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 void ext4_commit_super(struct super_block *sb,
55 struct ext4_super_block *es, int sync);
56 static void ext4_mark_recovery_complete(struct super_block *sb,
57 struct ext4_super_block *es);
58 static void ext4_clear_journal_err(struct super_block *sb,
59 struct ext4_super_block *es);
60 static int ext4_sync_fs(struct super_block *sb, int wait);
61 static const char *ext4_decode_error(struct super_block *sb, int errno,
63 static int ext4_remount(struct super_block *sb, int *flags, char *data);
64 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
65 static void ext4_unlockfs(struct super_block *sb);
66 static void ext4_write_super(struct super_block *sb);
67 static void ext4_write_super_lockfs(struct super_block *sb);
70 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
73 return le32_to_cpu(bg->bg_block_bitmap_lo) |
74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
78 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
81 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
86 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
89 return le32_to_cpu(bg->bg_inode_table_lo) |
90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
94 __u32 ext4_free_blks_count(struct super_block *sb,
95 struct ext4_group_desc *bg)
97 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
98 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
99 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
102 __u32 ext4_free_inodes_count(struct super_block *sb,
103 struct ext4_group_desc *bg)
105 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
107 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
110 __u32 ext4_used_dirs_count(struct super_block *sb,
111 struct ext4_group_desc *bg)
113 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
115 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
118 __u32 ext4_itable_unused_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
121 return le16_to_cpu(bg->bg_itable_unused_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
123 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
126 void ext4_block_bitmap_set(struct super_block *sb,
127 struct ext4_group_desc *bg, ext4_fsblk_t blk)
129 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
130 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
131 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
134 void ext4_inode_bitmap_set(struct super_block *sb,
135 struct ext4_group_desc *bg, ext4_fsblk_t blk)
137 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
138 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
139 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
142 void ext4_inode_table_set(struct super_block *sb,
143 struct ext4_group_desc *bg, ext4_fsblk_t blk)
145 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
146 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
147 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
150 void ext4_free_blks_set(struct super_block *sb,
151 struct ext4_group_desc *bg, __u32 count)
153 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
158 void ext4_free_inodes_set(struct super_block *sb,
159 struct ext4_group_desc *bg, __u32 count)
161 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
166 void ext4_used_dirs_set(struct super_block *sb,
167 struct ext4_group_desc *bg, __u32 count)
169 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
174 void ext4_itable_unused_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
177 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183 * Wrappers for jbd2_journal_start/end.
185 * The only special thing we need to do here is to make sure that all
186 * journal_end calls result in the superblock being marked dirty, so
187 * that sync() will call the filesystem's write_super callback if
190 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
194 if (sb->s_flags & MS_RDONLY)
195 return ERR_PTR(-EROFS);
197 /* Special case here: if the journal has aborted behind our
198 * backs (eg. EIO in the commit thread), then we still need to
199 * take the FS itself readonly cleanly. */
200 journal = EXT4_SB(sb)->s_journal;
202 if (is_journal_aborted(journal)) {
203 ext4_abort(sb, __func__,
204 "Detected aborted journal");
205 return ERR_PTR(-EROFS);
207 return jbd2_journal_start(journal, nblocks);
210 * We're not journaling, return the appropriate indication.
212 current->journal_info = EXT4_NOJOURNAL_HANDLE;
213 return current->journal_info;
217 * The only special thing we need to do here is to make sure that all
218 * jbd2_journal_stop calls result in the superblock being marked dirty, so
219 * that sync() will call the filesystem's write_super callback if
222 int __ext4_journal_stop(const char *where, handle_t *handle)
224 struct super_block *sb;
228 if (!ext4_handle_valid(handle)) {
230 * Do this here since we don't call jbd2_journal_stop() in
233 current->journal_info = NULL;
236 sb = handle->h_transaction->t_journal->j_private;
238 rc = jbd2_journal_stop(handle);
243 __ext4_std_error(sb, where, err);
247 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
248 struct buffer_head *bh, handle_t *handle, int err)
251 const char *errstr = ext4_decode_error(NULL, err, nbuf);
253 BUG_ON(!ext4_handle_valid(handle));
256 BUFFER_TRACE(bh, "abort");
261 if (is_handle_aborted(handle))
264 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
265 caller, errstr, err_fn);
267 jbd2_journal_abort_handle(handle);
270 /* Deal with the reporting of failure conditions on a filesystem such as
271 * inconsistencies detected or read IO failures.
273 * On ext2, we can store the error state of the filesystem in the
274 * superblock. That is not possible on ext4, because we may have other
275 * write ordering constraints on the superblock which prevent us from
276 * writing it out straight away; and given that the journal is about to
277 * be aborted, we can't rely on the current, or future, transactions to
278 * write out the superblock safely.
280 * We'll just use the jbd2_journal_abort() error code to record an error in
281 * the journal instead. On recovery, the journal will compain about
282 * that error until we've noted it down and cleared it.
285 static void ext4_handle_error(struct super_block *sb)
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
290 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
292 if (sb->s_flags & MS_RDONLY)
295 if (!test_opt(sb, ERRORS_CONT)) {
296 journal_t *journal = EXT4_SB(sb)->s_journal;
298 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
300 jbd2_journal_abort(journal, -EIO);
302 if (test_opt(sb, ERRORS_RO)) {
303 printk(KERN_CRIT "Remounting filesystem read-only\n");
304 sb->s_flags |= MS_RDONLY;
306 ext4_commit_super(sb, es, 1);
307 if (test_opt(sb, ERRORS_PANIC))
308 panic("EXT4-fs (device %s): panic forced after error\n",
312 void ext4_error(struct super_block *sb, const char *function,
313 const char *fmt, ...)
318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
323 ext4_handle_error(sb);
326 static const char *ext4_decode_error(struct super_block *sb, int errno,
333 errstr = "IO failure";
336 errstr = "Out of memory";
339 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
340 errstr = "Journal has aborted";
342 errstr = "Readonly filesystem";
345 /* If the caller passed in an extra buffer for unknown
346 * errors, textualise them now. Else we just return
349 /* Check for truncated error codes... */
350 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
359 /* __ext4_std_error decodes expected errors from journaling functions
360 * automatically and invokes the appropriate error response. */
362 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
367 /* Special case: if the error is EROFS, and we're not already
368 * inside a transaction, then there's really no point in logging
370 if (errno == -EROFS && journal_current_handle() == NULL &&
371 (sb->s_flags & MS_RDONLY))
374 errstr = ext4_decode_error(sb, errno, nbuf);
375 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
376 sb->s_id, function, errstr);
378 ext4_handle_error(sb);
382 * ext4_abort is a much stronger failure handler than ext4_error. The
383 * abort function may be used to deal with unrecoverable failures such
384 * as journal IO errors or ENOMEM at a critical moment in log management.
386 * We unconditionally force the filesystem into an ABORT|READONLY state,
387 * unless the error response on the fs has been set to panic in which
388 * case we take the easy way out and panic immediately.
391 void ext4_abort(struct super_block *sb, const char *function,
392 const char *fmt, ...)
396 printk(KERN_CRIT "ext4_abort called.\n");
399 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
404 if (test_opt(sb, ERRORS_PANIC))
405 panic("EXT4-fs panic from previous error\n");
407 if (sb->s_flags & MS_RDONLY)
410 printk(KERN_CRIT "Remounting filesystem read-only\n");
411 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
412 sb->s_flags |= MS_RDONLY;
413 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
414 if (EXT4_SB(sb)->s_journal)
415 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
418 void ext4_warning(struct super_block *sb, const char *function,
419 const char *fmt, ...)
424 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
431 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
432 const char *function, const char *fmt, ...)
437 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
445 if (test_opt(sb, ERRORS_CONT)) {
446 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
448 ext4_commit_super(sb, es, 0);
451 ext4_unlock_group(sb, grp);
452 ext4_handle_error(sb);
454 * We only get here in the ERRORS_RO case; relocking the group
455 * may be dangerous, but nothing bad will happen since the
456 * filesystem will have already been marked read/only and the
457 * journal has been aborted. We return 1 as a hint to callers
458 * who might what to use the return value from
459 * ext4_grp_locked_error() to distinguish beween the
460 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
461 * aggressively from the ext4 function in question, with a
462 * more appropriate error code.
464 ext4_lock_group(sb, grp);
469 void ext4_update_dynamic_rev(struct super_block *sb)
471 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
476 ext4_warning(sb, __func__,
477 "updating to rev %d because of new feature flag, "
478 "running e2fsck is recommended",
481 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
482 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
483 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
484 /* leave es->s_feature_*compat flags alone */
485 /* es->s_uuid will be set by e2fsck if empty */
488 * The rest of the superblock fields should be zero, and if not it
489 * means they are likely already in use, so leave them alone. We
490 * can leave it up to e2fsck to clean up any inconsistencies there.
495 * Open the external journal device
497 static struct block_device *ext4_blkdev_get(dev_t dev)
499 struct block_device *bdev;
500 char b[BDEVNAME_SIZE];
502 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
508 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
509 __bdevname(dev, b), PTR_ERR(bdev));
514 * Release the journal device
516 static int ext4_blkdev_put(struct block_device *bdev)
519 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
522 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
524 struct block_device *bdev;
527 bdev = sbi->journal_bdev;
529 ret = ext4_blkdev_put(bdev);
530 sbi->journal_bdev = NULL;
535 static inline struct inode *orphan_list_entry(struct list_head *l)
537 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
540 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
544 printk(KERN_ERR "sb orphan head is %d\n",
545 le32_to_cpu(sbi->s_es->s_last_orphan));
547 printk(KERN_ERR "sb_info orphan list:\n");
548 list_for_each(l, &sbi->s_orphan) {
549 struct inode *inode = orphan_list_entry(l);
551 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
552 inode->i_sb->s_id, inode->i_ino, inode,
553 inode->i_mode, inode->i_nlink,
558 static void ext4_put_super(struct super_block *sb)
560 struct ext4_sb_info *sbi = EXT4_SB(sb);
561 struct ext4_super_block *es = sbi->s_es;
565 ext4_ext_release(sb);
566 ext4_xattr_put_super(sb);
567 if (sbi->s_journal) {
568 err = jbd2_journal_destroy(sbi->s_journal);
569 sbi->s_journal = NULL;
571 ext4_abort(sb, __func__,
572 "Couldn't clean up the journal");
574 if (!(sb->s_flags & MS_RDONLY)) {
575 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
576 es->s_state = cpu_to_le16(sbi->s_mount_state);
577 ext4_commit_super(sb, es, 1);
580 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
581 remove_proc_entry(sb->s_id, ext4_proc_root);
584 for (i = 0; i < sbi->s_gdb_count; i++)
585 brelse(sbi->s_group_desc[i]);
586 kfree(sbi->s_group_desc);
587 kfree(sbi->s_flex_groups);
588 percpu_counter_destroy(&sbi->s_freeblocks_counter);
589 percpu_counter_destroy(&sbi->s_freeinodes_counter);
590 percpu_counter_destroy(&sbi->s_dirs_counter);
591 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
594 for (i = 0; i < MAXQUOTAS; i++)
595 kfree(sbi->s_qf_names[i]);
598 /* Debugging code just in case the in-memory inode orphan list
599 * isn't empty. The on-disk one can be non-empty if we've
600 * detected an error and taken the fs readonly, but the
601 * in-memory list had better be clean by this point. */
602 if (!list_empty(&sbi->s_orphan))
603 dump_orphan_list(sb, sbi);
604 J_ASSERT(list_empty(&sbi->s_orphan));
606 invalidate_bdev(sb->s_bdev);
607 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
609 * Invalidate the journal device's buffers. We don't want them
610 * floating about in memory - the physical journal device may
611 * hotswapped, and it breaks the `ro-after' testing code.
