opt, XATTR_USER);
1210 case Opt_user_xattr:
1211 case Opt_nouser_xattr:
1212 printk(KERN_ERR "EXT4 (no)user_xattr options "
1216 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1218 set_opt(sbi->s_mount_opt, POSIX_ACL);
1221 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1226 printk(KERN_ERR "EXT4 (no)acl options "
1230 case Opt_journal_update:
1232 /* Eventually we will want to be able to create
1233 a journal file here. For now, only allow the
1234 user to specify an existing inode to be the
1237 printk(KERN_ERR "EXT4-fs: cannot specify "
1238 "journal on remount\n");
1241 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1243 case Opt_journal_dev:
1245 printk(KERN_ERR "EXT4-fs: cannot specify "
1246 "journal on remount\n");
1249 if (match_int(&args[0], &option))
1251 *journal_devnum = option;
1253 case Opt_journal_checksum:
1254 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1256 case Opt_journal_async_commit:
1257 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1258 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1261 set_opt(sbi->s_mount_opt, NOLOAD);
1264 if (match_int(&args[0], &option))
1269 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1270 sbi->s_commit_interval = HZ * option;
1272 case Opt_max_batch_time:
1273 if (match_int(&args[0], &option))
1278 option = EXT4_DEF_MAX_BATCH_TIME;
1279 sbi->s_max_batch_time = option;
1281 case Opt_min_batch_time:
1282 if (match_int(&args[0], &option))
1286 sbi->s_min_batch_time = option;
1288 case Opt_data_journal:
1289 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1291 case Opt_data_ordered:
1292 data_opt = EXT4_MOUNT_ORDERED_DATA;
1294 case Opt_data_writeback:
1295 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1298 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1301 "EXT4-fs: cannot change data "
1302 "mode on remount\n");
1306 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1307 sbi->s_mount_opt |= data_opt;
1310 case Opt_data_err_abort:
1311 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1313 case Opt_data_err_ignore:
1314 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1323 if (sb_any_quota_loaded(sb) &&
1324 !sbi->s_qf_names[qtype]) {
1326 "EXT4-fs: Cannot change journaled "
1327 "quota options when quota turned on.\n");
1330 qname = match_strdup(&args[0]);
1333 "EXT4-fs: not enough memory for "
1334 "storing quotafile name.\n");
1337 if (sbi->s_qf_names[qtype] &&
1338 strcmp(sbi->s_qf_names[qtype], qname)) {
1340 "EXT4-fs: %s quota file already "
1341 "specified.\n", QTYPE2NAME(qtype));
1345 sbi->s_qf_names[qtype] = qname;
1346 if (strchr(sbi->s_qf_names[qtype], '/')) {
1348 "EXT4-fs: quotafile must be on "
1349 "filesystem root.\n");
1350 kfree(sbi->s_qf_names[qtype]);
1351 sbi->s_qf_names[qtype] = NULL;
1354 set_opt(sbi->s_mount_opt, QUOTA);
1356 case Opt_offusrjquota:
1359 case Opt_offgrpjquota:
1362 if (sb_any_quota_loaded(sb) &&
1363 sbi->s_qf_names[qtype]) {
1364 printk(KERN_ERR "EXT4-fs: Cannot change "
1365 "journaled quota options when "
1366 "quota turned on.\n");
1370 * The space will be released later when all options
1371 * are confirmed to be correct
1373 sbi->s_qf_names[qtype] = NULL;
1375 case Opt_jqfmt_vfsold:
1376 qfmt = QFMT_VFS_OLD;
1378 case Opt_jqfmt_vfsv0:
1381 if (sb_any_quota_loaded(sb) &&
1382 sbi->s_jquota_fmt != qfmt) {
1383 printk(KERN_ERR "EXT4-fs: Cannot change "
1384 "journaled quota options when "
1385 "quota turned on.\n");
1388 sbi->s_jquota_fmt = qfmt;
1392 set_opt(sbi->s_mount_opt, QUOTA);
1393 set_opt(sbi->s_mount_opt, USRQUOTA);
1396 set_opt(sbi->s_mount_opt, QUOTA);
1397 set_opt(sbi->s_mount_opt, GRPQUOTA);
1400 if (sb_any_quota_loaded(sb)) {
1401 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1402 "options when quota turned on.\n");
1405 clear_opt(sbi->s_mount_opt, QUOTA);
1406 clear_opt(sbi->s_mount_opt, USRQUOTA);
1407 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1414 "EXT4-fs: quota options not supported.\n");
1418 case Opt_offusrjquota:
1419 case Opt_offgrpjquota:
1420 case Opt_jqfmt_vfsold:
1421 case Opt_jqfmt_vfsv0:
1423 "EXT4-fs: journaled quota options not "
1430 set_opt(sbi->s_mount_opt, ABORT);
1433 clear_opt(sbi->s_mount_opt, BARRIER);
1436 if (match_int(&args[0], &option)) {
1437 set_opt(sbi->s_mount_opt, BARRIER);
1441 set_opt(sbi->s_mount_opt, BARRIER);
1443 clear_opt(sbi->s_mount_opt, BARRIER);
1449 printk("EXT4-fs: resize option only available "
1453 if (match_int(&args[0], &option) != 0)
1455 *n_blocks_count = option;
1458 set_opt(sbi->s_mount_opt, NOBH);
1461 clear_opt(sbi->s_mount_opt, NOBH);
1464 set_opt(sbi->s_mount_opt, I_VERSION);
1465 sb->s_flags |= MS_I_VERSION;
1467 case Opt_nodelalloc:
1468 clear_opt(sbi->s_mount_opt, DELALLOC);
1471 if (match_int(&args[0], &option))
1475 sbi->s_stripe = option;
1478 set_opt(sbi->s_mount_opt, DELALLOC);
1480 case Opt_inode_readahead_blks:
1481 if (match_int(&args[0], &option))
1483 if (option < 0 || option > (1 << 30))
1485 if (!is_power_of_2(option)) {
1486 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1487 " must be a power of 2\n");
1490 sbi->s_inode_readahead_blks = option;
1492 case Opt_journal_ioprio:
1493 if (match_int(&args[0], &option))
1495 if (option < 0 || option > 7)
1497 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1500 case Opt_noauto_da_alloc:
1501 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1503 case Opt_auto_da_alloc:
1504 if (match_int(&args[0], &option)) {
1505 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1509 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1511 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1515 "EXT4-fs: Unrecognized mount option \"%s\" "
1516 "or missing value\n", p);
1521 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1522 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1523 sbi->s_qf_names[USRQUOTA])
1524 clear_opt(sbi->s_mount_opt, USRQUOTA);
1526 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1527 sbi->s_qf_names[GRPQUOTA])
1528 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1530 if ((sbi->s_qf_names[USRQUOTA] &&
1531 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1532 (sbi->s_qf_names[GRPQUOTA] &&
1533 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1534 printk(KERN_ERR "EXT4-fs: old and new quota "
1535 "format mixing.\n");
1539 if (!sbi->s_jquota_fmt) {
1540 printk(KERN_ERR "EXT4-fs: journaled quota format "
1541 "not specified.\n");
1545 if (sbi->s_jquota_fmt) {
1546 printk(KERN_ERR "EXT4-fs: journaled quota format "
1547 "specified with no journaling "
1556 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1559 struct ext4_sb_info *sbi = EXT4_SB(sb);
1562 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1563 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1564 "forcing read-only mode\n");
1569 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1570 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1571 "running e2fsck is recommended\n");
1572 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1574 "EXT4-fs warning: mounting fs with errors, "
1575 "running e2fsck is recommended\n");
1576 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1577 le16_to_cpu(es->s_mnt_count) >=
1578 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1580 "EXT4-fs warning: maximal mount count reached, "
1581 "running e2fsck is recommended\n");
1582 else if (le32_to_cpu(es->s_checkinterval) &&
1583 (le32_to_cpu(es->s_lastcheck) +
1584 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1586 "EXT4-fs warning: checktime reached, "
1587 "running e2fsck is recommended\n");
1588 if (!sbi->s_journal)
1589 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1590 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1591 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1592 le16_add_cpu(&es->s_mnt_count, 1);
1593 es->s_mtime = cpu_to_le32(get_seconds());
1594 ext4_update_dynamic_rev(sb);
1596 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1598 ext4_commit_super(sb, 1);
1599 if (test_opt(sb, DEBUG))
1600 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1601 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1603 sbi->s_groups_count,
1604 EXT4_BLOCKS_PER_GROUP(sb),
1605 EXT4_INODES_PER_GROUP(sb),
1608 if (EXT4_SB(sb)->s_journal) {
