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-fs: 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, I_VERSION))
833 seq_puts(seq, ",i_version");
834 if (!test_opt(sb, DELALLOC))
835 seq_puts(seq, ",nodelalloc");
839 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
841 * journal mode get enabled in different ways
842 * So just print the value even if we didn't specify it
844 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
845 seq_puts(seq, ",data=journal");
846 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
847 seq_puts(seq, ",data=ordered");
848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
849 seq_puts(seq, ",data=writeback");
851 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
852 seq_printf(seq, ",inode_readahead_blks=%u",
853 sbi->s_inode_readahead_blks);
855 if (test_opt(sb, DATA_ERR_ABORT))
856 seq_puts(seq, ",data_err=abort");
858 ext4_show_quota_options(seq, sb);
863 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
864 u64 ino, u32 generation)
868 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
869 return ERR_PTR(-ESTALE);
870 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
871 return ERR_PTR(-ESTALE);
873 /* iget isn't really right if the inode is currently unallocated!!
875 * ext4_read_inode will return a bad_inode if the inode had been
876 * deleted, so we should be safe.
878 * Currently we don't know the generation for parent directory, so
879 * a generation of 0 means "accept any"
881 inode = ext4_iget(sb, ino);
883 return ERR_CAST(inode);
884 if (generation && inode->i_generation != generation) {
886 return ERR_PTR(-ESTALE);
892 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
893 int fh_len, int fh_type)
895 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
899 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
900 int fh_len, int fh_type)
902 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
907 * Try to release metadata pages (indirect blocks, directories) which are
908 * mapped via the block device. Since these pages could have journal heads
909 * which would prevent try_to_free_buffers() from freeing them, we must use
910 * jbd2 layer's try_to_free_buffers() function to release them.
912 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
914 journal_t *journal = EXT4_SB(sb)->s_journal;
916 WARN_ON(PageChecked(page));
917 if (!page_has_buffers(page))
920 return jbd2_journal_try_to_free_buffers(journal, page,
922 return try_to_free_buffers(page);
926 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
927 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
929 static int ext4_dquot_initialize(struct inode *inode, int type);
930 static int ext4_dquot_drop(struct inode *inode);
931 static int ext4_write_dquot(struct dquot *dquot);
932 static int ext4_acquire_dquot(struct dquot *dquot);
933 static int ext4_release_dquot(struct dquot *dquot);
934 static int ext4_mark_dquot_dirty(struct dquot *dquot);
935 static int ext4_write_info(struct super_block *sb, int type);
936 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
937 char *path, int remount);
938 static int ext4_quota_on_mount(struct super_block *sb, int type);
939 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
940 size_t len, loff_t off);
941 static ssize_t ext4_quota_write(struct super_block *sb, int type,
942 const char *data, size_t len, loff_t off);
944 static struct dquot_operations ext4_quota_operations = {
945 .initialize = ext4_dquot_initialize,
946 .drop = ext4_dquot_drop,
947 .alloc_space = dquot_alloc_space,
948 .alloc_inode = dquot_alloc_inode,
949 .free_space = dquot_free_space,
950 .free_inode = dquot_free_inode,
951 .transfer = dquot_transfer,
952 .write_dquot = ext4_write_dquot,
953 .acquire_dquot = ext4_acquire_dquot,
954 .release_dquot = ext4_release_dquot,
955 .mark_dirty = ext4_mark_dquot_dirty,
956 .write_info = ext4_write_info
959 static struct quotactl_ops ext4_qctl_operations = {
960 .quota_on = ext4_quota_on,
961 .quota_off = vfs_quota_off,
962 .quota_sync = vfs_quota_sync,
963 .get_info = vfs_get_dqinfo,
964 .set_info = vfs_set_dqinfo,
965 .get_dqblk = vfs_get_dqblk,
966 .set_dqblk = vfs_set_dqblk
970 static const struct super_operations ext4_sops = {
971 .alloc_inode = ext4_alloc_inode,
972 .destroy_inode = ext4_destroy_inode,
973 .write_inode = ext4_write_inode,
974 .dirty_inode = ext4_dirty_inode,
975 .delete_inode = ext4_delete_inode,
976 .put_super = ext4_put_super,
977 .write_super = ext4_write_super,
978 .sync_fs = ext4_sync_fs,
979 .write_super_lockfs = ext4_write_super_lockfs,
980 .unlockfs = ext4_unlockfs,
981 .statfs = ext4_statfs,
982 .remount_fs = ext4_remount,
983 .clear_inode = ext4_clear_inode,
984 .show_options = ext4_show_options,
986 .quota_read = ext4_quota_read,
987 .quota_write = ext4_quota_write,
989 .bdev_try_to_free_page = bdev_try_to_free_page,
992 static const struct export_operations ext4_export_ops = {
993 .fh_to_dentry = ext4_fh_to_dentry,
994 .fh_to_parent = ext4_fh_to_parent,
995 .get_parent = ext4_get_parent,
999 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1000 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1001 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1002 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1003 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
1004 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1005 Opt_journal_update, Opt_journal_dev,
1006 Opt_journal_checksum, Opt_journal_async_commit,
1007 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1008 Opt_data_err_abort, Opt_data_err_ignore,
1009 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1010 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1011 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1012 Opt_grpquota, Opt_i_version,
1013 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1014 Opt_inode_readahead_blks, Opt_journal_ioprio
1017 static const match_table_t tokens = {
1018 {Opt_bsd_df, "bsddf"},
1019 {Opt_minix_df, "minixdf"},
1020 {Opt_grpid, "grpid"},
1021 {Opt_grpid, "bsdgroups"},
1022 {Opt_nogrpid, "nogrpid"},
1023 {Opt_nogrpid, "sysvgroups"},
1024 {Opt_resgid, "resgid=%u"},
1025 {Opt_resuid, "resuid=%u"},
1027 {Opt_err_cont, "errors=continue"},
1028 {Opt_err_panic, "errors=panic"},
1029 {Opt_err_ro, "errors=remount-ro"},
1030 {Opt_nouid32, "nouid32"},
1031 {Opt_debug, "debug"},
1032 {Opt_oldalloc, "oldalloc"},
1033 {Opt_orlov, "orlov"},
1034 {Opt_user_xattr, "user_xattr"},
1035 {Opt_nouser_xattr, "nouser_xattr"},
1037 {Opt_noacl, "noacl"},
1038 {Opt_reservation, "reservation"},
1039 {Opt_noreservation, "noreservation"},
1040 {Opt_noload, "noload"},
1043 {Opt_commit, "commit=%u"},
1044 {Opt_min_batch_time, "min_batch_time=%u"},
1045 {Opt_max_batch_time, "max_batch_time=%u"},
1046 {Opt_journal_update, "journal=update"},
1047 {Opt_journal_dev, "journal_dev=%u"},
1048 {Opt_journal_checksum, "journal_checksum"},
1049 {Opt_journal_async_commit, "journal_async_commit"},
1050 {Opt_abort, "abort"},
1051 {Opt_data_journal, "data=journal"},
1052 {Opt_data_ordered, "data=ordered"},
1053 {Opt_data_writeback, "data=writeback"},
1054 {Opt_data_err_abort, "data_err=abort"},
1055 {Opt_data_err_ignore, "data_err=ignore"},
1056 {Opt_offusrjquota, "usrjquota="},
1057 {Opt_usrjquota, "usrjquota=%s"},
1058 {Opt_offgrpjquota, "grpjquota="},
1059 {Opt_grpjquota, "grpjquota=%s"},
1060 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1061 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1062 {Opt_grpquota, "grpquota"},
1063 {Opt_noquota, "noquota"},
1064 {Opt_quota, "quota"},
1065 {Opt_usrquota, "usrquota"},
1066 {Opt_barrier, "barrier=%u"},
1067 {Opt_i_version, "i_version"},
1068 {Opt_stripe, "stripe=%u"},
1069 {Opt_resize, "resize"},
1070 {Opt_delalloc, "delalloc"},
1071 {Opt_nodelalloc, "nodelalloc"},
1072 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1073 {Opt_journal_ioprio, "journal_ioprio=%u"},
1077 static ext4_fsblk_t get_sb_block(void **data)
1079 ext4_fsblk_t sb_block;
1080 char *options = (char *) *data;
1082 if (!options || strncmp(options, "sb=", 3) != 0)
1083 return 1; /* Default location */
1085 /*todo: use simple_strtoll with >32bit ext4 */
1086 sb_block = simple_strtoul(options, &options, 0);
1087 if (*options && *options != ',') {
1088 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1092 if (*options == ',')
1094 *data = (void *) options;
1098 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1100 static int parse_options(char *options, struct super_block *sb,
1101 unsigned long *journal_devnum,
1102 unsigned int *journal_ioprio,
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];
1118 while ((p = strsep(&options, ",")) != NULL) {