613 sync_blockdev(sbi->journal_bdev);
614 invalidate_bdev(sbi->journal_bdev);
615 ext4_blkdev_remove(sbi);
617 sb->s_fs_info = NULL;
622 static struct kmem_cache *ext4_inode_cachep;
625 * Called inside transaction, so use GFP_NOFS
627 static struct inode *ext4_alloc_inode(struct super_block *sb)
629 struct ext4_inode_info *ei;
631 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
634 #ifdef CONFIG_EXT4_FS_POSIX_ACL
635 ei->i_acl = EXT4_ACL_NOT_CACHED;
636 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
638 ei->vfs_inode.i_version = 1;
639 ei->vfs_inode.i_data.writeback_index = 0;
640 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
641 INIT_LIST_HEAD(&ei->i_prealloc_list);
642 spin_lock_init(&ei->i_prealloc_lock);
644 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
645 * therefore it can be null here. Don't check it, just initialize
648 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
649 ei->i_reserved_data_blocks = 0;
650 ei->i_reserved_meta_blocks = 0;
651 ei->i_allocated_meta_blocks = 0;
652 ei->i_delalloc_reserved_flag = 0;
653 spin_lock_init(&(ei->i_block_reservation_lock));
654 return &ei->vfs_inode;
657 static void ext4_destroy_inode(struct inode *inode)
659 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
660 printk("EXT4 Inode %p: orphan list check failed!\n",
662 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
663 EXT4_I(inode), sizeof(struct ext4_inode_info),
667 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
670 static void init_once(void *foo)
672 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
674 INIT_LIST_HEAD(&ei->i_orphan);
675 #ifdef CONFIG_EXT4_FS_XATTR
676 init_rwsem(&ei->xattr_sem);
678 init_rwsem(&ei->i_data_sem);
679 inode_init_once(&ei->vfs_inode);
682 static int init_inodecache(void)
684 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
685 sizeof(struct ext4_inode_info),
686 0, (SLAB_RECLAIM_ACCOUNT|
689 if (ext4_inode_cachep == NULL)
694 static void destroy_inodecache(void)
696 kmem_cache_destroy(ext4_inode_cachep);
699 static void ext4_clear_inode(struct inode *inode)
701 #ifdef CONFIG_EXT4_FS_POSIX_ACL
702 if (EXT4_I(inode)->i_acl &&
703 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
704 posix_acl_release(EXT4_I(inode)->i_acl);
705 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
707 if (EXT4_I(inode)->i_default_acl &&
708 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
709 posix_acl_release(EXT4_I(inode)->i_default_acl);
710 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
713 ext4_discard_preallocations(inode);
714 if (EXT4_JOURNAL(inode))
715 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
716 &EXT4_I(inode)->jinode);
719 static inline void ext4_show_quota_options(struct seq_file *seq,
720 struct super_block *sb)
722 #if defined(CONFIG_QUOTA)
723 struct ext4_sb_info *sbi = EXT4_SB(sb);
725 if (sbi->s_jquota_fmt)
726 seq_printf(seq, ",jqfmt=%s",
727 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
729 if (sbi->s_qf_names[USRQUOTA])
730 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
732 if (sbi->s_qf_names[GRPQUOTA])
733 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
735 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
736 seq_puts(seq, ",usrquota");
738 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
739 seq_puts(seq, ",grpquota");
745 * - it's set to a non-default value OR
746 * - if the per-sb default is different from the global default
748 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
751 unsigned long def_mount_opts;
752 struct super_block *sb = vfs->mnt_sb;
753 struct ext4_sb_info *sbi = EXT4_SB(sb);
754 struct ext4_super_block *es = sbi->s_es;
756 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
757 def_errors = le16_to_cpu(es->s_errors);
759 if (sbi->s_sb_block != 1)
760 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
761 if (test_opt(sb, MINIX_DF))
762 seq_puts(seq, ",minixdf");
763 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
764 seq_puts(seq, ",grpid");
765 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
766 seq_puts(seq, ",nogrpid");
767 if (sbi->s_resuid != EXT4_DEF_RESUID ||
768 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
769 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
771 if (sbi->s_resgid != EXT4_DEF_RESGID ||
772 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
773 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
775 if (test_opt(sb, ERRORS_RO)) {
776 if (def_errors == EXT4_ERRORS_PANIC ||
777 def_errors == EXT4_ERRORS_CONTINUE) {
778 seq_puts(seq, ",errors=remount-ro");
781 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
782 seq_puts(seq, ",errors=continue");
783 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
784 seq_puts(seq, ",errors=panic");
785 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
786 seq_puts(seq, ",nouid32");
787 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
788 seq_puts(seq, ",debug");
789 if (test_opt(sb, OLDALLOC))
790 seq_puts(seq, ",oldalloc");
791 #ifdef CONFIG_EXT4_FS_XATTR
792 if (test_opt(sb, XATTR_USER) &&
793 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
794 seq_puts(seq, ",user_xattr");
795 if (!test_opt(sb, XATTR_USER) &&
796 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
797 seq_puts(seq, ",nouser_xattr");
800 #ifdef CONFIG_EXT4_FS_POSIX_ACL
801 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
802 seq_puts(seq, ",acl");
803 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
804 seq_puts(seq, ",noacl");
806 if (!test_opt(sb, RESERVATION))
807 seq_puts(seq, ",noreservation");
808 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
809 seq_printf(seq, ",commit=%u",
810 (unsigned) (sbi->s_commit_interval / HZ));
812 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
813 seq_printf(seq, ",min_batch_time=%u",
814 (unsigned) sbi->s_min_batch_time);
816 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
817 seq_printf(seq, ",max_batch_time=%u",
818 (unsigned) sbi->s_min_batch_time);
822 * We're changing the default of barrier mount option, so
823 * let's always display its mount state so it's clear what its
826 seq_puts(seq, ",barrier=");
827 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
828 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
829 seq_puts(seq, ",journal_async_commit");
830 if (test_opt(sb, NOBH))
831 seq_puts(seq, ",nobh");
832 if (!test_opt(sb, EXTENTS))
833 seq_puts(seq, ",noextents");
834 if (test_opt(sb, I_VERSION))
835 seq_puts(seq, ",i_version");
836 if (!test_opt(sb, DELALLOC))
837 seq_puts(seq, ",nodelalloc");
841 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
843 * journal mode get enabled in different ways
844 * So just print the value even if we didn't specify it
846 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
847 seq_puts(seq, ",data=journal");
848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
849 seq_puts(seq, ",data=ordered");
850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
851 seq_puts(seq, ",data=writeback");
853 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
854 seq_printf(seq, ",inode_readahead_blks=%u",
855 sbi->s_inode_readahead_blks);
857 if (test_opt(sb, DATA_ERR_ABORT))
858 seq_puts(seq, ",data_err=abort");
860 ext4_show_quota_options(seq, sb);
865 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
866 u64 ino, u32 generation)
870 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
871 return ERR_PTR(-ESTALE);
872 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
873 return ERR_PTR(-ESTALE);
875 /* iget isn't really right if the inode is currently unallocated!!
877 * ext4_read_inode will return a bad_inode if the inode had been
878 * deleted, so we should be safe.
880 * Currently we don't know the generation for parent directory, so
881 * a generation of 0 means "accept any"
883 inode = ext4_iget(sb, ino);
885 return ERR_CAST(inode);
886 if (generation && inode->i_generation != generation) {
888 return ERR_PTR(-ESTALE);
894 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
895 int fh_len, int fh_type)
897 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
901 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
902 int fh_len, int fh_type)
904 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
909 * Try to release metadata pages (indirect blocks, directories) which are
910 * mapped via the block device. Since these pages could have journal heads
911 * which would prevent try_to_free_buffers() from freeing them, we must use
912 * jbd2 layer's try_to_free_buffers() function to release them.
