2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/marker.h>
41 #include <linux/log2.h>
42 #include <linux/crc16.h>
43 #include <asm/uaccess.h>
46 #include "ext4_jbd2.h"
50 static int default_mb_history_length = 1000;
52 module_param_named(default_mb_history_length, default_mb_history_length,
54 MODULE_PARM_DESC(default_mb_history_length,
55 "Default number of entries saved for mb_history");
57 struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_commit_super(struct super_block *sb, int sync);
63 static void ext4_mark_recovery_complete(struct super_block *sb,
64 struct ext4_super_block *es);
65 static void ext4_clear_journal_err(struct super_block *sb,
66 struct ext4_super_block *es);
67 static int ext4_sync_fs(struct super_block *sb, int wait);
68 static const char *ext4_decode_error(struct super_block *sb, int errno,
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static void ext4_write_super(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
77 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
80 return le32_to_cpu(bg->bg_block_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
85 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
86 struct ext4_group_desc *bg)
88 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
93 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
94 struct ext4_group_desc *bg)
96 return le32_to_cpu(bg->bg_inode_table_lo) |
97 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
98 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
101 __u32 ext4_free_blks_count(struct super_block *sb,
102 struct ext4_group_desc *bg)
104 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
105 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
106 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
109 __u32 ext4_free_inodes_count(struct super_block *sb,
110 struct ext4_group_desc *bg)
112 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
113 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
114 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
117 __u32 ext4_used_dirs_count(struct super_block *sb,
118 struct ext4_group_desc *bg)
120 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
121 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
122 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
125 __u32 ext4_itable_unused_count(struct super_block *sb,
126 struct ext4_group_desc *bg)
128 return le16_to_cpu(bg->bg_itable_unused_lo) |
129 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
130 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
133 void ext4_block_bitmap_set(struct super_block *sb,
134 struct ext4_group_desc *bg, ext4_fsblk_t blk)
136 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
137 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
138 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
141 void ext4_inode_bitmap_set(struct super_block *sb,
142 struct ext4_group_desc *bg, ext4_fsblk_t blk)
144 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
145 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
146 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
149 void ext4_inode_table_set(struct super_block *sb,
150 struct ext4_group_desc *bg, ext4_fsblk_t blk)
152 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
153 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
154 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
157 void ext4_free_blks_set(struct super_block *sb,
158 struct ext4_group_desc *bg, __u32 count)
160 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
161 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
162 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
165 void ext4_free_inodes_set(struct super_block *sb,
166 struct ext4_group_desc *bg, __u32 count)
168 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
169 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
170 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
173 void ext4_used_dirs_set(struct super_block *sb,
174 struct ext4_group_desc *bg, __u32 count)
176 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
177 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
178 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
181 void ext4_itable_unused_set(struct super_block *sb,
182 struct ext4_group_desc *bg, __u32 count)
184 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
185 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
186 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190 * Wrappers for jbd2_journal_start/end.
192 * The only special thing we need to do here is to make sure that all
193 * journal_end calls result in the superblock being marked dirty, so
194 * that sync() will call the filesystem's write_super callback if
197 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
201 if (sb->s_flags & MS_RDONLY)
202 return ERR_PTR(-EROFS);
204 /* Special case here: if the journal has aborted behind our
205 * backs (eg. EIO in the commit thread), then we still need to
206 * take the FS itself readonly cleanly. */
207 journal = EXT4_SB(sb)->s_journal;
209 if (is_journal_aborted(journal)) {
210 ext4_abort(sb, __func__,
211 "Detected aborted journal");
212 return ERR_PTR(-EROFS);
214 return jbd2_journal_start(journal, nblocks);
217 * We're not journaling, return the appropriate indication.
219 current->journal_info = EXT4_NOJOURNAL_HANDLE;
220 return current->journal_info;
224 * The only special thing we need to do here is to make sure that all
225 * jbd2_journal_stop calls result in the superblock being marked dirty, so
226 * that sync() will call the filesystem's write_super callback if
229 int __ext4_journal_stop(const char *where, handle_t *handle)
231 struct super_block *sb;
235 if (!ext4_handle_valid(handle)) {
237 * Do this here since we don't call jbd2_journal_stop() in
240 current->journal_info = NULL;
243 sb = handle->h_transaction->t_journal->j_private;
245 rc = jbd2_journal_stop(handle);
250 __ext4_std_error(sb, where, err);
254 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
255 struct buffer_head *bh, handle_t *handle, int err)
258 const char *errstr = ext4_decode_error(NULL, err, nbuf);
260 BUG_ON(!ext4_handle_valid(handle));
263 BUFFER_TRACE(bh, "abort");
268 if (is_handle_aborted(handle))
271 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
272 caller, errstr, err_fn);
274 jbd2_journal_abort_handle(handle);
277 /* Deal with the reporting of failure conditions on a filesystem such as
278 * inconsistencies detected or read IO failures.
280 * On ext2, we can store the error state of the filesystem in the
281 * superblock. That is not possible on ext4, because we may have other
282 * write ordering constraints on the superblock which prevent us from
283 * writing it out straight away; and given that the journal is about to
284 * be aborted, we can't rely on the current, or future, transactions to
285 * write out the superblock safely.
287 * We'll just use the jbd2_journal_abort() error code to record an error in
288 * the journal instead. On recovery, the journal will compain about
289 * that error until we've noted it down and cleared it.
292 static void ext4_handle_error(struct super_block *sb)
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 if (sb->s_flags & MS_RDONLY)
302 if (!test_opt(sb, ERRORS_CONT)) {
303 journal_t *journal = EXT4_SB(sb)->s_journal;
305 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
307 jbd2_journal_abort(journal, -EIO);
309 if (test_opt(sb, ERRORS_RO)) {
310 printk(KERN_CRIT "Remounting filesystem read-only\n");
311 sb->s_flags |= MS_RDONLY;
313 ext4_commit_super(sb, 1);
314 if (test_opt(sb, ERRORS_PANIC))
315 panic("EXT4-fs (device %s): panic forced after error\n",
319 void ext4_error(struct super_block *sb, const char *function,
320 const char *fmt, ...)