1609 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1610 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1611 "external", EXT4_SB(sb)->s_journal->j_devname);
1613 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1618 static int ext4_fill_flex_info(struct super_block *sb)
1620 struct ext4_sb_info *sbi = EXT4_SB(sb);
1621 struct ext4_group_desc *gdp = NULL;
1622 struct buffer_head *bh;
1623 ext4_group_t flex_group_count;
1624 ext4_group_t flex_group;
1625 int groups_per_flex = 0;
1629 if (!sbi->s_es->s_log_groups_per_flex) {
1630 sbi->s_log_groups_per_flex = 0;
1634 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1635 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1637 /* We allocate both existing and potentially added groups */
1638 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1639 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1640 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1641 size = flex_group_count * sizeof(struct flex_groups);
1642 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1643 if (sbi->s_flex_groups == NULL) {
1644 sbi->s_flex_groups = vmalloc(size);
1645 if (sbi->s_flex_groups)
1646 memset(sbi->s_flex_groups, 0, size);
1648 if (sbi->s_flex_groups == NULL) {
1649 printk(KERN_ERR "EXT4-fs: not enough memory for "
1650 "%u flex groups\n", flex_group_count);
1654 for (i = 0; i < sbi->s_groups_count; i++) {
1655 gdp = ext4_get_group_desc(sb, i, &bh);
1657 flex_group = ext4_flex_group(sbi, i);
1658 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1659 ext4_free_inodes_count(sb, gdp));
1660 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1661 ext4_free_blks_count(sb, gdp));
1662 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1663 ext4_used_dirs_count(sb, gdp));
1671 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1672 struct ext4_group_desc *gdp)
1676 if (sbi->s_es->s_feature_ro_compat &
1677 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1678 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1679 __le32 le_group = cpu_to_le32(block_group);
1681 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1682 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1683 crc = crc16(crc, (__u8 *)gdp, offset);
1684 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1685 /* for checksum of struct ext4_group_desc do the rest...*/
1686 if ((sbi->s_es->s_feature_incompat &
1687 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1688 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1689 crc = crc16(crc, (__u8 *)gdp + offset,
1690 le16_to_cpu(sbi->s_es->s_desc_size) -
1694 return cpu_to_le16(crc);
1697 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1698 struct ext4_group_desc *gdp)
1700 if ((sbi->s_es->s_feature_ro_compat &
1701 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1702 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1708 /* Called at mount-time, super-block is locked */
1709 static int ext4_check_descriptors(struct super_block *sb)
1711 struct ext4_sb_info *sbi = EXT4_SB(sb);
1712 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1713 ext4_fsblk_t last_block;
1714 ext4_fsblk_t block_bitmap;
1715 ext4_fsblk_t inode_bitmap;
1716 ext4_fsblk_t inode_table;
1717 int flexbg_flag = 0;
1720 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1723 ext4_debug("Checking group descriptors");
1725 for (i = 0; i < sbi->s_groups_count; i++) {
1726 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1728 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1729 last_block = ext4_blocks_count(sbi->s_es) - 1;
1731 last_block = first_block +
1732 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1734 block_bitmap = ext4_block_bitmap(sb, gdp);
1735 if (block_bitmap < first_block || block_bitmap > last_block) {
1736 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1737 "Block bitmap for group %u not in group "
1738 "(block %llu)!\n", i, block_bitmap);
1741 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1742 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1743 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1744 "Inode bitmap for group %u not in group "
1745 "(block %llu)!\n", i, inode_bitmap);
1748 inode_table = ext4_inode_table(sb, gdp);
1749 if (inode_table < first_block ||
1750 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1751 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1752 "Inode table for group %u not in group "
1753 "(block %llu)!\n", i, inode_table);
1756 spin_lock(sb_bgl_lock(sbi, i));
1757 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1758 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1759 "Checksum for group %u failed (%u!=%u)\n",
1760 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1761 gdp)), le16_to_cpu(gdp->bg_checksum));
1762 if (!(sb->s_flags & MS_RDONLY)) {
1763 spin_unlock(sb_bgl_lock(sbi, i));
1767 spin_unlock(sb_bgl_lock(sbi, i));
1769 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1772 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1773 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1777 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1778 * the superblock) which were deleted from all directories, but held open by
1779 * a process at the time of a crash. We walk the list and try to delete these
1780 * inodes at recovery time (only with a read-write filesystem).
1782 * In order to keep the orphan inode chain consistent during traversal (in
1783 * case of crash during recovery), we link each inode into the superblock
1784 * orphan list_head and handle it the same way as an inode deletion during
1785 * normal operation (which journals the operations for us).
1787 * We only do an iget() and an iput() on each inode, which is very safe if we
1788 * accidentally point at an in-use or already deleted inode. The worst that
1789 * can happen in this case is that we get a "bit already cleared" message from
1790 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1791 * e2fsck was run on this filesystem, and it must have already done the orphan
1792 * inode cleanup for us, so we can safely abort without any further action.
1794 static void ext4_orphan_cleanup(struct super_block *sb,
1795 struct ext4_super_block *es)
1797 unsigned int s_flags = sb->s_flags;
1798 int nr_orphans = 0, nr_truncates = 0;
1802 if (!es->s_last_orphan) {
1803 jbd_debug(4, "no orphan inodes to clean up\n");
1807 if (bdev_read_only(sb->s_bdev)) {
1808 printk(KERN_ERR "EXT4-fs: write access "
1809 "unavailable, skipping orphan cleanup.\n");
1813 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1814 if (es->s_last_orphan)
1815 jbd_debug(1, "Errors on filesystem, "
1816 "clearing orphan list.\n");
1817 es->s_last_orphan = 0;
1818 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1822 if (s_flags & MS_RDONLY) {
1823 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1825 sb->s_flags &= ~MS_RDONLY;
1828 /* Needed for iput() to work correctly and not trash data */
1829 sb->s_flags |= MS_ACTIVE;
1830 /* Turn on quotas so that they are updated correctly */
1831 for (i = 0; i < MAXQUOTAS; i++) {
1832 if (EXT4_SB(sb)->s_qf_names[i]) {
1833 int ret = ext4_quota_on_mount(sb, i);
1836 "EXT4-fs: Cannot turn on journaled "
1837 "quota: error %d\n", ret);
1842 while (es->s_last_orphan) {
1843 struct inode *inode;
1845 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1846 if (IS_ERR(inode)) {
1847 es->s_last_orphan = 0;
1851 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1853 if (inode->i_nlink) {
1855 "%s: truncating inode %lu to %lld bytes\n",
1856 __func__, inode->i_ino, inode->i_size);
1857 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1858 inode->i_ino, inode->i_size);
1859 ext4_truncate(inode);
1863 "%s: deleting unreferenced inode %lu\n",
1864 __func__, inode->i_ino);
1865 jbd_debug(2, "deleting unreferenced inode %lu\n",
1869 iput(inode); /* The delete magic happens here! */
1872 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1875 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1876 sb->s_id, PLURAL(nr_orphans));
1878 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1879 sb->s_id, PLURAL(nr_truncates));
1881 /* Turn quotas off */
1882 for (i = 0; i < MAXQUOTAS; i++) {
1883 if (sb_dqopt(sb)->files[i])
1884 vfs_quota_off(sb, i, 0);
1887 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1890 * Maximal extent format file size.