914 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
916 journal_t *journal = EXT4_SB(sb)->s_journal;
918 WARN_ON(PageChecked(page));
919 if (!page_has_buffers(page))
922 return jbd2_journal_try_to_free_buffers(journal, page,
924 return try_to_free_buffers(page);
928 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
929 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
931 static int ext4_dquot_initialize(struct inode *inode, int type);
932 static int ext4_dquot_drop(struct inode *inode);
933 static int ext4_write_dquot(struct dquot *dquot);
934 static int ext4_acquire_dquot(struct dquot *dquot);
935 static int ext4_release_dquot(struct dquot *dquot);
936 static int ext4_mark_dquot_dirty(struct dquot *dquot);
937 static int ext4_write_info(struct super_block *sb, int type);
938 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
939 char *path, int remount);
940 static int ext4_quota_on_mount(struct super_block *sb, int type);
941 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
942 size_t len, loff_t off);
943 static ssize_t ext4_quota_write(struct super_block *sb, int type,
944 const char *data, size_t len, loff_t off);
946 static struct dquot_operations ext4_quota_operations = {
947 .initialize = ext4_dquot_initialize,
948 .drop = ext4_dquot_drop,
949 .alloc_space = dquot_alloc_space,
950 .alloc_inode = dquot_alloc_inode,
951 .free_space = dquot_free_space,
952 .free_inode = dquot_free_inode,
953 .transfer = dquot_transfer,
954 .write_dquot = ext4_write_dquot,
955 .acquire_dquot = ext4_acquire_dquot,
956 .release_dquot = ext4_release_dquot,
957 .mark_dirty = ext4_mark_dquot_dirty,
958 .write_info = ext4_write_info
961 static struct quotactl_ops ext4_qctl_operations = {
962 .quota_on = ext4_quota_on,
963 .quota_off = vfs_quota_off,
964 .quota_sync = vfs_quota_sync,
965 .get_info = vfs_get_dqinfo,
966 .set_info = vfs_set_dqinfo,
967 .get_dqblk = vfs_get_dqblk,
968 .set_dqblk = vfs_set_dqblk
972 static const struct super_operations ext4_sops = {
973 .alloc_inode = ext4_alloc_inode,
974 .destroy_inode = ext4_destroy_inode,
975 .write_inode = ext4_write_inode,
976 .dirty_inode = ext4_dirty_inode,
977 .delete_inode = ext4_delete_inode,
978 .put_super = ext4_put_super,
979 .write_super = ext4_write_super,
980 .sync_fs = ext4_sync_fs,
981 .write_super_lockfs = ext4_write_super_lockfs,
982 .unlockfs = ext4_unlockfs,
983 .statfs = ext4_statfs,
984 .remount_fs = ext4_remount,
985 .clear_inode = ext4_clear_inode,
986 .show_options = ext4_show_options,
988 .quota_read = ext4_quota_read,
989 .quota_write = ext4_quota_write,
991 .bdev_try_to_free_page = bdev_try_to_free_page,
994 static const struct export_operations ext4_export_ops = {
995 .fh_to_dentry = ext4_fh_to_dentry,
996 .fh_to_parent = ext4_fh_to_parent,
997 .get_parent = ext4_get_parent,
1001 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1002 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1003 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1004 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1005 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
1006 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1007 Opt_journal_update, Opt_journal_dev,
1008 Opt_journal_checksum, Opt_journal_async_commit,
1009 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1010 Opt_data_err_abort, Opt_data_err_ignore,
1011 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1012 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1013 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1014 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
1015 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1016 Opt_inode_readahead_blks
1019 static const match_table_t tokens = {
1020 {Opt_bsd_df, "bsddf"},
1021 {Opt_minix_df, "minixdf"},
1022 {Opt_grpid, "grpid"},
1023 {Opt_grpid, "bsdgroups"},
1024 {Opt_nogrpid, "nogrpid"},
1025 {Opt_nogrpid, "sysvgroups"},
1026 {Opt_resgid, "resgid=%u"},
1027 {Opt_resuid, "resuid=%u"},
1029 {Opt_err_cont, "errors=continue"},
1030 {Opt_err_panic, "errors=panic"},
1031 {Opt_err_ro, "errors=remount-ro"},
1032 {Opt_nouid32, "nouid32"},
1033 {Opt_debug, "debug"},
1034 {Opt_oldalloc, "oldalloc"},
1035 {Opt_orlov, "orlov"},
1036 {Opt_user_xattr, "user_xattr"},
1037 {Opt_nouser_xattr, "nouser_xattr"},
1039 {Opt_noacl, "noacl"},
1040 {Opt_reservation, "reservation"},
1041 {Opt_noreservation, "noreservation"},
1042 {Opt_noload, "noload"},
1045 {Opt_commit, "commit=%u"},
1046 {Opt_min_batch_time, "min_batch_time=%u"},
1047 {Opt_max_batch_time, "max_batch_time=%u"},
1048 {Opt_journal_update, "journal=update"},
1049 {Opt_journal_dev, "journal_dev=%u"},
1050 {Opt_journal_checksum, "journal_checksum"},
1051 {Opt_journal_async_commit, "journal_async_commit"},
1052 {Opt_abort, "abort"},
1053 {Opt_data_journal, "data=journal"},
1054 {Opt_data_ordered, "data=ordered"},
1055 {Opt_data_writeback, "data=writeback"},
1056 {Opt_data_err_abort, "data_err=abort"},
1057 {Opt_data_err_ignore, "data_err=ignore"},
1058 {Opt_offusrjquota, "usrjquota="},
1059 {Opt_usrjquota, "usrjquota=%s"},
1060 {Opt_offgrpjquota, "grpjquota="},
1061 {Opt_grpjquota, "grpjquota=%s"},
1062 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1063 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1064 {Opt_grpquota, "grpquota"},
1065 {Opt_noquota, "noquota"},
1066 {Opt_quota, "quota"},
1067 {Opt_usrquota, "usrquota"},
1068 {Opt_barrier, "barrier=%u"},
1069 {Opt_extents, "extents"},
1070 {Opt_noextents, "noextents"},
1071 {Opt_i_version, "i_version"},
1072 {Opt_stripe, "stripe=%u"},
1073 {Opt_resize, "resize"},
1074 {Opt_delalloc, "delalloc"},
1075 {Opt_nodelalloc, "nodelalloc"},
1076 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1080 static ext4_fsblk_t get_sb_block(void **data)
1082 ext4_fsblk_t sb_block;
1083 char *options = (char *) *data;
1085 if (!options || strncmp(options, "sb=", 3) != 0)
1086 return 1; /* Default location */
1088 /*todo: use simple_strtoll with >32bit ext4 */
1089 sb_block = simple_strtoul(options, &options, 0);
1090 if (*options && *options != ',') {
1091 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1095 if (*options == ',')
1097 *data = (void *) options;
1101 static int parse_options(char *options, struct super_block *sb,
1102 unsigned long *journal_devnum,
1103 ext4_fsblk_t *n_blocks_count, int is_remount)
1105 struct ext4_sb_info *sbi = EXT4_SB(sb);
1107 substring_t args[MAX_OPT_ARGS];
1114 ext4_fsblk_t last_block;
1119 while ((p = strsep(&options, ",")) != NULL) {
1124 token = match_token(p, tokens, args);
1127 clear_opt(sbi->s_mount_opt, MINIX_DF);
1130 set_opt(sbi->s_mount_opt, MINIX_DF);
1133 set_opt(sbi->s_mount_opt, GRPID);
1136 clear_opt(sbi->s_mount_opt, GRPID);
1139 if (match_int(&args[0], &option))
1141 sbi->s_resuid = option;
1144 if (match_int(&args[0], &option))
1146 sbi->s_resgid = option;
1149 /* handled by get_sb_block() instead of here */
1150 /* *sb_block = match_int(&args[0]); */
1153 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1154 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1155 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1158 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1159 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1160 set_opt(sbi->s_mount_opt, ERRORS_RO);
1163 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1164 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1165 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1168 set_opt(sbi->s_mount_opt, NO_UID32);
1171 set_opt(sbi->s_mount_opt, DEBUG);
1174 set_opt(sbi->s_mount_opt, OLDALLOC);
1177 clear_opt(sbi->s_mount_opt, OLDALLOC);
1179 #ifdef CONFIG_EXT4_FS_XATTR
1180 case Opt_user_xattr:
1181 set_opt(sbi->s_mount_opt, XATTR_USER);
1183 case Opt_nouser_xattr:
1184 clear_opt(sbi->s_mount_opt, XATTR_USER);
1187 case Opt_user_xattr:
1188 case Opt_nouser_xattr:
1189 printk(KERN_ERR "EXT4 (no)user_xattr options "
1193 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1195 set_opt(sbi->s_mount_opt, POSIX_ACL);
1198 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1203 printk(KERN_ERR "EXT4 (no)acl options "
1207 case Opt_reservation:
1208 set_opt(sbi->s_mount_opt, RESERVATION);
1210 case Opt_noreservation:
1211 clear_opt(sbi->s_mount_opt, RESERVATION);
1213 case Opt_journal_update:
1215 /* Eventually we will want to be able to create
1216 a journal file here. For now, only allow the
1217 user to specify an existing inode to be the
1220 printk(KERN_ERR "EXT4-fs: cannot specify "
1221 "journal on remount\n");
1224 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1226 case Opt_journal_dev:
1228 printk(KERN_ERR "EXT4-fs: cannot specify "
1229 "journal on remount\n");
1232 if (match_int(&args[0], &option))
1234 *journal_devnum = option;
1236 case Opt_journal_checksum:
1237 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1239 case Opt_journal_async_commit:
1240 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1241 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1244 set_opt(sbi->s_mount_opt, NOLOAD);
1247 if (match_int(&args[0], &option))
1252 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1253 sbi->s_commit_interval = HZ * option;
1255 case Opt_max_batch_time:
1256 if (match_int(&args[0], &option))
1261 option = EXT4_DEF_MAX_BATCH_TIME;
1262 sbi->s_max_batch_time = option;
1264 case Opt_min_batch_time:
1265 if (match_int(&args[0], &option))
1269 sbi->s_min_batch_time = option;
1271 case Opt_data_journal:
1272 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1274 case Opt_data_ordered:
1275 data_opt = EXT4_MOUNT_ORDERED_DATA;
1277 case Opt_data_writeback:
1278 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1281 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1284 "EXT4-fs: cannot change data "
1285 "mode on remount\n");
1289 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1290 sbi->s_mount_opt |= data_opt;
1293 case Opt_data_err_abort:
1294 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1296 case Opt_data_err_ignore:
1297 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1306 if ((sb_any_quota_enabled(sb) ||
1307 sb_any_quota_suspended(sb)) &&
1308 !sbi->s_qf_names[qtype]) {
1310 "EXT4-fs: Cannot change journaled "
1311 "quota options when quota turned on.\n");
1314 qname = match_strdup(&args[0]);
1317 "EXT4-fs: not enough memory for "
1318 "storing quotafile name.\n");
1321 if (sbi->s_qf_names[qtype] &&
1322 strcmp(sbi->s_qf_names[qtype], qname)) {
1324 "EXT4-fs: %s quota file already "
1325 "specified.\n", QTYPE2NAME(qtype));
1329 sbi->s_qf_names[qtype] = qname;
1330 if (strchr(sbi->s_qf_names[qtype], '/')) {
1332 "EXT4-fs: quotafile must be on "
1333 "filesystem root.\n");
1334 kfree(sbi->s_qf_names[qtype]);
1335 sbi->s_qf_names[qtype] = NULL;
1338 set_opt(sbi->s_mount_opt, QUOTA);
1340 case Opt_offusrjquota:
1343 case Opt_offgrpjquota:
1346 if ((sb_any_quota_enabled(sb) ||
1347 sb_any_quota_suspended(sb)) &&
1348 sbi->s_qf_names[qtype]) {
1349 printk(KERN_ERR "EXT4-fs: Cannot change "
1350 "journaled quota options when "
1351 "quota turned on.\n");
1355 * The space will be released later when all options
1356 * are confirmed to be correct
1358 sbi->s_qf_names[qtype] = NULL;
1360 case Opt_jqfmt_vfsold:
1361 qfmt = QFMT_VFS_OLD;
1363 case Opt_jqfmt_vfsv0:
1366 if ((sb_any_quota_enabled(sb) ||
1367 sb_any_quota_suspended(sb)) &&
1368 sbi->s_jquota_fmt != qfmt) {
1369 printk(KERN_ERR "EXT4-fs: Cannot change "
1370 "journaled quota options when "
1371 "quota turned on.\n");
1374 sbi->s_jquota_fmt = qfmt;
1378 set_opt(sbi->s_mount_opt, QUOTA);
1379 set_opt(sbi->s_mount_opt, USRQUOTA);
1382 set_opt(sbi->s_mount_opt, QUOTA);
1383 set_opt(sbi->s_mount_opt, GRPQUOTA);
1386 if (sb_any_quota_enabled(sb)) {
1387 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1388 "options when quota turned on.\n");
1391 clear_opt(sbi->s_mount_opt, QUOTA);
1392 clear_opt(sbi->s_mount_opt, USRQUOTA);
1393 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1400 "EXT4-fs: quota options not supported.\n");
1404 case Opt_offusrjquota:
1405 case Opt_offgrpjquota:
1406 case Opt_jqfmt_vfsold:
1407 case Opt_jqfmt_vfsv0:
1409 "EXT4-fs: journaled quota options not "
1416 set_opt(sbi->s_mount_opt, ABORT);
1419 if (match_int(&args[0], &option))
