2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
50 struct proc_dir_entry *ext4_proc_root;
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static int ext4_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 int ext4_unfreeze(struct super_block *sb);
66 static void ext4_write_super(struct super_block *sb);
67 static int ext4_freeze(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_write_dquot(struct dquot *dquot);
930 static int ext4_acquire_dquot(struct dquot *dquot);
931 static int ext4_release_dquot(struct dquot *dquot);
932 static int ext4_mark_dquot_dirty(struct dquot *dquot);
933 static int ext4_write_info(struct super_block *sb, int type);
934 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
935 char *path, int remount);
936 static int ext4_quota_on_mount(struct super_block *sb, int type);
937 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
938 size_t len, loff_t off);
939 static ssize_t ext4_quota_write(struct super_block *sb, int type,
940 const char *data, size_t len, loff_t off);
942 static struct dquot_operations ext4_quota_operations = {
943 .initialize = dquot_initialize,
945 .alloc_space = dquot_alloc_space,
946 .alloc_inode = dquot_alloc_inode,
947 .free_space = dquot_free_space,
948 .free_inode = dquot_free_inode,
949 .transfer = dquot_transfer,
950 .write_dquot = ext4_write_dquot,
951 .acquire_dquot = ext4_acquire_dquot,
952 .release_dquot = ext4_release_dquot,
953 .mark_dirty = ext4_mark_dquot_dirty,
954 .write_info = ext4_write_info,
955 .alloc_dquot = dquot_alloc,
956 .destroy_dquot = dquot_destroy,
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 .freeze_fs = ext4_freeze,
980 .unfreeze_fs = ext4_unfreeze,
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_loaded(sb) &&
1306 !sbi->s_qf_names[qtype]) {
1308 "EXT4-fs: Cannot change journaled "
1309 "quota options when quota turned on.\n");
1312 qname = match_strdup(&args[0]);
1315 "EXT4-fs: not enough memory for "
1316 "storing quotafile name.\n");
1319 if (sbi->s_qf_names[qtype] &&
1320 strcmp(sbi->s_qf_names[qtype], qname)) {
1322 "EXT4-fs: %s quota file already "
1323 "specified.\n", QTYPE2NAME(qtype));
1327 sbi->s_qf_names[qtype] = qname;
1328 if (strchr(sbi->s_qf_names[qtype], '/')) {
1330 "EXT4-fs: quotafile must be on "
1331 "filesystem root.\n");
1332 kfree(sbi->s_qf_names[qtype]);
1333 sbi->s_qf_names[qtype] = NULL;
1336 set_opt(sbi->s_mount_opt, QUOTA);
1338 case Opt_offusrjquota:
1341 case Opt_offgrpjquota:
1344 if (sb_any_quota_loaded(sb) &&
1345 sbi->s_qf_names[qtype]) {
1346 printk(KERN_ERR "EXT4-fs: Cannot change "
1347 "journaled quota options when "
1348 "quota turned on.\n");
1352 * The space will be released later when all options
1353 * are confirmed to be correct
1355 sbi->s_qf_names[qtype] = NULL;
1357 case Opt_jqfmt_vfsold:
1358 qfmt = QFMT_VFS_OLD;
1360 case Opt_jqfmt_vfsv0:
1363 if (sb_any_quota_loaded(sb) &&
1364 sbi->s_jquota_fmt != qfmt) {
1365 printk(KERN_ERR "EXT4-fs: Cannot change "
1366 "journaled quota options when "
1367 "quota turned on.\n");
1370 sbi->s_jquota_fmt = qfmt;
1374 set_opt(sbi->s_mount_opt, QUOTA);
1375 set_opt(sbi->s_mount_opt, USRQUOTA);
1378 set_opt(sbi->s_mount_opt, QUOTA);
1379 set_opt(sbi->s_mount_opt, GRPQUOTA);
1382 if (sb_any_quota_loaded(sb)) {
1383 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1384 "options when quota turned on.\n");
1387 clear_opt(sbi->s_mount_opt, QUOTA);
1388 clear_opt(sbi->s_mount_opt, USRQUOTA);
1389 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1396 "EXT4-fs: quota options not supported.\n");
1400 case Opt_offusrjquota:
1401 case Opt_offgrpjquota:
1402 case Opt_jqfmt_vfsold:
1403 case Opt_jqfmt_vfsv0:
1405 "EXT4-fs: journaled quota options not "
1412 set_opt(sbi->s_mount_opt, ABORT);
1415 if (match_int(&args[0], &option))
1418 set_opt(sbi->s_mount_opt, BARRIER);
1420 clear_opt(sbi->s_mount_opt, BARRIER);
1426 printk("EXT4-fs: resize option only available "
1430 if (match_int(&args[0], &option) != 0)
1432 *n_blocks_count = option;
1435 set_opt(sbi->s_mount_opt, NOBH);
1438 clear_opt(sbi->s_mount_opt, NOBH);
1441 set_opt(sbi->s_mount_opt, I_VERSION);
1442 sb->s_flags |= MS_I_VERSION;
1444 case Opt_nodelalloc:
1445 clear_opt(sbi->s_mount_opt, DELALLOC);
1448 if (match_int(&args[0], &option))
1452 sbi->s_stripe = option;
1455 set_opt(sbi->s_mount_opt, DELALLOC);
1457 case Opt_inode_readahead_blks:
1458 if (match_int(&args[0], &option))
1460 if (option < 0 || option > (1 << 30))
1462 sbi->s_inode_readahead_blks = option;
1464 case Opt_journal_ioprio:
1465 if (match_int(&args[0], &option))
1467 if (option < 0 || option > 7)
1469 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1474 "EXT4-fs: Unrecognized mount option \"%s\" "
1475 "or missing value\n", p);
1480 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1481 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1482 sbi->s_qf_names[USRQUOTA])
1483 clear_opt(sbi->s_mount_opt, USRQUOTA);
1485 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1486 sbi->s_qf_names[GRPQUOTA])
1487 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1489 if ((sbi->s_qf_names[USRQUOTA] &&
1490 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1491 (sbi->s_qf_names[GRPQUOTA] &&
1492 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1493 printk(KERN_ERR "EXT4-fs: old and new quota "
1494 "format mixing.\n");
1498 if (!sbi->s_jquota_fmt) {
1499 printk(KERN_ERR "EXT4-fs: journaled quota format "
1500 "not specified.\n");
1504 if (sbi->s_jquota_fmt) {
1505 printk(KERN_ERR "EXT4-fs: journaled quota format "
1506 "specified with no journaling "
1515 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1518 struct ext4_sb_info *sbi = EXT4_SB(sb);
1521 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1522 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1523 "forcing read-only mode\n");
1528 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1529 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1530 "running e2fsck is recommended\n");
1531 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1533 "EXT4-fs warning: mounting fs with errors, "
1534 "running e2fsck is recommended\n");
1535 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1536 le16_to_cpu(es->s_mnt_count) >=
1537 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1539 "EXT4-fs warning: maximal mount count reached, "
1540 "running e2fsck is recommended\n");
1541 else if (le32_to_cpu(es->s_checkinterval) &&
1542 (le32_to_cpu(es->s_lastcheck) +
1543 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1545 "EXT4-fs warning: checktime reached, "
1546 "running e2fsck is recommended\n");
1547 if (!sbi->s_journal)
1548 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1549 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1550 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1551 le16_add_cpu(&es->s_mnt_count, 1);
1552 es->s_mtime = cpu_to_le32(get_seconds());
1553 ext4_update_dynamic_rev(sb);
1555 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1557 ext4_commit_super(sb, es, 1);
1558 if (test_opt(sb, DEBUG))
1559 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1560 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1562 sbi->s_groups_count,
1563 EXT4_BLOCKS_PER_GROUP(sb),
1564 EXT4_INODES_PER_GROUP(sb),
1567 if (EXT4_SB(sb)->s_journal) {
1568 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1569 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1570 "external", EXT4_SB(sb)->s_journal->j_devname);
1572 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1577 static int ext4_fill_flex_info(struct super_block *sb)
1579 struct ext4_sb_info *sbi = EXT4_SB(sb);
1580 struct ext4_group_desc *gdp = NULL;
1581 struct buffer_head *bh;