1123 token = match_token(p, tokens, args);
1126 clear_opt(sbi->s_mount_opt, MINIX_DF);
1129 set_opt(sbi->s_mount_opt, MINIX_DF);
1132 set_opt(sbi->s_mount_opt, GRPID);
1135 clear_opt(sbi->s_mount_opt, GRPID);
1138 if (match_int(&args[0], &option))
1140 sbi->s_resuid = option;
1143 if (match_int(&args[0], &option))
1145 sbi->s_resgid = option;
1148 /* handled by get_sb_block() instead of here */
1149 /* *sb_block = match_int(&args[0]); */
1152 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1153 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1154 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1157 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1158 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1159 set_opt(sbi->s_mount_opt, ERRORS_RO);
1162 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1163 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1164 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1167 set_opt(sbi->s_mount_opt, NO_UID32);
1170 set_opt(sbi->s_mount_opt, DEBUG);
1173 set_opt(sbi->s_mount_opt, OLDALLOC);
1176 clear_opt(sbi->s_mount_opt, OLDALLOC);
1178 #ifdef CONFIG_EXT4_FS_XATTR
1179 case Opt_user_xattr:
1180 set_opt(sbi->s_mount_opt, XATTR_USER);
1182 case Opt_nouser_xattr:
1183 clear_opt(sbi->s_mount_opt, XATTR_USER);
1186 case Opt_user_xattr:
1187 case Opt_nouser_xattr:
1188 printk(KERN_ERR "EXT4 (no)user_xattr options "
1192 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1194 set_opt(sbi->s_mount_opt, POSIX_ACL);
1197 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1202 printk(KERN_ERR "EXT4 (no)acl options "
1206 case Opt_reservation:
1207 set_opt(sbi->s_mount_opt, RESERVATION);
1209 case Opt_noreservation:
1210 clear_opt(sbi->s_mount_opt, RESERVATION);
1212 case Opt_journal_update:
1214 /* Eventually we will want to be able to create
1215 a journal file here. For now, only allow the
1216 user to specify an existing inode to be the
1219 printk(KERN_ERR "EXT4-fs: cannot specify "
1220 "journal on remount\n");
1223 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1225 case Opt_journal_dev:
1227 printk(KERN_ERR "EXT4-fs: cannot specify "
1228 "journal on remount\n");
1231 if (match_int(&args[0], &option))
1233 *journal_devnum = option;
1235 case Opt_journal_checksum:
1236 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1238 case Opt_journal_async_commit:
1239 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1240 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1243 set_opt(sbi->s_mount_opt, NOLOAD);
1246 if (match_int(&args[0], &option))
1251 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1252 sbi->s_commit_interval = HZ * option;
1254 case Opt_max_batch_time:
1255 if (match_int(&args[0], &option))
1260 option = EXT4_DEF_MAX_BATCH_TIME;
1261 sbi->s_max_batch_time = option;
1263 case Opt_min_batch_time:
1264 if (match_int(&args[0], &option))
1268 sbi->s_min_batch_time = option;
1270 case Opt_data_journal:
1271 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1273 case Opt_data_ordered:
1274 data_opt = EXT4_MOUNT_ORDERED_DATA;
1276 case Opt_data_writeback:
1277 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1280 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1283 "EXT4-fs: cannot change data "
1284 "mode on remount\n");
1288 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1289 sbi->s_mount_opt |= data_opt;
1292 case Opt_data_err_abort:
1293 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1295 case Opt_data_err_ignore:
1296 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1305 if ((sb_any_quota_enabled(sb) ||
1306 sb_any_quota_suspended(sb)) &&
1307 !sbi->s_qf_names[qtype]) {
1309 "EXT4-fs: Cannot change journaled "
1310 "quota options when quota turned on.\n");
1313 qname = match_strdup(&args[0]);
1316 "EXT4-fs: not enough memory for "
1317 "storing quotafile name.\n");
1320 if (sbi->s_qf_names[qtype] &&
1321 strcmp(sbi->s_qf_names[qtype], qname)) {
1323 "EXT4-fs: %s quota file already "
1324 "specified.\n", QTYPE2NAME(qtype));
1328 sbi->s_qf_names[qtype] = qname;
1329 if (strchr(sbi->s_qf_names[qtype], '/')) {
1331 "EXT4-fs: quotafile must be on "
1332 "filesystem root.\n");
1333 kfree(sbi->s_qf_names[qtype]);
1334 sbi->s_qf_names[qtype] = NULL;
1337 set_opt(sbi->s_mount_opt, QUOTA);
1339 case Opt_offusrjquota:
1342 case Opt_offgrpjquota:
1345 if ((sb_any_quota_enabled(sb) ||
1346 sb_any_quota_suspended(sb)) &&
1347 sbi->s_qf_names[qtype]) {
1348 printk(KERN_ERR "EXT4-fs: Cannot change "
1349 "journaled quota options when "
1350 "quota turned on.\n");
1354 * The space will be released later when all options
1355 * are confirmed to be correct
1357 sbi->s_qf_names[qtype] = NULL;
1359 case Opt_jqfmt_vfsold:
1360 qfmt = QFMT_VFS_OLD;
1362 case Opt_jqfmt_vfsv0:
1365 if ((sb_any_quota_enabled(sb) ||
1366 sb_any_quota_suspended(sb)) &&
1367 sbi->s_jquota_fmt != qfmt) {
1368 printk(KERN_ERR "EXT4-fs: Cannot change "
1369 "journaled quota options when "
1370 "quota turned on.\n");
1373 sbi->s_jquota_fmt = qfmt;
1377 set_opt(sbi->s_mount_opt, QUOTA);
1378 set_opt(sbi->s_mount_opt, USRQUOTA);
1381 set_opt(sbi->s_mount_opt, QUOTA);
1382 set_opt(sbi->s_mount_opt, GRPQUOTA);
1385 if (sb_any_quota_enabled(sb)) {
1386 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1387 "options when quota turned on.\n");
1390 clear_opt(sbi->s_mount_opt, QUOTA);
1391 clear_opt(sbi->s_mount_opt, USRQUOTA);
1392 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1399 "EXT4-fs: quota options not supported.\n");
1403 case Opt_offusrjquota:
1404 case Opt_offgrpjquota:
1405 case Opt_jqfmt_vfsold:
1406 case Opt_jqfmt_vfsv0:
1408 "EXT4-fs: journaled quota options not "
1415 set_opt(sbi->s_mount_opt, ABORT);
1418 if (match_int(&args[0], &option))
1421 set_opt(sbi->s_mount_opt, BARRIER);
1423 clear_opt(sbi->s_mount_opt, BARRIER);
1429 printk("EXT4-fs: resize option only available "
1433 if (match_int(&args[0], &option) != 0)
1435 *n_blocks_count = option;
1438 set_opt(sbi->s_mount_opt, NOBH);
1441 clear_opt(sbi->s_mount_opt, NOBH);
1444 set_opt(sbi->s_mount_opt, I_VERSION);
1445 sb->s_flags |= MS_I_VERSION;
1447 case Opt_nodelalloc:
1448 clear_opt(sbi->s_mount_opt, DELALLOC);
1451 if (match_int(&args[0], &option))
1455 sbi->s_stripe = option;
1458 set_opt(sbi->s_mount_opt, DELALLOC);
1460 case Opt_inode_readahead_blks:
1461 if (match_int(&args[0], &option))
1463 if (option < 0 || option > (1 << 30))
1465 sbi->s_inode_readahead_blks = option;
1467 case Opt_journal_ioprio:
1468 if (match_int(&args[0], &option))
1470 if (option < 0 || option > 7)
1472 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1477 "EXT4-fs: Unrecognized mount option \"%s\" "
1478 "or missing value\n", p);
1483 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1484 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1485 sbi->s_qf_names[USRQUOTA])
1486 clear_opt(sbi->s_mount_opt, USRQUOTA);
1488 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1489 sbi->s_qf_names[GRPQUOTA])
1490 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1492 if ((sbi->s_qf_names[USRQUOTA] &&
1493 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1494 (sbi->s_qf_names[GRPQUOTA] &&
1495 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1496 printk(KERN_ERR "EXT4-fs: old and new quota "
1497 "format mixing.\n");
1501 if (!sbi->s_jquota_fmt) {
1502 printk(KERN_ERR "EXT4-fs: journaled quota format "
1503 "not specified.\n");
1507 if (sbi->s_jquota_fmt) {
1508 printk(KERN_ERR "EXT4-fs: journaled quota format "
1509 "specified with no journaling "
1518 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1521 struct ext4_sb_info *sbi = EXT4_SB(sb);
1524 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1525 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1526 "forcing read-only mode\n");
1531 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1532 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1533 "running e2fsck is recommended\n");
1534 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1536 "EXT4-fs warning: mounting fs with errors, "
1537 "running e2fsck is recommended\n");
1538 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1539 le16_to_cpu(es->s_mnt_count) >=
1540 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1542 "EXT4-fs warning: maximal mount count reached, "
1543 "running e2fsck is recommended\n");
1544 else if (le32_to_cpu(es->s_checkinterval) &&
1545 (le32_to_cpu(es->s_lastcheck) +
1546 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1548 "EXT4-fs warning: checktime reached, "
1549 "running e2fsck is recommended\n");
1550 if (!sbi->s_journal)