325 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
330 ext4_handle_error(sb);
333 static const char *ext4_decode_error(struct super_block *sb, int errno,
340 errstr = "IO failure";
343 errstr = "Out of memory";
346 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
347 errstr = "Journal has aborted";
349 errstr = "Readonly filesystem";
352 /* If the caller passed in an extra buffer for unknown
353 * errors, textualise them now. Else we just return
356 /* Check for truncated error codes... */
357 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
366 /* __ext4_std_error decodes expected errors from journaling functions
367 * automatically and invokes the appropriate error response. */
369 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
374 /* Special case: if the error is EROFS, and we're not already
375 * inside a transaction, then there's really no point in logging
377 if (errno == -EROFS && journal_current_handle() == NULL &&
378 (sb->s_flags & MS_RDONLY))
381 errstr = ext4_decode_error(sb, errno, nbuf);
382 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
383 sb->s_id, function, errstr);
385 ext4_handle_error(sb);
389 * ext4_abort is a much stronger failure handler than ext4_error. The
390 * abort function may be used to deal with unrecoverable failures such
391 * as journal IO errors or ENOMEM at a critical moment in log management.
393 * We unconditionally force the filesystem into an ABORT|READONLY state,
394 * unless the error response on the fs has been set to panic in which
395 * case we take the easy way out and panic immediately.
398 void ext4_abort(struct super_block *sb, const char *function,
399 const char *fmt, ...)
403 printk(KERN_CRIT "ext4_abort called.\n");
406 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
411 if (test_opt(sb, ERRORS_PANIC))
412 panic("EXT4-fs panic from previous error\n");
414 if (sb->s_flags & MS_RDONLY)
417 printk(KERN_CRIT "Remounting filesystem read-only\n");
418 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
419 sb->s_flags |= MS_RDONLY;
420 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
421 if (EXT4_SB(sb)->s_journal)
422 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
425 void ext4_warning(struct super_block *sb, const char *function,
426 const char *fmt, ...)
431 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
438 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
439 const char *function, const char *fmt, ...)
444 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
447 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
452 if (test_opt(sb, ERRORS_CONT)) {
453 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
454 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
455 ext4_commit_super(sb, 0);
458 ext4_unlock_group(sb, grp);
459 ext4_handle_error(sb);
461 * We only get here in the ERRORS_RO case; relocking the group
462 * may be dangerous, but nothing bad will happen since the
463 * filesystem will have already been marked read/only and the
464 * journal has been aborted. We return 1 as a hint to callers
465 * who might what to use the return value from
466 * ext4_grp_locked_error() to distinguish beween the
467 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
468 * aggressively from the ext4 function in question, with a
469 * more appropriate error code.
471 ext4_lock_group(sb, grp);
476 void ext4_update_dynamic_rev(struct super_block *sb)
478 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
480 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
483 ext4_warning(sb, __func__,
484 "updating to rev %d because of new feature flag, "
485 "running e2fsck is recommended",
488 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
489 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
490 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
491 /* leave es->s_feature_*compat flags alone */
492 /* es->s_uuid will be set by e2fsck if empty */
495 * The rest of the superblock fields should be zero, and if not it
496 * means they are likely already in use, so leave them alone. We
497 * can leave it up to e2fsck to clean up any inconsistencies there.
502 * Open the external journal device
504 static struct block_device *ext4_blkdev_get(dev_t dev)
506 struct block_device *bdev;
507 char b[BDEVNAME_SIZE];
509 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
515 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
516 __bdevname(dev, b), PTR_ERR(bdev));
521 * Release the journal device
523 static int ext4_blkdev_put(struct block_device *bdev)
526 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
529 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
531 struct block_device *bdev;
534 bdev = sbi->journal_bdev;
536 ret = ext4_blkdev_put(bdev);
537 sbi->journal_bdev = NULL;
542 static inline struct inode *orphan_list_entry(struct list_head *l)
544 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
547 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
551 printk(KERN_ERR "sb orphan head is %d\n",
552 le32_to_cpu(sbi->s_es->s_last_orphan));
554 printk(KERN_ERR "sb_info orphan list:\n");
555 list_for_each(l, &sbi->s_orphan) {
556 struct inode *inode = orphan_list_entry(l);
558 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
559 inode->i_sb->s_id, inode->i_ino, inode,
560 inode->i_mode, inode->i_nlink,
565 static void ext4_put_super(struct super_block *sb)
567 struct ext4_sb_info *sbi = EXT4_SB(sb);
568 struct ext4_super_block *es = sbi->s_es;
571 ext4_release_system_zone(sb);
573 ext4_ext_release(sb);
574 ext4_xattr_put_super(sb);
575 if (sbi->s_journal) {
576 err = jbd2_journal_destroy(sbi->s_journal);
577 sbi->s_journal = NULL;
579 ext4_abort(sb, __func__,
580 "Couldn't clean up the journal");
582 if (!(sb->s_flags & MS_RDONLY)) {
583 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
584 es->s_state = cpu_to_le16(sbi->s_mount_state);
585 ext4_commit_super(sb, 1);
588 remove_proc_entry(sb->s_id, ext4_proc_root);
590 kobject_del(&sbi->s_kobj);
592 for (i = 0; i < sbi->s_gdb_count; i++)
593 brelse(sbi->s_group_desc[i]);
594 kfree(sbi->s_group_desc);
595 if (is_vmalloc_addr(sbi->s_flex_groups))
596 vfree(sbi->s_flex_groups);
598 kfree(sbi->s_flex_groups);
599 percpu_counter_destroy(&sbi->s_freeblocks_counter);
600 percpu_counter_destroy(&sbi->s_freeinodes_counter);
601 percpu_counter_destroy(&sbi->s_dirs_counter);
602 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
605 for (i = 0; i < MAXQUOTAS; i++)
606 kfree(sbi->s_qf_names[i]);
609 /* Debugging code just in case the in-memory inode orphan list
610 * isn't empty. The on-disk one can be non-empty if we've
611 * detected an error and taken the fs readonly, but the
612 * in-memory list had better be clean by this point. */
613 if (!list_empty(&sbi->s_orphan))
614 dump_orphan_list(sb, sbi);
615 J_ASSERT(list_empty(&sbi->s_orphan));
617 invalidate_bdev(sb->s_bdev);
618 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
620 * Invalidate the journal device's buffers. We don't want them
621 * floating about in memory - the physical journal device may
622 * hotswapped, and it breaks the `ro-after' testing code.