1582 ext4_group_t flex_group_count;
1583 ext4_group_t flex_group;
1584 int groups_per_flex = 0;
1587 if (!sbi->s_es->s_log_groups_per_flex) {
1588 sbi->s_log_groups_per_flex = 0;
1592 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1593 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1595 /* We allocate both existing and potentially added groups */
1596 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1597 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1598 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1599 sbi->s_flex_groups = kzalloc(flex_group_count *
1600 sizeof(struct flex_groups), GFP_KERNEL);
1601 if (sbi->s_flex_groups == NULL) {
1602 printk(KERN_ERR "EXT4-fs: not enough memory for "
1603 "%u flex groups\n", flex_group_count);
1607 for (i = 0; i < sbi->s_groups_count; i++) {
1608 gdp = ext4_get_group_desc(sb, i, &bh);
1610 flex_group = ext4_flex_group(sbi, i);
1611 sbi->s_flex_groups[flex_group].free_inodes +=
1612 ext4_free_inodes_count(sb, gdp);
1613 sbi->s_flex_groups[flex_group].free_blocks +=
1614 ext4_free_blks_count(sb, gdp);
1622 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1623 struct ext4_group_desc *gdp)
1627 if (sbi->s_es->s_feature_ro_compat &
1628 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1629 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1630 __le32 le_group = cpu_to_le32(block_group);
1632 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1633 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1634 crc = crc16(crc, (__u8 *)gdp, offset);
1635 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1636 /* for checksum of struct ext4_group_desc do the rest...*/
1637 if ((sbi->s_es->s_feature_incompat &
1638 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1639 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1640 crc = crc16(crc, (__u8 *)gdp + offset,
1641 le16_to_cpu(sbi->s_es->s_desc_size) -
1645 return cpu_to_le16(crc);
1648 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1649 struct ext4_group_desc *gdp)
1651 if ((sbi->s_es->s_feature_ro_compat &
1652 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1653 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1659 /* Called at mount-time, super-block is locked */
1660 static int ext4_check_descriptors(struct super_block *sb)
1662 struct ext4_sb_info *sbi = EXT4_SB(sb);
1663 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1664 ext4_fsblk_t last_block;
1665 ext4_fsblk_t block_bitmap;
1666 ext4_fsblk_t inode_bitmap;
1667 ext4_fsblk_t inode_table;
1668 int flexbg_flag = 0;
1671 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1674 ext4_debug("Checking group descriptors");
1676 for (i = 0; i < sbi->s_groups_count; i++) {
1677 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1679 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1680 last_block = ext4_blocks_count(sbi->s_es) - 1;
1682 last_block = first_block +
1683 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1685 block_bitmap = ext4_block_bitmap(sb, gdp);
1686 if (block_bitmap < first_block || block_bitmap > last_block) {
1687 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1688 "Block bitmap for group %u not in group "
1689 "(block %llu)!\n", i, block_bitmap);
1692 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1693 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1694 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1695 "Inode bitmap for group %u not in group "
1696 "(block %llu)!\n", i, inode_bitmap);
1699 inode_table = ext4_inode_table(sb, gdp);
1700 if (inode_table < first_block ||
1701 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1702 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1703 "Inode table for group %u not in group "
1704 "(block %llu)!\n", i, inode_table);
1707 spin_lock(sb_bgl_lock(sbi, i));
1708 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1709 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1710 "Checksum for group %u failed (%u!=%u)\n",
1711 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1712 gdp)), le16_to_cpu(gdp->bg_checksum));
1713 if (!(sb->s_flags & MS_RDONLY)) {
1714 spin_unlock(sb_bgl_lock(sbi, i));
1718 spin_unlock(sb_bgl_lock(sbi, i));
1720 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1723 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1724 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1728 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1729 * the superblock) which were deleted from all directories, but held open by
1730 * a process at the time of a crash. We walk the list and try to delete these
1731 * inodes at recovery time (only with a read-write filesystem).
1733 * In order to keep the orphan inode chain consistent during traversal (in
1734 * case of crash during recovery), we link each inode into the superblock
1735 * orphan list_head and handle it the same way as an inode deletion during
1736 * normal operation (which journals the operations for us).
1738 * We only do an iget() and an iput() on each inode, which is very safe if we
1739 * accidentally point at an in-use or already deleted inode. The worst that
1740 * can happen in this case is that we get a "bit already cleared" message from
1741 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1742 * e2fsck was run on this filesystem, and it must have already done the orphan
1743 * inode cleanup for us, so we can safely abort without any further action.
1745 static void ext4_orphan_cleanup(struct super_block *sb,
1746 struct ext4_super_block *es)
1748 unsigned int s_flags = sb->s_flags;
1749 int nr_orphans = 0, nr_truncates = 0;
1753 if (!es->s_last_orphan) {
1754 jbd_debug(4, "no orphan inodes to clean up\n");
1758 if (bdev_read_only(sb->s_bdev)) {
1759 printk(KERN_ERR "EXT4-fs: write access "
1760 "unavailable, skipping orphan cleanup.\n");
1764 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1765 if (es->s_last_orphan)
1766 jbd_debug(1, "Errors on filesystem, "
1767 "clearing orphan list.\n");
1768 es->s_last_orphan = 0;
1769 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1773 if (s_flags & MS_RDONLY) {
1774 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1776 sb->s_flags &= ~MS_RDONLY;
1779 /* Needed for iput() to work correctly and not trash data */
1780 sb->s_flags |= MS_ACTIVE;
1781 /* Turn on quotas so that they are updated correctly */
1782 for (i = 0; i < MAXQUOTAS; i++) {
1783 if (EXT4_SB(sb)->s_qf_names[i]) {
1784 int ret = ext4_quota_on_mount(sb, i);
1787 "EXT4-fs: Cannot turn on journaled "
1788 "quota: error %d\n", ret);
1793 while (es->s_last_orphan) {
1794 struct inode *inode;
1796 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1797 if (IS_ERR(inode)) {
1798 es->s_last_orphan = 0;
1802 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1804 if (inode->i_nlink) {
1806 "%s: truncating inode %lu to %lld bytes\n",
1807 __func__, inode->i_ino, inode->i_size);
1808 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1809 inode->i_ino, inode->i_size);
1810 ext4_truncate(inode);
1814 "%s: deleting unreferenced inode %lu\n",
1815 __func__, inode->i_ino);
1816 jbd_debug(2, "deleting unreferenced inode %lu\n",
1820 iput(inode); /* The delete magic happens here! */
1823 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1826 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1827 sb->s_id, PLURAL(nr_orphans));
1829 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1830 sb->s_id, PLURAL(nr_truncates));
1832 /* Turn quotas off */
1833 for (i = 0; i < MAXQUOTAS; i++) {
1834 if (sb_dqopt(sb)->files[i])
1835 vfs_quota_off(sb, i, 0);
1838 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1841 * Maximal extent format file size.
