2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
50 struct proc_dir_entry *ext4_proc_root;
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
56 static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
96 __u32 ext4_free_blks_count(struct super_block *sb,
97 struct ext4_group_desc *bg)
99 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
104 __u32 ext4_free_inodes_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
107 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
112 __u32 ext4_used_dirs_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
115 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
120 __u32 ext4_itable_unused_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
123 return le16_to_cpu(bg->bg_itable_unused_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
128 void ext4_block_bitmap_set(struct super_block *sb,
129 struct ext4_group_desc *bg, ext4_fsblk_t blk)
131 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
132 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
133 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
136 void ext4_inode_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
139 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
144 void ext4_inode_table_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
147 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
152 void ext4_free_blks_set(struct super_block *sb,
153 struct ext4_group_desc *bg, __u32 count)
155 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
160 void ext4_free_inodes_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
163 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
168 void ext4_used_dirs_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
171 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
176 void ext4_itable_unused_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
179 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
185 * Wrappers for jbd2_journal_start/end.
187 * The only special thing we need to do here is to make sure that all
188 * journal_end calls result in the superblock being marked dirty, so
189 * that sync() will call the filesystem's write_super callback if
192 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
196 if (sb->s_flags & MS_RDONLY)
197 return ERR_PTR(-EROFS);
199 /* Special case here: if the journal has aborted behind our
200 * backs (eg. EIO in the commit thread), then we still need to
201 * take the FS itself readonly cleanly. */
202 journal = EXT4_SB(sb)->s_journal;
204 if (is_journal_aborted(journal)) {
205 ext4_abort(sb, __func__,
206 "Detected aborted journal");
207 return ERR_PTR(-EROFS);
209 return jbd2_journal_start(journal, nblocks);
212 * We're not journaling, return the appropriate indication.
214 current->journal_info = EXT4_NOJOURNAL_HANDLE;
215 return current->journal_info;
219 * The only special thing we need to do here is to make sure that all
220 * jbd2_journal_stop calls result in the superblock being marked dirty, so
221 * that sync() will call the filesystem's write_super callback if
224 int __ext4_journal_stop(const char *where, handle_t *handle)
226 struct super_block *sb;
230 if (!ext4_handle_valid(handle)) {
232 * Do this here since we don't call jbd2_journal_stop() in
235 current->journal_info = NULL;
238 sb = handle->h_transaction->t_journal->j_private;
240 rc = jbd2_journal_stop(handle);
245 __ext4_std_error(sb, where, err);
249 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
250 struct buffer_head *bh, handle_t *handle, int err)
253 const char *errstr = ext4_decode_error(NULL, err, nbuf);
255 BUG_ON(!ext4_handle_valid(handle));
258 BUFFER_TRACE(bh, "abort");
263 if (is_handle_aborted(handle))
266 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
267 caller, errstr, err_fn);
269 jbd2_journal_abort_handle(handle);
272 /* Deal with the reporting of failure conditions on a filesystem such as
273 * inconsistencies detected or read IO failures.
275 * On ext2, we can store the error state of the filesystem in the
276 * superblock. That is not possible on ext4, because we may have other
277 * write ordering constraints on the superblock which prevent us from
278 * writing it out straight away; and given that the journal is about to
279 * be aborted, we can't rely on the current, or future, transactions to
280 * write out the superblock safely.
282 * We'll just use the jbd2_journal_abort() error code to record an error in
283 * the journal instead. On recovery, the journal will compain about
284 * that error until we've noted it down and cleared it.
287 static void ext4_handle_error(struct super_block *sb)
289 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
291 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
292 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
294 if (sb->s_flags & MS_RDONLY)
297 if (!test_opt(sb, ERRORS_CONT)) {
298 journal_t *journal = EXT4_SB(sb)->s_journal;
300 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
302 jbd2_journal_abort(journal, -EIO);
304 if (test_opt(sb, ERRORS_RO)) {
305 printk(KERN_CRIT "Remounting filesystem read-only\n");
306 sb->s_flags |= MS_RDONLY;
308 ext4_commit_super(sb, es, 1);
309 if (test_opt(sb, ERRORS_PANIC))
310 panic("EXT4-fs (device %s): panic forced after error\n",
314 void ext4_error(struct super_block *sb, const char *function,
315 const char *fmt, ...)
320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
325 ext4_handle_error(sb);
328 static const char *ext4_decode_error(struct super_block *sb, int errno,
335 errstr = "IO failure";
338 errstr = "Out of memory";
341 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
342 errstr = "Journal has aborted";
344 errstr = "Readonly filesystem";
347 /* If the caller passed in an extra buffer for unknown
348 * errors, textualise them now. Else we just return
351 /* Check for truncated error codes... */
352 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
361 /* __ext4_std_error decodes expected errors from journaling functions
362 * automatically and invokes the appropriate error response. */
364 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
369 /* Special case: if the error is EROFS, and we're not already
370 * inside a transaction, then there's really no point in logging
372 if (errno == -EROFS && journal_current_handle() == NULL &&
373 (sb->s_flags & MS_RDONLY))
376 errstr = ext4_decode_error(sb, errno, nbuf);
377 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
378 sb->s_id, function, errstr);
380 ext4_handle_error(sb);
384 * ext4_abort is a much stronger failure handler than ext4_error. The
385 * abort function may be used to deal with unrecoverable failures such
386 * as journal IO errors or ENOMEM at a critical moment in log management.
388 * We unconditionally force the filesystem into an ABORT|READONLY state,
389 * unless the error response on the fs has been set to panic in which
390 * case we take the easy way out and panic immediately.
393 void ext4_abort(struct super_block *sb, const char *function,
394 const char *fmt, ...)
398 printk(KERN_CRIT "ext4_abort called.\n");
401 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
406 if (test_opt(sb, ERRORS_PANIC))
407 panic("EXT4-fs panic from previous error\n");
409 if (sb->s_flags & MS_RDONLY)
412 printk(KERN_CRIT "Remounting filesystem read-only\n");
413 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
414 sb->s_flags |= MS_RDONLY;
415 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
416 if (EXT4_SB(sb)->s_journal)
417 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
420 void ext4_warning(struct super_block *sb, const char *function,
421 const char *fmt, ...)
426 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
433 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
434 const char *function, const char *fmt, ...)
439 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
442 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
447 if (test_opt(sb, ERRORS_CONT)) {
448 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
449 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
450 ext4_commit_super(sb, es, 0);
453 ext4_unlock_group(sb, grp);
454 ext4_handle_error(sb);
456 * We only get here in the ERRORS_RO case; relocking the group
457 * may be dangerous, but nothing bad will happen since the
458 * filesystem will have already been marked read/only and the
459 * journal has been aborted. We return 1 as a hint to callers
460 * who might what to use the return value from
461 * ext4_grp_locked_error() to distinguish beween the
462 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
463 * aggressively from the ext4 function in question, with a
464 * more appropriate error code.
466 ext4_lock_group(sb, grp);
471 void ext4_update_dynamic_rev(struct super_block *sb)
473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
475 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
478 ext4_warning(sb, __func__,
479 "updating to rev %d because of new feature flag, "
480 "running e2fsck is recommended",
483 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
484 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
485 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
486 /* leave es->s_feature_*compat flags alone */
487 /* es->s_uuid will be set by e2fsck if empty */
490 * The rest of the superblock fields should be zero, and if not it
491 * means they are likely already in use, so leave them alone. We
492 * can leave it up to e2fsck to clean up any inconsistencies there.
497 * Open the external journal device
499 static struct block_device *ext4_blkdev_get(dev_t dev)
501 struct block_device *bdev;
502 char b[BDEVNAME_SIZE];
504 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
510 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
511 __bdevname(dev, b), PTR_ERR(bdev));
516 * Release the journal device
518 static int ext4_blkdev_put(struct block_device *bdev)
521 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
524 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
526 struct block_device *bdev;
529 bdev = sbi->journal_bdev;
531 ret = ext4_blkdev_put(bdev);
532 sbi->journal_bdev = NULL;
537 static inline struct inode *orphan_list_entry(struct list_head *l)
539 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
542 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
546 printk(KERN_ERR "sb orphan head is %d\n",
547 le32_to_cpu(sbi->s_es->s_last_orphan));
549 printk(KERN_ERR "sb_info orphan list:\n");
550 list_for_each(l, &sbi->s_orphan) {
551 struct inode *inode = orphan_list_entry(l);
553 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
554 inode->i_sb->s_id, inode->i_ino, inode,
555 inode->i_mode, inode->i_nlink,
560 static void ext4_put_super(struct super_block *sb)
562 struct ext4_sb_info *sbi = EXT4_SB(sb);
563 struct ext4_super_block *es = sbi->s_es;
567 ext4_ext_release(sb);
568 ext4_xattr_put_super(sb);
569 if (sbi->s_journal) {
570 err = jbd2_journal_destroy(sbi->s_journal);
571 sbi->s_journal = NULL;
573 ext4_abort(sb, __func__,
574 "Couldn't clean up the journal");
576 if (!(sb->s_flags & MS_RDONLY)) {
577 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
578 es->s_state = cpu_to_le16(sbi->s_mount_state);
579 ext4_commit_super(sb, es, 1);
582 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
583 remove_proc_entry(sb->s_id, ext4_proc_root);
586 for (i = 0; i < sbi->s_gdb_count; i++)
587 brelse(sbi->s_group_desc[i]);
588 kfree(sbi->s_group_desc);
589 kfree(sbi->s_flex_groups);
590 percpu_counter_destroy(&sbi->s_freeblocks_counter);
591 percpu_counter_destroy(&sbi->s_freeinodes_counter);
592 percpu_counter_destroy(&sbi->s_dirs_counter);
593 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
596 for (i = 0; i < MAXQUOTAS; i++)
597 kfree(sbi->s_qf_names[i]);
600 /* Debugging code just in case the in-memory inode orphan list
601 * isn't empty. The on-disk one can be non-empty if we've
602 * detected an error and taken the fs readonly, but the
603 * in-memory list had better be clean by this point. */
604 if (!list_empty(&sbi->s_orphan))
605 dump_orphan_list(sb, sbi);
606 J_ASSERT(list_empty(&sbi->s_orphan));
608 invalidate_bdev(sb->s_bdev);
609 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
611 * Invalidate the journal device's buffers. We don't want them
612 * floating about in memory - the physical journal device may
613 * hotswapped, and it breaks the `ro-after' testing code.