1891 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1892 * extent format containers, within a sector_t, and within i_blocks
1893 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1894 * so that won't be a limiting factor.
1896 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1898 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1901 loff_t upper_limit = MAX_LFS_FILESIZE;
1903 /* small i_blocks in vfs inode? */
1904 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1906 * CONFIG_LBD is not enabled implies the inode
1907 * i_block represent total blocks in 512 bytes
1908 * 32 == size of vfs inode i_blocks * 8
1910 upper_limit = (1LL << 32) - 1;
1912 /* total blocks in file system block size */
1913 upper_limit >>= (blkbits - 9);
1914 upper_limit <<= blkbits;
1917 /* 32-bit extent-start container, ee_block */
1922 /* Sanity check against vm- & vfs- imposed limits */
1923 if (res > upper_limit)
1930 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1931 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1932 * We need to be 1 filesystem block less than the 2^48 sector limit.
1934 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1936 loff_t res = EXT4_NDIR_BLOCKS;
1939 /* This is calculated to be the largest file size for a
1940 * dense, bitmapped file such that the total number of
1941 * sectors in the file, including data and all indirect blocks,
1942 * does not exceed 2^48 -1
1943 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1944 * total number of 512 bytes blocks of the file
1947 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1949 * !has_huge_files or CONFIG_LBD is not enabled
1950 * implies the inode i_block represent total blocks in
1951 * 512 bytes 32 == size of vfs inode i_blocks * 8
1953 upper_limit = (1LL << 32) - 1;
1955 /* total blocks in file system block size */
1956 upper_limit >>= (bits - 9);
1960 * We use 48 bit ext4_inode i_blocks
1961 * With EXT4_HUGE_FILE_FL set the i_blocks
1962 * represent total number of blocks in
1963 * file system block size
1965 upper_limit = (1LL << 48) - 1;
1969 /* indirect blocks */
1971 /* double indirect blocks */
1972 meta_blocks += 1 + (1LL << (bits-2));
1973 /* tripple indirect blocks */
1974 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1976 upper_limit -= meta_blocks;
1977 upper_limit <<= bits;
1979 res += 1LL << (bits-2);
1980 res += 1LL << (2*(bits-2));
1981 res += 1LL << (3*(bits-2));
1983 if (res > upper_limit)
1986 if (res > MAX_LFS_FILESIZE)
1987 res = MAX_LFS_FILESIZE;
1992 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1993 ext4_fsblk_t logical_sb_block, int nr)
1995 struct ext4_sb_info *sbi = EXT4_SB(sb);
1996 ext4_group_t bg, first_meta_bg;
1999 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2001 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2003 return logical_sb_block + nr + 1;
2004 bg = sbi->s_desc_per_block * nr;
2005 if (ext4_bg_has_super(sb, bg))
2007 return (has_super + ext4_group_first_block_no(sb, bg));
2011 * ext4_get_stripe_size: Get the stripe size.
2012 * @sbi: In memory super block info
2014 * If we have specified it via mount option, then
2015 * use the mount option value. If the value specified at mount time is
2016 * greater than the blocks per group use the super block value.
2017 * If the super block value is greater than blocks per group return 0.
2018 * Allocator needs it be less than blocks per group.
2021 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2023 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2024 unsigned long stripe_width =
2025 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2027 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2028 return sbi->s_stripe;
2030 if (stripe_width <= sbi->s_blocks_per_group)
2031 return stripe_width;
2033 if (stride <= sbi->s_blocks_per_group)
2042 struct attribute attr;
2043 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2044 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2045 const char *, size_t);
2049 static int parse_strtoul(const char *buf,
2050 unsigned long max, unsigned long *value)
2054 while (*buf && isspace(*buf))
2056 *value = simple_strtoul(buf, &endp, 0);
2057 while (*endp && isspace(*endp))
2059 if (*endp || *value > max)
2065 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2066 struct ext4_sb_info *sbi,
2069 return snprintf(buf, PAGE_SIZE, "%llu\n",
2070 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2073 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2074 struct ext4_sb_info *sbi, char *buf)
2076 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2078 return snprintf(buf, PAGE_SIZE, "%lu\n",
2079 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2080 sbi->s_sectors_written_start) >> 1);
2083 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2084 struct ext4_sb_info *sbi, char *buf)
2086 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2088 return snprintf(buf, PAGE_SIZE, "%llu\n",
2089 sbi->s_kbytes_written +
2090 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2091 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2094 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2095 struct ext4_sb_info *sbi,
2096 const char *buf, size_t count)
2100 if (parse_strtoul(buf, 0x40000000, &t))
2103 if (!is_power_of_2(t))
2106 sbi->s_inode_readahead_blks = t;
2110 static ssize_t sbi_ui_show(struct ext4_attr *a,
2111 struct ext4_sb_info *sbi, char *buf)
2113 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2115 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2118 static ssize_t sbi_ui_store(struct ext4_attr *a,
2119 struct ext4_sb_info *sbi,
2120 const char *buf, size_t count)
2122 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2125 if (parse_strtoul(buf, 0xffffffff, &t))
2131 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2132 static struct ext4_attr ext4_attr_##_name = { \
2133 .attr = {.name = __stringify(_name), .mode = _mode }, \
2136 .offset = offsetof(struct ext4_sb_info, _elname), \
2138 #define EXT4_ATTR(name, mode, show, store) \
2139 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2141 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2142 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2143 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2144 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2145 #define ATTR_LIST(name) &ext4_attr_##name.attr
2147 EXT4_RO_ATTR(delayed_allocation_blocks);
2148 EXT4_RO_ATTR(session_write_kbytes);
2149 EXT4_RO_ATTR(lifetime_write_kbytes);
2150 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2151 inode_readahead_blks_store, s_inode_readahead_blks);
2152 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2153 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2154 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2155 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2156 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2157 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2159 static struct attribute *ext4_attrs[] = {
2160 ATTR_LIST(delayed_allocation_blocks),
2161 ATTR_LIST(session_write_kbytes),
2162 ATTR_LIST(lifetime_write_kbytes),
2163 ATTR_LIST(inode_readahead_blks),
2164 ATTR_LIST(mb_stats),
2165 ATTR_LIST(mb_max_to_scan),
2166 ATTR_LIST(mb_min_to_scan),
2167 ATTR_LIST(mb_order2_req),
2168 ATTR_LIST(mb_stream_req),
2169 ATTR_LIST(mb_group_prealloc),
2173 static ssize_t ext4_attr_show(struct kobject *kobj,
2174 struct attribute *attr, char *buf)
2176 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2178 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2180 return a->show ? a->show(a, sbi, buf) : 0;
2183 static ssize_t ext4_attr_store(struct kobject *kobj,
2184 struct attribute *attr,
2185 const char *buf, size_t len)
2187 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2189 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2191 return a->store ? a->store(a, sbi, buf, len) : 0;
2194 static void ext4_sb_release(struct kobject *kobj)
2196 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2198 complete(&sbi->s_kobj_unregister);
2202 static struct sysfs_ops ext4_attr_ops = {
2203 .show = ext4_attr_show,
2204 .store = ext4_attr_store,
2207 static struct kobj_type ext4_ktype = {
2208 .default_attrs = ext4_attrs,
2209 .sysfs_ops = &ext4_attr_ops,
2210 .release = ext4_sb_release,
2213 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2214 __releases(kernel_lock)
2215 __acquires(kernel_lock)
2218 struct buffer_head *bh;
2219 struct ext4_super_block *es = NULL;
2220 struct ext4_sb_info *sbi;
2222 ext4_fsblk_t sb_block = get_sb_block(&data);
2223 ext4_fsblk_t logical_sb_block;
2224 unsigned long offset = 0;
2225 unsigned long journal_devnum = 0;
2226 unsigned long def_mount_opts;
2232 unsigned int db_count;
2234 int needs_recovery, has_huge_files;
2238 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2240 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2244 sbi->s_blockgroup_lock =
2245 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2246 if (!sbi->s_blockgroup_lock) {
2250 sb->s_fs_info = sbi;
2251 sbi->s_mount_opt = 0;
2252 sbi->s_resuid = EXT4_DEF_RESUID;
2253 sbi->s_resgid = EXT4_DEF_RESGID;
2254 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2255 sbi->s_sb_block = sb_block;
2256 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2261 /* Cleanup superblock name */
2262 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2265 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2267 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2272 * The ext4 superblock will not be buffer aligned for other than 1kB
2273 * block sizes. We need to calculate the offset from buffer start.