1422 set_opt(sbi->s_mount_opt, BARRIER);
1424 clear_opt(sbi->s_mount_opt, BARRIER);
1430 printk("EXT4-fs: resize option only available "
1434 if (match_int(&args[0], &option) != 0)
1436 *n_blocks_count = option;
1439 set_opt(sbi->s_mount_opt, NOBH);
1442 clear_opt(sbi->s_mount_opt, NOBH);
1445 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1446 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1447 ext4_warning(sb, __func__,
1448 "extents feature not enabled "
1449 "on this filesystem, use tune2fs");
1452 set_opt(sbi->s_mount_opt, EXTENTS);
1456 * When e2fsprogs support resizing an already existing
1457 * ext3 file system to greater than 2**32 we need to
1458 * add support to block allocator to handle growing
1459 * already existing block mapped inode so that blocks
1460 * allocated for them fall within 2**32
1462 last_block = ext4_blocks_count(sbi->s_es) - 1;
1463 if (last_block > 0xffffffffULL) {
1464 printk(KERN_ERR "EXT4-fs: Filesystem too "
1465 "large to mount with "
1466 "-o noextents options\n");
1469 clear_opt(sbi->s_mount_opt, EXTENTS);
1472 set_opt(sbi->s_mount_opt, I_VERSION);
1473 sb->s_flags |= MS_I_VERSION;
1475 case Opt_nodelalloc:
1476 clear_opt(sbi->s_mount_opt, DELALLOC);
1479 if (match_int(&args[0], &option))
1483 sbi->s_stripe = option;
1486 set_opt(sbi->s_mount_opt, DELALLOC);
1488 case Opt_inode_readahead_blks:
1489 if (match_int(&args[0], &option))
1491 if (option < 0 || option > (1 << 30))
1493 sbi->s_inode_readahead_blks = option;
1497 "EXT4-fs: Unrecognized mount option \"%s\" "
1498 "or missing value\n", p);
1503 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1504 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1505 sbi->s_qf_names[USRQUOTA])
1506 clear_opt(sbi->s_mount_opt, USRQUOTA);
1508 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1509 sbi->s_qf_names[GRPQUOTA])
1510 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1512 if ((sbi->s_qf_names[USRQUOTA] &&
1513 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1514 (sbi->s_qf_names[GRPQUOTA] &&
1515 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1516 printk(KERN_ERR "EXT4-fs: old and new quota "
1517 "format mixing.\n");
1521 if (!sbi->s_jquota_fmt) {
1522 printk(KERN_ERR "EXT4-fs: journaled quota format "
1523 "not specified.\n");
1527 if (sbi->s_jquota_fmt) {
1528 printk(KERN_ERR "EXT4-fs: journaled quota format "
1529 "specified with no journaling "
1538 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1541 struct ext4_sb_info *sbi = EXT4_SB(sb);
1544 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1545 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1546 "forcing read-only mode\n");
1551 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1552 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1553 "running e2fsck is recommended\n");
1554 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1556 "EXT4-fs warning: mounting fs with errors, "
1557 "running e2fsck is recommended\n");
1558 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1559 le16_to_cpu(es->s_mnt_count) >=
1560 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1562 "EXT4-fs warning: maximal mount count reached, "
1563 "running e2fsck is recommended\n");
1564 else if (le32_to_cpu(es->s_checkinterval) &&
1565 (le32_to_cpu(es->s_lastcheck) +
1566 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1568 "EXT4-fs warning: checktime reached, "
1569 "running e2fsck is recommended\n");
1570 if (!sbi->s_journal)
1571 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1572 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1573 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1574 le16_add_cpu(&es->s_mnt_count, 1);
1575 es->s_mtime = cpu_to_le32(get_seconds());
1576 ext4_update_dynamic_rev(sb);
1578 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1580 ext4_commit_super(sb, es, 1);
1581 if (test_opt(sb, DEBUG))
1582 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1583 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1585 sbi->s_groups_count,
1586 EXT4_BLOCKS_PER_GROUP(sb),
1587 EXT4_INODES_PER_GROUP(sb),
1590 if (EXT4_SB(sb)->s_journal) {
1591 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1592 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1593 "external", EXT4_SB(sb)->s_journal->j_devname);
1595 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1600 static int ext4_fill_flex_info(struct super_block *sb)
1602 struct ext4_sb_info *sbi = EXT4_SB(sb);
1603 struct ext4_group_desc *gdp = NULL;
1604 struct buffer_head *bh;
1605 ext4_group_t flex_group_count;
1606 ext4_group_t flex_group;
1607 int groups_per_flex = 0;
1610 if (!sbi->s_es->s_log_groups_per_flex) {
1611 sbi->s_log_groups_per_flex = 0;
1615 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1616 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1618 /* We allocate both existing and potentially added groups */
1619 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1620 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1621 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1622 sbi->s_flex_groups = kzalloc(flex_group_count *
1623 sizeof(struct flex_groups), GFP_KERNEL);
1624 if (sbi->s_flex_groups == NULL) {
1625 printk(KERN_ERR "EXT4-fs: not enough memory for "
1626 "%u flex groups\n", flex_group_count);
1630 for (i = 0; i < sbi->s_groups_count; i++) {
1631 gdp = ext4_get_group_desc(sb, i, &bh);
1633 flex_group = ext4_flex_group(sbi, i);
1634 sbi->s_flex_groups[flex_group].free_inodes +=
1635 ext4_free_inodes_count(sb, gdp);
1636 sbi->s_flex_groups[flex_group].free_blocks +=
1637 ext4_free_blks_count(sb, gdp);
1645 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1646 struct ext4_group_desc *gdp)
1650 if (sbi->s_es->s_feature_ro_compat &
1651 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1652 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1653 __le32 le_group = cpu_to_le32(block_group);
1655 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1656 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1657 crc = crc16(crc, (__u8 *)gdp, offset);
1658 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1659 /* for checksum of struct ext4_group_desc do the rest...*/
1660 if ((sbi->s_es->s_feature_incompat &
1661 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1662 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1663 crc = crc16(crc, (__u8 *)gdp + offset,
1664 le16_to_cpu(sbi->s_es->s_desc_size) -
1668 return cpu_to_le16(crc);
1671 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1672 struct ext4_group_desc *gdp)
1674 if ((sbi->s_es->s_feature_ro_compat &
1675 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1676 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1682 /* Called at mount-time, super-block is locked */
1683 static int ext4_check_descriptors(struct super_block *sb)
1685 struct ext4_sb_info *sbi = EXT4_SB(sb);
1686 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1687 ext4_fsblk_t last_block;
1688 ext4_fsblk_t block_bitmap;
1689 ext4_fsblk_t inode_bitmap;
1690 ext4_fsblk_t inode_table;
1691 int flexbg_flag = 0;
1694 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1697 ext4_debug("Checking group descriptors");
1699 for (i = 0; i < sbi->s_groups_count; i++) {
1700 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1702 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1703 last_block = ext4_blocks_count(sbi->s_es) - 1;
1705 last_block = first_block +
1706 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1708 block_bitmap = ext4_block_bitmap(sb, gdp);
1709 if (block_bitmap < first_block || block_bitmap > last_block) {
1710 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1711 "Block bitmap for group %u not in group "
1712 "(block %llu)!\n", i, block_bitmap);
1715 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1716 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1717 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1718 "Inode bitmap for group %u not in group "
1719 "(block %llu)!\n", i, inode_bitmap);
1722 inode_table = ext4_inode_table(sb, gdp);
1723 if (inode_table < first_block ||
1724 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1725 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1726 "Inode table for group %u not in group "
1727 "(block %llu)!\n", i, inode_table);
1730 spin_lock(sb_bgl_lock(sbi, i));
1731 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1732 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1733 "Checksum for group %u failed (%u!=%u)\n",
1734 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1735 gdp)), le16_to_cpu(gdp->bg_checksum));
1736 if (!(sb->s_flags & MS_RDONLY)) {
1737 spin_unlock(sb_bgl_lock(sbi, i));
1741 spin_unlock(sb_bgl_lock(sbi, i));
1743 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1746 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1747 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1751 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1752 * the superblock) which were deleted from all directories, but held open by
1753 * a process at the time of a crash. We walk the list and try to delete these
1754 * inodes at recovery time (only with a read-write filesystem).
1756 * In order to keep the orphan inode chain consistent during traversal (in
1757 * case of crash during recovery), we link each inode into the superblock
1758 * orphan list_head and handle it the same way as an inode deletion during
1759 * normal operation (which journals the operations for us).
1761 * We only do an iget() and an iput() on each inode, which is very safe if we
1762 * accidentally point at an in-use or already deleted inode. The worst that
1763 * can happen in this case is that we get a "bit already cleared" message from
1764 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1765 * e2fsck was run on this filesystem, and it must have already done the orphan
1766 * inode cleanup for us, so we can safely abort without any further action.
1768 static void ext4_orphan_cleanup(struct super_block *sb,
1769 struct ext4_super_block *es)
1771 unsigned int s_flags = sb->s_flags;
1772 int nr_orphans = 0, nr_truncates = 0;
1776 if (!es->s_last_orphan) {
1777 jbd_debug(4, "no orphan inodes to clean up\n");
1781 if (bdev_read_only(sb->s_bdev)) {
1782 printk(KERN_ERR "EXT4-fs: write access "
1783 "unavailable, skipping orphan cleanup.\n");
1787 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1788 if (es->s_last_orphan)
1789 jbd_debug(1, "Errors on filesystem, "
1790 "clearing orphan list.\n");
1791 es->s_last_orphan = 0;
1792 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1796 if (s_flags & MS_RDONLY) {
1797 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1799 sb->s_flags &= ~MS_RDONLY;
1802 /* Needed for iput() to work correctly and not trash data */
1803 sb->s_flags |= MS_ACTIVE;
1804 /* Turn on quotas so that they are updated correctly */
1805 for (i = 0; i < MAXQUOTAS; i++) {
1806 if (EXT4_SB(sb)->s_qf_names[i]) {
1807 int ret = ext4_quota_on_mount(sb, i);
1810 "EXT4-fs: Cannot turn on journaled "
1811 "quota: error %d\n", ret);
1816 while (es->s_last_orphan) {
1817 struct inode *inode;
1819 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1820 if (IS_ERR(inode)) {
1821 es->s_last_orphan = 0;
1825 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1827 if (inode->i_nlink) {
1829 "%s: truncating inode %lu to %lld bytes\n",
1830 __func__, inode->i_ino, inode->i_size);
1831 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1832 inode->i_ino, inode->i_size);
1833 ext4_truncate(inode);
1837 "%s: deleting unreferenced inode %lu\n",
1838 __func__, inode->i_ino);
1839 jbd_debug(2, "deleting unreferenced inode %lu\n",
1843 iput(inode); /* The delete magic happens here! */
1846 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1849 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1850 sb->s_id, PLURAL(nr_orphans));
1852 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1853 sb->s_id, PLURAL(nr_truncates));
1855 /* Turn quotas off */
1856 for (i = 0; i < MAXQUOTAS; i++) {
1857 if (sb_dqopt(sb)->files[i])
1858 vfs_quota_off(sb, i, 0);
1861 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1864 * Maximal extent format file size.