624 sync_blockdev(sbi->journal_bdev);
625 invalidate_bdev(sbi->journal_bdev);
626 ext4_blkdev_remove(sbi);
628 sb->s_fs_info = NULL;
630 * Now that we are completely done shutting down the
631 * superblock, we need to actually destroy the kobject.
635 kobject_put(&sbi->s_kobj);
636 wait_for_completion(&sbi->s_kobj_unregister);
639 kfree(sbi->s_blockgroup_lock);
644 static struct kmem_cache *ext4_inode_cachep;
647 * Called inside transaction, so use GFP_NOFS
649 static struct inode *ext4_alloc_inode(struct super_block *sb)
651 struct ext4_inode_info *ei;
653 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
656 #ifdef CONFIG_EXT4_FS_POSIX_ACL
657 ei->i_acl = EXT4_ACL_NOT_CACHED;
658 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
660 ei->vfs_inode.i_version = 1;
661 ei->vfs_inode.i_data.writeback_index = 0;
662 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
663 INIT_LIST_HEAD(&ei->i_prealloc_list);
664 spin_lock_init(&ei->i_prealloc_lock);
666 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
667 * therefore it can be null here. Don't check it, just initialize
670 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
671 ei->i_reserved_data_blocks = 0;
672 ei->i_reserved_meta_blocks = 0;
673 ei->i_allocated_meta_blocks = 0;
674 ei->i_delalloc_reserved_flag = 0;
675 spin_lock_init(&(ei->i_block_reservation_lock));
676 return &ei->vfs_inode;
679 static void ext4_destroy_inode(struct inode *inode)
681 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
682 printk("EXT4 Inode %p: orphan list check failed!\n",
684 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
685 EXT4_I(inode), sizeof(struct ext4_inode_info),
689 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
692 static void init_once(void *foo)
694 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
696 INIT_LIST_HEAD(&ei->i_orphan);
697 #ifdef CONFIG_EXT4_FS_XATTR
698 init_rwsem(&ei->xattr_sem);
700 init_rwsem(&ei->i_data_sem);
701 inode_init_once(&ei->vfs_inode);
704 static int init_inodecache(void)
706 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
707 sizeof(struct ext4_inode_info),
708 0, (SLAB_RECLAIM_ACCOUNT|
711 if (ext4_inode_cachep == NULL)
716 static void destroy_inodecache(void)
718 kmem_cache_destroy(ext4_inode_cachep);
721 static void ext4_clear_inode(struct inode *inode)
723 #ifdef CONFIG_EXT4_FS_POSIX_ACL
724 if (EXT4_I(inode)->i_acl &&
725 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
726 posix_acl_release(EXT4_I(inode)->i_acl);
727 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
729 if (EXT4_I(inode)->i_default_acl &&
730 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
731 posix_acl_release(EXT4_I(inode)->i_default_acl);
732 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
735 ext4_discard_preallocations(inode);
736 if (EXT4_JOURNAL(inode))
737 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
738 &EXT4_I(inode)->jinode);
741 static inline void ext4_show_quota_options(struct seq_file *seq,
742 struct super_block *sb)
744 #if defined(CONFIG_QUOTA)
745 struct ext4_sb_info *sbi = EXT4_SB(sb);
747 if (sbi->s_jquota_fmt)
748 seq_printf(seq, ",jqfmt=%s",
749 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
751 if (sbi->s_qf_names[USRQUOTA])
752 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
754 if (sbi->s_qf_names[GRPQUOTA])
755 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
757 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
758 seq_puts(seq, ",usrquota");
760 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
761 seq_puts(seq, ",grpquota");
767 * - it's set to a non-default value OR
768 * - if the per-sb default is different from the global default
770 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
773 unsigned long def_mount_opts;
774 struct super_block *sb = vfs->mnt_sb;
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
776 struct ext4_super_block *es = sbi->s_es;
778 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
779 def_errors = le16_to_cpu(es->s_errors);
781 if (sbi->s_sb_block != 1)
782 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
783 if (test_opt(sb, MINIX_DF))
784 seq_puts(seq, ",minixdf");
785 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
786 seq_puts(seq, ",grpid");
787 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
788 seq_puts(seq, ",nogrpid");
789 if (sbi->s_resuid != EXT4_DEF_RESUID ||
790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
791 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
793 if (sbi->s_resgid != EXT4_DEF_RESGID ||
794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
795 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
797 if (test_opt(sb, ERRORS_RO)) {
798 if (def_errors == EXT4_ERRORS_PANIC ||
799 def_errors == EXT4_ERRORS_CONTINUE) {
800 seq_puts(seq, ",errors=remount-ro");
803 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
804 seq_puts(seq, ",errors=continue");
805 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
806 seq_puts(seq, ",errors=panic");
807 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
808 seq_puts(seq, ",nouid32");
809 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
810 seq_puts(seq, ",debug");
811 if (test_opt(sb, OLDALLOC))
812 seq_puts(seq, ",oldalloc");
813 #ifdef CONFIG_EXT4_FS_XATTR
814 if (test_opt(sb, XATTR_USER) &&
815 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
816 seq_puts(seq, ",user_xattr");
817 if (!test_opt(sb, XATTR_USER) &&
818 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
819 seq_puts(seq, ",nouser_xattr");
822 #ifdef CONFIG_EXT4_FS_POSIX_ACL
823 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
824 seq_puts(seq, ",acl");
825 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
826 seq_puts(seq, ",noacl");
828 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
829 seq_printf(seq, ",commit=%u",
830 (unsigned) (sbi->s_commit_interval / HZ));
832 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