2275 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2276 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2277 offset = do_div(logical_sb_block, blocksize);
2279 logical_sb_block = sb_block;
2282 if (!(bh = sb_bread(sb, logical_sb_block))) {
2283 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2287 * Note: s_es must be initialized as soon as possible because
2288 * some ext4 macro-instructions depend on its value
2290 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2292 sb->s_magic = le16_to_cpu(es->s_magic);
2293 if (sb->s_magic != EXT4_SUPER_MAGIC)
2295 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2297 /* Set defaults before we parse the mount options */
2298 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2299 if (def_mount_opts & EXT4_DEFM_DEBUG)
2300 set_opt(sbi->s_mount_opt, DEBUG);
2301 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2302 set_opt(sbi->s_mount_opt, GRPID);
2303 if (def_mount_opts & EXT4_DEFM_UID16)
2304 set_opt(sbi->s_mount_opt, NO_UID32);
2305 #ifdef CONFIG_EXT4_FS_XATTR
2306 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2307 set_opt(sbi->s_mount_opt, XATTR_USER);
2309 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2310 if (def_mount_opts & EXT4_DEFM_ACL)
2311 set_opt(sbi->s_mount_opt, POSIX_ACL);
2313 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2314 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2315 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2316 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2317 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2318 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2320 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2321 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2322 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2323 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2325 set_opt(sbi->s_mount_opt, ERRORS_RO);
2327 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2328 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2329 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2330 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2331 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2333 set_opt(sbi->s_mount_opt, BARRIER);
2336 * enable delayed allocation by default
2337 * Use -o nodelalloc to turn it off
2339 set_opt(sbi->s_mount_opt, DELALLOC);
2342 if (!parse_options((char *) data, sb, &journal_devnum,
2343 &journal_ioprio, NULL, 0))
2346 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2347 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2349 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2350 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2351 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2352 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2354 "EXT4-fs warning: feature flags set on rev 0 fs, "
2355 "running e2fsck is recommended\n");
2358 * Check feature flags regardless of the revision level, since we
2359 * previously didn't change the revision level when setting the flags,
2360 * so there is a chance incompat flags are set on a rev 0 filesystem.
2362 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2364 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2365 "unsupported optional features (%x).\n", sb->s_id,
2366 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2367 ~EXT4_FEATURE_INCOMPAT_SUPP));
2370 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2371 if (!(sb->s_flags & MS_RDONLY) && features) {
2372 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2373 "unsupported optional features (%x).\n", sb->s_id,
2374 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2375 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2378 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2379 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2380 if (has_huge_files) {
2382 * Large file size enabled file system can only be
2383 * mount if kernel is build with CONFIG_LBD
2385 if (sizeof(root->i_blocks) < sizeof(u64) &&
2386 !(sb->s_flags & MS_RDONLY)) {
2387 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2388 "files cannot be mounted read-write "
2389 "without CONFIG_LBD.\n", sb->s_id);
2393 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2395 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2396 blocksize > EXT4_MAX_BLOCK_SIZE) {
2398 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2399 blocksize, sb->s_id);
2403 if (sb->s_blocksize != blocksize) {
2405 /* Validate the filesystem blocksize */
2406 if (!sb_set_blocksize(sb, blocksize)) {
2407 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2413 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2414 offset = do_div(logical_sb_block, blocksize);
2415 bh = sb_bread(sb, logical_sb_block);
2418 "EXT4-fs: Can't read superblock on 2nd try.\n");
2421 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2423 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2425 "EXT4-fs: Magic mismatch, very weird !\n");
2430 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2432 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2434 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2435 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2436 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2438 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2439 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2440 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2441 (!is_power_of_2(sbi->s_inode_size)) ||
2442 (sbi->s_inode_size > blocksize)) {
2444 "EXT4-fs: unsupported inode size: %d\n",
2448 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2449 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2451 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2452 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2453 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2454 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2455 !is_power_of_2(sbi->s_desc_size)) {
2457 "EXT4-fs: unsupported descriptor size %lu\n",
2462 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2463 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2464 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2465 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2467 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2468 if (sbi->s_inodes_per_block == 0)
2470 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2471 sbi->s_inodes_per_block;
2472 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2474 sbi->s_mount_state = le16_to_cpu(es->s_state);
2475 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2476 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2477 for (i = 0; i < 4; i++)
2478 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2479 sbi->s_def_hash_version = es->s_def_hash_version;
2480 i = le32_to_cpu(es->s_flags);
2481 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2482 sbi->s_hash_unsigned = 3;
2483 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2484 #ifdef __CHAR_UNSIGNED__
2485 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2486 sbi->s_hash_unsigned = 3;
2488 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2493 if (sbi->s_blocks_per_group > blocksize * 8) {
2495 "EXT4-fs: #blocks per group too big: %lu\n",
2496 sbi->s_blocks_per_group);
2499 if (sbi->s_inodes_per_group > blocksize * 8) {
2501 "EXT4-fs: #inodes per group too big: %lu\n",
2502 sbi->s_inodes_per_group);
2506 if (ext4_blocks_count(es) >
2507 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2508 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2509 " too large to mount safely\n", sb->s_id);
2510 if (sizeof(sector_t) < 8)
2511 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2516 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2519 /* check blocks count against device size */
2520 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2521 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2522 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2523 "exceeds size of device (%llu blocks)\n",
2524 ext4_blocks_count(es), blocks_count);
2529 * It makes no sense for the first data block to be beyond the end
2530 * of the filesystem.