1842 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1843 * extent format containers, within a sector_t, and within i_blocks
1844 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1845 * so that won't be a limiting factor.
1847 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1849 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1852 loff_t upper_limit = MAX_LFS_FILESIZE;
1854 /* small i_blocks in vfs inode? */
1855 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1857 * CONFIG_LBD is not enabled implies the inode
1858 * i_block represent total blocks in 512 bytes
1859 * 32 == size of vfs inode i_blocks * 8
1861 upper_limit = (1LL << 32) - 1;
1863 /* total blocks in file system block size */
1864 upper_limit >>= (blkbits - 9);
1865 upper_limit <<= blkbits;
1868 /* 32-bit extent-start container, ee_block */
1873 /* Sanity check against vm- & vfs- imposed limits */
1874 if (res > upper_limit)
1881 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1882 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1883 * We need to be 1 filesystem block less than the 2^48 sector limit.
1885 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1887 loff_t res = EXT4_NDIR_BLOCKS;
1890 /* This is calculated to be the largest file size for a
1891 * dense, bitmapped file such that the total number of
1892 * sectors in the file, including data and all indirect blocks,
1893 * does not exceed 2^48 -1
1894 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1895 * total number of 512 bytes blocks of the file
1898 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1900 * !has_huge_files or CONFIG_LBD is not enabled
1901 * implies the inode i_block represent total blocks in
1902 * 512 bytes 32 == size of vfs inode i_blocks * 8
1904 upper_limit = (1LL << 32) - 1;
1906 /* total blocks in file system block size */
1907 upper_limit >>= (bits - 9);
1911 * We use 48 bit ext4_inode i_blocks
1912 * With EXT4_HUGE_FILE_FL set the i_blocks
1913 * represent total number of blocks in
1914 * file system block size
1916 upper_limit = (1LL << 48) - 1;
1920 /* indirect blocks */
1922 /* double indirect blocks */
1923 meta_blocks += 1 + (1LL << (bits-2));
1924 /* tripple indirect blocks */
1925 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1927 upper_limit -= meta_blocks;
1928 upper_limit <<= bits;
1930 res += 1LL << (bits-2);
1931 res += 1LL << (2*(bits-2));
1932 res += 1LL << (3*(bits-2));
1934 if (res > upper_limit)
1937 if (res > MAX_LFS_FILESIZE)
1938 res = MAX_LFS_FILESIZE;
1943 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1944 ext4_fsblk_t logical_sb_block, int nr)
1946 struct ext4_sb_info *sbi = EXT4_SB(sb);
1947 ext4_group_t bg, first_meta_bg;
1950 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1952 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1954 return logical_sb_block + nr + 1;
1955 bg = sbi->s_desc_per_block * nr;
1956 if (ext4_bg_has_super(sb, bg))
1958 return (has_super + ext4_group_first_block_no(sb, bg));
1962 * ext4_get_stripe_size: Get the stripe size.
1963 * @sbi: In memory super block info
1965 * If we have specified it via mount option, then
1966 * use the mount option value. If the value specified at mount time is
1967 * greater than the blocks per group use the super block value.
1968 * If the super block value is greater than blocks per group return 0.
1969 * Allocator needs it be less than blocks per group.
1972 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1974 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1975 unsigned long stripe_width =
1976 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1978 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1979 return sbi->s_stripe;
1981 if (stripe_width <= sbi->s_blocks_per_group)
1982 return stripe_width;
1984 if (stride <= sbi->s_blocks_per_group)
1990 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1991 __releases(kernel_lock)
1992 __acquires(kernel_lock)
1995 struct buffer_head *bh;
1996 struct ext4_super_block *es = NULL;
1997 struct ext4_sb_info *sbi;
1999 ext4_fsblk_t sb_block = get_sb_block(&data);
2000 ext4_fsblk_t logical_sb_block;
2001 unsigned long offset = 0;
2002 unsigned long journal_devnum = 0;
2003 unsigned long def_mount_opts;
2009 unsigned int db_count;
2011 int needs_recovery, has_huge_files;
2015 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2017 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2020 sb->s_fs_info = sbi;
2021 sbi->s_mount_opt = 0;
2022 sbi->s_resuid = EXT4_DEF_RESUID;
2023 sbi->s_resgid = EXT4_DEF_RESGID;
2024 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2025 sbi->s_sb_block = sb_block;
2029 /* Cleanup superblock name */
2030 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2033 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2035 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2040 * The ext4 superblock will not be buffer aligned for other than 1kB
2041 * block sizes. We need to calculate the offset from buffer start.
2043 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2044 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2045 offset = do_div(logical_sb_block, blocksize);
2047 logical_sb_block = sb_block;
2050 if (!(bh = sb_bread(sb, logical_sb_block))) {
2051 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2055 * Note: s_es must be initialized as soon as possible because
2056 * some ext4 macro-instructions depend on its value
2058 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2060 sb->s_magic = le16_to_cpu(es->s_magic);
2061 if (sb->s_magic != EXT4_SUPER_MAGIC)
2064 /* Set defaults before we parse the mount options */
2065 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2066 if (def_mount_opts & EXT4_DEFM_DEBUG)
2067 set_opt(sbi->s_mount_opt, DEBUG);
2068 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2069 set_opt(sbi->s_mount_opt, GRPID);
2070 if (def_mount_opts & EXT4_DEFM_UID16)
2071 set_opt(sbi->s_mount_opt, NO_UID32);
2072 #ifdef CONFIG_EXT4_FS_XATTR
2073 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2074 set_opt(sbi->s_mount_opt, XATTR_USER);
2076 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2077 if (def_mount_opts & EXT4_DEFM_ACL)
2078 set_opt(sbi->s_mount_opt, POSIX_ACL);
2080 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2081 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2082 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2083 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2084 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2085 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2087 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2088 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2089 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2090 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2092 set_opt(sbi->s_mount_opt, ERRORS_RO);
2094 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2095 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2096 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2097 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2098 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2100 set_opt(sbi->s_mount_opt, RESERVATION);
2101 set_opt(sbi->s_mount_opt, BARRIER);
2104 * enable delayed allocation by default
2105 * Use -o nodelalloc to turn it off
2107 set_opt(sbi->s_mount_opt, DELALLOC);
2110 if (!parse_options((char *) data, sb, &journal_devnum,
2111 &journal_ioprio, NULL, 0))
2114 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2115 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2117 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2118 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2119 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2120 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2122 "EXT4-fs warning: feature flags set on rev 0 fs, "
2123 "running e2fsck is recommended\n");
2126 * Check feature flags regardless of the revision level, since we
2127 * previously didn't change the revision level when setting the flags,
2128 * so there is a chance incompat flags are set on a rev 0 filesystem.