1551 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1552 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1553 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1554 le16_add_cpu(&es->s_mnt_count, 1);
1555 es->s_mtime = cpu_to_le32(get_seconds());
1556 ext4_update_dynamic_rev(sb);
1558 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1560 ext4_commit_super(sb, es, 1);
1561 if (test_opt(sb, DEBUG))
1562 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1563 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1565 sbi->s_groups_count,
1566 EXT4_BLOCKS_PER_GROUP(sb),
1567 EXT4_INODES_PER_GROUP(sb),
1570 if (EXT4_SB(sb)->s_journal) {
1571 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1572 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1573 "external", EXT4_SB(sb)->s_journal->j_devname);
1575 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1580 static int ext4_fill_flex_info(struct super_block *sb)
1582 struct ext4_sb_info *sbi = EXT4_SB(sb);
1583 struct ext4_group_desc *gdp = NULL;
1584 struct buffer_head *bh;
1585 ext4_group_t flex_group_count;
1586 ext4_group_t flex_group;
1587 int groups_per_flex = 0;
1590 if (!sbi->s_es->s_log_groups_per_flex) {
1591 sbi->s_log_groups_per_flex = 0;
1595 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1596 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1598 /* We allocate both existing and potentially added groups */
1599 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1600 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1601 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1602 sbi->s_flex_groups = kzalloc(flex_group_count *
1603 sizeof(struct flex_groups), GFP_KERNEL);
1604 if (sbi->s_flex_groups == NULL) {
1605 printk(KERN_ERR "EXT4-fs: not enough memory for "
1606 "%u flex groups\n", flex_group_count);
1610 for (i = 0; i < sbi->s_groups_count; i++) {
1611 gdp = ext4_get_group_desc(sb, i, &bh);
1613 flex_group = ext4_flex_group(sbi, i);
1614 sbi->s_flex_groups[flex_group].free_inodes +=
1615 ext4_free_inodes_count(sb, gdp);
1616 sbi->s_flex_groups[flex_group].free_blocks +=
1617 ext4_free_blks_count(sb, gdp);
1625 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1626 struct ext4_group_desc *gdp)
1630 if (sbi->s_es->s_feature_ro_compat &
1631 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1632 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1633 __le32 le_group = cpu_to_le32(block_group);
1635 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1636 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1637 crc = crc16(crc, (__u8 *)gdp, offset);
1638 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1639 /* for checksum of struct ext4_group_desc do the rest...*/
1640 if ((sbi->s_es->s_feature_incompat &
1641 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1642 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1643 crc = crc16(crc, (__u8 *)gdp + offset,
1644 le16_to_cpu(sbi->s_es->s_desc_size) -
1648 return cpu_to_le16(crc);
1651 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1652 struct ext4_group_desc *gdp)
1654 if ((sbi->s_es->s_feature_ro_compat &
1655 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1656 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1662 /* Called at mount-time, super-block is locked */
1663 static int ext4_check_descriptors(struct super_block *sb)
1665 struct ext4_sb_info *sbi = EXT4_SB(sb);
1666 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1667 ext4_fsblk_t last_block;
1668 ext4_fsblk_t block_bitmap;
1669 ext4_fsblk_t inode_bitmap;
1670 ext4_fsblk_t inode_table;
1671 int flexbg_flag = 0;
1674 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1677 ext4_debug("Checking group descriptors");
1679 for (i = 0; i < sbi->s_groups_count; i++) {
1680 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1682 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1683 last_block = ext4_blocks_count(sbi->s_es) - 1;
1685 last_block = first_block +
1686 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1688 block_bitmap = ext4_block_bitmap(sb, gdp);
1689 if (block_bitmap < first_block || block_bitmap > last_block) {
1690 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1691 "Block bitmap for group %u not in group "
1692 "(block %llu)!\n", i, block_bitmap);
1695 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1696 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1697 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1698 "Inode bitmap for group %u not in group "
1699 "(block %llu)!\n", i, inode_bitmap);
1702 inode_table = ext4_inode_table(sb, gdp);
1703 if (inode_table < first_block ||
1704 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1705 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1706 "Inode table for group %u not in group "
1707 "(block %llu)!\n", i, inode_table);
1710 spin_lock(sb_bgl_lock(sbi, i));
1711 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1712 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1713 "Checksum for group %u failed (%u!=%u)\n",
1714 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1715 gdp)), le16_to_cpu(gdp->bg_checksum));
1716 if (!(sb->s_flags & MS_RDONLY)) {
1717 spin_unlock(sb_bgl_lock(sbi, i));
1721 spin_unlock(sb_bgl_lock(sbi, i));
1723 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1726 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1727 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1731 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1732 * the superblock) which were deleted from all directories, but held open by
1733 * a process at the time of a crash. We walk the list and try to delete these
1734 * inodes at recovery time (only with a read-write filesystem).
1736 * In order to keep the orphan inode chain consistent during traversal (in
1737 * case of crash during recovery), we link each inode into the superblock
1738 * orphan list_head and handle it the same way as an inode deletion during
1739 * normal operation (which journals the operations for us).
1741 * We only do an iget() and an iput() on each inode, which is very safe if we
1742 * accidentally point at an in-use or already deleted inode. The worst that
1743 * can happen in this case is that we get a "bit already cleared" message from
1744 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1745 * e2fsck was run on this filesystem, and it must have already done the orphan
1746 * inode cleanup for us, so we can safely abort without any further action.
1748 static void ext4_orphan_cleanup(struct super_block *sb,
1749 struct ext4_super_block *es)
1751 unsigned int s_flags = sb->s_flags;
1752 int nr_orphans = 0, nr_truncates = 0;
1756 if (!es->s_last_orphan) {
1757 jbd_debug(4, "no orphan inodes to clean up\n");
1761 if (bdev_read_only(sb->s_bdev)) {
1762 printk(KERN_ERR "EXT4-fs: write access "
1763 "unavailable, skipping orphan cleanup.\n");
1767 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1768 if (es->s_last_orphan)
1769 jbd_debug(1, "Errors on filesystem, "
1770 "clearing orphan list.\n");
1771 es->s_last_orphan = 0;
1772 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1776 if (s_flags & MS_RDONLY) {
1777 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1779 sb->s_flags &= ~MS_RDONLY;
1782 /* Needed for iput() to work correctly and not trash data */
1783 sb->s_flags |= MS_ACTIVE;
1784 /* Turn on quotas so that they are updated correctly */
1785 for (i = 0; i < MAXQUOTAS; i++) {
1786 if (EXT4_SB(sb)->s_qf_names[i]) {
1787 int ret = ext4_quota_on_mount(sb, i);
1790 "EXT4-fs: Cannot turn on journaled "
1791 "quota: error %d\n", ret);
1796 while (es->s_last_orphan) {
1797 struct inode *inode;
1799 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1800 if (IS_ERR(inode)) {
1801 es->s_last_orphan = 0;
1805 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1807 if (inode->i_nlink) {
1809 "%s: truncating inode %lu to %lld bytes\n",
1810 __func__, inode->i_ino, inode->i_size);
1811 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1812 inode->i_ino, inode->i_size);
1813 ext4_truncate(inode);
1817 "%s: deleting unreferenced inode %lu\n",
1818 __func__, inode->i_ino);
1819 jbd_debug(2, "deleting unreferenced inode %lu\n",
1823 iput(inode); /* The delete magic happens here! */
1826 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1829 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1830 sb->s_id, PLURAL(nr_orphans));
1832 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1833 sb->s_id, PLURAL(nr_truncates));
1835 /* Turn quotas off */
1836 for (i = 0; i < MAXQUOTAS; i++) {
1837 if (sb_dqopt(sb)->files[i])
1838 vfs_quota_off(sb, i, 0);
1841 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1844 * Maximal extent format file size.