2532 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2533 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2534 "block %u is beyond end of filesystem (%llu)\n",
2535 le32_to_cpu(es->s_first_data_block),
2536 ext4_blocks_count(es));
2539 blocks_count = (ext4_blocks_count(es) -
2540 le32_to_cpu(es->s_first_data_block) +
2541 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2542 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2543 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2544 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2545 "(block count %llu, first data block %u, "
2546 "blocks per group %lu)\n", sbi->s_groups_count,
2547 ext4_blocks_count(es),
2548 le32_to_cpu(es->s_first_data_block),
2549 EXT4_BLOCKS_PER_GROUP(sb));
2552 sbi->s_groups_count = blocks_count;
2553 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2554 EXT4_DESC_PER_BLOCK(sb);
2555 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2557 if (sbi->s_group_desc == NULL) {
2558 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2562 #ifdef CONFIG_PROC_FS
2564 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2567 bgl_lock_init(sbi->s_blockgroup_lock);
2569 for (i = 0; i < db_count; i++) {
2570 block = descriptor_loc(sb, logical_sb_block, i);
2571 sbi->s_group_desc[i] = sb_bread(sb, block);
2572 if (!sbi->s_group_desc[i]) {
2573 printk(KERN_ERR "EXT4-fs: "
2574 "can't read group descriptor %d\n", i);
2579 if (!ext4_check_descriptors(sb)) {
2580 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2583 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2584 if (!ext4_fill_flex_info(sb)) {
2586 "EXT4-fs: unable to initialize "
2587 "flex_bg meta info!\n");
2591 sbi->s_gdb_count = db_count;
2592 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2593 spin_lock_init(&sbi->s_next_gen_lock);
2595 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2596 ext4_count_free_blocks(sb));
2598 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2599 ext4_count_free_inodes(sb));
2602 err = percpu_counter_init(&sbi->s_dirs_counter,
2603 ext4_count_dirs(sb));
2606 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2609 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2613 sbi->s_stripe = ext4_get_stripe_size(sbi);
2616 * set up enough so that it can read an inode
2618 sb->s_op = &ext4_sops;
2619 sb->s_export_op = &ext4_export_ops;
2620 sb->s_xattr = ext4_xattr_handlers;
2622 sb->s_qcop = &ext4_qctl_operations;
2623 sb->dq_op = &ext4_quota_operations;
2625 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2629 needs_recovery = (es->s_last_orphan != 0 ||
2630 EXT4_HAS_INCOMPAT_FEATURE(sb,
2631 EXT4_FEATURE_INCOMPAT_RECOVER));
2634 * The first inode we look at is the journal inode. Don't try
2635 * root first: it may be modified in the journal!
2637 if (!test_opt(sb, NOLOAD) &&
2638 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2639 if (ext4_load_journal(sb, es, journal_devnum))
2641 if (!(sb->s_flags & MS_RDONLY) &&
2642 EXT4_SB(sb)->s_journal->j_failed_commit) {
2643 printk(KERN_CRIT "EXT4-fs error (device %s): "
2644 "ext4_fill_super: Journal transaction "
2645 "%u is corrupt\n", sb->s_id,
2646 EXT4_SB(sb)->s_journal->j_failed_commit);
2647 if (test_opt(sb, ERRORS_RO)) {
2649 "Mounting filesystem read-only\n");
2650 sb->s_flags |= MS_RDONLY;
2651 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2652 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2654 if (test_opt(sb, ERRORS_PANIC)) {
2655 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2656 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2657 ext4_commit_super(sb, 1);
2661 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2662 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2663 printk(KERN_ERR "EXT4-fs: required journal recovery "
2664 "suppressed and not mounted read-only\n");
2667 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2668 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2669 sbi->s_journal = NULL;
2674 if (ext4_blocks_count(es) > 0xffffffffULL &&
2675 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2676 JBD2_FEATURE_INCOMPAT_64BIT)) {
2677 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2681 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2682 jbd2_journal_set_features(sbi->s_journal,
2683 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2684 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2685 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2686 jbd2_journal_set_features(sbi->s_journal,
2687 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2688 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2689 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2691 jbd2_journal_clear_features(sbi->s_journal,
2692 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2693 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2696 /* We have now updated the journal if required, so we can
2697 * validate the data journaling mode. */
2698 switch (test_opt(sb, DATA_FLAGS)) {
2700 /* No mode set, assume a default based on the journal
2701 * capabilities: ORDERED_DATA if the journal can
2702 * cope, else JOURNAL_DATA
2704 if (jbd2_journal_check_available_features
2705 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2706 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2708 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2711 case EXT4_MOUNT_ORDERED_DATA:
2712 case EXT4_MOUNT_WRITEBACK_DATA:
2713 if (!jbd2_journal_check_available_features
2714 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2715 printk(KERN_ERR "EXT4-fs: Journal does not support "
2716 "requested data journaling mode\n");
2722 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2726 if (test_opt(sb, NOBH)) {
2727 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2728 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2729 "its supported only with writeback mode\n");
2730 clear_opt(sbi->s_mount_opt, NOBH);
2734 * The jbd2_journal_load will have done any necessary log recovery,
2735 * so we can safely mount the rest of the filesystem now.
2738 root = ext4_iget(sb, EXT4_ROOT_INO);
2740 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2741 ret = PTR_ERR(root);
2744 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2746 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2749 sb->s_root = d_alloc_root(root);
2751 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2757 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2759 /* determine the minimum size of new large inodes, if present */
2760 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2761 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2762 EXT4_GOOD_OLD_INODE_SIZE;
2763 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2764 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2765 if (sbi->s_want_extra_isize <
2766 le16_to_cpu(es->s_want_extra_isize))
2767 sbi->s_want_extra_isize =
2768 le16_to_cpu(es->s_want_extra_isize);
2769 if (sbi->s_want_extra_isize <
2770 le16_to_cpu(es->s_min_extra_isize))
2771 sbi->s_want_extra_isize =
2772 le16_to_cpu(es->s_min_extra_isize);
2775 /* Check if enough inode space is available */
2776 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2777 sbi->s_inode_size) {
2778 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2779 EXT4_GOOD_OLD_INODE_SIZE;
2780 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2784 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2785 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2786 "requested data journaling mode\n");
2787 clear_opt(sbi->s_mount_opt, DELALLOC);
2788 } else if (test_opt(sb, DELALLOC))
2789 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2792 err = ext4_mb_init(sb, needs_recovery);
2794 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2799 sbi->s_kobj.kset = ext4_kset;
2800 init_completion(&sbi->s_kobj_unregister);
2801 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2804 ext4_mb_release(sb);
2805 ext4_ext_release(sb);
2810 * akpm: core read_super() calls in here with the superblock locked.
2811 * That deadlocks, because orphan cleanup needs to lock the superblock
2812 * in numerous places. Here we just pop the lock - it's relatively
2813 * harmless, because we are now ready to accept write_super() requests,
2814 * and aviro says that's the only reason for hanging onto the
2817 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2818 ext4_orphan_cleanup(sb, es);
2819 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2820 if (needs_recovery) {
2821 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2822 ext4_mark_recovery_complete(sb, es);
2824 if (EXT4_SB(sb)->s_journal) {
2825 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2826 descr = " journalled data mode";
2827 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2828 descr = " ordered data mode";
2830 descr = " writeback data mode";
2832 descr = "out journal";
2834 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2842 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2847 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2848 if (sbi->s_journal) {
2849 jbd2_journal_destroy(sbi->s_journal);
2850 sbi->s_journal = NULL;
2853 if (sbi->s_flex_groups) {
2854 if (is_vmalloc_addr(sbi->s_flex_groups))
2855 vfree(sbi->s_flex_groups);
2857 kfree(sbi->s_flex_groups);
2859 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2860 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2861 percpu_counter_destroy(&sbi->s_dirs_counter);
2862 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2864 for (i = 0; i < db_count; i++)
2865 brelse(sbi->s_group_desc[i]);
2866 kfree(sbi->s_group_desc);
2869 remove_proc_entry(sb->s_id, ext4_proc_root);
2872 for (i = 0; i < MAXQUOTAS; i++)
2873 kfree(sbi->s_qf_names[i]);
2875 ext4_blkdev_remove(sbi);
2878 sb->s_fs_info = NULL;
2885 * Setup any per-fs journal parameters now. We'll do this both on
2886 * initial mount, once the journal has been initialised but before we've
2887 * done any recovery; and again on any subsequent remount.