615 sync_blockdev(sbi->journal_bdev);
616 invalidate_bdev(sbi->journal_bdev);
617 ext4_blkdev_remove(sbi);
619 sb->s_fs_info = NULL;
624 static struct kmem_cache *ext4_inode_cachep;
627 * Called inside transaction, so use GFP_NOFS
629 static struct inode *ext4_alloc_inode(struct super_block *sb)
631 struct ext4_inode_info *ei;
633 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
636 #ifdef CONFIG_EXT4_FS_POSIX_ACL
637 ei->i_acl = EXT4_ACL_NOT_CACHED;
638 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
640 ei->vfs_inode.i_version = 1;
641 ei->vfs_inode.i_data.writeback_index = 0;
642 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
643 INIT_LIST_HEAD(&ei->i_prealloc_list);
644 spin_lock_init(&ei->i_prealloc_lock);
646 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
647 * therefore it can be null here. Don't check it, just initialize
650 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
651 ei->i_reserved_data_blocks = 0;
652 ei->i_reserved_meta_blocks = 0;
653 ei->i_allocated_meta_blocks = 0;
654 ei->i_delalloc_reserved_flag = 0;
655 spin_lock_init(&(ei->i_block_reservation_lock));
656 return &ei->vfs_inode;
659 static void ext4_destroy_inode(struct inode *inode)
661 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
662 printk("EXT4 Inode %p: orphan list check failed!\n",
664 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
665 EXT4_I(inode), sizeof(struct ext4_inode_info),
669 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
672 static void init_once(void *foo)
674 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
676 INIT_LIST_HEAD(&ei->i_orphan);
677 #ifdef CONFIG_EXT4_FS_XATTR
678 init_rwsem(&ei->xattr_sem);
680 init_rwsem(&ei->i_data_sem);
681 inode_init_once(&ei->vfs_inode);
684 static int init_inodecache(void)
686 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
687 sizeof(struct ext4_inode_info),
688 0, (SLAB_RECLAIM_ACCOUNT|
691 if (ext4_inode_cachep == NULL)
696 static void destroy_inodecache(void)
698 kmem_cache_destroy(ext4_inode_cachep);
701 static void ext4_clear_inode(struct inode *inode)
703 #ifdef CONFIG_EXT4_FS_POSIX_ACL
704 if (EXT4_I(inode)->i_acl &&
705 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
706 posix_acl_release(EXT4_I(inode)->i_acl);
707 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
709 if (EXT4_I(inode)->i_default_acl &&
710 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
711 posix_acl_release(EXT4_I(inode)->i_default_acl);
712 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
715 ext4_discard_preallocations(inode);
716 if (EXT4_JOURNAL(inode))
717 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
718 &EXT4_I(inode)->jinode);
721 static inline void ext4_show_quota_options(struct seq_file *seq,
722 struct super_block *sb)
724 #if defined(CONFIG_QUOTA)
725 struct ext4_sb_info *sbi = EXT4_SB(sb);
727 if (sbi->s_jquota_fmt)
728 seq_printf(seq, ",jqfmt=%s",
729 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
731 if (sbi->s_qf_names[USRQUOTA])
732 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
734 if (sbi->s_qf_names[GRPQUOTA])
735 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
737 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
738 seq_puts(seq, ",usrquota");
740 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
741 seq_puts(seq, ",grpquota");
747 * - it's set to a non-default value OR
748 * - if the per-sb default is different from the global default
750 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
753 unsigned long def_mount_opts;
754 struct super_block *sb = vfs->mnt_sb;
755 struct ext4_sb_info *sbi = EXT4_SB(sb);
756 struct ext4_super_block *es = sbi->s_es;
758 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
759 def_errors = le16_to_cpu(es->s_errors);
761 if (sbi->s_sb_block != 1)
762 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
763 if (test_opt(sb, MINIX_DF))
764 seq_puts(seq, ",minixdf");
765 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
766 seq_puts(seq, ",grpid");
767 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
768 seq_puts(seq, ",nogrpid");
769 if (sbi->s_resuid != EXT4_DEF_RESUID ||
770 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
771 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
773 if (sbi->s_resgid != EXT4_DEF_RESGID ||
774 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
775 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
777 if (test_opt(sb, ERRORS_RO)) {
778 if (def_errors == EXT4_ERRORS_PANIC ||
779 def_errors == EXT4_ERRORS_CONTINUE) {
780 seq_puts(seq, ",errors=remount-ro");
783 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
784 seq_puts(seq, ",errors=continue");
785 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
786 seq_puts(seq, ",errors=panic");
787 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
788 seq_puts(seq, ",nouid32");
789 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
790 seq_puts(seq, ",debug");
791 if (test_opt(sb, OLDALLOC))
792 seq_puts(seq, ",oldalloc");
793 #ifdef CONFIG_EXT4_FS_XATTR
794 if (test_opt(sb, XATTR_USER) &&
795 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
796 seq_puts(seq, ",user_xattr");
797 if (!test_opt(sb, XATTR_USER) &&
798 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
799 seq_puts(seq, ",nouser_xattr");
802 #ifdef CONFIG_EXT4_FS_POSIX_ACL
803 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
804 seq_puts(seq, ",acl");
805 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
806 seq_puts(seq, ",noacl");
808 if (!test_opt(sb, RESERVATION))
809 seq_puts(seq, ",noreservation");
810 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
811 seq_printf(seq, ",commit=%u",
812 (unsigned) (sbi->s_commit_interval / HZ));
814 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
815 seq_printf(seq, ",min_batch_time=%u",
816 (unsigned) sbi->s_min_batch_time);
818 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
819 seq_printf(seq, ",max_batch_time=%u",
820 (unsigned) sbi->s_min_batch_time);
824 * We're changing the default of barrier mount option, so
825 * let's always display its mount state so it's clear what its
828 seq_puts(seq, ",barrier=");
829 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
830 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
831 seq_puts(seq, ",journal_async_commit");
832 if (test_opt(sb, NOBH))
833 seq_puts(seq, ",nobh");
834 if (!test_opt(sb, EXTENTS))
835 seq_puts(seq, ",noextents");
836 if (test_opt(sb, I_VERSION))
837 seq_puts(seq, ",i_version");
838 if (!test_opt(sb, DELALLOC))
839 seq_puts(seq, ",nodelalloc");
843 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
845 * journal mode get enabled in different ways
846 * So just print the value even if we didn't specify it
848 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
849 seq_puts(seq, ",data=journal");
850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
851 seq_puts(seq, ",data=ordered");
852 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
853 seq_puts(seq, ",data=writeback");
855 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
856 seq_printf(seq, ",inode_readahead_blks=%u",
857 sbi->s_inode_readahead_blks);
859 if (test_opt(sb, DATA_ERR_ABORT))
860 seq_puts(seq, ",data_err=abort");
862 ext4_show_quota_options(seq, sb);
867 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
868 u64 ino, u32 generation)
872 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
873 return ERR_PTR(-ESTALE);
874 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
875 return ERR_PTR(-ESTALE);
877 /* iget isn't really right if the inode is currently unallocated!!
879 * ext4_read_inode will return a bad_inode if the inode had been
880 * deleted, so we should be safe.
882 * Currently we don't know the generation for parent directory, so
883 * a generation of 0 means "accept any"
885 inode = ext4_iget(sb, ino);
887 return ERR_CAST(inode);
888 if (generation && inode->i_generation != generation) {
890 return ERR_PTR(-ESTALE);
896 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
897 int fh_len, int fh_type)
899 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
903 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
904 int fh_len, int fh_type)
906 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
911 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
912 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
914 static int ext4_dquot_initialize(struct inode *inode, int type);
915 static int ext4_dquot_drop(struct inode *inode);
916 static int ext4_write_dquot(struct dquot *dquot);
917 static int ext4_acquire_dquot(struct dquot *dquot);
918 static int ext4_release_dquot(struct dquot *dquot);
919 static int ext4_mark_dquot_dirty(struct dquot *dquot);
920 static int ext4_write_info(struct super_block *sb, int type);
921 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
922 char *path, int remount);
923 static int ext4_quota_on_mount(struct super_block *sb, int type);
924 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
925 size_t len, loff_t off);
926 static ssize_t ext4_quota_write(struct super_block *sb, int type,
927 const char *data, size_t len, loff_t off);
929 static struct dquot_operations ext4_quota_operations = {
930 .initialize = ext4_dquot_initialize,
931 .drop = ext4_dquot_drop,
932 .alloc_space = dquot_alloc_space,
933 .alloc_inode = dquot_alloc_inode,
934 .free_space = dquot_free_space,
935 .free_inode = dquot_free_inode,
936 .transfer = dquot_transfer,
937 .write_dquot = ext4_write_dquot,
938 .acquire_dquot = ext4_acquire_dquot,
939 .release_dquot = ext4_release_dquot,
940 .mark_dirty = ext4_mark_dquot_dirty,
941 .write_info = ext4_write_info
944 static struct quotactl_ops ext4_qctl_operations = {
945 .quota_on = ext4_quota_on,
946 .quota_off = vfs_quota_off,
947 .quota_sync = vfs_quota_sync,
948 .get_info = vfs_get_dqinfo,
949 .set_info = vfs_set_dqinfo,
950 .get_dqblk = vfs_get_dqblk,
951 .set_dqblk = vfs_set_dqblk
955 static const struct super_operations ext4_sops = {
956 .alloc_inode = ext4_alloc_inode,
957 .destroy_inode = ext4_destroy_inode,
958 .write_inode = ext4_write_inode,
959 .dirty_inode = ext4_dirty_inode,
960 .delete_inode = ext4_delete_inode,
961 .put_super = ext4_put_super,
962 .write_super = ext4_write_super,
963 .sync_fs = ext4_sync_fs,
964 .write_super_lockfs = ext4_write_super_lockfs,
965 .unlockfs = ext4_unlockfs,
966 .statfs = ext4_statfs,
967 .remount_fs = ext4_remount,
968 .clear_inode = ext4_clear_inode,
969 .show_options = ext4_show_options,
971 .quota_read = ext4_quota_read,
972 .quota_write = ext4_quota_write,
976 static const struct export_operations ext4_export_ops = {
977 .fh_to_dentry = ext4_fh_to_dentry,
978 .fh_to_parent = ext4_fh_to_parent,
979 .get_parent = ext4_get_parent,
983 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
984 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
985 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