1845 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1846 * extent format containers, within a sector_t, and within i_blocks
1847 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1848 * so that won't be a limiting factor.
1850 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1852 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1855 loff_t upper_limit = MAX_LFS_FILESIZE;
1857 /* small i_blocks in vfs inode? */
1858 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1860 * CONFIG_LBD is not enabled implies the inode
1861 * i_block represent total blocks in 512 bytes
1862 * 32 == size of vfs inode i_blocks * 8
1864 upper_limit = (1LL << 32) - 1;
1866 /* total blocks in file system block size */
1867 upper_limit >>= (blkbits - 9);
1868 upper_limit <<= blkbits;
1871 /* 32-bit extent-start container, ee_block */
1876 /* Sanity check against vm- & vfs- imposed limits */
1877 if (res > upper_limit)
1884 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1885 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1886 * We need to be 1 filesystem block less than the 2^48 sector limit.
1888 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1890 loff_t res = EXT4_NDIR_BLOCKS;
1893 /* This is calculated to be the largest file size for a
1894 * dense, bitmapped file such that the total number of
1895 * sectors in the file, including data and all indirect blocks,
1896 * does not exceed 2^48 -1
1897 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1898 * total number of 512 bytes blocks of the file
1901 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1903 * !has_huge_files or CONFIG_LBD is not enabled
1904 * implies the inode i_block represent total blocks in
1905 * 512 bytes 32 == size of vfs inode i_blocks * 8
1907 upper_limit = (1LL << 32) - 1;
1909 /* total blocks in file system block size */
1910 upper_limit >>= (bits - 9);
1914 * We use 48 bit ext4_inode i_blocks
1915 * With EXT4_HUGE_FILE_FL set the i_blocks
1916 * represent total number of blocks in
1917 * file system block size
1919 upper_limit = (1LL << 48) - 1;
1923 /* indirect blocks */
1925 /* double indirect blocks */
1926 meta_blocks += 1 + (1LL << (bits-2));
1927 /* tripple indirect blocks */
1928 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1930 upper_limit -= meta_blocks;
1931 upper_limit <<= bits;
1933 res += 1LL << (bits-2);
1934 res += 1LL << (2*(bits-2));
1935 res += 1LL << (3*(bits-2));
1937 if (res > upper_limit)
1940 if (res > MAX_LFS_FILESIZE)
1941 res = MAX_LFS_FILESIZE;
1946 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1947 ext4_fsblk_t logical_sb_block, int nr)
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 ext4_group_t bg, first_meta_bg;
1953 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1955 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1957 return logical_sb_block + nr + 1;
1958 bg = sbi->s_desc_per_block * nr;
1959 if (ext4_bg_has_super(sb, bg))
1961 return (has_super + ext4_group_first_block_no(sb, bg));
1965 * ext4_get_stripe_size: Get the stripe size.
1966 * @sbi: In memory super block info
1968 * If we have specified it via mount option, then
1969 * use the mount option value. If the value specified at mount time is
1970 * greater than the blocks per group use the super block value.
1971 * If the super block value is greater than blocks per group return 0.
1972 * Allocator needs it be less than blocks per group.
1975 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1977 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1978 unsigned long stripe_width =
1979 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1981 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1982 return sbi->s_stripe;
1984 if (stripe_width <= sbi->s_blocks_per_group)
1985 return stripe_width;
1987 if (stride <= sbi->s_blocks_per_group)
1993 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1994 __releases(kernel_lock)
1995 __acquires(kernel_lock)
1998 struct buffer_head *bh;
1999 struct ext4_super_block *es = NULL;
2000 struct ext4_sb_info *sbi;
2002 ext4_fsblk_t sb_block = get_sb_block(&data);
2003 ext4_fsblk_t logical_sb_block;
2004 unsigned long offset = 0;
2005 unsigned long journal_devnum = 0;
2006 unsigned long def_mount_opts;
2012 unsigned int db_count;
2014 int needs_recovery, has_huge_files;
2018 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2020 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2023 sb->s_fs_info = sbi;
2024 sbi->s_mount_opt = 0;
2025 sbi->s_resuid = EXT4_DEF_RESUID;
2026 sbi->s_resgid = EXT4_DEF_RESGID;
2027 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2028 sbi->s_sb_block = sb_block;
2032 /* Cleanup superblock name */
2033 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2036 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2038 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2043 * The ext4 superblock will not be buffer aligned for other than 1kB
2044 * block sizes. We need to calculate the offset from buffer start.
2046 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2047 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2048 offset = do_div(logical_sb_block, blocksize);
2050 logical_sb_block = sb_block;
2053 if (!(bh = sb_bread(sb, logical_sb_block))) {
2054 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2058 * Note: s_es must be initialized as soon as possible because
2059 * some ext4 macro-instructions depend on its value
2061 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2063 sb->s_magic = le16_to_cpu(es->s_magic);
2064 if (sb->s_magic != EXT4_SUPER_MAGIC)
2067 /* Set defaults before we parse the mount options */
2068 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2069 if (def_mount_opts & EXT4_DEFM_DEBUG)
2070 set_opt(sbi->s_mount_opt, DEBUG);
2071 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2072 set_opt(sbi->s_mount_opt, GRPID);
2073 if (def_mount_opts & EXT4_DEFM_UID16)
2074 set_opt(sbi->s_mount_opt, NO_UID32);
2075 #ifdef CONFIG_EXT4_FS_XATTR
2076 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2077 set_opt(sbi->s_mount_opt, XATTR_USER);
2079 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2080 if (def_mount_opts & EXT4_DEFM_ACL)
2081 set_opt(sbi->s_mount_opt, POSIX_ACL);
2083 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2084 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2085 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2086 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2087 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2088 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2090 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2091 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2092 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2093 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2095 set_opt(sbi->s_mount_opt, ERRORS_RO);
2097 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2098 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2099 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2100 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2101 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2103 set_opt(sbi->s_mount_opt, RESERVATION);
2104 set_opt(sbi->s_mount_opt, BARRIER);
2107 * enable delayed allocation by default
2108 * Use -o nodelalloc to turn it off
2110 set_opt(sbi->s_mount_opt, DELALLOC);
2113 if (!parse_options((char *) data, sb, &journal_devnum,
2114 &journal_ioprio, NULL, 0))
2117 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2118 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2120 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2121 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2122 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2123 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2125 "EXT4-fs warning: feature flags set on rev 0 fs, "
2126 "running e2fsck is recommended\n");
2129 * Check feature flags regardless of the revision level, since we
2130 * previously didn't change the revision level when setting the flags,
2131 * so there is a chance incompat flags are set on a rev 0 filesystem.
2133 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2135 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2136 "unsupported optional features (%x).\n", sb->s_id,
2137 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2138 ~EXT4_FEATURE_INCOMPAT_SUPP));
2141 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2142 if (!(sb->s_flags & MS_RDONLY) && features) {
2143 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2144 "unsupported optional features (%x).\n", sb->s_id,
2145 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2146 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2149 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2150 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2151 if (has_huge_files) {
2153 * Large file size enabled file system can only be
2154 * mount if kernel is build with CONFIG_LBD
2156 if (sizeof(root->i_blocks) < sizeof(u64) &&
2157 !(sb->s_flags & MS_RDONLY)) {
2158 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2159 "files cannot be mounted read-write "
2160 "without CONFIG_LBD.\n", sb->s_id);
2164 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2166 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2167 blocksize > EXT4_MAX_BLOCK_SIZE) {
2169 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2170 blocksize, sb->s_id);
2174 if (sb->s_blocksize != blocksize) {
2176 /* Validate the filesystem blocksize */
2177 if (!sb_set_blocksize(sb, blocksize)) {
2178 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2184 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2185 offset = do_div(logical_sb_block, blocksize);
2186 bh = sb_bread(sb, logical_sb_block);
2189 "EXT4-fs: Can't read superblock on 2nd try.\n");
2192 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2194 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2196 "EXT4-fs: Magic mismatch, very weird !\n");
2201 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2203 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2205 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2206 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2207 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2209 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2210 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2211 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2212 (!is_power_of_2(sbi->s_inode_size)) ||
2213 (sbi->s_inode_size > blocksize)) {
2215 "EXT4-fs: unsupported inode size: %d\n",
2219 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2220 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2222 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2223 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2224 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2225 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2226 !is_power_of_2(sbi->s_desc_size)) {
2228 "EXT4-fs: unsupported descriptor size %lu\n",
2233 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2234 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2235 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2236 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2238 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2239 if (sbi->s_inodes_per_block == 0)
2241 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2242 sbi->s_inodes_per_block;
2243 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2245 sbi->s_mount_state = le16_to_cpu(es->s_state);
2246 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2247 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2248 for (i = 0; i < 4; i++)
2249 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2250 sbi->s_def_hash_version = es->s_def_hash_version;
2251 i = le32_to_cpu(es->s_flags);
2252 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2253 sbi->s_hash_unsigned = 3;
2254 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2255 #ifdef __CHAR_UNSIGNED__
2256 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2257 sbi->s_hash_unsigned = 3;
2259 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2264 if (sbi->s_blocks_per_group > blocksize * 8) {
2266 "EXT4-fs: #blocks per group too big: %lu\n",
2267 sbi->s_blocks_per_group);
2270 if (sbi->s_inodes_per_group > blocksize * 8) {
2272 "EXT4-fs: #inodes per group too big: %lu\n",
2273 sbi->s_inodes_per_group);
2277 if (ext4_blocks_count(es) >
2278 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2279 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2280 " too large to mount safely\n", sb->s_id);
2281 if (sizeof(sector_t) < 8)
2282 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2287 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2291 * It makes no sense for the first data block to be beyond the end
2292 * of the filesystem.