2889 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2891 struct ext4_sb_info *sbi = EXT4_SB(sb);
2893 journal->j_commit_interval = sbi->s_commit_interval;
2894 journal->j_min_batch_time = sbi->s_min_batch_time;
2895 journal->j_max_batch_time = sbi->s_max_batch_time;
2897 spin_lock(&journal->j_state_lock);
2898 if (test_opt(sb, BARRIER))
2899 journal->j_flags |= JBD2_BARRIER;
2901 journal->j_flags &= ~JBD2_BARRIER;
2902 if (test_opt(sb, DATA_ERR_ABORT))
2903 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2905 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2906 spin_unlock(&journal->j_state_lock);
2909 static journal_t *ext4_get_journal(struct super_block *sb,
2910 unsigned int journal_inum)
2912 struct inode *journal_inode;
2915 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2917 /* First, test for the existence of a valid inode on disk. Bad
2918 * things happen if we iget() an unused inode, as the subsequent
2919 * iput() will try to delete it. */
2921 journal_inode = ext4_iget(sb, journal_inum);
2922 if (IS_ERR(journal_inode)) {
2923 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2926 if (!journal_inode->i_nlink) {
2927 make_bad_inode(journal_inode);
2928 iput(journal_inode);
2929 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2933 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2934 journal_inode, journal_inode->i_size);
2935 if (!S_ISREG(journal_inode->i_mode)) {
2936 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2937 iput(journal_inode);
2941 journal = jbd2_journal_init_inode(journal_inode);
2943 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2944 iput(journal_inode);
2947 journal->j_private = sb;
2948 ext4_init_journal_params(sb, journal);
2952 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2955 struct buffer_head *bh;
2959 int hblock, blocksize;
2960 ext4_fsblk_t sb_block;
2961 unsigned long offset;
2962 struct ext4_super_block *es;
2963 struct block_device *bdev;
2965 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2967 bdev = ext4_blkdev_get(j_dev);
2971 if (bd_claim(bdev, sb)) {
2973 "EXT4-fs: failed to claim external journal device.\n");
2974 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2978 blocksize = sb->s_blocksize;
2979 hblock = bdev_hardsect_size(bdev);
2980 if (blocksize < hblock) {
2982 "EXT4-fs: blocksize too small for journal device.\n");
2986 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2987 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2988 set_blocksize(bdev, blocksize);
2989 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2990 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2991 "external journal\n");
2995 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2996 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2997 !(le32_to_cpu(es->s_feature_incompat) &
2998 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2999 printk(KERN_ERR "EXT4-fs: external journal has "
3000 "bad superblock\n");
3005 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3006 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3011 len = ext4_blocks_count(es);
3012 start = sb_block + 1;
3013 brelse(bh); /* we're done with the superblock */
3015 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3016 start, len, blocksize);
3018 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3021 journal->j_private = sb;
3022 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3023 wait_on_buffer(journal->j_sb_buffer);
3024 if (!buffer_uptodate(journal->j_sb_buffer)) {
3025 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3028 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3029 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3030 "user (unsupported) - %d\n",
3031 be32_to_cpu(journal->j_superblock->s_nr_users));
3034 EXT4_SB(sb)->journal_bdev = bdev;
3035 ext4_init_journal_params(sb, journal);
3038 jbd2_journal_destroy(journal);
3040 ext4_blkdev_put(bdev);
3044 static int ext4_load_journal(struct super_block *sb,
3045 struct ext4_super_block *es,
3046 unsigned long journal_devnum)
3049 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3052 int really_read_only;
3054 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3056 if (journal_devnum &&
3057 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3058 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3059 "numbers have changed\n");
3060 journal_dev = new_decode_dev(journal_devnum);
3062 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3064 really_read_only = bdev_read_only(sb->s_bdev);
3067 * Are we loading a blank journal or performing recovery after a
3068 * crash? For recovery, we need to check in advance whether we
3069 * can get read-write access to the device.
3072 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3073 if (sb->s_flags & MS_RDONLY) {
3074 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3075 "required on readonly filesystem.\n");
3076 if (really_read_only) {
3077 printk(KERN_ERR "EXT4-fs: write access "
3078 "unavailable, cannot proceed.\n");
3081 printk(KERN_INFO "EXT4-fs: write access will "
3082 "be enabled during recovery.\n");
3086 if (journal_inum && journal_dev) {
3087 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3088 "and inode journals!\n");
3093 if (!(journal = ext4_get_journal(sb, journal_inum)))
3096 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3100 if (journal->j_flags & JBD2_BARRIER)
3101 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3103 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3105 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3106 err = jbd2_journal_update_format(journal);
3108 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3109 jbd2_journal_destroy(journal);
3114 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3115 err = jbd2_journal_wipe(journal, !really_read_only);
3117 err = jbd2_journal_load(journal);
3120 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3121 jbd2_journal_destroy(journal);
3125 EXT4_SB(sb)->s_journal = journal;
3126 ext4_clear_journal_err(sb, es);
3128 if (journal_devnum &&
3129 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3130 es->s_journal_dev = cpu_to_le32(journal_devnum);
3132 /* Make sure we flush the recovery flag to disk. */
3133 ext4_commit_super(sb, 1);
3139 static int ext4_commit_super(struct super_block *sb, int sync)
3141 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3142 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3147 if (buffer_write_io_error(sbh)) {
3149 * Oh, dear. A previous attempt to write the
3150 * superblock failed. This could happen because the
3151 * USB device was yanked out. Or it could happen to
3152 * be a transient write error and maybe the block will
3153 * be remapped. Nothing we can do but to retry the
3154 * write and hope for the best.
3156 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3157 "superblock detected for %s.\n", sb->s_id);
3158 clear_buffer_write_io_error(sbh);
3159 set_buffer_uptodate(sbh);
3161 es->s_wtime = cpu_to_le32(get_seconds());
3162 es->s_kbytes_written =
3163 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3164 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3165 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3166 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3167 &EXT4_SB(sb)->s_freeblocks_counter));
3168 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3169 &EXT4_SB(sb)->s_freeinodes_counter));
3171 BUFFER_TRACE(sbh, "marking dirty");
3172 mark_buffer_dirty(sbh);
3174 error = sync_dirty_buffer(sbh);
3178 error = buffer_write_io_error(sbh);
3180 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3181 "superblock for %s.\n", sb->s_id);
3182 clear_buffer_write_io_error(sbh);
3183 set_buffer_uptodate(sbh);
3191 * Have we just finished recovery? If so, and if we are mounting (or
3192 * remounting) the filesystem readonly, then we will end up with a
3193 * consistent fs on disk. Record that fact.
3195 static void ext4_mark_recovery_complete(struct super_block *sb,
3196 struct ext4_super_block *es)
3198 journal_t *journal = EXT4_SB(sb)->s_journal;
3200 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3201 BUG_ON(journal != NULL);
3204 jbd2_journal_lock_updates(journal);
3205 if (jbd2_journal_flush(journal) < 0)
3209 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3210 sb->s_flags & MS_RDONLY) {
3211 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3212 ext4_commit_super(sb, 1);
3217 jbd2_journal_unlock_updates(journal);
3221 * If we are mounting (or read-write remounting) a filesystem whose journal
3222 * has recorded an error from a previous lifetime, move that error to the
3223 * main filesystem now.
3225 static void ext4_clear_journal_err(struct super_block *sb,
3226 struct ext4_super_block *es)
3232 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3234 journal = EXT4_SB(sb)->s_journal;
3237 * Now check for any error status which may have been recorded in the
3238 * journal by a prior ext4_error() or ext4_abort()
3241 j_errno = jbd2_journal_errno(journal);
3245 errstr = ext4_decode_error(sb, j_errno, nbuf);
3246 ext4_warning(sb, __func__, "Filesystem error recorded "
3247 "from previous mount: %s", errstr);
3248 ext4_warning(sb, __func__, "Marking fs in need of "
3249 "filesystem check.");
3251 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3252 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3253 ext4_commit_super(sb, 1);
3255 jbd2_journal_clear_err(journal);
3260 * Force the running and committing transactions to commit,
3261 * and wait on the commit.