833 seq_printf(seq, ",min_batch_time=%u",
834 (unsigned) sbi->s_min_batch_time);
836 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
837 seq_printf(seq, ",max_batch_time=%u",
838 (unsigned) sbi->s_min_batch_time);
842 * We're changing the default of barrier mount option, so
843 * let's always display its mount state so it's clear what its
846 seq_puts(seq, ",barrier=");
847 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
848 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
849 seq_puts(seq, ",journal_async_commit");
850 if (test_opt(sb, NOBH))
851 seq_puts(seq, ",nobh");
852 if (test_opt(sb, I_VERSION))
853 seq_puts(seq, ",i_version");
854 if (!test_opt(sb, DELALLOC))
855 seq_puts(seq, ",nodelalloc");
859 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
861 * journal mode get enabled in different ways
862 * So just print the value even if we didn't specify it
864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
865 seq_puts(seq, ",data=journal");
866 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
867 seq_puts(seq, ",data=ordered");
868 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
869 seq_puts(seq, ",data=writeback");
871 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
872 seq_printf(seq, ",inode_readahead_blks=%u",
873 sbi->s_inode_readahead_blks);
875 if (test_opt(sb, DATA_ERR_ABORT))
876 seq_puts(seq, ",data_err=abort");
878 if (test_opt(sb, NO_AUTO_DA_ALLOC))
879 seq_puts(seq, ",noauto_da_alloc");
881 ext4_show_quota_options(seq, sb);
886 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
887 u64 ino, u32 generation)
891 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
892 return ERR_PTR(-ESTALE);
893 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
894 return ERR_PTR(-ESTALE);
896 /* iget isn't really right if the inode is currently unallocated!!
898 * ext4_read_inode will return a bad_inode if the inode had been
899 * deleted, so we should be safe.
901 * Currently we don't know the generation for parent directory, so
902 * a generation of 0 means "accept any"
904 inode = ext4_iget(sb, ino);
906 return ERR_CAST(inode);
907 if (generation && inode->i_generation != generation) {
909 return ERR_PTR(-ESTALE);
915 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
916 int fh_len, int fh_type)
918 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
922 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
923 int fh_len, int fh_type)
925 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
930 * Try to release metadata pages (indirect blocks, directories) which are
931 * mapped via the block device. Since these pages could have journal heads
932 * which would prevent try_to_free_buffers() from freeing them, we must use
933 * jbd2 layer's try_to_free_buffers() function to release them.
935 static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
937 journal_t *journal = EXT4_SB(sb)->s_journal;
939 WARN_ON(PageChecked(page));
940 if (!page_has_buffers(page))
943 return jbd2_journal_try_to_free_buffers(journal, page,
945 return try_to_free_buffers(page);
949 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
950 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
952 static int ext4_write_dquot(struct dquot *dquot);
953 static int ext4_acquire_dquot(struct dquot *dquot);
954 static int ext4_release_dquot(struct dquot *dquot);
955 static int ext4_mark_dquot_dirty(struct dquot *dquot);
956 static int ext4_write_info(struct super_block *sb, int type);
957 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount);
959 static int ext4_quota_on_mount(struct super_block *sb, int type);
960 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
961 size_t len, loff_t off);
962 static ssize_t ext4_quota_write(struct super_block *sb, int type,
963 const char *data, size_t len, loff_t off);
965 static struct dquot_operations ext4_quota_operations = {
966 .initialize = dquot_initialize,
968 .alloc_space = dquot_alloc_space,
969 .reserve_space = dquot_reserve_space,
970 .claim_space = dquot_claim_space,
971 .release_rsv = dquot_release_reserved_space,
972 .get_reserved_space = ext4_get_reserved_space,
973 .alloc_inode = dquot_alloc_inode,
974 .free_space = dquot_free_space,
975 .free_inode = dquot_free_inode,
976 .transfer = dquot_transfer,
977 .write_dquot = ext4_write_dquot,
978 .acquire_dquot = ext4_acquire_dquot,
979 .release_dquot = ext4_release_dquot,
980 .mark_dirty = ext4_mark_dquot_dirty,
981 .write_info = ext4_write_info,
982 .alloc_dquot = dquot_alloc,
983 .destroy_dquot = dquot_destroy,
986 static struct quotactl_ops ext4_qctl_operations = {
987 .quota_on = ext4_quota_on,
988 .quota_off = vfs_quota_off,
989 .quota_sync = vfs_quota_sync,
990 .get_info = vfs_get_dqinfo,
991 .set_info = vfs_set_dqinfo,
992 .get_dqblk = vfs_get_dqblk,
993 .set_dqblk = vfs_set_dqblk
997 static const struct super_operations ext4_sops = {
998 .alloc_inode = ext4_alloc_inode,
999 .destroy_inode = ext4_destroy_inode,
1000 .write_inode = ext4_write_inode,
1001 .dirty_inode = ext4_dirty_inode,
1002 .delete_inode = ext4_delete_inode,
1003 .put_super = ext4_put_super,
1004 .sync_fs = ext4_sync_fs,
1005 .freeze_fs = ext4_freeze,
1006 .unfreeze_fs = ext4_unfreeze,
1007 .statfs = ext4_statfs,
1008 .remount_fs = ext4_remount,
1009 .clear_inode = ext4_clear_inode,
1010 .show_options = ext4_show_options,
1012 .quota_read = ext4_quota_read,
1013 .quota_write = ext4_quota_write,
1015 .bdev_try_to_free_page = bdev_try_to_free_page,
1018 static const struct super_operations ext4_nojournal_sops = {
1019 .alloc_inode = ext4_alloc_inode,
1020 .destroy_inode = ext4_destroy_inode,
1021 .write_inode = ext4_write_inode,
1022 .dirty_inode = ext4_dirty_inode,
1023 .delete_inode = ext4_delete_inode,
1024 .write_super = ext4_write_super,
1025 .put_super = ext4_put_super,
1026 .statfs = ext4_statfs,
1027 .remount_fs = ext4_remount,
1028 .clear_inode = ext4_clear_inode,
1029 .show_options = ext4_show_options,