2294 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2295 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2296 "block %u is beyond end of filesystem (%llu)\n",
2297 le32_to_cpu(es->s_first_data_block),
2298 ext4_blocks_count(es));
2301 blocks_count = (ext4_blocks_count(es) -
2302 le32_to_cpu(es->s_first_data_block) +
2303 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2304 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2305 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2306 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2307 "(block count %llu, first data block %u, "
2308 "blocks per group %lu)\n", sbi->s_groups_count,
2309 ext4_blocks_count(es),
2310 le32_to_cpu(es->s_first_data_block),
2311 EXT4_BLOCKS_PER_GROUP(sb));
2314 sbi->s_groups_count = blocks_count;
2315 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2316 EXT4_DESC_PER_BLOCK(sb);
2317 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2319 if (sbi->s_group_desc == NULL) {
2320 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2324 #ifdef CONFIG_PROC_FS
2326 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2329 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2331 &sbi->s_inode_readahead_blks);
2334 bgl_lock_init(&sbi->s_blockgroup_lock);
2336 for (i = 0; i < db_count; i++) {
2337 block = descriptor_loc(sb, logical_sb_block, i);
2338 sbi->s_group_desc[i] = sb_bread(sb, block);
2339 if (!sbi->s_group_desc[i]) {
2340 printk(KERN_ERR "EXT4-fs: "
2341 "can't read group descriptor %d\n", i);
2346 if (!ext4_check_descriptors(sb)) {
2347 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2350 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2351 if (!ext4_fill_flex_info(sb)) {
2353 "EXT4-fs: unable to initialize "
2354 "flex_bg meta info!\n");
2358 sbi->s_gdb_count = db_count;
2359 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2360 spin_lock_init(&sbi->s_next_gen_lock);
2362 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2363 ext4_count_free_blocks(sb));
2365 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2366 ext4_count_free_inodes(sb));
2369 err = percpu_counter_init(&sbi->s_dirs_counter,
2370 ext4_count_dirs(sb));
2373 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2376 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2380 sbi->s_stripe = ext4_get_stripe_size(sbi);
2383 * set up enough so that it can read an inode
2385 sb->s_op = &ext4_sops;
2386 sb->s_export_op = &ext4_export_ops;
2387 sb->s_xattr = ext4_xattr_handlers;
2389 sb->s_qcop = &ext4_qctl_operations;
2390 sb->dq_op = &ext4_quota_operations;
2392 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2396 needs_recovery = (es->s_last_orphan != 0 ||
2397 EXT4_HAS_INCOMPAT_FEATURE(sb,
2398 EXT4_FEATURE_INCOMPAT_RECOVER));
2401 * The first inode we look at is the journal inode. Don't try
2402 * root first: it may be modified in the journal!
2404 if (!test_opt(sb, NOLOAD) &&
2405 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2406 if (ext4_load_journal(sb, es, journal_devnum))
2408 if (!(sb->s_flags & MS_RDONLY) &&
2409 EXT4_SB(sb)->s_journal->j_failed_commit) {
2410 printk(KERN_CRIT "EXT4-fs error (device %s): "
2411 "ext4_fill_super: Journal transaction "
2412 "%u is corrupt\n", sb->s_id,
2413 EXT4_SB(sb)->s_journal->j_failed_commit);
2414 if (test_opt(sb, ERRORS_RO)) {
2416 "Mounting filesystem read-only\n");
2417 sb->s_flags |= MS_RDONLY;
2418 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2419 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2421 if (test_opt(sb, ERRORS_PANIC)) {
2422 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2423 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2424 ext4_commit_super(sb, es, 1);
2428 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2429 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2430 printk(KERN_ERR "EXT4-fs: required journal recovery "
2431 "suppressed and not mounted read-only\n");
2434 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2435 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2436 sbi->s_journal = NULL;
2441 if (ext4_blocks_count(es) > 0xffffffffULL &&
2442 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2443 JBD2_FEATURE_INCOMPAT_64BIT)) {
2444 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2448 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2449 jbd2_journal_set_features(sbi->s_journal,
2450 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2451 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2452 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2453 jbd2_journal_set_features(sbi->s_journal,
2454 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2455 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2456 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2458 jbd2_journal_clear_features(sbi->s_journal,
2459 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2460 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2463 /* We have now updated the journal if required, so we can
2464 * validate the data journaling mode. */
2465 switch (test_opt(sb, DATA_FLAGS)) {
2467 /* No mode set, assume a default based on the journal
2468 * capabilities: ORDERED_DATA if the journal can
2469 * cope, else JOURNAL_DATA
2471 if (jbd2_journal_check_available_features
2472 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2473 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2475 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2478 case EXT4_MOUNT_ORDERED_DATA:
2479 case EXT4_MOUNT_WRITEBACK_DATA:
2480 if (!jbd2_journal_check_available_features
2481 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2482 printk(KERN_ERR "EXT4-fs: Journal does not support "
2483 "requested data journaling mode\n");
2489 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2493 if (test_opt(sb, NOBH)) {
2494 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2495 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2496 "its supported only with writeback mode\n");
2497 clear_opt(sbi->s_mount_opt, NOBH);
2501 * The jbd2_journal_load will have done any necessary log recovery,
2502 * so we can safely mount the rest of the filesystem now.
2505 root = ext4_iget(sb, EXT4_ROOT_INO);
2507 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2508 ret = PTR_ERR(root);
2511 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2513 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2516 sb->s_root = d_alloc_root(root);
2518 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2524 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2526 /* determine the minimum size of new large inodes, if present */
2527 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2528 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2529 EXT4_GOOD_OLD_INODE_SIZE;
2530 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2531 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2532 if (sbi->s_want_extra_isize <
2533 le16_to_cpu(es->s_want_extra_isize))
2534 sbi->s_want_extra_isize =
2535 le16_to_cpu(es->s_want_extra_isize);
2536 if (sbi->s_want_extra_isize <
2537 le16_to_cpu(es->s_min_extra_isize))
2538 sbi->s_want_extra_isize =
2539 le16_to_cpu(es->s_min_extra_isize);
2542 /* Check if enough inode space is available */
2543 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2544 sbi->s_inode_size) {
2545 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2546 EXT4_GOOD_OLD_INODE_SIZE;
2547 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2551 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2552 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2553 "requested data journaling mode\n");
2554 clear_opt(sbi->s_mount_opt, DELALLOC);
2555 } else if (test_opt(sb, DELALLOC))
2556 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2559 err = ext4_mb_init(sb, needs_recovery);
2561 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2567 * akpm: core read_super() calls in here with the superblock locked.
2568 * That deadlocks, because orphan cleanup needs to lock the superblock
2569 * in numerous places. Here we just pop the lock - it's relatively
2570 * harmless, because we are now ready to accept write_super() requests,
2571 * and aviro says that's the only reason for hanging onto the
2574 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2575 ext4_orphan_cleanup(sb, es);
2576 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2577 if (needs_recovery) {
2578 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2579 ext4_mark_recovery_complete(sb, es);
2581 if (EXT4_SB(sb)->s_journal) {
2582 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2583 descr = " journalled data mode";
2584 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2585 descr = " ordered data mode";
2587 descr = " writeback data mode";
2589 descr = "out journal";
2591 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2599 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2604 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2605 if (sbi->s_journal) {
2606 jbd2_journal_destroy(sbi->s_journal);
2607 sbi->s_journal = NULL;
2610 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2611 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2612 percpu_counter_destroy(&sbi->s_dirs_counter);
2613 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2615 for (i = 0; i < db_count; i++)
2616 brelse(sbi->s_group_desc[i]);
2617 kfree(sbi->s_group_desc);
2620 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2621 remove_proc_entry(sb->s_id, ext4_proc_root);
2624 for (i = 0; i < MAXQUOTAS; i++)
2625 kfree(sbi->s_qf_names[i]);
2627 ext4_blkdev_remove(sbi);
2630 sb->s_fs_info = NULL;
2637 * Setup any per-fs journal parameters now. We'll do this both on
2638 * initial mount, once the journal has been initialised but before we've
2639 * done any recovery; and again on any subsequent remount.