3263 int ext4_force_commit(struct super_block *sb)
3268 if (sb->s_flags & MS_RDONLY)
3271 journal = EXT4_SB(sb)->s_journal;
3273 ret = ext4_journal_force_commit(journal);
3279 * Ext4 always journals updates to the superblock itself, so we don't
3280 * have to propagate any other updates to the superblock on disk at this
3281 * point if the journalling is enabled.
3283 static void ext4_write_super(struct super_block *sb)
3285 if (EXT4_SB(sb)->s_journal) {
3286 if (mutex_trylock(&sb->s_lock) != 0)
3289 ext4_commit_super(sb, 1);
3293 static int ext4_sync_fs(struct super_block *sb, int wait)
3298 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3299 if (EXT4_SB(sb)->s_journal) {
3300 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3303 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3307 ext4_commit_super(sb, wait);
3313 * LVM calls this function before a (read-only) snapshot is created. This
3314 * gives us a chance to flush the journal completely and mark the fs clean.
3316 static int ext4_freeze(struct super_block *sb)
3321 if (!(sb->s_flags & MS_RDONLY)) {
3322 journal = EXT4_SB(sb)->s_journal;
3325 /* Now we set up the journal barrier. */
3326 jbd2_journal_lock_updates(journal);
3329 * We don't want to clear needs_recovery flag when we
3330 * failed to flush the journal.
3332 error = jbd2_journal_flush(journal);
3337 /* Journal blocked and flushed, clear needs_recovery flag. */
3338 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3339 error = ext4_commit_super(sb, 1);
3345 jbd2_journal_unlock_updates(journal);
3350 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3351 * flag here, even though the filesystem is not technically dirty yet.
3353 static int ext4_unfreeze(struct super_block *sb)
3355 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3357 /* Reser the needs_recovery flag before the fs is unlocked. */
3358 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3359 ext4_commit_super(sb, 1);
3361 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3366 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3368 struct ext4_super_block *es;
3369 struct ext4_sb_info *sbi = EXT4_SB(sb);
3370 ext4_fsblk_t n_blocks_count = 0;
3371 unsigned long old_sb_flags;
3372 struct ext4_mount_options old_opts;
3374 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3380 /* Store the original options */
3381 old_sb_flags = sb->s_flags;
3382 old_opts.s_mount_opt = sbi->s_mount_opt;
3383 old_opts.s_resuid = sbi->s_resuid;
3384 old_opts.s_resgid = sbi->s_resgid;
3385 old_opts.s_commit_interval = sbi->s_commit_interval;
3386 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3387 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3389 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3390 for (i = 0; i < MAXQUOTAS; i++)
3391 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3393 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3394 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3397 * Allow the "check" option to be passed as a remount option.
3399 if (!parse_options(data, sb, NULL, &journal_ioprio,
3400 &n_blocks_count, 1)) {
3405 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3406 ext4_abort(sb, __func__, "Abort forced by user");
3408 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3409 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3413 if (sbi->s_journal) {
3414 ext4_init_journal_params(sb, sbi->s_journal);
3415 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3418 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3419 n_blocks_count > ext4_blocks_count(es)) {
3420 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3425 if (*flags & MS_RDONLY) {
3427 * First of all, the unconditional stuff we have to do
3428 * to disable replay of the journal when we next remount
3430 sb->s_flags |= MS_RDONLY;
3433 * OK, test if we are remounting a valid rw partition
3434 * readonly, and if so set the rdonly flag and then
3435 * mark the partition as valid again.
3437 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3438 (sbi->s_mount_state & EXT4_VALID_FS))
3439 es->s_state = cpu_to_le16(sbi->s_mount_state);
3442 * We have to unlock super so that we can wait for
3445 if (sbi->s_journal) {
3447 ext4_mark_recovery_complete(sb, es);
3452 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3453 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3454 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3455 "remount RDWR because of unsupported "
3456 "optional features (%x).\n", sb->s_id,
3457 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3458 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3464 * Make sure the group descriptor checksums
3465 * are sane. If they aren't, refuse to
3468 for (g = 0; g < sbi->s_groups_count; g++) {
3469 struct ext4_group_desc *gdp =
3470 ext4_get_group_desc(sb, g, NULL);
3472 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3474 "EXT4-fs: ext4_remount: "
3475 "Checksum for group %u failed (%u!=%u)\n",
3476 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3477 le16_to_cpu(gdp->bg_checksum));
3484 * If we have an unprocessed orphan list hanging
3485 * around from a previously readonly bdev mount,
3486 * require a full umount/remount for now.
3488 if (es->s_last_orphan) {
3489 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3490 "remount RDWR because of unprocessed "
3491 "orphan inode list. Please "
3492 "umount/remount instead.\n",
3499 * Mounting a RDONLY partition read-write, so reread
3500 * and store the current valid flag. (It may have
3501 * been changed by e2fsck since we originally mounted
3505 ext4_clear_journal_err(sb, es);
3506 sbi->s_mount_state = le16_to_cpu(es->s_state);
3507 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3509 if (!ext4_setup_super(sb, es, 0))
3510 sb->s_flags &= ~MS_RDONLY;
3513 if (sbi->s_journal == NULL)
3514 ext4_commit_super(sb, 1);
3517 /* Release old quota file names */
3518 for (i = 0; i < MAXQUOTAS; i++)
3519 if (old_opts.s_qf_names[i] &&
3520 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3521 kfree(old_opts.s_qf_names[i]);
3525 sb->s_flags = old_sb_flags;
3526 sbi->s_mount_opt = old_opts.s_mount_opt;
3527 sbi->s_resuid = old_opts.s_resuid;
3528 sbi->s_resgid = old_opts.s_resgid;
3529 sbi->s_commit_interval = old_opts.s_commit_interval;
3530 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3531 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3533 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3534 for (i = 0; i < MAXQUOTAS; i++) {
3535 if (sbi->s_qf_names[i] &&
3536 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3537 kfree(sbi->s_qf_names[i]);
3538 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3544 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3546 struct super_block *sb = dentry->d_sb;
3547 struct ext4_sb_info *sbi = EXT4_SB(sb);
3548 struct ext4_super_block *es = sbi->s_es;
3551 if (test_opt(sb, MINIX_DF)) {
3552 sbi->s_overhead_last = 0;
3553 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3554 ext4_group_t ngroups = sbi->s_groups_count, i;
3555 ext4_fsblk_t overhead = 0;
3559 * Compute the overhead (FS structures). This is constant
3560 * for a given filesystem unless the number of block groups
3561 * changes so we cache the previous value until it does.
3565 * All of the blocks before first_data_block are
3568 overhead = le32_to_cpu(es->s_first_data_block);
3571 * Add the overhead attributed to the superblock and
3572 * block group descriptors. If the sparse superblocks
3573 * feature is turned on, then not all groups have this.
3575 for (i = 0; i < ngroups; i++) {
3576 overhead += ext4_bg_has_super(sb, i) +
3577 ext4_bg_num_gdb(sb, i);
3582 * Every block group has an inode bitmap, a block
3583 * bitmap, and an inode table.