2130 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2132 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2133 "unsupported optional features (%x).\n", sb->s_id,
2134 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2135 ~EXT4_FEATURE_INCOMPAT_SUPP));
2138 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2139 if (!(sb->s_flags & MS_RDONLY) && features) {
2140 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2141 "unsupported optional features (%x).\n", sb->s_id,
2142 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2143 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2146 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2147 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2148 if (has_huge_files) {
2150 * Large file size enabled file system can only be
2151 * mount if kernel is build with CONFIG_LBD
2153 if (sizeof(root->i_blocks) < sizeof(u64) &&
2154 !(sb->s_flags & MS_RDONLY)) {
2155 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2156 "files cannot be mounted read-write "
2157 "without CONFIG_LBD.\n", sb->s_id);
2161 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2163 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2164 blocksize > EXT4_MAX_BLOCK_SIZE) {
2166 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2167 blocksize, sb->s_id);
2171 if (sb->s_blocksize != blocksize) {
2173 /* Validate the filesystem blocksize */
2174 if (!sb_set_blocksize(sb, blocksize)) {
2175 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2181 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2182 offset = do_div(logical_sb_block, blocksize);
2183 bh = sb_bread(sb, logical_sb_block);
2186 "EXT4-fs: Can't read superblock on 2nd try.\n");
2189 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2191 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2193 "EXT4-fs: Magic mismatch, very weird !\n");
2198 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2200 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2202 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2203 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2204 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2206 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2207 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2208 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2209 (!is_power_of_2(sbi->s_inode_size)) ||
2210 (sbi->s_inode_size > blocksize)) {
2212 "EXT4-fs: unsupported inode size: %d\n",
2216 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2217 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2219 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2220 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2221 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2222 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2223 !is_power_of_2(sbi->s_desc_size)) {
2225 "EXT4-fs: unsupported descriptor size %lu\n",
2230 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2231 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2232 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2233 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2235 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2236 if (sbi->s_inodes_per_block == 0)
2238 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2239 sbi->s_inodes_per_block;
2240 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2242 sbi->s_mount_state = le16_to_cpu(es->s_state);
2243 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2244 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2245 for (i = 0; i < 4; i++)
2246 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2247 sbi->s_def_hash_version = es->s_def_hash_version;
2248 i = le32_to_cpu(es->s_flags);
2249 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2250 sbi->s_hash_unsigned = 3;
2251 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2252 #ifdef __CHAR_UNSIGNED__
2253 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2254 sbi->s_hash_unsigned = 3;
2256 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2261 if (sbi->s_blocks_per_group > blocksize * 8) {
2263 "EXT4-fs: #blocks per group too big: %lu\n",
2264 sbi->s_blocks_per_group);
2267 if (sbi->s_inodes_per_group > blocksize * 8) {
2269 "EXT4-fs: #inodes per group too big: %lu\n",
2270 sbi->s_inodes_per_group);
2274 if (ext4_blocks_count(es) >
2275 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2276 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2277 " too large to mount safely\n", sb->s_id);
2278 if (sizeof(sector_t) < 8)
2279 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2284 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2288 * It makes no sense for the first data block to be beyond the end
2289 * of the filesystem.
2291 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2292 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2293 "block %u is beyond end of filesystem (%llu)\n",
2294 le32_to_cpu(es->s_first_data_block),
2295 ext4_blocks_count(es));
2298 blocks_count = (ext4_blocks_count(es) -
2299 le32_to_cpu(es->s_first_data_block) +
2300 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2301 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2302 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2303 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2304 "(block count %llu, first data block %u, "
2305 "blocks per group %lu)\n", sbi->s_groups_count,
2306 ext4_blocks_count(es),
2307 le32_to_cpu(es->s_first_data_block),
2308 EXT4_BLOCKS_PER_GROUP(sb));
2311 sbi->s_groups_count = blocks_count;
2312 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2313 EXT4_DESC_PER_BLOCK(sb);
2314 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2316 if (sbi->s_group_desc == NULL) {
2317 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2321 #ifdef CONFIG_PROC_FS
2323 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2326 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2328 &sbi->s_inode_readahead_blks);
2331 bgl_lock_init(&sbi->s_blockgroup_lock);
2333 for (i = 0; i < db_count; i++) {
2334 block = descriptor_loc(sb, logical_sb_block, i);
2335 sbi->s_group_desc[i] = sb_bread(sb, block);
2336 if (!sbi->s_group_desc[i]) {
2337 printk(KERN_ERR "EXT4-fs: "
2338 "can't read group descriptor %d\n", i);
2343 if (!ext4_check_descriptors(sb)) {
2344 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2347 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2348 if (!ext4_fill_flex_info(sb)) {
2350 "EXT4-fs: unable to initialize "
2351 "flex_bg meta info!\n");
2355 sbi->s_gdb_count = db_count;
2356 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2357 spin_lock_init(&sbi->s_next_gen_lock);
2359 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2360 ext4_count_free_blocks(sb));
2362 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2363 ext4_count_free_inodes(sb));
2366 err = percpu_counter_init(&sbi->s_dirs_counter,
2367 ext4_count_dirs(sb));
2370 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2373 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2377 sbi->s_stripe = ext4_get_stripe_size(sbi);
2380 * set up enough so that it can read an inode
2382 sb->s_op = &ext4_sops;
2383 sb->s_export_op = &ext4_export_ops;
2384 sb->s_xattr = ext4_xattr_handlers;
2386 sb->s_qcop = &ext4_qctl_operations;
2387 sb->dq_op = &ext4_quota_operations;
2389 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2393 needs_recovery = (es->s_last_orphan != 0 ||
2394 EXT4_HAS_INCOMPAT_FEATURE(sb,
2395 EXT4_FEATURE_INCOMPAT_RECOVER));
2398 * The first inode we look at is the journal inode. Don't try
2399 * root first: it may be modified in the journal!