986 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
987 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
988 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
989 Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
990 Opt_journal_checksum, Opt_journal_async_commit,
991 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
992 Opt_data_err_abort, Opt_data_err_ignore,
993 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
994 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
995 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
996 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
997 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
998 Opt_inode_readahead_blks
1001 static const match_table_t tokens = {
1002 {Opt_bsd_df, "bsddf"},
1003 {Opt_minix_df, "minixdf"},
1004 {Opt_grpid, "grpid"},
1005 {Opt_grpid, "bsdgroups"},
1006 {Opt_nogrpid, "nogrpid"},
1007 {Opt_nogrpid, "sysvgroups"},
1008 {Opt_resgid, "resgid=%u"},
1009 {Opt_resuid, "resuid=%u"},
1011 {Opt_err_cont, "errors=continue"},
1012 {Opt_err_panic, "errors=panic"},
1013 {Opt_err_ro, "errors=remount-ro"},
1014 {Opt_nouid32, "nouid32"},
1015 {Opt_debug, "debug"},
1016 {Opt_oldalloc, "oldalloc"},
1017 {Opt_orlov, "orlov"},
1018 {Opt_user_xattr, "user_xattr"},
1019 {Opt_nouser_xattr, "nouser_xattr"},
1021 {Opt_noacl, "noacl"},
1022 {Opt_reservation, "reservation"},
1023 {Opt_noreservation, "noreservation"},
1024 {Opt_noload, "noload"},
1027 {Opt_commit, "commit=%u"},
1028 {Opt_min_batch_time, "min_batch_time=%u"},
1029 {Opt_max_batch_time, "max_batch_time=%u"},
1030 {Opt_journal_update, "journal=update"},
1031 {Opt_journal_inum, "journal=%u"},
1032 {Opt_journal_dev, "journal_dev=%u"},
1033 {Opt_journal_checksum, "journal_checksum"},
1034 {Opt_journal_async_commit, "journal_async_commit"},
1035 {Opt_abort, "abort"},
1036 {Opt_data_journal, "data=journal"},
1037 {Opt_data_ordered, "data=ordered"},
1038 {Opt_data_writeback, "data=writeback"},
1039 {Opt_data_err_abort, "data_err=abort"},
1040 {Opt_data_err_ignore, "data_err=ignore"},
1041 {Opt_offusrjquota, "usrjquota="},
1042 {Opt_usrjquota, "usrjquota=%s"},
1043 {Opt_offgrpjquota, "grpjquota="},
1044 {Opt_grpjquota, "grpjquota=%s"},
1045 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1046 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1047 {Opt_grpquota, "grpquota"},
1048 {Opt_noquota, "noquota"},
1049 {Opt_quota, "quota"},
1050 {Opt_usrquota, "usrquota"},
1051 {Opt_barrier, "barrier=%u"},
1052 {Opt_extents, "extents"},
1053 {Opt_noextents, "noextents"},
1054 {Opt_i_version, "i_version"},
1055 {Opt_stripe, "stripe=%u"},
1056 {Opt_resize, "resize"},
1057 {Opt_delalloc, "delalloc"},
1058 {Opt_nodelalloc, "nodelalloc"},
1059 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1063 static ext4_fsblk_t get_sb_block(void **data)
1065 ext4_fsblk_t sb_block;
1066 char *options = (char *) *data;
1068 if (!options || strncmp(options, "sb=", 3) != 0)
1069 return 1; /* Default location */
1071 /*todo: use simple_strtoll with >32bit ext4 */
1072 sb_block = simple_strtoul(options, &options, 0);
1073 if (*options && *options != ',') {
1074 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1078 if (*options == ',')
1080 *data = (void *) options;
1084 static int parse_options(char *options, struct super_block *sb,
1085 unsigned int *inum, unsigned long *journal_devnum,
1086 ext4_fsblk_t *n_blocks_count, int is_remount)
1088 struct ext4_sb_info *sbi = EXT4_SB(sb);
1090 substring_t args[MAX_OPT_ARGS];
1097 ext4_fsblk_t last_block;
1102 while ((p = strsep(&options, ",")) != NULL) {
1107 token = match_token(p, tokens, args);
1110 clear_opt(sbi->s_mount_opt, MINIX_DF);
1113 set_opt(sbi->s_mount_opt, MINIX_DF);
1116 set_opt(sbi->s_mount_opt, GRPID);
1119 clear_opt(sbi->s_mount_opt, GRPID);
1122 if (match_int(&args[0], &option))
1124 sbi->s_resuid = option;
1127 if (match_int(&args[0], &option))
1129 sbi->s_resgid = option;
1132 /* handled by get_sb_block() instead of here */
1133 /* *sb_block = match_int(&args[0]); */
1136 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1137 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1138 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1141 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1142 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1143 set_opt(sbi->s_mount_opt, ERRORS_RO);
1146 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1147 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1148 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1151 set_opt(sbi->s_mount_opt, NO_UID32);
1154 set_opt(sbi->s_mount_opt, DEBUG);
1157 set_opt(sbi->s_mount_opt, OLDALLOC);
1160 clear_opt(sbi->s_mount_opt, OLDALLOC);
1162 #ifdef CONFIG_EXT4_FS_XATTR
1163 case Opt_user_xattr:
1164 set_opt(sbi->s_mount_opt, XATTR_USER);
1166 case Opt_nouser_xattr:
1167 clear_opt(sbi->s_mount_opt, XATTR_USER);
1170 case Opt_user_xattr:
1171 case Opt_nouser_xattr:
1172 printk(KERN_ERR "EXT4 (no)user_xattr options "
1176 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1178 set_opt(sbi->s_mount_opt, POSIX_ACL);
1181 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1186 printk(KERN_ERR "EXT4 (no)acl options "
1190 case Opt_reservation:
1191 set_opt(sbi->s_mount_opt, RESERVATION);
1193 case Opt_noreservation:
1194 clear_opt(sbi->s_mount_opt, RESERVATION);
1196 case Opt_journal_update:
1198 /* Eventually we will want to be able to create
1199 a journal file here. For now, only allow the
1200 user to specify an existing inode to be the
1203 printk(KERN_ERR "EXT4-fs: cannot specify "
1204 "journal on remount\n");
1207 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1209 case Opt_journal_inum:
1211 printk(KERN_ERR "EXT4-fs: cannot specify "
1212 "journal on remount\n");
1215 if (match_int(&args[0], &option))
1219 case Opt_journal_dev:
1221 printk(KERN_ERR "EXT4-fs: cannot specify "
1222 "journal on remount\n");
1225 if (match_int(&args[0], &option))
1227 *journal_devnum = option;
1229 case Opt_journal_checksum:
1230 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1232 case Opt_journal_async_commit:
1233 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1234 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1237 set_opt(sbi->s_mount_opt, NOLOAD);
1240 if (match_int(&args[0], &option))
1245 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1246 sbi->s_commit_interval = HZ * option;
1248 case Opt_max_batch_time:
1249 if (match_int(&args[0], &option))
1254 option = EXT4_DEF_MAX_BATCH_TIME;
1255 sbi->s_max_batch_time = option;
1257 case Opt_min_batch_time:
1258 if (match_int(&args[0], &option))
1262 sbi->s_min_batch_time = option;
1264 case Opt_data_journal:
1265 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1267 case Opt_data_ordered:
1268 data_opt = EXT4_MOUNT_ORDERED_DATA;
1270 case Opt_data_writeback:
1271 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1274 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1277 "EXT4-fs: cannot change data "
1278 "mode on remount\n");
1282 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1283 sbi->s_mount_opt |= data_opt;
1286 case Opt_data_err_abort:
1287 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1289 case Opt_data_err_ignore:
1290 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1299 if ((sb_any_quota_enabled(sb) ||
1300 sb_any_quota_suspended(sb)) &&
1301 !sbi->s_qf_names[qtype]) {
1303 "EXT4-fs: Cannot change journaled "
1304 "quota options when quota turned on.\n");
1307 qname = match_strdup(&args[0]);
1310 "EXT4-fs: not enough memory for "
1311 "storing quotafile name.\n");
1314 if (sbi->s_qf_names[qtype] &&
1315 strcmp(sbi->s_qf_names[qtype], qname)) {
1317 "EXT4-fs: %s quota file already "
1318 "specified.\n", QTYPE2NAME(qtype));
1322 sbi->s_qf_names[qtype] = qname;
1323 if (strchr(sbi->s_qf_names[qtype], '/')) {
1325 "EXT4-fs: quotafile must be on "
1326 "filesystem root.\n");
1327 kfree(sbi->s_qf_names[qtype]);
1328 sbi->s_qf_names[qtype] = NULL;
1331 set_opt(sbi->s_mount_opt, QUOTA);
1333 case Opt_offusrjquota:
1336 case Opt_offgrpjquota:
1339 if ((sb_any_quota_enabled(sb) ||
1340 sb_any_quota_suspended(sb)) &&
1341 sbi->s_qf_names[qtype]) {
1342 printk(KERN_ERR "EXT4-fs: Cannot change "
1343 "journaled quota options when "
1344 "quota turned on.\n");
1348 * The space will be released later when all options
1349 * are confirmed to be correct
1351 sbi->s_qf_names[qtype] = NULL;
1353 case Opt_jqfmt_vfsold:
1354 qfmt = QFMT_VFS_OLD;
1356 case Opt_jqfmt_vfsv0:
1359 if ((sb_any_quota_enabled(sb) ||
1360 sb_any_quota_suspended(sb)) &&
1361 sbi->s_jquota_fmt != qfmt) {
1362 printk(KERN_ERR "EXT4-fs: Cannot change "
1363 "journaled quota options when "
1364 "quota turned on.\n");
1367 sbi->s_jquota_fmt = qfmt;
1371 set_opt(sbi->s_mount_opt, QUOTA);
1372 set_opt(sbi->s_mount_opt, USRQUOTA);
1375 set_opt(sbi->s_mount_opt, QUOTA);
1376 set_opt(sbi->s_mount_opt, GRPQUOTA);
1379 if (sb_any_quota_enabled(sb)) {
1380 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1381 "options when quota turned on.\n");
1384 clear_opt(sbi->s_mount_opt, QUOTA);
1385 clear_opt(sbi->s_mount_opt, USRQUOTA);
1386 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1393 "EXT4-fs: quota options not supported.\n");
1397 case Opt_offusrjquota:
1398 case Opt_offgrpjquota:
1399 case Opt_jqfmt_vfsold:
1400 case Opt_jqfmt_vfsv0:
1402 "EXT4-fs: journaled quota options not "
1409 set_opt(sbi->s_mount_opt, ABORT);
1412 if (match_int(&args[0], &option))
1415 set_opt(sbi->s_mount_opt, BARRIER);
1417 clear_opt(sbi->s_mount_opt, BARRIER);
1423 printk("EXT4-fs: resize option only available "
1427 if (match_int(&args[0], &option) != 0)
1429 *n_blocks_count = option;
1432 set_opt(sbi->s_mount_opt, NOBH);
1435 clear_opt(sbi->s_mount_opt, NOBH);
1438 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1439 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1440 ext4_warning(sb, __func__,
1441 "extents feature not enabled "
1442 "on this filesystem, use tune2fs");
1445 set_opt(sbi->s_mount_opt, EXTENTS);
1449 * When e2fsprogs support resizing an already existing
1450 * ext3 file system to greater than 2**32 we need to
1451 * add support to block allocator to handle growing
1452 * already existing block mapped inode so that blocks
1453 * allocated for them fall within 2**32
1455 last_block = ext4_blocks_count(sbi->s_es) - 1;
1456 if (last_block > 0xffffffffULL) {
1457 printk(KERN_ERR "EXT4-fs: Filesystem too "
1458 "large to mount with "
1459 "-o noextents options\n");
1462 clear_opt(sbi->s_mount_opt, EXTENTS);
1465 set_opt(sbi->s_mount_opt, I_VERSION);
1466 sb->s_flags |= MS_I_VERSION;
1468 case Opt_nodelalloc:
1469 clear_opt(sbi->s_mount_opt, DELALLOC);
1472 if (match_int(&args[0], &option))
1476 sbi->s_stripe = option;
1479 set_opt(sbi->s_mount_opt, DELALLOC);
1481 case Opt_inode_readahead_blks:
1482 if (match_int(&args[0], &option))
1484 if (option < 0 || option > (1 << 30))
1486 sbi->s_inode_readahead_blks = option;