3585 overhead += ngroups * (2 + sbi->s_itb_per_group);
3586 sbi->s_overhead_last = overhead;
3588 sbi->s_blocks_last = ext4_blocks_count(es);
3591 buf->f_type = EXT4_SUPER_MAGIC;
3592 buf->f_bsize = sb->s_blocksize;
3593 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3594 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3595 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3596 ext4_free_blocks_count_set(es, buf->f_bfree);
3597 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3598 if (buf->f_bfree < ext4_r_blocks_count(es))
3600 buf->f_files = le32_to_cpu(es->s_inodes_count);
3601 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3602 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3603 buf->f_namelen = EXT4_NAME_LEN;
3604 fsid = le64_to_cpup((void *)es->s_uuid) ^
3605 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3606 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3607 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3611 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3612 * is locked for write. Otherwise the are possible deadlocks:
3613 * Process 1 Process 2
3614 * ext4_create() quota_sync()
3615 * jbd2_journal_start() write_dquot()
3616 * vfs_dq_init() down(dqio_mutex)
3617 * down(dqio_mutex) jbd2_journal_start()
3623 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3625 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3628 static int ext4_write_dquot(struct dquot *dquot)
3632 struct inode *inode;
3634 inode = dquot_to_inode(dquot);
3635 handle = ext4_journal_start(inode,
3636 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3638 return PTR_ERR(handle);
3639 ret = dquot_commit(dquot);
3640 err = ext4_journal_stop(handle);
3646 static int ext4_acquire_dquot(struct dquot *dquot)
3651 handle = ext4_journal_start(dquot_to_inode(dquot),
3652 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3654 return PTR_ERR(handle);
3655 ret = dquot_acquire(dquot);
3656 err = ext4_journal_stop(handle);
3662 static int ext4_release_dquot(struct dquot *dquot)
3667 handle = ext4_journal_start(dquot_to_inode(dquot),
3668 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3669 if (IS_ERR(handle)) {
3670 /* Release dquot anyway to avoid endless cycle in dqput() */
3671 dquot_release(dquot);
3672 return PTR_ERR(handle);
3674 ret = dquot_release(dquot);
3675 err = ext4_journal_stop(handle);
3681 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3683 /* Are we journaling quotas? */
3684 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3685 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3686 dquot_mark_dquot_dirty(dquot);
3687 return ext4_write_dquot(dquot);
3689 return dquot_mark_dquot_dirty(dquot);
3693 static int ext4_write_info(struct super_block *sb, int type)
3698 /* Data block + inode block */
3699 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3701 return PTR_ERR(handle);
3702 ret = dquot_commit_info(sb, type);
3703 err = ext4_journal_stop(handle);
3710 * Turn on quotas during mount time - we need to find
3711 * the quota file and such...
3713 static int ext4_quota_on_mount(struct super_block *sb, int type)
3715 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3716 EXT4_SB(sb)->s_jquota_fmt, type);
3720 * Standard function to be called on quota_on
3722 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3723 char *name, int remount)
3728 if (!test_opt(sb, QUOTA))
3730 /* When remounting, no checks are needed and in fact, name is NULL */
3732 return vfs_quota_on(sb, type, format_id, name, remount);
3734 err = kern_path(name, LOOKUP_FOLLOW, &path);
3738 /* Quotafile not on the same filesystem? */
3739 if (path.mnt->mnt_sb != sb) {
3743 /* Journaling quota? */
3744 if (EXT4_SB(sb)->s_qf_names[type]) {
3745 /* Quotafile not in fs root? */
3746 if (path.dentry->d_parent != sb->s_root)
3748 "EXT4-fs: Quota file not on filesystem root. "
3749 "Journaled quota will not work.\n");
3753 * When we journal data on quota file, we have to flush journal to see
3754 * all updates to the file when we bypass pagecache...
3756 if (EXT4_SB(sb)->s_journal &&
3757 ext4_should_journal_data(path.dentry->d_inode)) {
3759 * We don't need to lock updates but journal_flush() could
3760 * otherwise be livelocked...
3762 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3763 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3764 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3771 err = vfs_quota_on_path(sb, type, format_id, &path);
3776 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3777 * acquiring the locks... As quota files are never truncated and quota code
3778 * itself serializes the operations (and noone else should touch the files)
3779 * we don't have to be afraid of races */
3780 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3781 size_t len, loff_t off)
3783 struct inode *inode = sb_dqopt(sb)->files[type];
3784 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3786 int offset = off & (sb->s_blocksize - 1);
3789 struct buffer_head *bh;
3790 loff_t i_size = i_size_read(inode);
3794 if (off+len > i_size)
3797 while (toread > 0) {
3798 tocopy = sb->s_blocksize - offset < toread ?
3799 sb->s_blocksize - offset : toread;
3800 bh = ext4_bread(NULL, inode, blk, 0, &err);
3803 if (!bh) /* A hole? */
3804 memset(data, 0, tocopy);
3806 memcpy(data, bh->b_data+offset, tocopy);
3816 /* Write to quotafile (we know the transaction is already started and has
3817 * enough credits) */
3818 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3819 const char *data, size_t len, loff_t off)
3821 struct inode *inode = sb_dqopt(sb)->files[type];
3822 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3824 int offset = off & (sb->s_blocksize - 1);
3826 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3827 size_t towrite = len;
3828 struct buffer_head *bh;
3829 handle_t *handle = journal_current_handle();
3831 if (EXT4_SB(sb)->s_journal && !handle) {
3832 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3833 " cancelled because transaction is not started.\n",
3834 (unsigned long long)off, (unsigned long long)len);
3837 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3838 while (towrite > 0) {
3839 tocopy = sb->s_blocksize - offset < towrite ?
3840 sb->s_blocksize - offset : towrite;
3841 bh = ext4_bread(handle, inode, blk, 1, &err);
3844 if (journal_quota) {
3845 err = ext4_journal_get_write_access(handle, bh);
3852 memcpy(bh->b_data+offset, data, tocopy);
3853 flush_dcache_page(bh->b_page);
3856 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3858 /* Always do at least ordered writes for quotas */
3859 err = ext4_jbd2_file_inode(handle, inode);
3860 mark_buffer_dirty(bh);
3871 if (len == towrite) {
3872 mutex_unlock(&inode->i_mutex);
3875 if (inode->i_size < off+len-towrite) {
3876 i_size_write(inode, off+len-towrite);
3877 EXT4_I(inode)->i_disksize = inode->i_size;
3879 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3880 ext4_mark_inode_dirty(handle, inode);
3881 mutex_unlock(&inode->i_mutex);
3882 return len - towrite;
3887 static int ext4_get_sb(struct file_system_type *fs_type,
3888 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3890 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3893 static struct file_system_type ext4_fs_type = {
3894 .owner = THIS_MODULE,
3896 .get_sb = ext4_get_sb,
3897 .kill_sb = kill_block_super,
3898 .fs_flags = FS_REQUIRES_DEV,
3901 #ifdef CONFIG_EXT4DEV_COMPAT
3902 static int ext4dev_get_sb(struct file_system_type *fs_type,
3903 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3905 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3906 "to mount using ext4\n");
3907 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3908 "will go away by 2.6.31\n");
3909 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3912 static struct file_system_type ext4dev_fs_type = {
3913 .owner = THIS_MODULE,
3915 .get_sb = ext4dev_get_sb,
3916 .kill_sb = kill_block_super,
3917 .fs_flags = FS_REQUIRES_DEV,
3919 MODULE_ALIAS("ext4dev");
3922 static int __init init_ext4_fs(void)
3926 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3929 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3930 err = init_ext4_mballoc();
3934 err = init_ext4_xattr();
3937 err = init_inodecache();
3940 err = register_filesystem(&ext4_fs_type);
3943 #ifdef CONFIG_EXT4DEV_COMPAT
3944 err = register_filesystem(&ext4dev_fs_type);
3946 unregister_filesystem(&ext4_fs_type);
3952 destroy_inodecache();
3956 exit_ext4_mballoc();
3960 static void __exit exit_ext4_fs(void)
3962 unregister_filesystem(&ext4_fs_type);
3963 #ifdef CONFIG_EXT4DEV_COMPAT
3964 unregister_filesystem(&ext4dev_fs_type);
3966 destroy_inodecache();
3968 exit_ext4_mballoc();
3969 remove_proc_entry("fs/ext4", NULL);
3970 kset_unregister(ext4_kset);
3973 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3974 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3975 MODULE_LICENSE("GPL");
3976 module_init(init_ext4_fs)
3977 module_exit(exit_ext4_fs)