2401 if (!test_opt(sb, NOLOAD) &&
2402 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2403 if (ext4_load_journal(sb, es, journal_devnum))
2405 if (!(sb->s_flags & MS_RDONLY) &&
2406 EXT4_SB(sb)->s_journal->j_failed_commit) {
2407 printk(KERN_CRIT "EXT4-fs error (device %s): "
2408 "ext4_fill_super: Journal transaction "
2409 "%u is corrupt\n", sb->s_id,
2410 EXT4_SB(sb)->s_journal->j_failed_commit);
2411 if (test_opt(sb, ERRORS_RO)) {
2413 "Mounting filesystem read-only\n");
2414 sb->s_flags |= MS_RDONLY;
2415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2418 if (test_opt(sb, ERRORS_PANIC)) {
2419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2420 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2421 ext4_commit_super(sb, es, 1);
2425 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2426 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2427 printk(KERN_ERR "EXT4-fs: required journal recovery "
2428 "suppressed and not mounted read-only\n");
2431 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2432 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2433 sbi->s_journal = NULL;
2438 if (ext4_blocks_count(es) > 0xffffffffULL &&
2439 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2440 JBD2_FEATURE_INCOMPAT_64BIT)) {
2441 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2445 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2446 jbd2_journal_set_features(sbi->s_journal,
2447 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2448 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2449 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2450 jbd2_journal_set_features(sbi->s_journal,
2451 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2452 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2453 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2455 jbd2_journal_clear_features(sbi->s_journal,
2456 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2457 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2460 /* We have now updated the journal if required, so we can
2461 * validate the data journaling mode. */
2462 switch (test_opt(sb, DATA_FLAGS)) {
2464 /* No mode set, assume a default based on the journal
2465 * capabilities: ORDERED_DATA if the journal can
2466 * cope, else JOURNAL_DATA
2468 if (jbd2_journal_check_available_features
2469 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2470 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2472 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2475 case EXT4_MOUNT_ORDERED_DATA:
2476 case EXT4_MOUNT_WRITEBACK_DATA:
2477 if (!jbd2_journal_check_available_features
2478 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2479 printk(KERN_ERR "EXT4-fs: Journal does not support "
2480 "requested data journaling mode\n");
2486 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2490 if (test_opt(sb, NOBH)) {
2491 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2492 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2493 "its supported only with writeback mode\n");
2494 clear_opt(sbi->s_mount_opt, NOBH);
2498 * The jbd2_journal_load will have done any necessary log recovery,
2499 * so we can safely mount the rest of the filesystem now.
2502 root = ext4_iget(sb, EXT4_ROOT_INO);
2504 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2505 ret = PTR_ERR(root);
2508 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2510 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2513 sb->s_root = d_alloc_root(root);
2515 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2521 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2523 /* determine the minimum size of new large inodes, if present */
2524 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2525 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2526 EXT4_GOOD_OLD_INODE_SIZE;
2527 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2528 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2529 if (sbi->s_want_extra_isize <
2530 le16_to_cpu(es->s_want_extra_isize))
2531 sbi->s_want_extra_isize =
2532 le16_to_cpu(es->s_want_extra_isize);
2533 if (sbi->s_want_extra_isize <
2534 le16_to_cpu(es->s_min_extra_isize))
2535 sbi->s_want_extra_isize =
2536 le16_to_cpu(es->s_min_extra_isize);
2539 /* Check if enough inode space is available */
2540 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2541 sbi->s_inode_size) {
2542 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2543 EXT4_GOOD_OLD_INODE_SIZE;
2544 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2548 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2549 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2550 "requested data journaling mode\n");
2551 clear_opt(sbi->s_mount_opt, DELALLOC);
2552 } else if (test_opt(sb, DELALLOC))
2553 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2556 err = ext4_mb_init(sb, needs_recovery);
2558 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2564 * akpm: core read_super() calls in here with the superblock locked.
2565 * That deadlocks, because orphan cleanup needs to lock the superblock
2566 * in numerous places. Here we just pop the lock - it's relatively
2567 * harmless, because we are now ready to accept write_super() requests,
2568 * and aviro says that's the only reason for hanging onto the
2571 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2572 ext4_orphan_cleanup(sb, es);
2573 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2574 if (needs_recovery) {
2575 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2576 ext4_mark_recovery_complete(sb, es);
2578 if (EXT4_SB(sb)->s_journal) {
2579 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2580 descr = " journalled data mode";
2581 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2582 descr = " ordered data mode";
2584 descr = " writeback data mode";
2586 descr = "out journal";
2588 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2596 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2601 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2602 if (sbi->s_journal) {
2603 jbd2_journal_destroy(sbi->s_journal);
2604 sbi->s_journal = NULL;
2607 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2608 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2609 percpu_counter_destroy(&sbi->s_dirs_counter);
2610 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2612 for (i = 0; i < db_count; i++)
2613 brelse(sbi->s_group_desc[i]);
2614 kfree(sbi->s_group_desc);
2617 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2618 remove_proc_entry(sb->s_id, ext4_proc_root);
2621 for (i = 0; i < MAXQUOTAS; i++)
2622 kfree(sbi->s_qf_names[i]);
2624 ext4_blkdev_remove(sbi);
2627 sb->s_fs_info = NULL;
2634 * Setup any per-fs journal parameters now. We'll do this both on
2635 * initial mount, once the journal has been initialised but before we've
2636 * done any recovery; and again on any subsequent remount.
2638 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2640 struct ext4_sb_info *sbi = EXT4_SB(sb);
2642 journal->j_commit_interval = sbi->s_commit_interval;
2643 journal->j_min_batch_time = sbi->s_min_batch_time;
2644 journal->j_max_batch_time = sbi->s_max_batch_time;
2646 spin_lock(&journal->j_state_lock);
2647 if (test_opt(sb, BARRIER))
2648 journal->j_flags |= JBD2_BARRIER;
2650 journal->j_flags &= ~JBD2_BARRIER;
2651 if (test_opt(sb, DATA_ERR_ABORT))
2652 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2654 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2655 spin_unlock(&journal->j_state_lock);
2658 static journal_t *ext4_get_journal(struct super_block *sb,
2659 unsigned int journal_inum)
2661 struct inode *journal_inode;
2664 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2666 /* First, test for the existence of a valid inode on disk. Bad
2667 * things happen if we iget() an unused inode, as the subsequent
2668 * iput() will try to delete it. */
2670 journal_inode = ext4_iget(sb, journal_inum);
2671 if (IS_ERR(journal_inode)) {
2672 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2675 if (!journal_inode->i_nlink) {
2676 make_bad_inode(journal_inode);
2677 iput(journal_inode);
2678 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2682 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2683 journal_inode, journal_inode->i_size);
2684 if (!S_ISREG(journal_inode->i_mode)) {
2685 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2686 iput(journal_inode);
2690 journal = jbd2_journal_init_inode(journal_inode);
2692 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2693 iput(journal_inode);
2696 journal->j_private = sb;
2697 ext4_init_journal_params(sb, journal);
2701 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2704 struct buffer_head *bh;
2708 int hblock, blocksize;
2709 ext4_fsblk_t sb_block;
2710 unsigned long offset;
2711 struct ext4_super_block *es;
2712 struct block_device *bdev;
2714 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2716 bdev = ext4_blkdev_get(j_dev);
2720 if (bd_claim(bdev, sb)) {
2722 "EXT4-fs: failed to claim external journal device.\n");
2723 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2727 blocksize = sb->s_blocksize;
2728 hblock = bdev_hardsect_size(bdev);
2729 if (blocksize < hblock) {
2731 "EXT4-fs: blocksize too small for journal device.\n");
2735 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2736 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2737 set_blocksize(bdev, blocksize);
2738 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2739 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2740 "external journal\n");
2744 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2745 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2746 !(le32_to_cpu(es->s_feature_incompat) &
2747 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2748 printk(KERN_ERR "EXT4-fs: external journal has "
2749 "bad superblock\n");
2754 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2755 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2760 len = ext4_blocks_count(es);
2761 start = sb_block + 1;
2762 brelse(bh); /* we're done with the superblock */
2764 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2765 start, len, blocksize);
2767 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2770 journal->j_private = sb;
2771 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2772 wait_on_buffer(journal->j_sb_buffer);
2773 if (!buffer_uptodate(journal->j_sb_buffer)) {
2774 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2777 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2778 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2779 "user (unsupported) - %d\n",
2780 be32_to_cpu(journal->j_superblock->s_nr_users));
2783 EXT4_SB(sb)->journal_bdev = bdev;
2784 ext4_init_journal_params(sb, journal);
2787 jbd2_journal_destroy(journal);
2789 ext4_blkdev_put(bdev);
2793 static int ext4_load_journal(struct super_block *sb,
2794 struct ext4_super_block *es,
2795 unsigned long journal_devnum)
2798 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2801 int really_read_only;
2803 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2805 if (journal_devnum &&
2806 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2807 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2808 "numbers have changed\n");
2809 journal_dev = new_decode_dev(journal_devnum);
2811 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2813 really_read_only = bdev_read_only(sb->s_bdev);
2816 * Are we loading a blank journal or performing recovery after a
2817 * crash? For recovery, we need to check in advance whether we
2818 * can get read-write access to the device.