1490 "EXT4-fs: Unrecognized mount option \"%s\" "
1491 "or missing value\n", p);
1496 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1497 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1498 sbi->s_qf_names[USRQUOTA])
1499 clear_opt(sbi->s_mount_opt, USRQUOTA);
1501 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1502 sbi->s_qf_names[GRPQUOTA])
1503 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1505 if ((sbi->s_qf_names[USRQUOTA] &&
1506 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1507 (sbi->s_qf_names[GRPQUOTA] &&
1508 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1509 printk(KERN_ERR "EXT4-fs: old and new quota "
1510 "format mixing.\n");
1514 if (!sbi->s_jquota_fmt) {
1515 printk(KERN_ERR "EXT4-fs: journaled quota format "
1516 "not specified.\n");
1520 if (sbi->s_jquota_fmt) {
1521 printk(KERN_ERR "EXT4-fs: journaled quota format "
1522 "specified with no journaling "
1531 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1534 struct ext4_sb_info *sbi = EXT4_SB(sb);
1537 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1538 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1539 "forcing read-only mode\n");
1544 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1545 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1546 "running e2fsck is recommended\n");
1547 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1549 "EXT4-fs warning: mounting fs with errors, "
1550 "running e2fsck is recommended\n");
1551 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1552 le16_to_cpu(es->s_mnt_count) >=
1553 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1555 "EXT4-fs warning: maximal mount count reached, "
1556 "running e2fsck is recommended\n");
1557 else if (le32_to_cpu(es->s_checkinterval) &&
1558 (le32_to_cpu(es->s_lastcheck) +
1559 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1561 "EXT4-fs warning: checktime reached, "
1562 "running e2fsck is recommended\n");
1563 if (!sbi->s_journal)
1564 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1565 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1566 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1567 le16_add_cpu(&es->s_mnt_count, 1);
1568 es->s_mtime = cpu_to_le32(get_seconds());
1569 ext4_update_dynamic_rev(sb);
1571 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1573 ext4_commit_super(sb, es, 1);
1574 if (test_opt(sb, DEBUG))
1575 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1576 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1578 sbi->s_groups_count,
1579 EXT4_BLOCKS_PER_GROUP(sb),
1580 EXT4_INODES_PER_GROUP(sb),
1583 if (EXT4_SB(sb)->s_journal) {
1584 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1585 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1586 "external", EXT4_SB(sb)->s_journal->j_devname);
1588 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1593 static int ext4_fill_flex_info(struct super_block *sb)
1595 struct ext4_sb_info *sbi = EXT4_SB(sb);
1596 struct ext4_group_desc *gdp = NULL;
1597 struct buffer_head *bh;
1598 ext4_group_t flex_group_count;
1599 ext4_group_t flex_group;
1600 int groups_per_flex = 0;
1603 if (!sbi->s_es->s_log_groups_per_flex) {
1604 sbi->s_log_groups_per_flex = 0;
1608 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1609 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1611 /* We allocate both existing and potentially added groups */
1612 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1613 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1614 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1615 sbi->s_flex_groups = kzalloc(flex_group_count *
1616 sizeof(struct flex_groups), GFP_KERNEL);
1617 if (sbi->s_flex_groups == NULL) {
1618 printk(KERN_ERR "EXT4-fs: not enough memory for "
1619 "%u flex groups\n", flex_group_count);
1623 for (i = 0; i < sbi->s_groups_count; i++) {
1624 gdp = ext4_get_group_desc(sb, i, &bh);
1626 flex_group = ext4_flex_group(sbi, i);
1627 sbi->s_flex_groups[flex_group].free_inodes +=
1628 ext4_free_inodes_count(sb, gdp);
1629 sbi->s_flex_groups[flex_group].free_blocks +=
1630 ext4_free_blks_count(sb, gdp);
1638 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1639 struct ext4_group_desc *gdp)
1643 if (sbi->s_es->s_feature_ro_compat &
1644 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1645 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1646 __le32 le_group = cpu_to_le32(block_group);
1648 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1649 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1650 crc = crc16(crc, (__u8 *)gdp, offset);
1651 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1652 /* for checksum of struct ext4_group_desc do the rest...*/
1653 if ((sbi->s_es->s_feature_incompat &
1654 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1655 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1656 crc = crc16(crc, (__u8 *)gdp + offset,
1657 le16_to_cpu(sbi->s_es->s_desc_size) -
1661 return cpu_to_le16(crc);
1664 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1665 struct ext4_group_desc *gdp)
1667 if ((sbi->s_es->s_feature_ro_compat &
1668 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1669 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1675 /* Called at mount-time, super-block is locked */
1676 static int ext4_check_descriptors(struct super_block *sb)
1678 struct ext4_sb_info *sbi = EXT4_SB(sb);
1679 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1680 ext4_fsblk_t last_block;
1681 ext4_fsblk_t block_bitmap;
1682 ext4_fsblk_t inode_bitmap;
1683 ext4_fsblk_t inode_table;
1684 int flexbg_flag = 0;
1687 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1690 ext4_debug("Checking group descriptors");
1692 for (i = 0; i < sbi->s_groups_count; i++) {
1693 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1695 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1696 last_block = ext4_blocks_count(sbi->s_es) - 1;
1698 last_block = first_block +
1699 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1701 block_bitmap = ext4_block_bitmap(sb, gdp);
1702 if (block_bitmap < first_block || block_bitmap > last_block) {
1703 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1704 "Block bitmap for group %u not in group "
1705 "(block %llu)!\n", i, block_bitmap);
1708 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1709 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1710 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1711 "Inode bitmap for group %u not in group "
1712 "(block %llu)!\n", i, inode_bitmap);
1715 inode_table = ext4_inode_table(sb, gdp);
1716 if (inode_table < first_block ||
1717 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1718 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1719 "Inode table for group %u not in group "
1720 "(block %llu)!\n", i, inode_table);
1723 spin_lock(sb_bgl_lock(sbi, i));
1724 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1725 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1726 "Checksum for group %u failed (%u!=%u)\n",
1727 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1728 gdp)), le16_to_cpu(gdp->bg_checksum));
1729 if (!(sb->s_flags & MS_RDONLY)) {
1730 spin_unlock(sb_bgl_lock(sbi, i));
1734 spin_unlock(sb_bgl_lock(sbi, i));
1736 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1739 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1740 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1744 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1745 * the superblock) which were deleted from all directories, but held open by
1746 * a process at the time of a crash. We walk the list and try to delete these
1747 * inodes at recovery time (only with a read-write filesystem).
1749 * In order to keep the orphan inode chain consistent during traversal (in
1750 * case of crash during recovery), we link each inode into the superblock
1751 * orphan list_head and handle it the same way as an inode deletion during
1752 * normal operation (which journals the operations for us).
1754 * We only do an iget() and an iput() on each inode, which is very safe if we
1755 * accidentally point at an in-use or already deleted inode. The worst that
1756 * can happen in this case is that we get a "bit already cleared" message from
1757 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1758 * e2fsck was run on this filesystem, and it must have already done the orphan
1759 * inode cleanup for us, so we can safely abort without any further action.
1761 static void ext4_orphan_cleanup(struct super_block *sb,
1762 struct ext4_super_block *es)
1764 unsigned int s_flags = sb->s_flags;
1765 int nr_orphans = 0, nr_truncates = 0;
1769 if (!es->s_last_orphan) {
1770 jbd_debug(4, "no orphan inodes to clean up\n");
1774 if (bdev_read_only(sb->s_bdev)) {
1775 printk(KERN_ERR "EXT4-fs: write access "
1776 "unavailable, skipping orphan cleanup.\n");
1780 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1781 if (es->s_last_orphan)
1782 jbd_debug(1, "Errors on filesystem, "
1783 "clearing orphan list.\n");
1784 es->s_last_orphan = 0;
1785 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1789 if (s_flags & MS_RDONLY) {
1790 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1792 sb->s_flags &= ~MS_RDONLY;
1795 /* Needed for iput() to work correctly and not trash data */
1796 sb->s_flags |= MS_ACTIVE;
1797 /* Turn on quotas so that they are updated correctly */
1798 for (i = 0; i < MAXQUOTAS; i++) {
1799 if (EXT4_SB(sb)->s_qf_names[i]) {
1800 int ret = ext4_quota_on_mount(sb, i);
1803 "EXT4-fs: Cannot turn on journaled "
1804 "quota: error %d\n", ret);
1809 while (es->s_last_orphan) {
1810 struct inode *inode;
1812 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1813 if (IS_ERR(inode)) {
1814 es->s_last_orphan = 0;
1818 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1820 if (inode->i_nlink) {
1822 "%s: truncating inode %lu to %lld bytes\n",
1823 __func__, inode->i_ino, inode->i_size);
1824 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1825 inode->i_ino, inode->i_size);
1826 ext4_truncate(inode);
1830 "%s: deleting unreferenced inode %lu\n",
1831 __func__, inode->i_ino);
1832 jbd_debug(2, "deleting unreferenced inode %lu\n",
1836 iput(inode); /* The delete magic happens here! */
1839 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1842 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1843 sb->s_id, PLURAL(nr_orphans));
1845 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1846 sb->s_id, PLURAL(nr_truncates));
1848 /* Turn quotas off */
1849 for (i = 0; i < MAXQUOTAS; i++) {
1850 if (sb_dqopt(sb)->files[i])
1851 vfs_quota_off(sb, i, 0);
1854 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1857 * Maximal extent format file size.
1858 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1859 * extent format containers, within a sector_t, and within i_blocks
1860 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1861 * so that won't be a limiting factor.