1031 .quota_read = ext4_quota_read,
1032 .quota_write = ext4_quota_write,
1034 .bdev_try_to_free_page = bdev_try_to_free_page,
1037 static const struct export_operations ext4_export_ops = {
1038 .fh_to_dentry = ext4_fh_to_dentry,
1039 .fh_to_parent = ext4_fh_to_parent,
1040 .get_parent = ext4_get_parent,
1044 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1045 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1046 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1047 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1048 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1049 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1050 Opt_journal_update, Opt_journal_dev,
1051 Opt_journal_checksum, Opt_journal_async_commit,
1052 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1053 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1054 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1055 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1056 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1057 Opt_usrquota, Opt_grpquota, Opt_i_version,
1058 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1059 Opt_block_validity, Opt_noblock_validity,
1060 Opt_inode_readahead_blks, Opt_journal_ioprio
1063 static const match_table_t tokens = {
1064 {Opt_bsd_df, "bsddf"},
1065 {Opt_minix_df, "minixdf"},
1066 {Opt_grpid, "grpid"},
1067 {Opt_grpid, "bsdgroups"},
1068 {Opt_nogrpid, "nogrpid"},
1069 {Opt_nogrpid, "sysvgroups"},
1070 {Opt_resgid, "resgid=%u"},
1071 {Opt_resuid, "resuid=%u"},
1073 {Opt_err_cont, "errors=continue"},
1074 {Opt_err_panic, "errors=panic"},
1075 {Opt_err_ro, "errors=remount-ro"},
1076 {Opt_nouid32, "nouid32"},
1077 {Opt_debug, "debug"},
1078 {Opt_oldalloc, "oldalloc"},
1079 {Opt_orlov, "orlov"},
1080 {Opt_user_xattr, "user_xattr"},
1081 {Opt_nouser_xattr, "nouser_xattr"},
1083 {Opt_noacl, "noacl"},
1084 {Opt_noload, "noload"},
1087 {Opt_commit, "commit=%u"},
1088 {Opt_min_batch_time, "min_batch_time=%u"},
1089 {Opt_max_batch_time, "max_batch_time=%u"},
1090 {Opt_journal_update, "journal=update"},
1091 {Opt_journal_dev, "journal_dev=%u"},
1092 {Opt_journal_checksum, "journal_checksum"},
1093 {Opt_journal_async_commit, "journal_async_commit"},
1094 {Opt_abort, "abort"},
1095 {Opt_data_journal, "data=journal"},
1096 {Opt_data_ordered, "data=ordered"},
1097 {Opt_data_writeback, "data=writeback"},
1098 {Opt_data_err_abort, "data_err=abort"},
1099 {Opt_data_err_ignore, "data_err=ignore"},
1100 {Opt_mb_history_length, "mb_history_length=%u"},
1101 {Opt_offusrjquota, "usrjquota="},
1102 {Opt_usrjquota, "usrjquota=%s"},
1103 {Opt_offgrpjquota, "grpjquota="},
1104 {Opt_grpjquota, "grpjquota=%s"},
1105 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1106 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1107 {Opt_grpquota, "grpquota"},
1108 {Opt_noquota, "noquota"},
1109 {Opt_quota, "quota"},
1110 {Opt_usrquota, "usrquota"},
1111 {Opt_barrier, "barrier=%u"},
1112 {Opt_barrier, "barrier"},
1113 {Opt_nobarrier, "nobarrier"},
1114 {Opt_i_version, "i_version"},
1115 {Opt_stripe, "stripe=%u"},
1116 {Opt_resize, "resize"},
1117 {Opt_delalloc, "delalloc"},
1118 {Opt_nodelalloc, "nodelalloc"},
1119 {Opt_block_validity, "block_validity"},
1120 {Opt_noblock_validity, "noblock_validity"},
1121 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1122 {Opt_journal_ioprio, "journal_ioprio=%u"},
1123 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1124 {Opt_auto_da_alloc, "auto_da_alloc"},
1125 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1129 static ext4_fsblk_t get_sb_block(void **data)
1131 ext4_fsblk_t sb_block;
1132 char *options = (char *) *data;
1134 if (!options || strncmp(options, "sb=", 3) != 0)
1135 return 1; /* Default location */
1137 /*todo: use simple_strtoll with >32bit ext4 */
1138 sb_block = simple_strtoul(options, &options, 0);
1139 if (*options && *options != ',') {
1140 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1144 if (*options == ',')
1146 *data = (void *) options;
1150 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1152 static int parse_options(char *options, struct super_block *sb,
1153 unsigned long *journal_devnum,
1154 unsigned int *journal_ioprio,
1155 ext4_fsblk_t *n_blocks_count, int is_remount)
1157 struct ext4_sb_info *sbi = EXT4_SB(sb);
1159 substring_t args[MAX_OPT_ARGS];
1170 while ((p = strsep(&options, ",")) != NULL) {
1175 token = match_token(p, tokens, args);
1178 clear_opt(sbi->s_mount_opt, MINIX_DF);
1181 set_opt(sbi->s_mount_opt, MINIX_DF);
1184 set_opt(sbi->s_mount_opt, GRPID);
1187 clear_opt(sbi->s_mount_opt, GRPID);
1190 if (match_int(&args[0], &option))
1192 sbi->s_resuid = option;
1195 if (match_int(&args[0], &option))
1197 sbi->s_resgid = option;
1200 /* handled by get_sb_block() instead of here */
1201 /* *sb_block = match_int(&args[0]); */
1204 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1205 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1206 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1209 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1210 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1211 set_opt(sbi->s_mount_opt, ERRORS_RO);
1214 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1215 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1216 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1219 set_opt(sbi->s_mount_opt, NO_UID32);
1222 set_opt(sbi->s_mount_opt, DEBUG);
1225 set_opt(sbi->s_mount_opt, OLDALLOC);
1228 clear_opt(sbi->s_mount_opt, OLDALLOC);
1230 #ifdef CONFIG_EXT4_FS_XATTR
1231 case Opt_user_xattr:
1232 set_opt(sbi->s_mount_opt, XATTR_USER);
1234 case Opt_nouser_xattr:
1235 clear_opt(sbi->s_mount_opt, XATTR_USER);
1238 case Opt_user_xattr:
1239 case Opt_nouser_xattr:
1240 printk(KERN_ERR "EXT4 (no)user_xattr options "
1244 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1246 set_opt(sbi->s_mount_opt, POSIX_ACL);
1249 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1254 printk(KERN_ERR "EXT4 (no)acl options "