2821 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2822 if (sb->s_flags & MS_RDONLY) {
2823 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2824 "required on readonly filesystem.\n");
2825 if (really_read_only) {
2826 printk(KERN_ERR "EXT4-fs: write access "
2827 "unavailable, cannot proceed.\n");
2830 printk(KERN_INFO "EXT4-fs: write access will "
2831 "be enabled during recovery.\n");
2835 if (journal_inum && journal_dev) {
2836 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2837 "and inode journals!\n");
2842 if (!(journal = ext4_get_journal(sb, journal_inum)))
2845 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2849 if (journal->j_flags & JBD2_BARRIER)
2850 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2852 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2854 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2855 err = jbd2_journal_update_format(journal);
2857 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2858 jbd2_journal_destroy(journal);
2863 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2864 err = jbd2_journal_wipe(journal, !really_read_only);
2866 err = jbd2_journal_load(journal);
2869 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2870 jbd2_journal_destroy(journal);
2874 EXT4_SB(sb)->s_journal = journal;
2875 ext4_clear_journal_err(sb, es);
2877 if (journal_devnum &&
2878 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2879 es->s_journal_dev = cpu_to_le32(journal_devnum);
2882 /* Make sure we flush the recovery flag to disk. */
2883 ext4_commit_super(sb, es, 1);
2889 static int ext4_commit_super(struct super_block *sb,
2890 struct ext4_super_block *es, int sync)
2892 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2897 if (buffer_write_io_error(sbh)) {
2899 * Oh, dear. A previous attempt to write the
2900 * superblock failed. This could happen because the
2901 * USB device was yanked out. Or it could happen to
2902 * be a transient write error and maybe the block will
2903 * be remapped. Nothing we can do but to retry the
2904 * write and hope for the best.
2906 printk(KERN_ERR "EXT4-fs: previous I/O error to "
2907 "superblock detected for %s.\n", sb->s_id);
2908 clear_buffer_write_io_error(sbh);
2909 set_buffer_uptodate(sbh);
2911 es->s_wtime = cpu_to_le32(get_seconds());
2912 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2913 &EXT4_SB(sb)->s_freeblocks_counter));
2914 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2915 &EXT4_SB(sb)->s_freeinodes_counter));
2917 BUFFER_TRACE(sbh, "marking dirty");
2918 mark_buffer_dirty(sbh);
2920 error = sync_dirty_buffer(sbh);
2924 error = buffer_write_io_error(sbh);
2926 printk(KERN_ERR "EXT4-fs: I/O error while writing "
2927 "superblock for %s.\n", sb->s_id);
2928 clear_buffer_write_io_error(sbh);
2929 set_buffer_uptodate(sbh);
2937 * Have we just finished recovery? If so, and if we are mounting (or
2938 * remounting) the filesystem readonly, then we will end up with a
2939 * consistent fs on disk. Record that fact.
2941 static void ext4_mark_recovery_complete(struct super_block *sb,
2942 struct ext4_super_block *es)
2944 journal_t *journal = EXT4_SB(sb)->s_journal;
2946 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2947 BUG_ON(journal != NULL);
2950 jbd2_journal_lock_updates(journal);
2951 if (jbd2_journal_flush(journal) < 0)
2955 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2956 sb->s_flags & MS_RDONLY) {
2957 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2959 ext4_commit_super(sb, es, 1);
2964 jbd2_journal_unlock_updates(journal);
2968 * If we are mounting (or read-write remounting) a filesystem whose journal
2969 * has recorded an error from a previous lifetime, move that error to the
2970 * main filesystem now.
2972 static void ext4_clear_journal_err(struct super_block *sb,
2973 struct ext4_super_block *es)
2979 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2981 journal = EXT4_SB(sb)->s_journal;
2984 * Now check for any error status which may have been recorded in the
2985 * journal by a prior ext4_error() or ext4_abort()
2988 j_errno = jbd2_journal_errno(journal);
2992 errstr = ext4_decode_error(sb, j_errno, nbuf);
2993 ext4_warning(sb, __func__, "Filesystem error recorded "
2994 "from previous mount: %s", errstr);
2995 ext4_warning(sb, __func__, "Marking fs in need of "
2996 "filesystem check.");
2998 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2999 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3000 ext4_commit_super(sb, es, 1);
3002 jbd2_journal_clear_err(journal);
3007 * Force the running and committing transactions to commit,
3008 * and wait on the commit.
3010 int ext4_force_commit(struct super_block *sb)
3015 if (sb->s_flags & MS_RDONLY)
3018 journal = EXT4_SB(sb)->s_journal;
3021 ret = ext4_journal_force_commit(journal);
3028 * Ext4 always journals updates to the superblock itself, so we don't
3029 * have to propagate any other updates to the superblock on disk at this
3030 * point. (We can probably nuke this function altogether, and remove
3031 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
3033 static void ext4_write_super(struct super_block *sb)
3035 if (EXT4_SB(sb)->s_journal) {
3036 if (mutex_trylock(&sb->s_lock) != 0)
3040 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3044 static int ext4_sync_fs(struct super_block *sb, int wait)
3049 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3051 if (EXT4_SB(sb)->s_journal) {
3052 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3055 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3059 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3065 * LVM calls this function before a (read-only) snapshot is created. This
3066 * gives us a chance to flush the journal completely and mark the fs clean.
3068 static int ext4_freeze(struct super_block *sb)
3074 if (!(sb->s_flags & MS_RDONLY)) {
3075 journal = EXT4_SB(sb)->s_journal;
3078 /* Now we set up the journal barrier. */
3079 jbd2_journal_lock_updates(journal);
3082 * We don't want to clear needs_recovery flag when we
3083 * failed to flush the journal.
3085 error = jbd2_journal_flush(journal);
3090 /* Journal blocked and flushed, clear needs_recovery flag. */
3091 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3092 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3098 jbd2_journal_unlock_updates(journal);
3103 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3104 * flag here, even though the filesystem is not technically dirty yet.