1865 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1866 * extent format containers, within a sector_t, and within i_blocks
1867 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1868 * so that won't be a limiting factor.
1870 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1872 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1875 loff_t upper_limit = MAX_LFS_FILESIZE;
1877 /* small i_blocks in vfs inode? */
1878 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1880 * CONFIG_LBD is not enabled implies the inode
1881 * i_block represent total blocks in 512 bytes
1882 * 32 == size of vfs inode i_blocks * 8
1884 upper_limit = (1LL << 32) - 1;
1886 /* total blocks in file system block size */
1887 upper_limit >>= (blkbits - 9);
1888 upper_limit <<= blkbits;
1891 /* 32-bit extent-start container, ee_block */
1896 /* Sanity check against vm- & vfs- imposed limits */
1897 if (res > upper_limit)
1904 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1905 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1906 * We need to be 1 filesystem block less than the 2^48 sector limit.
1908 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1910 loff_t res = EXT4_NDIR_BLOCKS;
1913 /* This is calculated to be the largest file size for a
1914 * dense, bitmapped file such that the total number of
1915 * sectors in the file, including data and all indirect blocks,
1916 * does not exceed 2^48 -1
1917 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1918 * total number of 512 bytes blocks of the file
1921 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1923 * !has_huge_files or CONFIG_LBD is not enabled
1924 * implies the inode i_block represent total blocks in
1925 * 512 bytes 32 == size of vfs inode i_blocks * 8
1927 upper_limit = (1LL << 32) - 1;
1929 /* total blocks in file system block size */
1930 upper_limit >>= (bits - 9);
1934 * We use 48 bit ext4_inode i_blocks
1935 * With EXT4_HUGE_FILE_FL set the i_blocks
1936 * represent total number of blocks in
1937 * file system block size
1939 upper_limit = (1LL << 48) - 1;
1943 /* indirect blocks */
1945 /* double indirect blocks */
1946 meta_blocks += 1 + (1LL << (bits-2));
1947 /* tripple indirect blocks */
1948 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1950 upper_limit -= meta_blocks;
1951 upper_limit <<= bits;
1953 res += 1LL << (bits-2);
1954 res += 1LL << (2*(bits-2));
1955 res += 1LL << (3*(bits-2));
1957 if (res > upper_limit)
1960 if (res > MAX_LFS_FILESIZE)
1961 res = MAX_LFS_FILESIZE;
1966 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1967 ext4_fsblk_t logical_sb_block, int nr)
1969 struct ext4_sb_info *sbi = EXT4_SB(sb);
1970 ext4_group_t bg, first_meta_bg;
1973 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1975 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1977 return logical_sb_block + nr + 1;
1978 bg = sbi->s_desc_per_block * nr;
1979 if (ext4_bg_has_super(sb, bg))
1981 return (has_super + ext4_group_first_block_no(sb, bg));
1985 * ext4_get_stripe_size: Get the stripe size.
1986 * @sbi: In memory super block info
1988 * If we have specified it via mount option, then
1989 * use the mount option value. If the value specified at mount time is
1990 * greater than the blocks per group use the super block value.
1991 * If the super block value is greater than blocks per group return 0.
1992 * Allocator needs it be less than blocks per group.
1995 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1997 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1998 unsigned long stripe_width =
1999 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2001 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2002 return sbi->s_stripe;
2004 if (stripe_width <= sbi->s_blocks_per_group)
2005 return stripe_width;
2007 if (stride <= sbi->s_blocks_per_group)
2013 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2014 __releases(kernel_lock)
2015 __acquires(kernel_lock)
2018 struct buffer_head *bh;
2019 struct ext4_super_block *es = NULL;
2020 struct ext4_sb_info *sbi;
2022 ext4_fsblk_t sb_block = get_sb_block(&data);
2023 ext4_fsblk_t logical_sb_block;
2024 unsigned long offset = 0;
2025 unsigned long journal_devnum = 0;
2026 unsigned long def_mount_opts;
2034 int needs_recovery, has_huge_files;
2039 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2042 sb->s_fs_info = sbi;
2043 sbi->s_mount_opt = 0;
2044 sbi->s_resuid = EXT4_DEF_RESUID;
2045 sbi->s_resgid = EXT4_DEF_RESGID;
2046 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2047 sbi->s_sb_block = sb_block;
2051 /* Cleanup superblock name */
2052 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2055 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2057 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2062 * The ext4 superblock will not be buffer aligned for other than 1kB
2063 * block sizes. We need to calculate the offset from buffer start.
2065 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2066 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2067 offset = do_div(logical_sb_block, blocksize);
2069 logical_sb_block = sb_block;
2072 if (!(bh = sb_bread(sb, logical_sb_block))) {
2073 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2077 * Note: s_es must be initialized as soon as possible because
2078 * some ext4 macro-instructions depend on its value
2080 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2082 sb->s_magic = le16_to_cpu(es->s_magic);
2083 if (sb->s_magic != EXT4_SUPER_MAGIC)
2086 /* Set defaults before we parse the mount options */
2087 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2088 if (def_mount_opts & EXT4_DEFM_DEBUG)
2089 set_opt(sbi->s_mount_opt, DEBUG);
2090 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2091 set_opt(sbi->s_mount_opt, GRPID);
2092 if (def_mount_opts & EXT4_DEFM_UID16)
2093 set_opt(sbi->s_mount_opt, NO_UID32);
2094 #ifdef CONFIG_EXT4_FS_XATTR
2095 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2096 set_opt(sbi->s_mount_opt, XATTR_USER);
2098 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2099 if (def_mount_opts & EXT4_DEFM_ACL)
2100 set_opt(sbi->s_mount_opt, POSIX_ACL);
2102 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2103 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2104 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2105 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2106 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2107 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2109 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2110 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2111 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2112 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2114 set_opt(sbi->s_mount_opt, ERRORS_RO);
2116 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2117 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2118 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2119 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2120 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2122 set_opt(sbi->s_mount_opt, RESERVATION);
2123 set_opt(sbi->s_mount_opt, BARRIER);
2126 * turn on extents feature by default in ext4 filesystem
2127 * only if feature flag already set by mkfs or tune2fs.
2128 * Use -o noextents to turn it off
2130 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2131 set_opt(sbi->s_mount_opt, EXTENTS);
2133 ext4_warning(sb, __func__,
2134 "extents feature not enabled on this filesystem, "
2138 * enable delayed allocation by default
2139 * Use -o nodelalloc to turn it off
2141 set_opt(sbi->s_mount_opt, DELALLOC);
2144 if (!parse_options((char *) data, sb, &journal_devnum, NULL, 0))
2147 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2148 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2150 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2151 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2152 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2153 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2155 "EXT4-fs warning: feature flags set on rev 0 fs, "
2156 "running e2fsck is recommended\n");
2159 * Check feature flags regardless of the revision level, since we
2160 * previously didn't change the revision level when setting the flags,
2161 * so there is a chance incompat flags are set on a rev 0 filesystem.
2163 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2165 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2166 "unsupported optional features (%x).\n", sb->s_id,
2167 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2168 ~EXT4_FEATURE_INCOMPAT_SUPP));
2171 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2172 if (!(sb->s_flags & MS_RDONLY) && features) {
2173 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2174 "unsupported optional features (%x).\n", sb->s_id,
2175 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2176 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2179 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2180 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2181 if (has_huge_files) {
2183 * Large file size enabled file system can only be
2184 * mount if kernel is build with CONFIG_LBD
2186 if (sizeof(root->i_blocks) < sizeof(u64) &&
2187 !(sb->s_flags & MS_RDONLY)) {
2188 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2189 "files cannot be mounted read-write "
2190 "without CONFIG_LBD.\n", sb->s_id);
2194 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2196 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2197 blocksize > EXT4_MAX_BLOCK_SIZE) {
2199 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2200 blocksize, sb->s_id);
2204 if (sb->s_blocksize != blocksize) {
2206 /* Validate the filesystem blocksize */
2207 if (!sb_set_blocksize(sb, blocksize)) {
2208 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2214 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2215 offset = do_div(logical_sb_block, blocksize);
2216 bh = sb_bread(sb, logical_sb_block);
2219 "EXT4-fs: Can't read superblock on 2nd try.\n");
2222 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2224 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2226 "EXT4-fs: Magic mismatch, very weird !\n");
2231 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2233 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2235 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2236 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2237 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2239 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2240 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2241 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2242 (!is_power_of_2(sbi->s_inode_size)) ||
2243 (sbi->s_inode_size > blocksize)) {
2245 "EXT4-fs: unsupported inode size: %d\n",
2249 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2250 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2252 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2253 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2254 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2255 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2256 !is_power_of_2(sbi->s_desc_size)) {
2258 "EXT4-fs: unsupported descriptor size %lu\n",
2263 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2264 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2265 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2266 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2268 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2269 if (sbi->s_inodes_per_block == 0)
2271 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2272 sbi->s_inodes_per_block;
2273 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2275 sbi->s_mount_state = le16_to_cpu(es->s_state);
2276 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2277 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2278 for (i = 0; i < 4; i++)
2279 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2280 sbi->s_def_hash_version = es->s_def_hash_version;
2281 i = le32_to_cpu(es->s_flags);
2282 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2283 sbi->s_hash_unsigned = 3;
2284 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2285 #ifdef __CHAR_UNSIGNED__
2286 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2287 sbi->s_hash_unsigned = 3;
2289 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2294 if (sbi->s_blocks_per_group > blocksize * 8) {
2296 "EXT4-fs: #blocks per group too big: %lu\n",
2297 sbi->s_blocks_per_group);
2300 if (sbi->s_inodes_per_group > blocksize * 8) {
2302 "EXT4-fs: #inodes per group too big: %lu\n",
2303 sbi->s_inodes_per_group);
2307 if (ext4_blocks_count(es) >
2308 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2309 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2310 " too large to mount safely\n", sb->s_id);
2311 if (sizeof(sector_t) < 8)
2312 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2317 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2320 /* ensure blocks_count calculation below doesn't sign-extend */
2321 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2322 le32_to_cpu(es->s_first_data_block) + 1) {
2323 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2324 "first data block %u, blocks per group %lu\n",
2325 ext4_blocks_count(es),
2326 le32_to_cpu(es->s_first_data_block),
2327 EXT4_BLOCKS_PER_GROUP(sb));
2330 blocks_count = (ext4_blocks_count(es) -
2331 le32_to_cpu(es->s_first_data_block) +
2332 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2333 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2334 sbi->s_groups_count = blocks_count;
2335 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2336 EXT4_DESC_PER_BLOCK(sb);
2337 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2339 if (sbi->s_group_desc == NULL) {
2340 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2344 #ifdef CONFIG_PROC_FS
2346 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2349 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2351 &sbi->s_inode_readahead_blks);
2354 bgl_lock_init(&sbi->s_blockgroup_lock);
2356 for (i = 0; i < db_count; i++) {
2357 block = descriptor_loc(sb, logical_sb_block, i);
2358 sbi->s_group_desc[i] = sb_bread(sb, block);
2359 if (!sbi->s_group_desc[i]) {
2360 printk(KERN_ERR "EXT4-fs: "
2361 "can't read group descriptor %d\n", i);
2366 if (!ext4_check_descriptors(sb)) {
2367 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2370 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2371 if (!ext4_fill_flex_info(sb)) {
2373 "EXT4-fs: unable to initialize "
2374 "flex_bg meta info!\n");
2378 sbi->s_gdb_count = db_count;
2379 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2380 spin_lock_init(&sbi->s_next_gen_lock);
2382 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2383 ext4_count_free_blocks(sb));
2385 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2386 ext4_count_free_inodes(sb));
2389 err = percpu_counter_init(&sbi->s_dirs_counter,
2390 ext4_count_dirs(sb));
2393 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2396 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2400 sbi->s_stripe = ext4_get_stripe_size(sbi);
2403 * set up enough so that it can read an inode
2405 sb->s_op = &ext4_sops;
2406 sb->s_export_op = &ext4_export_ops;
2407 sb->s_xattr = ext4_xattr_handlers;
2409 sb->s_qcop = &ext4_qctl_operations;
2410 sb->dq_op = &ext4_quota_operations;
2412 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2416 needs_recovery = (es->s_last_orphan != 0 ||
2417 EXT4_HAS_INCOMPAT_FEATURE(sb,
2418 EXT4_FEATURE_INCOMPAT_RECOVER));
2421 * The first inode we look at is the journal inode. Don't try
2422 * root first: it may be modified in the journal!