1863 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1865 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1868 loff_t upper_limit = MAX_LFS_FILESIZE;
1870 /* small i_blocks in vfs inode? */
1871 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1873 * CONFIG_LBD is not enabled implies the inode
1874 * i_block represent total blocks in 512 bytes
1875 * 32 == size of vfs inode i_blocks * 8
1877 upper_limit = (1LL << 32) - 1;
1879 /* total blocks in file system block size */
1880 upper_limit >>= (blkbits - 9);
1881 upper_limit <<= blkbits;
1884 /* 32-bit extent-start container, ee_block */
1889 /* Sanity check against vm- & vfs- imposed limits */
1890 if (res > upper_limit)
1897 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1898 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1899 * We need to be 1 filesystem block less than the 2^48 sector limit.
1901 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1903 loff_t res = EXT4_NDIR_BLOCKS;
1906 /* This is calculated to be the largest file size for a
1907 * dense, bitmapped file such that the total number of
1908 * sectors in the file, including data and all indirect blocks,
1909 * does not exceed 2^48 -1
1910 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1911 * total number of 512 bytes blocks of the file
1914 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1916 * !has_huge_files or CONFIG_LBD is not enabled
1917 * implies the inode i_block represent total blocks in
1918 * 512 bytes 32 == size of vfs inode i_blocks * 8
1920 upper_limit = (1LL << 32) - 1;
1922 /* total blocks in file system block size */
1923 upper_limit >>= (bits - 9);
1927 * We use 48 bit ext4_inode i_blocks
1928 * With EXT4_HUGE_FILE_FL set the i_blocks
1929 * represent total number of blocks in
1930 * file system block size
1932 upper_limit = (1LL << 48) - 1;
1936 /* indirect blocks */
1938 /* double indirect blocks */
1939 meta_blocks += 1 + (1LL << (bits-2));
1940 /* tripple indirect blocks */
1941 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1943 upper_limit -= meta_blocks;
1944 upper_limit <<= bits;
1946 res += 1LL << (bits-2);
1947 res += 1LL << (2*(bits-2));
1948 res += 1LL << (3*(bits-2));
1950 if (res > upper_limit)
1953 if (res > MAX_LFS_FILESIZE)
1954 res = MAX_LFS_FILESIZE;
1959 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1960 ext4_fsblk_t logical_sb_block, int nr)
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 ext4_group_t bg, first_meta_bg;
1966 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1968 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1970 return logical_sb_block + nr + 1;
1971 bg = sbi->s_desc_per_block * nr;
1972 if (ext4_bg_has_super(sb, bg))
1974 return (has_super + ext4_group_first_block_no(sb, bg));
1978 * ext4_get_stripe_size: Get the stripe size.
1979 * @sbi: In memory super block info
1981 * If we have specified it via mount option, then
1982 * use the mount option value. If the value specified at mount time is
1983 * greater than the blocks per group use the super block value.
1984 * If the super block value is greater than blocks per group return 0.
1985 * Allocator needs it be less than blocks per group.
1988 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1990 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1991 unsigned long stripe_width =
1992 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1994 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1995 return sbi->s_stripe;
1997 if (stripe_width <= sbi->s_blocks_per_group)
1998 return stripe_width;
2000 if (stride <= sbi->s_blocks_per_group)
2006 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2007 __releases(kernel_lock)
2008 __acquires(kernel_lock)
2011 struct buffer_head *bh;
2012 struct ext4_super_block *es = NULL;
2013 struct ext4_sb_info *sbi;
2015 ext4_fsblk_t sb_block = get_sb_block(&data);
2016 ext4_fsblk_t logical_sb_block;
2017 unsigned long offset = 0;
2018 unsigned int journal_inum = 0;
2019 unsigned long journal_devnum = 0;
2020 unsigned long def_mount_opts;
2028 int needs_recovery, has_huge_files;
2033 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2036 sb->s_fs_info = sbi;
2037 sbi->s_mount_opt = 0;
2038 sbi->s_resuid = EXT4_DEF_RESUID;
2039 sbi->s_resgid = EXT4_DEF_RESGID;
2040 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2041 sbi->s_sb_block = sb_block;
2045 /* Cleanup superblock name */
2046 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2049 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2051 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2056 * The ext4 superblock will not be buffer aligned for other than 1kB
2057 * block sizes. We need to calculate the offset from buffer start.
2059 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2060 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2061 offset = do_div(logical_sb_block, blocksize);
2063 logical_sb_block = sb_block;
2066 if (!(bh = sb_bread(sb, logical_sb_block))) {
2067 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2071 * Note: s_es must be initialized as soon as possible because
2072 * some ext4 macro-instructions depend on its value
2074 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2076 sb->s_magic = le16_to_cpu(es->s_magic);
2077 if (sb->s_magic != EXT4_SUPER_MAGIC)
2080 /* Set defaults before we parse the mount options */
2081 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2082 if (def_mount_opts & EXT4_DEFM_DEBUG)
2083 set_opt(sbi->s_mount_opt, DEBUG);
2084 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2085 set_opt(sbi->s_mount_opt, GRPID);
2086 if (def_mount_opts & EXT4_DEFM_UID16)
2087 set_opt(sbi->s_mount_opt, NO_UID32);
2088 #ifdef CONFIG_EXT4_FS_XATTR
2089 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2090 set_opt(sbi->s_mount_opt, XATTR_USER);
2092 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2093 if (def_mount_opts & EXT4_DEFM_ACL)
2094 set_opt(sbi->s_mount_opt, POSIX_ACL);
2096 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2097 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2098 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2099 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2100 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2101 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2103 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2104 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2105 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2106 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2108 set_opt(sbi->s_mount_opt, ERRORS_RO);
2110 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2111 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2112 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2113 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2114 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2116 set_opt(sbi->s_mount_opt, RESERVATION);
2117 set_opt(sbi->s_mount_opt, BARRIER);
2120 * turn on extents feature by default in ext4 filesystem
2121 * only if feature flag already set by mkfs or tune2fs.
2122 * Use -o noextents to turn it off
2124 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2125 set_opt(sbi->s_mount_opt, EXTENTS);
2127 ext4_warning(sb, __func__,
2128 "extents feature not enabled on this filesystem, "
2132 * enable delayed allocation by default
2133 * Use -o nodelalloc to turn it off
2135 set_opt(sbi->s_mount_opt, DELALLOC);
2138 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2142 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2143 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2145 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2146 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2147 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2148 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2150 "EXT4-fs warning: feature flags set on rev 0 fs, "
2151 "running e2fsck is recommended\n");
2154 * Check feature flags regardless of the revision level, since we
2155 * previously didn't change the revision level when setting the flags,
2156 * so there is a chance incompat flags are set on a rev 0 filesystem.
2158 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2160 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2161 "unsupported optional features (%x).\n", sb->s_id,
2162 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2163 ~EXT4_FEATURE_INCOMPAT_SUPP));
2166 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2167 if (!(sb->s_flags & MS_RDONLY) && features) {
2168 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2169 "unsupported optional features (%x).\n", sb->s_id,
2170 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2171 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2174 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2175 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2176 if (has_huge_files) {
2178 * Large file size enabled file system can only be
2179 * mount if kernel is build with CONFIG_LBD
2181 if (sizeof(root->i_blocks) < sizeof(u64) &&
2182 !(sb->s_flags & MS_RDONLY)) {
2183 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2184 "files cannot be mounted read-write "
2185 "without CONFIG_LBD.\n", sb->s_id);
2189 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2191 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2192 blocksize > EXT4_MAX_BLOCK_SIZE) {
2194 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2195 blocksize, sb->s_id);
2199 if (sb->s_blocksize != blocksize) {
2201 /* Validate the filesystem blocksize */
2202 if (!sb_set_blocksize(sb, blocksize)) {
2203 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2209 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2210 offset = do_div(logical_sb_block, blocksize);
2211 bh = sb_bread(sb, logical_sb_block);
2214 "EXT4-fs: Can't read superblock on 2nd try.\n");
2217 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2219 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2221 "EXT4-fs: Magic mismatch, very weird !\n");
2226 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2228 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2230 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2231 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2232 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2234 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2235 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2236 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2237 (!is_power_of_2(sbi->s_inode_size)) ||
2238 (sbi->s_inode_size > blocksize)) {
2240 "EXT4-fs: unsupported inode size: %d\n",
2244 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2245 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2247 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2248 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2249 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2250 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2251 !is_power_of_2(sbi->s_desc_size)) {
2253 "EXT4-fs: unsupported descriptor size %lu\n",
2258 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2259 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2260 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2261 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2263 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2264 if (sbi->s_inodes_per_block == 0)
2266 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2267 sbi->s_inodes_per_block;
2268 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2270 sbi->s_mount_state = le16_to_cpu(es->s_state);
2271 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2272 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2273 for (i = 0; i < 4; i++)
2274 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2275 sbi->s_def_hash_version = es->s_def_hash_version;
2276 i = le32_to_cpu(es->s_flags);
2277 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2278 sbi->s_hash_unsigned = 3;
2279 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2280 #ifdef __CHAR_UNSIGNED__
2281 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2282 sbi->s_hash_unsigned = 3;
2284 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2289 if (sbi->s_blocks_per_group > blocksize * 8) {
2291 "EXT4-fs: #blocks per group too big: %lu\n",
2292 sbi->s_blocks_per_group);
2295 if (sbi->s_inodes_per_group > blocksize * 8) {
2297 "EXT4-fs: #inodes per group too big: %lu\n",
2298 sbi->s_inodes_per_group);
2302 if (ext4_blocks_count(es) >
2303 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2304 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2305 " too large to mount safely\n", sb->s_id);
2306 if (sizeof(sector_t) < 8)
2307 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2312 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2315 /* ensure blocks_count calculation below doesn't sign-extend */
2316 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2317 le32_to_cpu(es->s_first_data_block) + 1) {
2318 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2319 "first data block %u, blocks per group %lu\n",
2320 ext4_blocks_count(es),
2321 le32_to_cpu(es->s_first_data_block),
2322 EXT4_BLOCKS_PER_GROUP(sb));
2325 blocks_count = (ext4_blocks_count(es) -
2326 le32_to_cpu(es->s_first_data_block) +
2327 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2328 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2329 sbi->s_groups_count = blocks_count;
2330 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2331 EXT4_DESC_PER_BLOCK(sb);
2332 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2334 if (sbi->s_group_desc == NULL) {
2335 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2339 #ifdef CONFIG_PROC_FS
2341 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2344 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2346 &sbi->s_inode_readahead_blks);
2349 bgl_lock_init(&sbi->s_blockgroup_lock);
2351 for (i = 0; i < db_count; i++) {
2352 block = descriptor_loc(sb, logical_sb_block, i);
2353 sbi->s_group_desc[i] = sb_bread(sb, block);
2354 if (!sbi->s_group_desc[i]) {
2355 printk(KERN_ERR "EXT4-fs: "
2356 "can't read group descriptor %d\n", i);
2361 if (!ext4_check_descriptors(sb)) {
2362 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2365 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2366 if (!ext4_fill_flex_info(sb)) {
2368 "EXT4-fs: unable to initialize "
2369 "flex_bg meta info!\n");
2373 sbi->s_gdb_count = db_count;
2374 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2375 spin_lock_init(&sbi->s_next_gen_lock);
2377 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2378 ext4_count_free_blocks(sb));
2380 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2381 ext4_count_free_inodes(sb));
2384 err = percpu_counter_init(&sbi->s_dirs_counter,
2385 ext4_count_dirs(sb));
2388 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2391 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2395 sbi->s_stripe = ext4_get_stripe_size(sbi);
2398 * set up enough so that it can read an inode
2400 sb->s_op = &ext4_sops;
2401 sb->s_export_op = &ext4_export_ops;
2402 sb->s_xattr = ext4_xattr_handlers;
2404 sb->s_qcop = &ext4_qctl_operations;
2405 sb->dq_op = &ext4_quota_operations;
2407 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2411 needs_recovery = (es->s_last_orphan != 0 ||
2412 EXT4_HAS_INCOMPAT_FEATURE(sb,
2413 EXT4_FEATURE_INCOMPAT_RECOVER));
2416 * The first inode we look at is the journal inode. Don't try
2417 * root first: it may be modified in the journal!