1258 case Opt_journal_update:
1260 /* Eventually we will want to be able to create
1261 a journal file here. For now, only allow the
1262 user to specify an existing inode to be the
1265 printk(KERN_ERR "EXT4-fs: cannot specify "
1266 "journal on remount\n");
1269 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1271 case Opt_journal_dev:
1273 printk(KERN_ERR "EXT4-fs: cannot specify "
1274 "journal on remount\n");
1277 if (match_int(&args[0], &option))
1279 *journal_devnum = option;
1281 case Opt_journal_checksum:
1282 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1284 case Opt_journal_async_commit:
1285 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1286 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1289 set_opt(sbi->s_mount_opt, NOLOAD);
1292 if (match_int(&args[0], &option))
1297 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1298 sbi->s_commit_interval = HZ * option;
1300 case Opt_max_batch_time:
1301 if (match_int(&args[0], &option))
1306 option = EXT4_DEF_MAX_BATCH_TIME;
1307 sbi->s_max_batch_time = option;
1309 case Opt_min_batch_time:
1310 if (match_int(&args[0], &option))
1314 sbi->s_min_batch_time = option;
1316 case Opt_data_journal:
1317 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1319 case Opt_data_ordered:
1320 data_opt = EXT4_MOUNT_ORDERED_DATA;
1322 case Opt_data_writeback:
1323 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1326 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1329 "EXT4-fs: cannot change data "
1330 "mode on remount\n");
1334 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1335 sbi->s_mount_opt |= data_opt;
1338 case Opt_data_err_abort:
1339 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1341 case Opt_data_err_ignore:
1342 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1344 case Opt_mb_history_length:
1345 if (match_int(&args[0], &option))
1349 sbi->s_mb_history_max = option;
1358 if (sb_any_quota_loaded(sb) &&
1359 !sbi->s_qf_names[qtype]) {
1361 "EXT4-fs: Cannot change journaled "
1362 "quota options when quota turned on.\n");
1365 qname = match_strdup(&args[0]);
1368 "EXT4-fs: not enough memory for "
1369 "storing quotafile name.\n");
1372 if (sbi->s_qf_names[qtype] &&
1373 strcmp(sbi->s_qf_names[qtype], qname)) {
1375 "EXT4-fs: %s quota file already "
1376 "specified.\n", QTYPE2NAME(qtype));
1380 sbi->s_qf_names[qtype] = qname;
1381 if (strchr(sbi->s_qf_names[qtype], '/')) {
1383 "EXT4-fs: quotafile must be on "
1384 "filesystem root.\n");
1385 kfree(sbi->s_qf_names[qtype]);
1386 sbi->s_qf_names[qtype] = NULL;
1389 set_opt(sbi->s_mount_opt, QUOTA);
1391 case Opt_offusrjquota:
1394 case Opt_offgrpjquota:
1397 if (sb_any_quota_loaded(sb) &&
1398 sbi->s_qf_names[qtype]) {
1399 printk(KERN_ERR "EXT4-fs: Cannot change "
1400 "journaled quota options when "
1401 "quota turned on.\n");
1405 * The space will be released later when all options
1406 * are confirmed to be correct
1408 sbi->s_qf_names[qtype] = NULL;
1410 case Opt_jqfmt_vfsold:
1411 qfmt = QFMT_VFS_OLD;
1413 case Opt_jqfmt_vfsv0:
1416 if (sb_any_quota_loaded(sb) &&
1417 sbi->s_jquota_fmt != qfmt) {
1418 printk(KERN_ERR "EXT4-fs: Cannot change "
1419 "journaled quota options when "
1420 "quota turned on.\n");
1423 sbi->s_jquota_fmt = qfmt;
1427 set_opt(sbi->s_mount_opt, QUOTA);
1428 set_opt(sbi->s_mount_opt, USRQUOTA);
1431 set_opt(sbi->s_mount_opt, QUOTA);
1432 set_opt(sbi->s_mount_opt, GRPQUOTA);
1435 if (sb_any_quota_loaded(sb)) {
1436 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1437 "options when quota turned on.\n");
1440 clear_opt(sbi->s_mount_opt, QUOTA);
1441 clear_opt(sbi->s_mount_opt, USRQUOTA);
1442 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1449 "EXT4-fs: quota options not supported.\n");
1453 case Opt_offusrjquota:
1454 case Opt_offgrpjquota:
1455 case Opt_jqfmt_vfsold:
1456 case Opt_jqfmt_vfsv0:
1458 "EXT4-fs: journaled quota options not "
1465 set_opt(sbi->s_mount_opt, ABORT);
1468 clear_opt(sbi->s_mount_opt, BARRIER);
1471 if (match_int(&args[0], &option)) {
1472 set_opt(sbi->s_mount_opt, BARRIER);
1476 set_opt(sbi->s_mount_opt, BARRIER);
1478 clear_opt(sbi->s_mount_opt, BARRIER);
1484 printk("EXT4-fs: resize option only available "
1488 if (match_int(&args[0], &option) != 0)
1490 *n_blocks_count = option;
1493 set_opt(sbi->s_mount_opt, NOBH);
1496 clear_opt(sbi->s_mount_opt, NOBH);
1499 set_opt(sbi->s_mount_opt, I_VERSION);
1500 sb->s_flags |= MS_I_VERSION;
1502 case Opt_nodelalloc:
1503 clear_opt(sbi->s_mount_opt, DELALLOC);
1506 if (match_int(&args[0], &option))
1510 sbi->s_stripe = option;
1513 set_opt(sbi->s_mount_opt, DELALLOC);
1515 case Opt_block_validity:
1516 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1518 case Opt_noblock_validity:
1519 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1521 case Opt_inode_readahead_blks:
1522 if (match_int(&args[0], &option))
1524 if (option < 0 || option > (1 << 30))
1526 if (!is_power_of_2(option)) {
1527 printk(KERN_ERR "EXT4-fs: inode_readahead_blks"
1528 " must be a power of 2\n");
1531 sbi->s_inode_readahead_blks = option;
1533 case Opt_journal_ioprio:
1534 if (match_int(&args[0], &option))
1536 if (option < 0 || option > 7)
1538 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1541 case Opt_noauto_da_alloc:
1542 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1544 case Opt_auto_da_alloc:
1545 if (match_int(&args[0], &option)) {
1546 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1550 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1552 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1556 "EXT4-fs: Unrecognized mount option \"%s\" "
1557 "or missing value\n", p);
1562 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1563 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1564 sbi->s_qf_names[USRQUOTA])
1565 clear_opt(sbi->s_mount_opt, USRQUOTA);
1567 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1568 sbi->s_qf_names[GRPQUOTA])