2641 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2643 struct ext4_sb_info *sbi = EXT4_SB(sb);
2645 journal->j_commit_interval = sbi->s_commit_interval;
2646 journal->j_min_batch_time = sbi->s_min_batch_time;
2647 journal->j_max_batch_time = sbi->s_max_batch_time;
2649 spin_lock(&journal->j_state_lock);
2650 if (test_opt(sb, BARRIER))
2651 journal->j_flags |= JBD2_BARRIER;
2653 journal->j_flags &= ~JBD2_BARRIER;
2654 if (test_opt(sb, DATA_ERR_ABORT))
2655 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2657 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2658 spin_unlock(&journal->j_state_lock);
2661 static journal_t *ext4_get_journal(struct super_block *sb,
2662 unsigned int journal_inum)
2664 struct inode *journal_inode;
2667 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2669 /* First, test for the existence of a valid inode on disk. Bad
2670 * things happen if we iget() an unused inode, as the subsequent
2671 * iput() will try to delete it. */
2673 journal_inode = ext4_iget(sb, journal_inum);
2674 if (IS_ERR(journal_inode)) {
2675 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2678 if (!journal_inode->i_nlink) {
2679 make_bad_inode(journal_inode);
2680 iput(journal_inode);
2681 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2685 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2686 journal_inode, journal_inode->i_size);
2687 if (!S_ISREG(journal_inode->i_mode)) {
2688 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2689 iput(journal_inode);
2693 journal = jbd2_journal_init_inode(journal_inode);
2695 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2696 iput(journal_inode);
2699 journal->j_private = sb;
2700 ext4_init_journal_params(sb, journal);
2704 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2707 struct buffer_head *bh;
2711 int hblock, blocksize;
2712 ext4_fsblk_t sb_block;
2713 unsigned long offset;
2714 struct ext4_super_block *es;
2715 struct block_device *bdev;
2717 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2719 bdev = ext4_blkdev_get(j_dev);
2723 if (bd_claim(bdev, sb)) {
2725 "EXT4-fs: failed to claim external journal device.\n");
2726 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2730 blocksize = sb->s_blocksize;
2731 hblock = bdev_hardsect_size(bdev);
2732 if (blocksize < hblock) {
2734 "EXT4-fs: blocksize too small for journal device.\n");
2738 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2739 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2740 set_blocksize(bdev, blocksize);
2741 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2742 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2743 "external journal\n");
2747 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2748 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2749 !(le32_to_cpu(es->s_feature_incompat) &
2750 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2751 printk(KERN_ERR "EXT4-fs: external journal has "
2752 "bad superblock\n");
2757 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2758 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2763 len = ext4_blocks_count(es);
2764 start = sb_block + 1;
2765 brelse(bh); /* we're done with the superblock */
2767 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2768 start, len, blocksize);
2770 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2773 journal->j_private = sb;
2774 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2775 wait_on_buffer(journal->j_sb_buffer);
2776 if (!buffer_uptodate(journal->j_sb_buffer)) {
2777 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2780 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2781 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2782 "user (unsupported) - %d\n",
2783 be32_to_cpu(journal->j_superblock->s_nr_users));
2786 EXT4_SB(sb)->journal_bdev = bdev;
2787 ext4_init_journal_params(sb, journal);
2790 jbd2_journal_destroy(journal);
2792 ext4_blkdev_put(bdev);
2796 static int ext4_load_journal(struct super_block *sb,
2797 struct ext4_super_block *es,
2798 unsigned long journal_devnum)
2801 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2804 int really_read_only;
2806 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2808 if (journal_devnum &&
2809 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2810 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2811 "numbers have changed\n");
2812 journal_dev = new_decode_dev(journal_devnum);
2814 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2816 really_read_only = bdev_read_only(sb->s_bdev);
2819 * Are we loading a blank journal or performing recovery after a
2820 * crash? For recovery, we need to check in advance whether we
2821 * can get read-write access to the device.
2824 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2825 if (sb->s_flags & MS_RDONLY) {
2826 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2827 "required on readonly filesystem.\n");
2828 if (really_read_only) {
2829 printk(KERN_ERR "EXT4-fs: write access "
2830 "unavailable, cannot proceed.\n");
2833 printk(KERN_INFO "EXT4-fs: write access will "
2834 "be enabled during recovery.\n");
2838 if (journal_inum && journal_dev) {
2839 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2840 "and inode journals!\n");
2845 if (!(journal = ext4_get_journal(sb, journal_inum)))
2848 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2852 if (journal->j_flags & JBD2_BARRIER)
2853 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2855 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2857 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2858 err = jbd2_journal_update_format(journal);
2860 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2861 jbd2_journal_destroy(journal);
2866 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2867 err = jbd2_journal_wipe(journal, !really_read_only);
2869 err = jbd2_journal_load(journal);
2872 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2873 jbd2_journal_destroy(journal);
2877 EXT4_SB(sb)->s_journal = journal;
2878 ext4_clear_journal_err(sb, es);
2880 if (journal_devnum &&
2881 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2882 es->s_journal_dev = cpu_to_le32(journal_devnum);
2885 /* Make sure we flush the recovery flag to disk. */
2886 ext4_commit_super(sb, es, 1);
2892 static void ext4_commit_super(struct super_block *sb,
2893 struct ext4_super_block *es, int sync)
2895 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2899 if (buffer_write_io_error(sbh)) {
2901 * Oh, dear. A previous attempt to write the
2902 * superblock failed. This could happen because the
2903 * USB device was yanked out. Or it could happen to
2904 * be a transient write error and maybe the block will
2905 * be remapped. Nothing we can do but to retry the
2906 * write and hope for the best.
2908 printk(KERN_ERR "EXT4-fs: previous I/O error to "
2909 "superblock detected for %s.\n", sb->s_id);
2910 clear_buffer_write_io_error(sbh);
2911 set_buffer_uptodate(sbh);
2913 es->s_wtime = cpu_to_le32(get_seconds());
2914 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2915 &EXT4_SB(sb)->s_freeblocks_counter));
2916 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2917 &EXT4_SB(sb)->s_freeinodes_counter));
2919 BUFFER_TRACE(sbh, "marking dirty");
2920 mark_buffer_dirty(sbh);
2922 sync_dirty_buffer(sbh);
2923 if (buffer_write_io_error(sbh)) {
2924 printk(KERN_ERR "EXT4-fs: I/O error while writing "
2925 "superblock for %s.\n", sb->s_id);
2926 clear_buffer_write_io_error(sbh);
2927 set_buffer_uptodate(sbh);
2934 * Have we just finished recovery? If so, and if we are mounting (or
2935 * remounting) the filesystem readonly, then we will end up with a
2936 * consistent fs on disk. Record that fact.
2938 static void ext4_mark_recovery_complete(struct super_block *sb,
2939 struct ext4_super_block *es)
2941 journal_t *journal = EXT4_SB(sb)->s_journal;
2943 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2944 BUG_ON(journal != NULL);
2947 jbd2_journal_lock_updates(journal);
2948 if (jbd2_journal_flush(journal) < 0)
2952 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2953 sb->s_flags & MS_RDONLY) {
2954 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2956 ext4_commit_super(sb, es, 1);
2961 jbd2_journal_unlock_updates(journal);
2965 * If we are mounting (or read-write remounting) a filesystem whose journal
2966 * has recorded an error from a previous lifetime, move that error to the
2967 * main filesystem now.
2969 static void ext4_clear_journal_err(struct super_block *sb,
2970 struct ext4_super_block *es)
2976 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2978 journal = EXT4_SB(sb)->s_journal;
2981 * Now check for any error status which may have been recorded in the
2982 * journal by a prior ext4_error() or ext4_abort()
2985 j_errno = jbd2_journal_errno(journal);
2989 errstr = ext4_decode_error(sb, j_errno, nbuf);
2990 ext4_warning(sb, __func__, "Filesystem error recorded "
2991 "from previous mount: %s", errstr);
2992 ext4_warning(sb, __func__, "Marking fs in need of "
2993 "filesystem check.");
2995 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2996 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2997 ext4_commit_super(sb, es, 1);
2999 jbd2_journal_clear_err(journal);
3004 * Force the running and committing transactions to commit,
3005 * and wait on the commit.