2419 if (!test_opt(sb, NOLOAD) &&
2420 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2421 if (ext4_load_journal(sb, es, journal_devnum))
2423 if (!(sb->s_flags & MS_RDONLY) &&
2424 EXT4_SB(sb)->s_journal->j_failed_commit) {
2425 printk(KERN_CRIT "EXT4-fs error (device %s): "
2426 "ext4_fill_super: Journal transaction "
2427 "%u is corrupt\n", sb->s_id,
2428 EXT4_SB(sb)->s_journal->j_failed_commit);
2429 if (test_opt(sb, ERRORS_RO)) {
2431 "Mounting filesystem read-only\n");
2432 sb->s_flags |= MS_RDONLY;
2433 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2434 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2436 if (test_opt(sb, ERRORS_PANIC)) {
2437 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2438 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2439 ext4_commit_super(sb, es, 1);
2443 } else if (journal_inum) {
2444 if (ext4_create_journal(sb, es, journal_inum))
2446 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2447 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2448 printk(KERN_ERR "EXT4-fs: required journal recovery "
2449 "suppressed and not mounted read-only\n");
2452 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2453 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2454 sbi->s_journal = NULL;
2459 if (ext4_blocks_count(es) > 0xffffffffULL &&
2460 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2461 JBD2_FEATURE_INCOMPAT_64BIT)) {
2462 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2466 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2467 jbd2_journal_set_features(sbi->s_journal,
2468 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2469 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2470 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2471 jbd2_journal_set_features(sbi->s_journal,
2472 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2473 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2474 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2476 jbd2_journal_clear_features(sbi->s_journal,
2477 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2478 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2481 /* We have now updated the journal if required, so we can
2482 * validate the data journaling mode. */
2483 switch (test_opt(sb, DATA_FLAGS)) {
2485 /* No mode set, assume a default based on the journal
2486 * capabilities: ORDERED_DATA if the journal can
2487 * cope, else JOURNAL_DATA
2489 if (jbd2_journal_check_available_features
2490 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2491 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2493 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2496 case EXT4_MOUNT_ORDERED_DATA:
2497 case EXT4_MOUNT_WRITEBACK_DATA:
2498 if (!jbd2_journal_check_available_features
2499 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2500 printk(KERN_ERR "EXT4-fs: Journal does not support "
2501 "requested data journaling mode\n");
2510 if (test_opt(sb, NOBH)) {
2511 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2512 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2513 "its supported only with writeback mode\n");
2514 clear_opt(sbi->s_mount_opt, NOBH);
2518 * The jbd2_journal_load will have done any necessary log recovery,
2519 * so we can safely mount the rest of the filesystem now.
2522 root = ext4_iget(sb, EXT4_ROOT_INO);
2524 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2525 ret = PTR_ERR(root);
2528 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2530 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2533 sb->s_root = d_alloc_root(root);
2535 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2541 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2543 /* determine the minimum size of new large inodes, if present */
2544 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2545 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2546 EXT4_GOOD_OLD_INODE_SIZE;
2547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2548 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2549 if (sbi->s_want_extra_isize <
2550 le16_to_cpu(es->s_want_extra_isize))
2551 sbi->s_want_extra_isize =
2552 le16_to_cpu(es->s_want_extra_isize);
2553 if (sbi->s_want_extra_isize <
2554 le16_to_cpu(es->s_min_extra_isize))
2555 sbi->s_want_extra_isize =
2556 le16_to_cpu(es->s_min_extra_isize);
2559 /* Check if enough inode space is available */
2560 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2561 sbi->s_inode_size) {
2562 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2563 EXT4_GOOD_OLD_INODE_SIZE;
2564 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2568 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2569 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2570 "requested data journaling mode\n");
2571 clear_opt(sbi->s_mount_opt, DELALLOC);
2572 } else if (test_opt(sb, DELALLOC))
2573 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2576 err = ext4_mb_init(sb, needs_recovery);
2578 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2584 * akpm: core read_super() calls in here with the superblock locked.
2585 * That deadlocks, because orphan cleanup needs to lock the superblock
2586 * in numerous places. Here we just pop the lock - it's relatively
2587 * harmless, because we are now ready to accept write_super() requests,
2588 * and aviro says that's the only reason for hanging onto the
2591 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2592 ext4_orphan_cleanup(sb, es);
2593 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2594 if (needs_recovery) {
2595 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2596 ext4_mark_recovery_complete(sb, es);
2598 if (EXT4_SB(sb)->s_journal) {
2599 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2600 descr = " journalled data mode";
2601 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2602 descr = " ordered data mode";
2604 descr = " writeback data mode";
2606 descr = "out journal";
2608 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2616 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2621 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2622 if (sbi->s_journal) {
2623 jbd2_journal_destroy(sbi->s_journal);
2624 sbi->s_journal = NULL;
2627 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2628 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2629 percpu_counter_destroy(&sbi->s_dirs_counter);
2630 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2632 for (i = 0; i < db_count; i++)
2633 brelse(sbi->s_group_desc[i]);
2634 kfree(sbi->s_group_desc);
2637 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2638 remove_proc_entry(sb->s_id, ext4_proc_root);
2641 for (i = 0; i < MAXQUOTAS; i++)
2642 kfree(sbi->s_qf_names[i]);
2644 ext4_blkdev_remove(sbi);
2647 sb->s_fs_info = NULL;
2654 * Setup any per-fs journal parameters now. We'll do this both on
2655 * initial mount, once the journal has been initialised but before we've
2656 * done any recovery; and again on any subsequent remount.
2658 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2660 struct ext4_sb_info *sbi = EXT4_SB(sb);
2662 journal->j_commit_interval = sbi->s_commit_interval;
2663 journal->j_min_batch_time = sbi->s_min_batch_time;
2664 journal->j_max_batch_time = sbi->s_max_batch_time;
2666 spin_lock(&journal->j_state_lock);
2667 if (test_opt(sb, BARRIER))
2668 journal->j_flags |= JBD2_BARRIER;
2670 journal->j_flags &= ~JBD2_BARRIER;
2671 if (test_opt(sb, DATA_ERR_ABORT))
2672 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2674 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2675 spin_unlock(&journal->j_state_lock);
2678 static journal_t *ext4_get_journal(struct super_block *sb,
2679 unsigned int journal_inum)
2681 struct inode *journal_inode;
2684 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2686 /* First, test for the existence of a valid inode on disk. Bad
2687 * things happen if we iget() an unused inode, as the subsequent
2688 * iput() will try to delete it. */
2690 journal_inode = ext4_iget(sb, journal_inum);
2691 if (IS_ERR(journal_inode)) {
2692 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2695 if (!journal_inode->i_nlink) {
2696 make_bad_inode(journal_inode);
2697 iput(journal_inode);
2698 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2702 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2703 journal_inode, journal_inode->i_size);
2704 if (!S_ISREG(journal_inode->i_mode)) {
2705 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2706 iput(journal_inode);
2710 journal = jbd2_journal_init_inode(journal_inode);
2712 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2713 iput(journal_inode);
2716 journal->j_private = sb;
2717 ext4_init_journal_params(sb, journal);
2721 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2724 struct buffer_head *bh;
2728 int hblock, blocksize;
2729 ext4_fsblk_t sb_block;
2730 unsigned long offset;
2731 struct ext4_super_block *es;
2732 struct block_device *bdev;
2734 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2736 bdev = ext4_blkdev_get(j_dev);
2740 if (bd_claim(bdev, sb)) {
2742 "EXT4: failed to claim external journal device.\n");
2743 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2747 blocksize = sb->s_blocksize;
2748 hblock = bdev_hardsect_size(bdev);
2749 if (blocksize < hblock) {
2751 "EXT4-fs: blocksize too small for journal device.\n");
2755 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2756 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2757 set_blocksize(bdev, blocksize);
2758 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2759 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2760 "external journal\n");
2764 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2765 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2766 !(le32_to_cpu(es->s_feature_incompat) &
2767 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2768 printk(KERN_ERR "EXT4-fs: external journal has "
2769 "bad superblock\n");
2774 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2775 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2780 len = ext4_blocks_count(es);
2781 start = sb_block + 1;
2782 brelse(bh); /* we're done with the superblock */
2784 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2785 start, len, blocksize);
2787 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2790 journal->j_private = sb;
2791 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2792 wait_on_buffer(journal->j_sb_buffer);
2793 if (!buffer_uptodate(journal->j_sb_buffer)) {
2794 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2797 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2798 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2799 "user (unsupported) - %d\n",
2800 be32_to_cpu(journal->j_superblock->s_nr_users));
2803 EXT4_SB(sb)->journal_bdev = bdev;
2804 ext4_init_journal_params(sb, journal);
2807 jbd2_journal_destroy(journal);
2809 ext4_blkdev_put(bdev);
2813 static int ext4_load_journal(struct super_block *sb,
2814 struct ext4_super_block *es,
2815 unsigned long journal_devnum)
2818 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2821 int really_read_only;
2823 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2825 if (journal_devnum &&
2826 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2827 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2828 "numbers have changed\n");
2829 journal_dev = new_decode_dev(journal_devnum);
2831 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2833 really_read_only = bdev_read_only(sb->s_bdev);
2836 * Are we loading a blank journal or performing recovery after a
2837 * crash? For recovery, we need to check in advance whether we
2838 * can get read-write access to the device.