2424 if (!test_opt(sb, NOLOAD) &&
2425 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2426 if (ext4_load_journal(sb, es, journal_devnum))
2428 if (!(sb->s_flags & MS_RDONLY) &&
2429 EXT4_SB(sb)->s_journal->j_failed_commit) {
2430 printk(KERN_CRIT "EXT4-fs error (device %s): "
2431 "ext4_fill_super: Journal transaction "
2432 "%u is corrupt\n", sb->s_id,
2433 EXT4_SB(sb)->s_journal->j_failed_commit);
2434 if (test_opt(sb, ERRORS_RO)) {
2436 "Mounting filesystem read-only\n");
2437 sb->s_flags |= MS_RDONLY;
2438 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2439 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2441 if (test_opt(sb, ERRORS_PANIC)) {
2442 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2443 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2444 ext4_commit_super(sb, es, 1);
2448 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2449 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2450 printk(KERN_ERR "EXT4-fs: required journal recovery "
2451 "suppressed and not mounted read-only\n");
2454 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2455 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2456 sbi->s_journal = NULL;
2461 if (ext4_blocks_count(es) > 0xffffffffULL &&
2462 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2463 JBD2_FEATURE_INCOMPAT_64BIT)) {
2464 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2468 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2469 jbd2_journal_set_features(sbi->s_journal,
2470 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2471 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2472 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2473 jbd2_journal_set_features(sbi->s_journal,
2474 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2475 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2476 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2478 jbd2_journal_clear_features(sbi->s_journal,
2479 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2480 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2483 /* We have now updated the journal if required, so we can
2484 * validate the data journaling mode. */
2485 switch (test_opt(sb, DATA_FLAGS)) {
2487 /* No mode set, assume a default based on the journal
2488 * capabilities: ORDERED_DATA if the journal can
2489 * cope, else JOURNAL_DATA
2491 if (jbd2_journal_check_available_features
2492 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2493 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2495 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2498 case EXT4_MOUNT_ORDERED_DATA:
2499 case EXT4_MOUNT_WRITEBACK_DATA:
2500 if (!jbd2_journal_check_available_features
2501 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2502 printk(KERN_ERR "EXT4-fs: Journal does not support "
2503 "requested data journaling mode\n");
2512 if (test_opt(sb, NOBH)) {
2513 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2514 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2515 "its supported only with writeback mode\n");
2516 clear_opt(sbi->s_mount_opt, NOBH);
2520 * The jbd2_journal_load will have done any necessary log recovery,
2521 * so we can safely mount the rest of the filesystem now.
2524 root = ext4_iget(sb, EXT4_ROOT_INO);
2526 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2527 ret = PTR_ERR(root);
2530 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2532 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2535 sb->s_root = d_alloc_root(root);
2537 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2543 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2545 /* determine the minimum size of new large inodes, if present */
2546 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2547 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2548 EXT4_GOOD_OLD_INODE_SIZE;
2549 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2550 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2551 if (sbi->s_want_extra_isize <
2552 le16_to_cpu(es->s_want_extra_isize))
2553 sbi->s_want_extra_isize =
2554 le16_to_cpu(es->s_want_extra_isize);
2555 if (sbi->s_want_extra_isize <
2556 le16_to_cpu(es->s_min_extra_isize))
2557 sbi->s_want_extra_isize =
2558 le16_to_cpu(es->s_min_extra_isize);
2561 /* Check if enough inode space is available */
2562 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2563 sbi->s_inode_size) {
2564 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2565 EXT4_GOOD_OLD_INODE_SIZE;
2566 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2570 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2571 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2572 "requested data journaling mode\n");
2573 clear_opt(sbi->s_mount_opt, DELALLOC);
2574 } else if (test_opt(sb, DELALLOC))
2575 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2578 err = ext4_mb_init(sb, needs_recovery);
2580 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2586 * akpm: core read_super() calls in here with the superblock locked.
2587 * That deadlocks, because orphan cleanup needs to lock the superblock
2588 * in numerous places. Here we just pop the lock - it's relatively
2589 * harmless, because we are now ready to accept write_super() requests,
2590 * and aviro says that's the only reason for hanging onto the
2593 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2594 ext4_orphan_cleanup(sb, es);
2595 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2596 if (needs_recovery) {
2597 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2598 ext4_mark_recovery_complete(sb, es);
2600 if (EXT4_SB(sb)->s_journal) {
2601 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2602 descr = " journalled data mode";
2603 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2604 descr = " ordered data mode";
2606 descr = " writeback data mode";
2608 descr = "out journal";
2610 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2618 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2623 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2624 if (sbi->s_journal) {
2625 jbd2_journal_destroy(sbi->s_journal);
2626 sbi->s_journal = NULL;
2629 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2630 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2631 percpu_counter_destroy(&sbi->s_dirs_counter);
2632 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2634 for (i = 0; i < db_count; i++)
2635 brelse(sbi->s_group_desc[i]);
2636 kfree(sbi->s_group_desc);
2639 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2640 remove_proc_entry(sb->s_id, ext4_proc_root);
2643 for (i = 0; i < MAXQUOTAS; i++)
2644 kfree(sbi->s_qf_names[i]);
2646 ext4_blkdev_remove(sbi);
2649 sb->s_fs_info = NULL;
2656 * Setup any per-fs journal parameters now. We'll do this both on
2657 * initial mount, once the journal has been initialised but before we've
2658 * done any recovery; and again on any subsequent remount.
2660 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2662 struct ext4_sb_info *sbi = EXT4_SB(sb);
2664 journal->j_commit_interval = sbi->s_commit_interval;
2665 journal->j_min_batch_time = sbi->s_min_batch_time;
2666 journal->j_max_batch_time = sbi->s_max_batch_time;
2668 spin_lock(&journal->j_state_lock);
2669 if (test_opt(sb, BARRIER))
2670 journal->j_flags |= JBD2_BARRIER;
2672 journal->j_flags &= ~JBD2_BARRIER;
2673 if (test_opt(sb, DATA_ERR_ABORT))
2674 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2676 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2677 spin_unlock(&journal->j_state_lock);
2680 static journal_t *ext4_get_journal(struct super_block *sb,
2681 unsigned int journal_inum)
2683 struct inode *journal_inode;
2686 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2688 /* First, test for the existence of a valid inode on disk. Bad
2689 * things happen if we iget() an unused inode, as the subsequent
2690 * iput() will try to delete it. */
2692 journal_inode = ext4_iget(sb, journal_inum);
2693 if (IS_ERR(journal_inode)) {
2694 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2697 if (!journal_inode->i_nlink) {
2698 make_bad_inode(journal_inode);
2699 iput(journal_inode);
2700 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2704 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2705 journal_inode, journal_inode->i_size);
2706 if (!S_ISREG(journal_inode->i_mode)) {
2707 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2708 iput(journal_inode);
2712 journal = jbd2_journal_init_inode(journal_inode);
2714 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2715 iput(journal_inode);
2718 journal->j_private = sb;
2719 ext4_init_journal_params(sb, journal);
2723 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2726 struct buffer_head *bh;
2730 int hblock, blocksize;
2731 ext4_fsblk_t sb_block;
2732 unsigned long offset;
2733 struct ext4_super_block *es;
2734 struct block_device *bdev;
2736 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2738 bdev = ext4_blkdev_get(j_dev);
2742 if (bd_claim(bdev, sb)) {
2744 "EXT4: failed to claim external journal device.\n");
2745 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2749 blocksize = sb->s_blocksize;
2750 hblock = bdev_hardsect_size(bdev);
2751 if (blocksize < hblock) {
2753 "EXT4-fs: blocksize too small for journal device.\n");
2757 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2758 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2759 set_blocksize(bdev, blocksize);
2760 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2761 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2762 "external journal\n");
2766 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2767 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2768 !(le32_to_cpu(es->s_feature_incompat) &
2769 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2770 printk(KERN_ERR "EXT4-fs: external journal has "
2771 "bad superblock\n");
2776 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2777 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2782 len = ext4_blocks_count(es);
2783 start = sb_block + 1;
2784 brelse(bh); /* we're done with the superblock */
2786 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2787 start, len, blocksize);
2789 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2792 journal->j_private = sb;
2793 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2794 wait_on_buffer(journal->j_sb_buffer);
2795 if (!buffer_uptodate(journal->j_sb_buffer)) {
2796 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2799 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2800 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2801 "user (unsupported) - %d\n",
2802 be32_to_cpu(journal->j_superblock->s_nr_users));
2805 EXT4_SB(sb)->journal_bdev = bdev;
2806 ext4_init_journal_params(sb, journal);
2809 jbd2_journal_destroy(journal);
2811 ext4_blkdev_put(bdev);
2815 static int ext4_load_journal(struct super_block *sb,
2816 struct ext4_super_block *es,
2817 unsigned long journal_devnum)
2820 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2823 int really_read_only;
2825 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2827 if (journal_devnum &&
2828 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2829 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2830 "numbers have changed\n");
2831 journal_dev = new_decode_dev(journal_devnum);
2833 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2835 really_read_only = bdev_read_only(sb->s_bdev);
2838 * Are we loading a blank journal or performing recovery after a
2839 * crash? For recovery, we need to check in advance whether we
2840 * can get read-write access to the device.