3007 int ext4_force_commit(struct super_block *sb)
3012 if (sb->s_flags & MS_RDONLY)
3015 journal = EXT4_SB(sb)->s_journal;
3018 ret = ext4_journal_force_commit(journal);
3025 * Ext4 always journals updates to the superblock itself, so we don't
3026 * have to propagate any other updates to the superblock on disk at this
3027 * point. (We can probably nuke this function altogether, and remove
3028 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3030 static void ext4_write_super(struct super_block *sb)
3032 if (EXT4_SB(sb)->s_journal) {
3033 if (mutex_trylock(&sb->s_lock) != 0)
3037 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3041 static int ext4_sync_fs(struct super_block *sb, int wait)
3045 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3047 if (EXT4_SB(sb)->s_journal) {
3049 ret = ext4_force_commit(sb);
3051 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3053 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3059 * LVM calls this function before a (read-only) snapshot is created. This
3060 * gives us a chance to flush the journal completely and mark the fs clean.
3062 static void ext4_write_super_lockfs(struct super_block *sb)
3066 if (!(sb->s_flags & MS_RDONLY)) {
3067 journal_t *journal = EXT4_SB(sb)->s_journal;
3070 /* Now we set up the journal barrier. */
3071 jbd2_journal_lock_updates(journal);
3074 * We don't want to clear needs_recovery flag when we
3075 * failed to flush the journal.
3077 if (jbd2_journal_flush(journal) < 0)
3081 /* Journal blocked and flushed, clear needs_recovery flag. */
3082 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3083 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3088 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3089 * flag here, even though the filesystem is not technically dirty yet.
3091 static void ext4_unlockfs(struct super_block *sb)
3093 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3095 /* Reser the needs_recovery flag before the fs is unlocked. */
3096 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3097 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3099 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3103 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3105 struct ext4_super_block *es;
3106 struct ext4_sb_info *sbi = EXT4_SB(sb);
3107 ext4_fsblk_t n_blocks_count = 0;
3108 unsigned long old_sb_flags;
3109 struct ext4_mount_options old_opts;
3111 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3117 /* Store the original options */
3118 old_sb_flags = sb->s_flags;
3119 old_opts.s_mount_opt = sbi->s_mount_opt;
3120 old_opts.s_resuid = sbi->s_resuid;
3121 old_opts.s_resgid = sbi->s_resgid;
3122 old_opts.s_commit_interval = sbi->s_commit_interval;
3123 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3124 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3126 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3127 for (i = 0; i < MAXQUOTAS; i++)
3128 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3130 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3131 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3134 * Allow the "check" option to be passed as a remount option.
3136 if (!parse_options(data, sb, NULL, &journal_ioprio,
3137 &n_blocks_count, 1)) {
3142 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3143 ext4_abort(sb, __func__, "Abort forced by user");
3145 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3146 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3150 if (sbi->s_journal) {
3151 ext4_init_journal_params(sb, sbi->s_journal);
3152 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3155 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3156 n_blocks_count > ext4_blocks_count(es)) {
3157 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3162 if (*flags & MS_RDONLY) {
3164 * First of all, the unconditional stuff we have to do
3165 * to disable replay of the journal when we next remount
3167 sb->s_flags |= MS_RDONLY;
3170 * OK, test if we are remounting a valid rw partition
3171 * readonly, and if so set the rdonly flag and then
3172 * mark the partition as valid again.
3174 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3175 (sbi->s_mount_state & EXT4_VALID_FS))
3176 es->s_state = cpu_to_le16(sbi->s_mount_state);
3179 * We have to unlock super so that we can wait for
3182 if (sbi->s_journal) {
3184 ext4_mark_recovery_complete(sb, es);
3189 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3190 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3191 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3192 "remount RDWR because of unsupported "
3193 "optional features (%x).\n", sb->s_id,
3194 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3195 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3201 * Make sure the group descriptor checksums
3202 * are sane. If they aren't, refuse to
3205 for (g = 0; g < sbi->s_groups_count; g++) {
3206 struct ext4_group_desc *gdp =
3207 ext4_get_group_desc(sb, g, NULL);
3209 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3211 "EXT4-fs: ext4_remount: "
3212 "Checksum for group %u failed (%u!=%u)\n",
3213 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3214 le16_to_cpu(gdp->bg_checksum));
3221 * If we have an unprocessed orphan list hanging
3222 * around from a previously readonly bdev mount,
3223 * require a full umount/remount for now.
3225 if (es->s_last_orphan) {
3226 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3227 "remount RDWR because of unprocessed "
3228 "orphan inode list. Please "
3229 "umount/remount instead.\n",
3236 * Mounting a RDONLY partition read-write, so reread
3237 * and store the current valid flag. (It may have
3238 * been changed by e2fsck since we originally mounted
3242 ext4_clear_journal_err(sb, es);
3243 sbi->s_mount_state = le16_to_cpu(es->s_state);
3244 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3246 if (!ext4_setup_super(sb, es, 0))
3247 sb->s_flags &= ~MS_RDONLY;
3250 if (sbi->s_journal == NULL)
3251 ext4_commit_super(sb, es, 1);
3254 /* Release old quota file names */
3255 for (i = 0; i < MAXQUOTAS; i++)
3256 if (old_opts.s_qf_names[i] &&
3257 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3258 kfree(old_opts.s_qf_names[i]);
3262 sb->s_flags = old_sb_flags;
3263 sbi->s_mount_opt = old_opts.s_mount_opt;
3264 sbi->s_resuid = old_opts.s_resuid;
3265 sbi->s_resgid = old_opts.s_resgid;
3266 sbi->s_commit_interval = old_opts.s_commit_interval;
3267 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3268 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3270 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3271 for (i = 0; i < MAXQUOTAS; i++) {
3272 if (sbi->s_qf_names[i] &&
3273 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3274 kfree(sbi->s_qf_names[i]);
3275 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3281 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3283 struct super_block *sb = dentry->d_sb;
3284 struct ext4_sb_info *sbi = EXT4_SB(sb);
3285 struct ext4_super_block *es = sbi->s_es;
3288 if (test_opt(sb, MINIX_DF)) {
3289 sbi->s_overhead_last = 0;
3290 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3291 ext4_group_t ngroups = sbi->s_groups_count, i;
3292 ext4_fsblk_t overhead = 0;
3296 * Compute the overhead (FS structures). This is constant
3297 * for a given filesystem unless the number of block groups
3298 * changes so we cache the previous value until it does.
3302 * All of the blocks before first_data_block are
3305 overhead = le32_to_cpu(es->s_first_data_block);
3308 * Add the overhead attributed to the superblock and
3309 * block group descriptors. If the sparse superblocks
3310 * feature is turned on, then not all groups have this.
3312 for (i = 0; i < ngroups; i++) {
3313 overhead += ext4_bg_has_super(sb, i) +
3314 ext4_bg_num_gdb(sb, i);
3319 * Every block group has an inode bitmap, a block
3320 * bitmap, and an inode table.
3322 overhead += ngroups * (2 + sbi->s_itb_per_group);
3323 sbi->s_overhead_last = overhead;
3325 sbi->s_blocks_last = ext4_blocks_count(es);
3328 buf->f_type = EXT4_SUPER_MAGIC;
3329 buf->f_bsize = sb->s_blocksize;
3330 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3331 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3332 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3333 ext4_free_blocks_count_set(es, buf->f_bfree);
3334 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3335 if (buf->f_bfree < ext4_r_blocks_count(es))
3337 buf->f_files = le32_to_cpu(es->s_inodes_count);
3338 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3339 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3340 buf->f_namelen = EXT4_NAME_LEN;
3341 fsid = le64_to_cpup((void *)es->s_uuid) ^
3342 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3343 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3344 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3348 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3349 * is locked for write. Otherwise the are possible deadlocks:
3350 * Process 1 Process 2
3351 * ext4_create() quota_sync()
3352 * jbd2_journal_start() write_dquot()
3353 * DQUOT_INIT() down(dqio_mutex)
3354 * down(dqio_mutex) jbd2_journal_start()
3360 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3362 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3365 static int ext4_dquot_initialize(struct inode *inode, int type)
3370 /* We may create quota structure so we need to reserve enough blocks */
3371 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3373 return PTR_ERR(handle);
3374 ret = dquot_initialize(inode, type);
3375 err = ext4_journal_stop(handle);
3381 static int ext4_dquot_drop(struct inode *inode)
3386 /* We may delete quota structure so we need to reserve enough blocks */
3387 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3388 if (IS_ERR(handle)) {
3390 * We call dquot_drop() anyway to at least release references
3391 * to quota structures so that umount does not hang.