2841 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2842 if (sb->s_flags & MS_RDONLY) {
2843 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2844 "required on readonly filesystem.\n");
2845 if (really_read_only) {
2846 printk(KERN_ERR "EXT4-fs: write access "
2847 "unavailable, cannot proceed.\n");
2850 printk(KERN_INFO "EXT4-fs: write access will "
2851 "be enabled during recovery.\n");
2855 if (journal_inum && journal_dev) {
2856 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2857 "and inode journals!\n");
2862 if (!(journal = ext4_get_journal(sb, journal_inum)))
2865 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2869 if (journal->j_flags & JBD2_BARRIER)
2870 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2872 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2874 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2875 err = jbd2_journal_update_format(journal);
2877 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2878 jbd2_journal_destroy(journal);
2883 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2884 err = jbd2_journal_wipe(journal, !really_read_only);
2886 err = jbd2_journal_load(journal);
2889 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2890 jbd2_journal_destroy(journal);
2894 EXT4_SB(sb)->s_journal = journal;
2895 ext4_clear_journal_err(sb, es);
2897 if (journal_devnum &&
2898 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2899 es->s_journal_dev = cpu_to_le32(journal_devnum);
2902 /* Make sure we flush the recovery flag to disk. */
2903 ext4_commit_super(sb, es, 1);
2909 static int ext4_create_journal(struct super_block *sb,
2910 struct ext4_super_block *es,
2911 unsigned int journal_inum)
2916 if (sb->s_flags & MS_RDONLY) {
2917 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2918 "create journal.\n");
2922 journal = ext4_get_journal(sb, journal_inum);
2926 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2929 err = jbd2_journal_create(journal);
2931 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2932 jbd2_journal_destroy(journal);
2936 EXT4_SB(sb)->s_journal = journal;
2938 ext4_update_dynamic_rev(sb);
2939 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2940 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2942 es->s_journal_inum = cpu_to_le32(journal_inum);
2945 /* Make sure we flush the recovery flag to disk. */
2946 ext4_commit_super(sb, es, 1);
2951 static void ext4_commit_super(struct super_block *sb,
2952 struct ext4_super_block *es, int sync)
2954 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2958 if (buffer_write_io_error(sbh)) {
2960 * Oh, dear. A previous attempt to write the
2961 * superblock failed. This could happen because the
2962 * USB device was yanked out. Or it could happen to
2963 * be a transient write error and maybe the block will
2964 * be remapped. Nothing we can do but to retry the
2965 * write and hope for the best.
2967 printk(KERN_ERR "ext4: previous I/O error to "
2968 "superblock detected for %s.\n", sb->s_id);
2969 clear_buffer_write_io_error(sbh);
2970 set_buffer_uptodate(sbh);
2972 es->s_wtime = cpu_to_le32(get_seconds());
2973 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2974 &EXT4_SB(sb)->s_freeblocks_counter));
2975 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2976 &EXT4_SB(sb)->s_freeinodes_counter));
2978 BUFFER_TRACE(sbh, "marking dirty");
2979 mark_buffer_dirty(sbh);
2981 sync_dirty_buffer(sbh);
2982 if (buffer_write_io_error(sbh)) {
2983 printk(KERN_ERR "ext4: I/O error while writing "
2984 "superblock for %s.\n", sb->s_id);
2985 clear_buffer_write_io_error(sbh);
2986 set_buffer_uptodate(sbh);
2993 * Have we just finished recovery? If so, and if we are mounting (or
2994 * remounting) the filesystem readonly, then we will end up with a
2995 * consistent fs on disk. Record that fact.
2997 static void ext4_mark_recovery_complete(struct super_block *sb,
2998 struct ext4_super_block *es)
3000 journal_t *journal = EXT4_SB(sb)->s_journal;
3002 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3003 BUG_ON(journal != NULL);
3006 jbd2_journal_lock_updates(journal);
3007 if (jbd2_journal_flush(journal) < 0)
3011 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3012 sb->s_flags & MS_RDONLY) {
3013 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3015 ext4_commit_super(sb, es, 1);
3020 jbd2_journal_unlock_updates(journal);
3024 * If we are mounting (or read-write remounting) a filesystem whose journal
3025 * has recorded an error from a previous lifetime, move that error to the
3026 * main filesystem now.
3028 static void ext4_clear_journal_err(struct super_block *sb,
3029 struct ext4_super_block *es)
3035 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3037 journal = EXT4_SB(sb)->s_journal;
3040 * Now check for any error status which may have been recorded in the
3041 * journal by a prior ext4_error() or ext4_abort()
3044 j_errno = jbd2_journal_errno(journal);
3048 errstr = ext4_decode_error(sb, j_errno, nbuf);
3049 ext4_warning(sb, __func__, "Filesystem error recorded "
3050 "from previous mount: %s", errstr);
3051 ext4_warning(sb, __func__, "Marking fs in need of "
3052 "filesystem check.");
3054 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3055 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3056 ext4_commit_super(sb, es, 1);
3058 jbd2_journal_clear_err(journal);
3063 * Force the running and committing transactions to commit,
3064 * and wait on the commit.
3066 int ext4_force_commit(struct super_block *sb)
3071 if (sb->s_flags & MS_RDONLY)
3074 journal = EXT4_SB(sb)->s_journal;
3077 ret = ext4_journal_force_commit(journal);
3084 * Ext4 always journals updates to the superblock itself, so we don't
3085 * have to propagate any other updates to the superblock on disk at this
3086 * point. (We can probably nuke this function altogether, and remove
3087 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3089 static void ext4_write_super(struct super_block *sb)
3091 if (EXT4_SB(sb)->s_journal) {
3092 if (mutex_trylock(&sb->s_lock) != 0)
3096 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3100 static int ext4_sync_fs(struct super_block *sb, int wait)
3104 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3106 if (EXT4_SB(sb)->s_journal) {
3108 ret = ext4_force_commit(sb);
3110 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3112 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3118 * LVM calls this function before a (read-only) snapshot is created. This
3119 * gives us a chance to flush the journal completely and mark the fs clean.
3121 static void ext4_write_super_lockfs(struct super_block *sb)
3125 if (!(sb->s_flags & MS_RDONLY)) {
3126 journal_t *journal = EXT4_SB(sb)->s_journal;
3129 /* Now we set up the journal barrier. */
3130 jbd2_journal_lock_updates(journal);
3133 * We don't want to clear needs_recovery flag when we
3134 * failed to flush the journal.
3136 if (jbd2_journal_flush(journal) < 0)
3140 /* Journal blocked and flushed, clear needs_recovery flag. */
3141 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3142 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3147 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3148 * flag here, even though the filesystem is not technically dirty yet.
3150 static void ext4_unlockfs(struct super_block *sb)
3152 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3154 /* Reser the needs_recovery flag before the fs is unlocked. */
3155 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3156 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3158 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3162 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3164 struct ext4_super_block *es;
3165 struct ext4_sb_info *sbi = EXT4_SB(sb);
3166 ext4_fsblk_t n_blocks_count = 0;
3167 unsigned long old_sb_flags;
3168 struct ext4_mount_options old_opts;
3175 /* Store the original options */
3176 old_sb_flags = sb->s_flags;
3177 old_opts.s_mount_opt = sbi->s_mount_opt;
3178 old_opts.s_resuid = sbi->s_resuid;
3179 old_opts.s_resgid = sbi->s_resgid;
3180 old_opts.s_commit_interval = sbi->s_commit_interval;
3181 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3182 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3184 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3185 for (i = 0; i < MAXQUOTAS; i++)
3186 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3190 * Allow the "check" option to be passed as a remount option.
3192 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3197 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3198 ext4_abort(sb, __func__, "Abort forced by user");
3200 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3201 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3206 ext4_init_journal_params(sb, sbi->s_journal);
3208 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3209 n_blocks_count > ext4_blocks_count(es)) {
3210 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3215 if (*flags & MS_RDONLY) {
3217 * First of all, the unconditional stuff we have to do
3218 * to disable replay of the journal when we next remount
3220 sb->s_flags |= MS_RDONLY;
3223 * OK, test if we are remounting a valid rw partition
3224 * readonly, and if so set the rdonly flag and then
3225 * mark the partition as valid again.
3227 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3228 (sbi->s_mount_state & EXT4_VALID_FS))
3229 es->s_state = cpu_to_le16(sbi->s_mount_state);
3232 * We have to unlock super so that we can wait for
3235 if (sbi->s_journal) {
3237 ext4_mark_recovery_complete(sb, es);
3242 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3243 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3244 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3245 "remount RDWR because of unsupported "
3246 "optional features (%x).\n", sb->s_id,
3247 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3248 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3254 * Make sure the group descriptor checksums
3255 * are sane. If they aren't, refuse to
3258 for (g = 0; g < sbi->s_groups_count; g++) {
3259 struct ext4_group_desc *gdp =
3260 ext4_get_group_desc(sb, g, NULL);
3262 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3264 "EXT4-fs: ext4_remount: "
3265 "Checksum for group %u failed (%u!=%u)\n",
3266 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3267 le16_to_cpu(gdp->bg_checksum));
3274 * If we have an unprocessed orphan list hanging
3275 * around from a previously readonly bdev mount,
3276 * require a full umount/remount for now.
3278 if (es->s_last_orphan) {
3279 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3280 "remount RDWR because of unprocessed "
3281 "orphan inode list. Please "
3282 "umount/remount instead.\n",
3289 * Mounting a RDONLY partition read-write, so reread
3290 * and store the current valid flag. (It may have
3291 * been changed by e2fsck since we originally mounted
3295 ext4_clear_journal_err(sb, es);
3296 sbi->s_mount_state = le16_to_cpu(es->s_state);
3297 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3299 if (!ext4_setup_super(sb, es, 0))
3300 sb->s_flags &= ~MS_RDONLY;
3303 if (sbi->s_journal == NULL)
3304 ext4_commit_super(sb, es, 1);
3307 /* Release old quota file names */
3308 for (i = 0; i < MAXQUOTAS; i++)
3309 if (old_opts.s_qf_names[i] &&
3310 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3311 kfree(old_opts.s_qf_names[i]);
3315 sb->s_flags = old_sb_flags;
3316 sbi->s_mount_opt = old_opts.s_mount_opt;
3317 sbi->s_resuid = old_opts.s_resuid;
3318 sbi->s_resgid = old_opts.s_resgid;
3319 sbi->s_commit_interval = old_opts.s_commit_interval;
3320 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3321 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3323 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3324 for (i = 0; i < MAXQUOTAS; i++) {
3325 if (sbi->s_qf_names[i] &&
3326 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3327 kfree(sbi->s_qf_names[i]);
3328 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3334 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3336 struct super_block *sb = dentry->d_sb;
3337 struct ext4_sb_info *sbi = EXT4_SB(sb);
3338 struct ext4_super_block *es = sbi->s_es;
3341 if (test_opt(sb, MINIX_DF)) {
3342 sbi->s_overhead_last = 0;
3343 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3344 ext4_group_t ngroups = sbi->s_groups_count, i;
3345 ext4_fsblk_t overhead = 0;
3349 * Compute the overhead (FS structures). This is constant
3350 * for a given filesystem unless the number of block groups
3351 * changes so we cache the previous value until it does.