3106 static int ext4_unfreeze(struct super_block *sb)
3108 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3110 /* Reser the needs_recovery flag before the fs is unlocked. */
3111 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3112 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3114 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3119 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3121 struct ext4_super_block *es;
3122 struct ext4_sb_info *sbi = EXT4_SB(sb);
3123 ext4_fsblk_t n_blocks_count = 0;
3124 unsigned long old_sb_flags;
3125 struct ext4_mount_options old_opts;
3127 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3133 /* Store the original options */
3134 old_sb_flags = sb->s_flags;
3135 old_opts.s_mount_opt = sbi->s_mount_opt;
3136 old_opts.s_resuid = sbi->s_resuid;
3137 old_opts.s_resgid = sbi->s_resgid;
3138 old_opts.s_commit_interval = sbi->s_commit_interval;
3139 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3140 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3142 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3143 for (i = 0; i < MAXQUOTAS; i++)
3144 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3146 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3147 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3150 * Allow the "check" option to be passed as a remount option.
3152 if (!parse_options(data, sb, NULL, &journal_ioprio,
3153 &n_blocks_count, 1)) {
3158 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3159 ext4_abort(sb, __func__, "Abort forced by user");
3161 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3162 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3166 if (sbi->s_journal) {
3167 ext4_init_journal_params(sb, sbi->s_journal);
3168 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3171 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3172 n_blocks_count > ext4_blocks_count(es)) {
3173 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3178 if (*flags & MS_RDONLY) {
3180 * First of all, the unconditional stuff we have to do
3181 * to disable replay of the journal when we next remount
3183 sb->s_flags |= MS_RDONLY;
3186 * OK, test if we are remounting a valid rw partition
3187 * readonly, and if so set the rdonly flag and then
3188 * mark the partition as valid again.
3190 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3191 (sbi->s_mount_state & EXT4_VALID_FS))
3192 es->s_state = cpu_to_le16(sbi->s_mount_state);
3195 * We have to unlock super so that we can wait for
3198 if (sbi->s_journal) {
3200 ext4_mark_recovery_complete(sb, es);
3205 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3206 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3207 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3208 "remount RDWR because of unsupported "
3209 "optional features (%x).\n", sb->s_id,
3210 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3211 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3217 * Make sure the group descriptor checksums
3218 * are sane. If they aren't, refuse to
3221 for (g = 0; g < sbi->s_groups_count; g++) {
3222 struct ext4_group_desc *gdp =
3223 ext4_get_group_desc(sb, g, NULL);
3225 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3227 "EXT4-fs: ext4_remount: "
3228 "Checksum for group %u failed (%u!=%u)\n",
3229 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3230 le16_to_cpu(gdp->bg_checksum));
3237 * If we have an unprocessed orphan list hanging
3238 * around from a previously readonly bdev mount,
3239 * require a full umount/remount for now.
3241 if (es->s_last_orphan) {
3242 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3243 "remount RDWR because of unprocessed "
3244 "orphan inode list. Please "
3245 "umount/remount instead.\n",
3252 * Mounting a RDONLY partition read-write, so reread
3253 * and store the current valid flag. (It may have
3254 * been changed by e2fsck since we originally mounted
3258 ext4_clear_journal_err(sb, es);
3259 sbi->s_mount_state = le16_to_cpu(es->s_state);
3260 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3262 if (!ext4_setup_super(sb, es, 0))
3263 sb->s_flags &= ~MS_RDONLY;
3266 if (sbi->s_journal == NULL)
3267 ext4_commit_super(sb, es, 1);
3270 /* Release old quota file names */
3271 for (i = 0; i < MAXQUOTAS; i++)
3272 if (old_opts.s_qf_names[i] &&
3273 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3274 kfree(old_opts.s_qf_names[i]);
3278 sb->s_flags = old_sb_flags;
3279 sbi->s_mount_opt = old_opts.s_mount_opt;
3280 sbi->s_resuid = old_opts.s_resuid;
3281 sbi->s_resgid = old_opts.s_resgid;
3282 sbi->s_commit_interval = old_opts.s_commit_interval;
3283 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3284 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3286 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3287 for (i = 0; i < MAXQUOTAS; i++) {
3288 if (sbi->s_qf_names[i] &&
3289 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3290 kfree(sbi->s_qf_names[i]);
3291 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3297 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3299 struct super_block *sb = dentry->d_sb;
3300 struct ext4_sb_info *sbi = EXT4_SB(sb);
3301 struct ext4_super_block *es = sbi->s_es;
3304 if (test_opt(sb, MINIX_DF)) {
3305 sbi->s_overhead_last = 0;
3306 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3307 ext4_group_t ngroups = sbi->s_groups_count, i;
3308 ext4_fsblk_t overhead = 0;
3312 * Compute the overhead (FS structures). This is constant
3313 * for a given filesystem unless the number of block groups
3314 * changes so we cache the previous value until it does.
3318 * All of the blocks before first_data_block are
3321 overhead = le32_to_cpu(es->s_first_data_block);
3324 * Add the overhead attributed to the superblock and
3325 * block group descriptors. If the sparse superblocks
3326 * feature is turned on, then not all groups have this.
3328 for (i = 0; i < ngroups; i++) {
3329 overhead += ext4_bg_has_super(sb, i) +
3330 ext4_bg_num_gdb(sb, i);
3335 * Every block group has an inode bitmap, a block
3336 * bitmap, and an inode table.
3338 overhead += ngroups * (2 + sbi->s_itb_per_group);
3339 sbi->s_overhead_last = overhead;
3341 sbi->s_blocks_last = ext4_blocks_count(es);
3344 buf->f_type = EXT4_SUPER_MAGIC;
3345 buf->f_bsize = sb->s_blocksize;
3346 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3347 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3348 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3349 ext4_free_blocks_count_set(es, buf->f_bfree);
3350 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3351 if (buf->f_bfree < ext4_r_blocks_count(es))
3353 buf->f_files = le32_to_cpu(es->s_inodes_count);
3354 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3355 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3356 buf->f_namelen = EXT4_NAME_LEN;
3357 fsid = le64_to_cpup((void *)es->s_uuid) ^
3358 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3359 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3360 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3364 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3365 * is locked for write. Otherwise the are possible deadlocks:
3366 * Process 1 Process 2
3367 * ext4_create() quota_sync()
3368 * jbd2_journal_start() write_dquot()
3369 * DQUOT_INIT() down(dqio_mutex)
3370 * down(dqio_mutex) jbd2_journal_start()
3376 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3378 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3381 static int ext4_write_dquot(struct dquot *dquot)
3385 struct inode *inode;
3387 inode = dquot_to_inode(dquot);
3388 handle = ext4_journal_start(inode,
3389 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3391 return PTR_ERR(handle);
3392 ret = dquot_commit(dquot);
3393 err = ext4_journal_stop(handle);
3399 static int ext4_acquire_dquot(struct dquot *dquot)
3404 handle = ext4_journal_start(dquot_to_inode(dquot),
3405 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3407 return PTR_ERR(handle);
3408 ret = dquot_acquire(dquot);
3409 err = ext4_journal_stop(handle);
3415 static int ext4_release_dquot(struct dquot *dquot)
3420 handle = ext4_journal_start(dquot_to_inode(dquot),
3421 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3422 if (IS_ERR(handle)) {
3423 /* Release dquot anyway to avoid endless cycle in dqput() */
3424 dquot_release(dquot);
3425 return PTR_ERR(handle);
3427 ret = dquot_release(dquot);
3428 err = ext4_journal_stop(handle);
3434 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3436 /* Are we journaling quotas? */
3437 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3438 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3439 dquot_mark_dquot_dirty(dquot);
3440 return ext4_write_dquot(dquot);
3442 return dquot_mark_dquot_dirty(dquot);
3446 static int ext4_write_info(struct super_block *sb, int type)
3451 /* Data block + inode block */
3452 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3454 return PTR_ERR(handle);
3455 ret = dquot_commit_info(sb, type);
3456 err = ext4_journal_stop(handle);
3463 * Turn on quotas during mount time - we need to find
3464 * the quota file and such...