3394 return PTR_ERR(handle);
3396 ret = dquot_drop(inode);
3397 err = ext4_journal_stop(handle);
3403 static int ext4_write_dquot(struct dquot *dquot)
3407 struct inode *inode;
3409 inode = dquot_to_inode(dquot);
3410 handle = ext4_journal_start(inode,
3411 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3413 return PTR_ERR(handle);
3414 ret = dquot_commit(dquot);
3415 err = ext4_journal_stop(handle);
3421 static int ext4_acquire_dquot(struct dquot *dquot)
3426 handle = ext4_journal_start(dquot_to_inode(dquot),
3427 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3429 return PTR_ERR(handle);
3430 ret = dquot_acquire(dquot);
3431 err = ext4_journal_stop(handle);
3437 static int ext4_release_dquot(struct dquot *dquot)
3442 handle = ext4_journal_start(dquot_to_inode(dquot),
3443 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3444 if (IS_ERR(handle)) {
3445 /* Release dquot anyway to avoid endless cycle in dqput() */
3446 dquot_release(dquot);
3447 return PTR_ERR(handle);
3449 ret = dquot_release(dquot);
3450 err = ext4_journal_stop(handle);
3456 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3458 /* Are we journaling quotas? */
3459 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3460 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3461 dquot_mark_dquot_dirty(dquot);
3462 return ext4_write_dquot(dquot);
3464 return dquot_mark_dquot_dirty(dquot);
3468 static int ext4_write_info(struct super_block *sb, int type)
3473 /* Data block + inode block */
3474 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3476 return PTR_ERR(handle);
3477 ret = dquot_commit_info(sb, type);
3478 err = ext4_journal_stop(handle);
3485 * Turn on quotas during mount time - we need to find
3486 * the quota file and such...
3488 static int ext4_quota_on_mount(struct super_block *sb, int type)
3490 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3491 EXT4_SB(sb)->s_jquota_fmt, type);
3495 * Standard function to be called on quota_on
3497 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3498 char *name, int remount)
3503 if (!test_opt(sb, QUOTA))
3505 /* When remounting, no checks are needed and in fact, name is NULL */
3507 return vfs_quota_on(sb, type, format_id, name, remount);
3509 err = kern_path(name, LOOKUP_FOLLOW, &path);
3513 /* Quotafile not on the same filesystem? */
3514 if (path.mnt->mnt_sb != sb) {
3518 /* Journaling quota? */
3519 if (EXT4_SB(sb)->s_qf_names[type]) {
3520 /* Quotafile not in fs root? */
3521 if (path.dentry->d_parent != sb->s_root)
3523 "EXT4-fs: Quota file not on filesystem root. "
3524 "Journaled quota will not work.\n");
3528 * When we journal data on quota file, we have to flush journal to see
3529 * all updates to the file when we bypass pagecache...
3531 if (EXT4_SB(sb)->s_journal &&
3532 ext4_should_journal_data(path.dentry->d_inode)) {
3534 * We don't need to lock updates but journal_flush() could
3535 * otherwise be livelocked...
3537 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3538 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3539 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3546 err = vfs_quota_on_path(sb, type, format_id, &path);
3551 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3552 * acquiring the locks... As quota files are never truncated and quota code
3553 * itself serializes the operations (and noone else should touch the files)
3554 * we don't have to be afraid of races */
3555 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3556 size_t len, loff_t off)
3558 struct inode *inode = sb_dqopt(sb)->files[type];
3559 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3561 int offset = off & (sb->s_blocksize - 1);
3564 struct buffer_head *bh;
3565 loff_t i_size = i_size_read(inode);
3569 if (off+len > i_size)
3572 while (toread > 0) {
3573 tocopy = sb->s_blocksize - offset < toread ?
3574 sb->s_blocksize - offset : toread;
3575 bh = ext4_bread(NULL, inode, blk, 0, &err);
3578 if (!bh) /* A hole? */
3579 memset(data, 0, tocopy);
3581 memcpy(data, bh->b_data+offset, tocopy);
3591 /* Write to quotafile (we know the transaction is already started and has
3592 * enough credits) */
3593 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3594 const char *data, size_t len, loff_t off)
3596 struct inode *inode = sb_dqopt(sb)->files[type];
3597 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3599 int offset = off & (sb->s_blocksize - 1);
3601 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3602 size_t towrite = len;
3603 struct buffer_head *bh;
3604 handle_t *handle = journal_current_handle();
3606 if (EXT4_SB(sb)->s_journal && !handle) {
3607 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3608 " cancelled because transaction is not started.\n",
3609 (unsigned long long)off, (unsigned long long)len);
3612 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3613 while (towrite > 0) {
3614 tocopy = sb->s_blocksize - offset < towrite ?
3615 sb->s_blocksize - offset : towrite;
3616 bh = ext4_bread(handle, inode, blk, 1, &err);
3619 if (journal_quota) {
3620 err = ext4_journal_get_write_access(handle, bh);
3627 memcpy(bh->b_data+offset, data, tocopy);
3628 flush_dcache_page(bh->b_page);
3631 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3633 /* Always do at least ordered writes for quotas */
3634 err = ext4_jbd2_file_inode(handle, inode);
3635 mark_buffer_dirty(bh);
3646 if (len == towrite) {
3647 mutex_unlock(&inode->i_mutex);
3650 if (inode->i_size < off+len-towrite) {
3651 i_size_write(inode, off+len-towrite);
3652 EXT4_I(inode)->i_disksize = inode->i_size;
3654 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3655 ext4_mark_inode_dirty(handle, inode);
3656 mutex_unlock(&inode->i_mutex);
3657 return len - towrite;
3662 static int ext4_get_sb(struct file_system_type *fs_type,
3663 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3665 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3668 #ifdef CONFIG_PROC_FS
3669 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3671 unsigned int *p = m->private;
3673 seq_printf(m, "%u\n", *p);
3677 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3679 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3682 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3683 size_t cnt, loff_t *ppos)
3685 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3688 if (cnt >= sizeof(str))
3690 if (copy_from_user(str, buf, cnt))
3693 *p = simple_strtoul(str, NULL, 0);
3697 const struct file_operations ext4_ui_proc_fops = {
3698 .owner = THIS_MODULE,
3699 .open = ext4_ui_proc_open,
3701 .llseek = seq_lseek,
3702 .release = single_release,
3703 .write = ext4_ui_proc_write,
3707 static struct file_system_type ext4_fs_type = {
3708 .owner = THIS_MODULE,
3710 .get_sb = ext4_get_sb,
3711 .kill_sb = kill_block_super,
3712 .fs_flags = FS_REQUIRES_DEV,
3715 #ifdef CONFIG_EXT4DEV_COMPAT
3716 static int ext4dev_get_sb(struct file_system_type *fs_type,
3717 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3719 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3720 "to mount using ext4\n");
3721 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3722 "will go away by 2.6.31\n");
3723 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3726 static struct file_system_type ext4dev_fs_type = {
3727 .owner = THIS_MODULE,
3729 .get_sb = ext4dev_get_sb,
3730 .kill_sb = kill_block_super,
3731 .fs_flags = FS_REQUIRES_DEV,
3733 MODULE_ALIAS("ext4dev");
3736 static int __init init_ext4_fs(void)
3740 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3741 err = init_ext4_mballoc();
3745 err = init_ext4_xattr();
3748 err = init_inodecache();
3751 err = register_filesystem(&ext4_fs_type);
3754 #ifdef CONFIG_EXT4DEV_COMPAT
3755 err = register_filesystem(&ext4dev_fs_type);
3757 unregister_filesystem(&ext4_fs_type);
3763 destroy_inodecache();
3767 exit_ext4_mballoc();
3771 static void __exit exit_ext4_fs(void)
3773 unregister_filesystem(&ext4_fs_type);
3774 #ifdef CONFIG_EXT4DEV_COMPAT
3775 unregister_filesystem(&ext4dev_fs_type);
3777 destroy_inodecache();
3779 exit_ext4_mballoc();
3780 remove_proc_entry("fs/ext4", NULL);
3783 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3784 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3785 MODULE_LICENSE("GPL");
3786 module_init(init_ext4_fs)
3787 module_exit(exit_ext4_fs)