3355 * All of the blocks before first_data_block are
3358 overhead = le32_to_cpu(es->s_first_data_block);
3361 * Add the overhead attributed to the superblock and
3362 * block group descriptors. If the sparse superblocks
3363 * feature is turned on, then not all groups have this.
3365 for (i = 0; i < ngroups; i++) {
3366 overhead += ext4_bg_has_super(sb, i) +
3367 ext4_bg_num_gdb(sb, i);
3372 * Every block group has an inode bitmap, a block
3373 * bitmap, and an inode table.
3375 overhead += ngroups * (2 + sbi->s_itb_per_group);
3376 sbi->s_overhead_last = overhead;
3378 sbi->s_blocks_last = ext4_blocks_count(es);
3381 buf->f_type = EXT4_SUPER_MAGIC;
3382 buf->f_bsize = sb->s_blocksize;
3383 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3384 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3385 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3386 ext4_free_blocks_count_set(es, buf->f_bfree);
3387 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3388 if (buf->f_bfree < ext4_r_blocks_count(es))
3390 buf->f_files = le32_to_cpu(es->s_inodes_count);
3391 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3392 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3393 buf->f_namelen = EXT4_NAME_LEN;
3394 fsid = le64_to_cpup((void *)es->s_uuid) ^
3395 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3396 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3397 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3401 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3402 * is locked for write. Otherwise the are possible deadlocks:
3403 * Process 1 Process 2
3404 * ext4_create() quota_sync()
3405 * jbd2_journal_start() write_dquot()
3406 * DQUOT_INIT() down(dqio_mutex)
3407 * down(dqio_mutex) jbd2_journal_start()
3413 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3415 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3418 static int ext4_dquot_initialize(struct inode *inode, int type)
3423 /* We may create quota structure so we need to reserve enough blocks */
3424 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3426 return PTR_ERR(handle);
3427 ret = dquot_initialize(inode, type);
3428 err = ext4_journal_stop(handle);
3434 static int ext4_dquot_drop(struct inode *inode)
3439 /* We may delete quota structure so we need to reserve enough blocks */
3440 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3441 if (IS_ERR(handle)) {
3443 * We call dquot_drop() anyway to at least release references
3444 * to quota structures so that umount does not hang.
3447 return PTR_ERR(handle);
3449 ret = dquot_drop(inode);
3450 err = ext4_journal_stop(handle);
3456 static int ext4_write_dquot(struct dquot *dquot)
3460 struct inode *inode;
3462 inode = dquot_to_inode(dquot);
3463 handle = ext4_journal_start(inode,
3464 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3466 return PTR_ERR(handle);
3467 ret = dquot_commit(dquot);
3468 err = ext4_journal_stop(handle);
3474 static int ext4_acquire_dquot(struct dquot *dquot)
3479 handle = ext4_journal_start(dquot_to_inode(dquot),
3480 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3482 return PTR_ERR(handle);
3483 ret = dquot_acquire(dquot);
3484 err = ext4_journal_stop(handle);
3490 static int ext4_release_dquot(struct dquot *dquot)
3495 handle = ext4_journal_start(dquot_to_inode(dquot),
3496 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3497 if (IS_ERR(handle)) {
3498 /* Release dquot anyway to avoid endless cycle in dqput() */
3499 dquot_release(dquot);
3500 return PTR_ERR(handle);
3502 ret = dquot_release(dquot);
3503 err = ext4_journal_stop(handle);
3509 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3511 /* Are we journaling quotas? */
3512 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3513 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3514 dquot_mark_dquot_dirty(dquot);
3515 return ext4_write_dquot(dquot);
3517 return dquot_mark_dquot_dirty(dquot);
3521 static int ext4_write_info(struct super_block *sb, int type)
3526 /* Data block + inode block */
3527 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3529 return PTR_ERR(handle);
3530 ret = dquot_commit_info(sb, type);
3531 err = ext4_journal_stop(handle);
3538 * Turn on quotas during mount time - we need to find
3539 * the quota file and such...
3541 static int ext4_quota_on_mount(struct super_block *sb, int type)
3543 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3544 EXT4_SB(sb)->s_jquota_fmt, type);
3548 * Standard function to be called on quota_on
3550 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3551 char *name, int remount)
3556 if (!test_opt(sb, QUOTA))
3558 /* When remounting, no checks are needed and in fact, name is NULL */
3560 return vfs_quota_on(sb, type, format_id, name, remount);
3562 err = kern_path(name, LOOKUP_FOLLOW, &path);
3566 /* Quotafile not on the same filesystem? */
3567 if (path.mnt->mnt_sb != sb) {
3571 /* Journaling quota? */
3572 if (EXT4_SB(sb)->s_qf_names[type]) {
3573 /* Quotafile not in fs root? */
3574 if (path.dentry->d_parent != sb->s_root)
3576 "EXT4-fs: Quota file not on filesystem root. "
3577 "Journaled quota will not work.\n");
3581 * When we journal data on quota file, we have to flush journal to see
3582 * all updates to the file when we bypass pagecache...
3584 if (EXT4_SB(sb)->s_journal &&
3585 ext4_should_journal_data(path.dentry->d_inode)) {
3587 * We don't need to lock updates but journal_flush() could
3588 * otherwise be livelocked...
3590 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3591 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3592 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3599 err = vfs_quota_on_path(sb, type, format_id, &path);
3604 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3605 * acquiring the locks... As quota files are never truncated and quota code
3606 * itself serializes the operations (and noone else should touch the files)
3607 * we don't have to be afraid of races */
3608 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3609 size_t len, loff_t off)
3611 struct inode *inode = sb_dqopt(sb)->files[type];
3612 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3614 int offset = off & (sb->s_blocksize - 1);
3617 struct buffer_head *bh;
3618 loff_t i_size = i_size_read(inode);
3622 if (off+len > i_size)
3625 while (toread > 0) {
3626 tocopy = sb->s_blocksize - offset < toread ?
3627 sb->s_blocksize - offset : toread;
3628 bh = ext4_bread(NULL, inode, blk, 0, &err);
3631 if (!bh) /* A hole? */
3632 memset(data, 0, tocopy);
3634 memcpy(data, bh->b_data+offset, tocopy);
3644 /* Write to quotafile (we know the transaction is already started and has
3645 * enough credits) */
3646 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3647 const char *data, size_t len, loff_t off)
3649 struct inode *inode = sb_dqopt(sb)->files[type];
3650 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3652 int offset = off & (sb->s_blocksize - 1);
3654 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3655 size_t towrite = len;
3656 struct buffer_head *bh;
3657 handle_t *handle = journal_current_handle();
3659 if (EXT4_SB(sb)->s_journal && !handle) {
3660 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3661 " cancelled because transaction is not started.\n",
3662 (unsigned long long)off, (unsigned long long)len);
3665 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3666 while (towrite > 0) {
3667 tocopy = sb->s_blocksize - offset < towrite ?
3668 sb->s_blocksize - offset : towrite;
3669 bh = ext4_bread(handle, inode, blk, 1, &err);
3672 if (journal_quota) {
3673 err = ext4_journal_get_write_access(handle, bh);
3680 memcpy(bh->b_data+offset, data, tocopy);
3681 flush_dcache_page(bh->b_page);
3684 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3686 /* Always do at least ordered writes for quotas */
3687 err = ext4_jbd2_file_inode(handle, inode);
3688 mark_buffer_dirty(bh);
3699 if (len == towrite) {
3700 mutex_unlock(&inode->i_mutex);
3703 if (inode->i_size < off+len-towrite) {
3704 i_size_write(inode, off+len-towrite);
3705 EXT4_I(inode)->i_disksize = inode->i_size;
3707 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3708 ext4_mark_inode_dirty(handle, inode);
3709 mutex_unlock(&inode->i_mutex);
3710 return len - towrite;
3715 static int ext4_get_sb(struct file_system_type *fs_type,
3716 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3718 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3721 #ifdef CONFIG_PROC_FS
3722 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3724 unsigned int *p = m->private;
3726 seq_printf(m, "%u\n", *p);
3730 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3732 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3735 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3736 size_t cnt, loff_t *ppos)
3738 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3741 if (cnt >= sizeof(str))
3743 if (copy_from_user(str, buf, cnt))
3746 *p = simple_strtoul(str, NULL, 0);
3750 const struct file_operations ext4_ui_proc_fops = {
3751 .owner = THIS_MODULE,
3752 .open = ext4_ui_proc_open,
3754 .llseek = seq_lseek,
3755 .release = single_release,
3756 .write = ext4_ui_proc_write,
3760 static struct file_system_type ext4_fs_type = {
3761 .owner = THIS_MODULE,
3763 .get_sb = ext4_get_sb,
3764 .kill_sb = kill_block_super,
3765 .fs_flags = FS_REQUIRES_DEV,
3768 #ifdef CONFIG_EXT4DEV_COMPAT
3769 static int ext4dev_get_sb(struct file_system_type *fs_type,
3770 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3772 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3773 "to mount using ext4\n");
3774 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3775 "will go away by 2.6.31\n");
3776 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3779 static struct file_system_type ext4dev_fs_type = {
3780 .owner = THIS_MODULE,
3782 .get_sb = ext4dev_get_sb,
3783 .kill_sb = kill_block_super,
3784 .fs_flags = FS_REQUIRES_DEV,
3786 MODULE_ALIAS("ext4dev");
3789 static int __init init_ext4_fs(void)
3793 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3794 err = init_ext4_mballoc();
3798 err = init_ext4_xattr();
3801 err = init_inodecache();
3804 err = register_filesystem(&ext4_fs_type);
3807 #ifdef CONFIG_EXT4DEV_COMPAT
3808 err = register_filesystem(&ext4dev_fs_type);
3810 unregister_filesystem(&ext4_fs_type);
3816 destroy_inodecache();
3820 exit_ext4_mballoc();
3824 static void __exit exit_ext4_fs(void)
3826 unregister_filesystem(&ext4_fs_type);
3827 #ifdef CONFIG_EXT4DEV_COMPAT
3828 unregister_filesystem(&ext4dev_fs_type);
3830 destroy_inodecache();
3832 exit_ext4_mballoc();
3833 remove_proc_entry("fs/ext4", NULL);
3836 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3837 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3838 MODULE_LICENSE("GPL");
3839 module_init(init_ext4_fs)
3840 module_exit(exit_ext4_fs)