2843 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2844 if (sb->s_flags & MS_RDONLY) {
2845 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2846 "required on readonly filesystem.\n");
2847 if (really_read_only) {
2848 printk(KERN_ERR "EXT4-fs: write access "
2849 "unavailable, cannot proceed.\n");
2852 printk(KERN_INFO "EXT4-fs: write access will "
2853 "be enabled during recovery.\n");
2857 if (journal_inum && journal_dev) {
2858 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2859 "and inode journals!\n");
2864 if (!(journal = ext4_get_journal(sb, journal_inum)))
2867 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2871 if (journal->j_flags & JBD2_BARRIER)
2872 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2874 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2876 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2877 err = jbd2_journal_update_format(journal);
2879 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2880 jbd2_journal_destroy(journal);
2885 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2886 err = jbd2_journal_wipe(journal, !really_read_only);
2888 err = jbd2_journal_load(journal);
2891 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2892 jbd2_journal_destroy(journal);
2896 EXT4_SB(sb)->s_journal = journal;
2897 ext4_clear_journal_err(sb, es);
2899 if (journal_devnum &&
2900 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2901 es->s_journal_dev = cpu_to_le32(journal_devnum);
2904 /* Make sure we flush the recovery flag to disk. */
2905 ext4_commit_super(sb, es, 1);
2911 static void ext4_commit_super(struct super_block *sb,
2912 struct ext4_super_block *es, int sync)
2914 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2918 if (buffer_write_io_error(sbh)) {
2920 * Oh, dear. A previous attempt to write the
2921 * superblock failed. This could happen because the
2922 * USB device was yanked out. Or it could happen to
2923 * be a transient write error and maybe the block will
2924 * be remapped. Nothing we can do but to retry the
2925 * write and hope for the best.
2927 printk(KERN_ERR "ext4: previous I/O error to "
2928 "superblock detected for %s.\n", sb->s_id);
2929 clear_buffer_write_io_error(sbh);
2930 set_buffer_uptodate(sbh);
2932 es->s_wtime = cpu_to_le32(get_seconds());
2933 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2934 &EXT4_SB(sb)->s_freeblocks_counter));
2935 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2936 &EXT4_SB(sb)->s_freeinodes_counter));
2938 BUFFER_TRACE(sbh, "marking dirty");
2939 mark_buffer_dirty(sbh);
2941 sync_dirty_buffer(sbh);
2942 if (buffer_write_io_error(sbh)) {
2943 printk(KERN_ERR "ext4: I/O error while writing "
2944 "superblock for %s.\n", sb->s_id);
2945 clear_buffer_write_io_error(sbh);
2946 set_buffer_uptodate(sbh);
2953 * Have we just finished recovery? If so, and if we are mounting (or
2954 * remounting) the filesystem readonly, then we will end up with a
2955 * consistent fs on disk. Record that fact.
2957 static void ext4_mark_recovery_complete(struct super_block *sb,
2958 struct ext4_super_block *es)
2960 journal_t *journal = EXT4_SB(sb)->s_journal;
2962 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2963 BUG_ON(journal != NULL);
2966 jbd2_journal_lock_updates(journal);
2967 if (jbd2_journal_flush(journal) < 0)
2971 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2972 sb->s_flags & MS_RDONLY) {
2973 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2975 ext4_commit_super(sb, es, 1);
2980 jbd2_journal_unlock_updates(journal);
2984 * If we are mounting (or read-write remounting) a filesystem whose journal
2985 * has recorded an error from a previous lifetime, move that error to the
2986 * main filesystem now.
2988 static void ext4_clear_journal_err(struct super_block *sb,
2989 struct ext4_super_block *es)
2995 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2997 journal = EXT4_SB(sb)->s_journal;
3000 * Now check for any error status which may have been recorded in the
3001 * journal by a prior ext4_error() or ext4_abort()
3004 j_errno = jbd2_journal_errno(journal);
3008 errstr = ext4_decode_error(sb, j_errno, nbuf);
3009 ext4_warning(sb, __func__, "Filesystem error recorded "
3010 "from previous mount: %s", errstr);
3011 ext4_warning(sb, __func__, "Marking fs in need of "
3012 "filesystem check.");
3014 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3015 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3016 ext4_commit_super(sb, es, 1);
3018 jbd2_journal_clear_err(journal);
3023 * Force the running and committing transactions to commit,
3024 * and wait on the commit.
3026 int ext4_force_commit(struct super_block *sb)
3031 if (sb->s_flags & MS_RDONLY)
3034 journal = EXT4_SB(sb)->s_journal;
3037 ret = ext4_journal_force_commit(journal);
3044 * Ext4 always journals updates to the superblock itself, so we don't
3045 * have to propagate any other updates to the superblock on disk at this
3046 * point. (We can probably nuke this function altogether, and remove
3047 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3049 static void ext4_write_super(struct super_block *sb)
3051 if (EXT4_SB(sb)->s_journal) {
3052 if (mutex_trylock(&sb->s_lock) != 0)
3056 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3060 static int ext4_sync_fs(struct super_block *sb, int wait)
3064 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3066 if (EXT4_SB(sb)->s_journal) {
3068 ret = ext4_force_commit(sb);
3070 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3072 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3078 * LVM calls this function before a (read-only) snapshot is created. This
3079 * gives us a chance to flush the journal completely and mark the fs clean.
3081 static void ext4_write_super_lockfs(struct super_block *sb)
3085 if (!(sb->s_flags & MS_RDONLY)) {
3086 journal_t *journal = EXT4_SB(sb)->s_journal;
3089 /* Now we set up the journal barrier. */
3090 jbd2_journal_lock_updates(journal);
3093 * We don't want to clear needs_recovery flag when we
3094 * failed to flush the journal.
3096 if (jbd2_journal_flush(journal) < 0)
3100 /* Journal blocked and flushed, clear needs_recovery flag. */
3101 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3102 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3107 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3108 * flag here, even though the filesystem is not technically dirty yet.
3110 static void ext4_unlockfs(struct super_block *sb)
3112 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3114 /* Reser the needs_recovery flag before the fs is unlocked. */
3115 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3116 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3118 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3122 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3124 struct ext4_super_block *es;
3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 ext4_fsblk_t n_blocks_count = 0;
3127 unsigned long old_sb_flags;
3128 struct ext4_mount_options old_opts;
3135 /* Store the original options */
3136 old_sb_flags = sb->s_flags;
3137 old_opts.s_mount_opt = sbi->s_mount_opt;
3138 old_opts.s_resuid = sbi->s_resuid;
3139 old_opts.s_resgid = sbi->s_resgid;
3140 old_opts.s_commit_interval = sbi->s_commit_interval;
3141 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3142 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3144 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3145 for (i = 0; i < MAXQUOTAS; i++)
3146 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3150 * Allow the "check" option to be passed as a remount option.
3152 if (!parse_options(data, sb, NULL, &n_blocks_count, 1)) {
3157 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3158 ext4_abort(sb, __func__, "Abort forced by user");
3160 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3161 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3166 ext4_init_journal_params(sb, sbi->s_journal);
3168 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3169 n_blocks_count > ext4_blocks_count(es)) {
3170 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3175 if (*flags & MS_RDONLY) {
3177 * First of all, the unconditional stuff we have to do
3178 * to disable replay of the journal when we next remount
3180 sb->s_flags |= MS_RDONLY;
3183 * OK, test if we are remounting a valid rw partition
3184 * readonly, and if so set the rdonly flag and then
3185 * mark the partition as valid again.
3187 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3188 (sbi->s_mount_state & EXT4_VALID_FS))
3189 es->s_state = cpu_to_le16(sbi->s_mount_state);
3192 * We have to unlock super so that we can wait for
3195 if (sbi->s_journal) {
3197 ext4_mark_recovery_complete(sb, es);
3202 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3203 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3204 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3205 "remount RDWR because of unsupported "
3206 "optional features (%x).\n", sb->s_id,
3207 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3208 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3214 * Make sure the group descriptor checksums
3215 * are sane. If they aren't, refuse to
3218 for (g = 0; g < sbi->s_groups_count; g++) {
3219 struct ext4_group_desc *gdp =
3220 ext4_get_group_desc(sb, g, NULL);
3222 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3224 "EXT4-fs: ext4_remount: "
3225 "Checksum for group %u failed (%u!=%u)\n",
3226 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3227 le16_to_cpu(gdp->bg_checksum));
3234 * If we have an unprocessed orphan list hanging
3235 * around from a previously readonly bdev mount,
3236 * require a full umount/remount for now.
3238 if (es->s_last_orphan) {
3239 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3240 "remount RDWR because of unprocessed "
3241 "orphan inode list. Please "
3242 "umount/remount instead.\n",
3249 * Mounting a RDONLY partition read-write, so reread
3250 * and store the current valid flag. (It may have
3251 * been changed by e2fsck since we originally mounted
3255 ext4_clear_journal_err(sb, es);
3256 sbi->s_mount_state = le16_to_cpu(es->s_state);
3257 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3259 if (!ext4_setup_super(sb, es, 0))
3260 sb->s_flags &= ~MS_RDONLY;
3263 if (sbi->s_journal == NULL)
3264 ext4_commit_super(sb, es, 1);
3267 /* Release old quota file names */
3268 for (i = 0; i < MAXQUOTAS; i++)
3269 if (old_opts.s_qf_names[i] &&
3270 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3271 kfree(old_opts.s_qf_names[i]);
3275 sb->s_flags = old_sb_flags;
3276 sbi->s_mount_opt = old_opts.s_mount_opt;
3277 sbi->s_resuid = old_opts.s_resuid;
3278 sbi->s_resgid = old_opts.s_resgid;
3279 sbi->s_commit_interval = old_opts.s_commit_interval;
3280 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3281 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3283 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3284 for (i = 0; i < MAXQUOTAS; i++) {
3285 if (sbi->s_qf_names[i] &&
3286 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3287 kfree(sbi->s_qf_names[i]);
3288 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3294 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3296 struct super_block *sb = dentry->d_sb;
3297 struct ext4_sb_info *sbi = EXT4_SB(sb);
3298 struct ext4_super_block *es = sbi->s_es;
3301 if (test_opt(sb, MINIX_DF)) {
3302 sbi->s_overhead_last = 0;
3303 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3304 ext4_group_t ngroups = sbi->s_groups_count, i;
3305 ext4_fsblk_t overhead = 0;
3309 * Compute the overhead (FS structures). This is constant
3310 * for a given filesystem unless the number of block groups
3311 * changes so we cache the previous value until it does.