1569 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1571 if ((sbi->s_qf_names[USRQUOTA] &&
1572 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1573 (sbi->s_qf_names[GRPQUOTA] &&
1574 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1575 printk(KERN_ERR "EXT4-fs: old and new quota "
1576 "format mixing.\n");
1580 if (!sbi->s_jquota_fmt) {
1581 printk(KERN_ERR "EXT4-fs: journaled quota format "
1582 "not specified.\n");
1586 if (sbi->s_jquota_fmt) {
1587 printk(KERN_ERR "EXT4-fs: journaled quota format "
1588 "specified with no journaling "
1597 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
1603 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1604 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1605 "forcing read-only mode\n");
1610 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1611 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1612 "running e2fsck is recommended\n");
1613 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1615 "EXT4-fs warning: mounting fs with errors, "
1616 "running e2fsck is recommended\n");
1617 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1618 le16_to_cpu(es->s_mnt_count) >=
1619 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1621 "EXT4-fs warning: maximal mount count reached, "
1622 "running e2fsck is recommended\n");
1623 else if (le32_to_cpu(es->s_checkinterval) &&
1624 (le32_to_cpu(es->s_lastcheck) +
1625 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1627 "EXT4-fs warning: checktime reached, "
1628 "running e2fsck is recommended\n");
1629 if (!sbi->s_journal)
1630 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1631 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1632 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1633 le16_add_cpu(&es->s_mnt_count, 1);
1634 es->s_mtime = cpu_to_le32(get_seconds());
1635 ext4_update_dynamic_rev(sb);
1637 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1639 ext4_commit_super(sb, 1);
1640 if (test_opt(sb, DEBUG))
1641 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1642 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1644 sbi->s_groups_count,
1645 EXT4_BLOCKS_PER_GROUP(sb),
1646 EXT4_INODES_PER_GROUP(sb),
1649 if (EXT4_SB(sb)->s_journal) {
1650 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1651 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1652 "external", EXT4_SB(sb)->s_journal->j_devname);
1654 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1659 static int ext4_fill_flex_info(struct super_block *sb)
1661 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 struct ext4_group_desc *gdp = NULL;
1663 struct buffer_head *bh;
1664 ext4_group_t flex_group_count;
1665 ext4_group_t flex_group;
1666 int groups_per_flex = 0;
1670 if (!sbi->s_es->s_log_groups_per_flex) {
1671 sbi->s_log_groups_per_flex = 0;
1675 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1676 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1678 /* We allocate both existing and potentially added groups */
1679 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1680 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1681 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1682 size = flex_group_count * sizeof(struct flex_groups);
1683 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1684 if (sbi->s_flex_groups == NULL) {
1685 sbi->s_flex_groups = vmalloc(size);
1686 if (sbi->s_flex_groups)
1687 memset(sbi->s_flex_groups, 0, size);
1689 if (sbi->s_flex_groups == NULL) {
1690 printk(KERN_ERR "EXT4-fs: not enough memory for "
1691 "%u flex groups\n", flex_group_count);
1695 for (i = 0; i < sbi->s_groups_count; i++) {
1696 gdp = ext4_get_group_desc(sb, i, &bh);
1698 flex_group = ext4_flex_group(sbi, i);
1699 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1700 ext4_free_inodes_count(sb, gdp));
1701 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1702 ext4_free_blks_count(sb, gdp));
1703 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1704 ext4_used_dirs_count(sb, gdp));
1712 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1713 struct ext4_group_desc *gdp)
1717 if (sbi->s_es->s_feature_ro_compat &
1718 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1719 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1720 __le32 le_group = cpu_to_le32(block_group);
1722 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1723 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1724 crc = crc16(crc, (__u8 *)gdp, offset);
1725 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1726 /* for checksum of struct ext4_group_desc do the rest...*/
1727 if ((sbi->s_es->s_feature_incompat &
1728 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1729 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1730 crc = crc16(crc, (__u8 *)gdp + offset,
1731 le16_to_cpu(sbi->s_es->s_desc_size) -
1735 return cpu_to_le16(crc);
1738 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1739 struct ext4_group_desc *gdp)
1741 if ((sbi->s_es->s_feature_ro_compat &
1742 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1743 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1749 /* Called at mount-time, super-block is locked */
1750 static int ext4_check_descriptors(struct super_block *sb)
1752 struct ext4_sb_info *sbi = EXT4_SB(sb);
1753 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1754 ext4_fsblk_t last_block;
1755 ext4_fsblk_t block_bitmap;
1756 ext4_fsblk_t inode_bitmap;
1757 ext4_fsblk_t inode_table;
1758 int flexbg_flag = 0;
1761 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1764 ext4_debug("Checking group descriptors");
1766 for (i = 0; i < sbi->s_groups_count; i++) {
1767 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1769 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1770 last_block = ext4_blocks_count(sbi->s_es) - 1;
1772 last_block = first_block +
1773 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1775 block_bitmap = ext4_block_bitmap(sb, gdp);