3466 static int ext4_quota_on_mount(struct super_block *sb, int type)
3468 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3469 EXT4_SB(sb)->s_jquota_fmt, type);
3473 * Standard function to be called on quota_on
3475 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3476 char *name, int remount)
3481 if (!test_opt(sb, QUOTA))
3483 /* When remounting, no checks are needed and in fact, name is NULL */
3485 return vfs_quota_on(sb, type, format_id, name, remount);
3487 err = kern_path(name, LOOKUP_FOLLOW, &path);
3491 /* Quotafile not on the same filesystem? */
3492 if (path.mnt->mnt_sb != sb) {
3496 /* Journaling quota? */
3497 if (EXT4_SB(sb)->s_qf_names[type]) {
3498 /* Quotafile not in fs root? */
3499 if (path.dentry->d_parent != sb->s_root)
3501 "EXT4-fs: Quota file not on filesystem root. "
3502 "Journaled quota will not work.\n");
3506 * When we journal data on quota file, we have to flush journal to see
3507 * all updates to the file when we bypass pagecache...
3509 if (EXT4_SB(sb)->s_journal &&
3510 ext4_should_journal_data(path.dentry->d_inode)) {
3512 * We don't need to lock updates but journal_flush() could
3513 * otherwise be livelocked...
3515 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3516 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3517 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3524 err = vfs_quota_on_path(sb, type, format_id, &path);
3529 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3530 * acquiring the locks... As quota files are never truncated and quota code
3531 * itself serializes the operations (and noone else should touch the files)
3532 * we don't have to be afraid of races */
3533 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3534 size_t len, loff_t off)
3536 struct inode *inode = sb_dqopt(sb)->files[type];
3537 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3539 int offset = off & (sb->s_blocksize - 1);
3542 struct buffer_head *bh;
3543 loff_t i_size = i_size_read(inode);
3547 if (off+len > i_size)
3550 while (toread > 0) {
3551 tocopy = sb->s_blocksize - offset < toread ?
3552 sb->s_blocksize - offset : toread;
3553 bh = ext4_bread(NULL, inode, blk, 0, &err);
3556 if (!bh) /* A hole? */
3557 memset(data, 0, tocopy);
3559 memcpy(data, bh->b_data+offset, tocopy);
3569 /* Write to quotafile (we know the transaction is already started and has
3570 * enough credits) */
3571 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3572 const char *data, size_t len, loff_t off)
3574 struct inode *inode = sb_dqopt(sb)->files[type];
3575 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3577 int offset = off & (sb->s_blocksize - 1);
3579 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3580 size_t towrite = len;
3581 struct buffer_head *bh;
3582 handle_t *handle = journal_current_handle();
3584 if (EXT4_SB(sb)->s_journal && !handle) {
3585 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3586 " cancelled because transaction is not started.\n",
3587 (unsigned long long)off, (unsigned long long)len);
3590 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3591 while (towrite > 0) {
3592 tocopy = sb->s_blocksize - offset < towrite ?
3593 sb->s_blocksize - offset : towrite;
3594 bh = ext4_bread(handle, inode, blk, 1, &err);
3597 if (journal_quota) {
3598 err = ext4_journal_get_write_access(handle, bh);
3605 memcpy(bh->b_data+offset, data, tocopy);
3606 flush_dcache_page(bh->b_page);
3609 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3611 /* Always do at least ordered writes for quotas */
3612 err = ext4_jbd2_file_inode(handle, inode);
3613 mark_buffer_dirty(bh);
3624 if (len == towrite) {
3625 mutex_unlock(&inode->i_mutex);
3628 if (inode->i_size < off+len-towrite) {
3629 i_size_write(inode, off+len-towrite);
3630 EXT4_I(inode)->i_disksize = inode->i_size;
3632 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3633 ext4_mark_inode_dirty(handle, inode);
3634 mutex_unlock(&inode->i_mutex);
3635 return len - towrite;
3640 static int ext4_get_sb(struct file_system_type *fs_type,
3641 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3643 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3646 #ifdef CONFIG_PROC_FS
3647 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3649 unsigned int *p = m->private;
3651 seq_printf(m, "%u\n", *p);
3655 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3657 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3660 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3661 size_t cnt, loff_t *ppos)
3663 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3666 if (cnt >= sizeof(str))
3668 if (copy_from_user(str, buf, cnt))
3671 *p = simple_strtoul(str, NULL, 0);
3675 const struct file_operations ext4_ui_proc_fops = {
3676 .owner = THIS_MODULE,
3677 .open = ext4_ui_proc_open,
3679 .llseek = seq_lseek,
3680 .release = single_release,
3681 .write = ext4_ui_proc_write,
3685 static struct file_system_type ext4_fs_type = {
3686 .owner = THIS_MODULE,
3688 .get_sb = ext4_get_sb,
3689 .kill_sb = kill_block_super,
3690 .fs_flags = FS_REQUIRES_DEV,
3693 #ifdef CONFIG_EXT4DEV_COMPAT
3694 static int ext4dev_get_sb(struct file_system_type *fs_type,
3695 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3697 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3698 "to mount using ext4\n");
3699 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3700 "will go away by 2.6.31\n");
3701 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3704 static struct file_system_type ext4dev_fs_type = {
3705 .owner = THIS_MODULE,
3707 .get_sb = ext4dev_get_sb,
3708 .kill_sb = kill_block_super,
3709 .fs_flags = FS_REQUIRES_DEV,
3711 MODULE_ALIAS("ext4dev");
3714 static int __init init_ext4_fs(void)
3718 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3719 err = init_ext4_mballoc();
3723 err = init_ext4_xattr();
3726 err = init_inodecache();
3729 err = register_filesystem(&ext4_fs_type);
3732 #ifdef CONFIG_EXT4DEV_COMPAT
3733 err = register_filesystem(&ext4dev_fs_type);
3735 unregister_filesystem(&ext4_fs_type);
3741 destroy_inodecache();
3745 exit_ext4_mballoc();
3749 static void __exit exit_ext4_fs(void)
3751 unregister_filesystem(&ext4_fs_type);
3752 #ifdef CONFIG_EXT4DEV_COMPAT
3753 unregister_filesystem(&ext4dev_fs_type);
3755 destroy_inodecache();
3757 exit_ext4_mballoc();
3758 remove_proc_entry("fs/ext4", NULL);
3761 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3762 MODULE_DESCRIPTION("Fourth Extended Filesystem");
3763 MODULE_LICENSE("GPL");
3764 module_init(init_ext4_fs)
3765 module_exit(exit_ext4_fs)