3315 * All of the blocks before first_data_block are
3318 overhead = le32_to_cpu(es->s_first_data_block);
3321 * Add the overhead attributed to the superblock and
3322 * block group descriptors. If the sparse superblocks
3323 * feature is turned on, then not all groups have this.
3325 for (i = 0; i < ngroups; i++) {
3326 overhead += ext4_bg_has_super(sb, i) +
3327 ext4_bg_num_gdb(sb, i);
3332 * Every block group has an inode bitmap, a block
3333 * bitmap, and an inode table.
3335 overhead += ngroups * (2 + sbi->s_itb_per_group);
3336 sbi->s_overhead_last = overhead;
3338 sbi->s_blocks_last = ext4_blocks_count(es);
3341 buf->f_type = EXT4_SUPER_MAGIC;
3342 buf->f_bsize = sb->s_blocksize;
3343 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3344 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3345 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3346 ext4_free_blocks_count_set(es, buf->f_bfree);
3347 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3348 if (buf->f_bfree < ext4_r_blocks_count(es))
3350 buf->f_files = le32_to_cpu(es->s_inodes_count);
3351 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3352 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3353 buf->f_namelen = EXT4_NAME_LEN;
3354 fsid = le64_to_cpup((void *)es->s_uuid) ^
3355 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3356 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3357 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3361 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3362 * is locked for write. Otherwise the are possible deadlocks:
3363 * Process 1 Process 2
3364 * ext4_create() quota_sync()
3365 * jbd2_journal_start() write_dquot()
3366 * DQUOT_INIT() down(dqio_mutex)
3367 * down(dqio_mutex) jbd2_journal_start()
3373 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3375 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3378 static int ext4_dquot_initialize(struct inode *inode, int type)
3383 /* We may create quota structure so we need to reserve enough blocks */
3384 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3386 return PTR_ERR(handle);
3387 ret = dquot_initialize(inode, type);
3388 err = ext4_journal_stop(handle);
3394 static int ext4_dquot_drop(struct inode *inode)
3399 /* We may delete quota structure so we need to reserve enough blocks */
3400 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3401 if (IS_ERR(handle)) {
3403 * We call dquot_drop() anyway to at least release references
3404 * to quota structures so that umount does not hang.
3407 return PTR_ERR(handle);
3409 ret = dquot_drop(inode);
3410 err = ext4_journal_stop(handle);
3416 static int ext4_write_dquot(struct dquot *dquot)
3420 struct inode *inode;
3422 inode = dquot_to_inode(dquot);
3423 handle = ext4_journal_start(inode,
3424 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3426 return PTR_ERR(handle);
3427 ret = dquot_commit(dquot);
3428 err = ext4_journal_stop(handle);
3434 static int ext4_acquire_dquot(struct dquot *dquot)
3439 handle = ext4_journal_start(dquot_to_inode(dquot),
3440 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3442 return PTR_ERR(handle);
3443 ret = dquot_acquire(dquot);
3444 err = ext4_journal_stop(handle);
3450 static int ext4_release_dquot(struct dquot *dquot)
3455 handle = ext4_journal_start(dquot_to_inode(dquot),
3456 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3457 if (IS_ERR(handle)) {
3458 /* Release dquot anyway to avoid endless cycle in dqput() */
3459 dquot_release(dquot);
3460 return PTR_ERR(handle);
3462 ret = dquot_release(dquot);
3463 err = ext4_journal_stop(handle);
3469 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3471 /* Are we journaling quotas? */
3472 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3473 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3474 dquot_mark_dquot_dirty(dquot);
3475 return ext4_write_dquot(dquot);
3477 return dquot_mark_dquot_dirty(dquot);
3481 static int ext4_write_info(struct super_block *sb, int type)
3486 /* Data block + inode block */
3487 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3489 return PTR_ERR(handle);
3490 ret = dquot_commit_info(sb, type);
3491 err = ext4_journal_stop(handle);
3498 * Turn on quotas during mount time - we need to find
3499 * the quota file and such...
3501 static int ext4_quota_on_mount(struct super_block *sb, int type)
3503 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3504 EXT4_SB(sb)->s_jquota_fmt, type);
3508 * Standard function to be called on quota_on
3510 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3511 char *name, int remount)
3516 if (!test_opt(sb, QUOTA))
3518 /* When remounting, no checks are needed and in fact, name is NULL */
3520 return vfs_quota_on(sb, type, format_id, name, remount);
3522 err = kern_path(name, LOOKUP_FOLLOW, &path);
3526 /* Quotafile not on the same filesystem? */
3527 if (path.mnt->mnt_sb != sb) {
3531 /* Journaling quota? */
3532 if (EXT4_SB(sb)->s_qf_names[type]) {
3533 /* Quotafile not in fs root? */
3534 if (path.dentry->d_parent != sb->s_root)
3536 "EXT4-fs: Quota file not on filesystem root. "
3537 "Journaled quota will not work.\n");
3541 * When we journal data on quota file, we have to flush journal to see
3542 * all updates to the file when we bypass pagecache...
3544 if (EXT4_SB(sb)->s_journal &&
3545 ext4_should_journal_data(path.dentry->d_inode)) {
3547 * We don't need to lock updates but journal_flush() could
3548 * otherwise be livelocked...
3550 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3551 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3552 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3559 err = vfs_quota_on_path(sb, type, format_id, &path);
3564 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3565 * acquiring the locks... As quota files are never truncated and quota code
3566 * itself serializes the operations (and noone else should touch the files)
3567 * we don't have to be afraid of races */
3568 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3569 size_t len, loff_t off)
3571 struct inode *inode = sb_dqopt(sb)->files[type];
3572 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3574 int offset = off & (sb->s_blocksize - 1);
3577 struct buffer_head *bh;
3578 loff_t i_size = i_size_read(inode);
3582 if (off+len > i_size)
3585 while (toread > 0) {
3586 tocopy = sb->s_blocksize - offset < toread ?
3587 sb->s_blocksize - offset : toread;
3588 bh = ext4_bread(NULL, inode, blk, 0, &err);
3591 if (!bh) /* A hole? */
3592 memset(data, 0, tocopy);
3594 memcpy(data, bh->b_data+offset, tocopy);
3604 /* Write to quotafile (we know the transaction is already started and has
3605 * enough credits) */
3606 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3607 const char *data, size_t len, loff_t off)
3609 struct inode *inode = sb_dqopt(sb)->files[type];
3610 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3612 int offset = off & (sb->s_blocksize - 1);
3614 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3615 size_t towrite = len;
3616 struct buffer_head *bh;
3617 handle_t *handle = journal_current_handle();
3619 if (EXT4_SB(sb)->s_journal && !handle) {
3620 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3621 " cancelled because transaction is not started.\n",
3622 (unsigned long long)off, (unsigned long long)len);
3625 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3626 while (towrite > 0) {
3627 tocopy = sb->s_blocksize - offset < towrite ?
3628 sb->s_blocksize - offset : towrite;
3629 bh = ext4_bread(handle, inode, blk, 1, &err);
3632 if (journal_quota) {
3633 err = ext4_journal_get_write_access(handle, bh);
3640 memcpy(bh->b_data+offset, data, tocopy);
3641 flush_dcache_page(bh->b_page);
3644 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3646 /* Always do at least ordered writes for quotas */
3647 err = ext4_jbd2_file_inode(handle, inode);
3648 mark_buffer_dirty(bh);
3659 if (len == towrite) {
3660 mutex_unlock(&inode->i_mutex);
3663 if (inode->i_size < off+len-towrite) {
3664 i_size_write(inode, off+len-towrite);
3665 EXT4_I(inode)->i_disksize = inode->i_size;
3667 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3668 ext4_mark_inode_dirty(handle, inode);
3669 mutex_unlock(&inode->i_mutex);
3670 return len - towrite;
3675 static int ext4_get_sb(struct file_system_type *fs_type,
3676 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3678 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3681 #ifdef CONFIG_PROC_FS
3682 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3684 unsigned int *p = m->private;
3686 seq_printf(m, "%u\n", *p);
3690 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3692 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3695 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3696 size_t cnt, loff_t *ppos)
3698 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3701 if (cnt >= sizeof(str))
3703 if (copy_from_user(str, buf, cnt))
3706 *p = simple_strtoul(str, NULL, 0);
3710 const struct file_operations ext4_ui_proc_fops = {
3711 .owner = THIS_MODULE,
3712 .open = ext4_ui_proc_open,
3714 .llseek = seq_lseek,
3715 .release = single_release,
3716 .write = ext4_ui_proc_write,
3720 static struct file_system_type ext4_fs_type = {
3721 .owner = THIS_MODULE,
3723 .get_sb = ext4_get_sb,
3724 .kill_sb = kill_block_super,
3725 .fs_flags = FS_REQUIRES_DEV,
3728 #ifdef CONFIG_EXT4DEV_COMPAT
3729 static int ext4dev_get_sb(struct file_system_type *fs_type,
3730 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3732 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3733 "to mount using ext4\n");
3734 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3735 "will go away by 2.6.31\n");
3736 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3739 static struct file_system_type ext4dev_fs_type = {
3740 .owner = THIS_MODULE,
3742 .get_sb = ext4dev_get_sb,
3743 .kill_sb = kill_block_super,
3744 .fs_flags = FS_REQUIRES_DEV,
3746 MODULE_ALIAS("ext4dev");
3749 static int __init init_ext4_fs(void)
3753 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3754 err = init_ext4_mballoc();
3758 err = init_ext4_xattr();
3761 err = init_inodecache();
3764 err = register_filesystem(&ext4_fs_type);
3767 #ifdef CONFIG_EXT4DEV_COMPAT
3768 err = register_filesystem(&ext4dev_fs_type);
3770 unregister_filesystem(&ext4_fs_type);
3776 destroy_inodecache();
3780 exit_ext4_mballoc();
3784 static void __exit exit_ext4_fs(void)
3786 unregister_filesystem(&ext4_fs_type);
3787 #ifdef CONFIG_EXT4DEV_COMPAT
3788 unregister_filesystem(&ext4dev_fs_type);
3790 destroy_inodecache();
3792 exit_ext4_mballoc();
3793 remove_proc_entry("fs/ext4", NULL);
3796 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3797 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3798 MODULE_LICENSE("GPL");
3799 module_init(init_ext4_fs)
3800 module_exit(exit_ext4_fs)