1776 if (block_bitmap < first_block || block_bitmap > last_block) {
1777 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1778 "Block bitmap for group %u not in group "
1779 "(block %llu)!\n", i, block_bitmap);
1782 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1783 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1784 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1785 "Inode bitmap for group %u not in group "
1786 "(block %llu)!\n", i, inode_bitmap);
1789 inode_table = ext4_inode_table(sb, gdp);
1790 if (inode_table < first_block ||
1791 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1792 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1793 "Inode table for group %u not in group "
1794 "(block %llu)!\n", i, inode_table);
1797 ext4_lock_group(sb, i);
1798 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1799 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1800 "Checksum for group %u failed (%u!=%u)\n",
1801 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1802 gdp)), le16_to_cpu(gdp->bg_checksum));
1803 if (!(sb->s_flags & MS_RDONLY)) {
1804 ext4_unlock_group(sb, i);
1808 ext4_unlock_group(sb, i);
1810 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1813 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1814 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1818 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1819 * the superblock) which were deleted from all directories, but held open by
1820 * a process at the time of a crash. We walk the list and try to delete these
1821 * inodes at recovery time (only with a read-write filesystem).
1823 * In order to keep the orphan inode chain consistent during traversal (in
1824 * case of crash during recovery), we link each inode into the superblock
1825 * orphan list_head and handle it the same way as an inode deletion during
1826 * normal operation (which journals the operations for us).
1828 * We only do an iget() and an iput() on each inode, which is very safe if we
1829 * accidentally point at an in-use or already deleted inode. The worst that
1830 * can happen in this case is that we get a "bit already cleared" message from
1831 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1832 * e2fsck was run on this filesystem, and it must have already done the orphan
1833 * inode cleanup for us, so we can safely abort without any further action.
1835 static void ext4_orphan_cleanup(struct super_block *sb,
1836 struct ext4_super_block *es)
1838 unsigned int s_flags = sb->s_flags;
1839 int nr_orphans = 0, nr_truncates = 0;
1843 if (!es->s_last_orphan) {
1844 jbd_debug(4, "no orphan inodes to clean up\n");
1848 if (bdev_read_only(sb->s_bdev)) {
1849 printk(KERN_ERR "EXT4-fs: write access "
1850 "unavailable, skipping orphan cleanup.\n");
1854 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1855 if (es->s_last_orphan)
1856 jbd_debug(1, "Errors on filesystem, "
1857 "clearing orphan list.\n");
1858 es->s_last_orphan = 0;
1859 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1863 if (s_flags & MS_RDONLY) {
1864 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1866 sb->s_flags &= ~MS_RDONLY;
1869 /* Needed for iput() to work correctly and not trash data */
1870 sb->s_flags |= MS_ACTIVE;
1871 /* Turn on quotas so that they are updated correctly */
1872 for (i = 0; i < MAXQUOTAS; i++) {
1873 if (EXT4_SB(sb)->s_qf_names[i]) {
1874 int ret = ext4_quota_on_mount(sb, i);
1877 "EXT4-fs: Cannot turn on journaled "
1878 "quota: error %d\n", ret);
1883 while (es->s_last_orphan) {
1884 struct inode *inode;
1886 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1887 if (IS_ERR(inode)) {
1888 es->s_last_orphan = 0;
1892 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1894 if (inode->i_nlink) {
1896 "%s: truncating inode %lu to %lld bytes\n",
1897 __func__, inode->i_ino, inode->i_size);
1898 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1899 inode->i_ino, inode->i_size);
1900 ext4_truncate(inode);
1904 "%s: deleting unreferenced inode %lu\n",
1905 __func__, inode->i_ino);
1906 jbd_debug(2, "deleting unreferenced inode %lu\n",
1910 iput(inode); /* The delete magic happens here! */
1913 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1916 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1917 sb->s_id, PLURAL(nr_orphans));
1919 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1920 sb->s_id, PLURAL(nr_truncates));
1922 /* Turn quotas off */
1923 for (i = 0; i < MAXQUOTAS; i++) {
1924 if (sb_dqopt(sb)->files[i])
1925 vfs_quota_off(sb, i, 0);
1928 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1931 * Maximal extent format file size.
1932 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1933 * extent format containers, within a sector_t, and within i_blocks
1934 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1935 * so that won't be a limiting factor.
1937 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1939 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1942 loff_t upper_limit = MAX_LFS_FILESIZE;
1944 /* small i_blocks in vfs inode? */
1945 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1947 * CONFIG_LBD is not enabled implies the inode
1948 * i_block represent total blocks in 512 bytes
1949 * 32 == size of vfs inode i_blocks * 8
1951 upper_limit = (1LL << 32) - 1;
1953 /* total blocks in file system block size */
1954 upper_limit >>= (blkbits - 9);
1955 upper_limit <<= blkbits;
1958 /* 32-bit extent-start container, ee_block */
1963 /* Sanity check against vm- & vfs- imposed limits */
1964 if (res > upper_limit)
1971 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1972 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1973 * We need to be 1 filesystem block less than the 2^48 sector limit.
1975 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1977 loff_t res = EXT4_NDIR_BLOCKS;
1980 /* This is calculated to be the largest file size for a
1981 * dense, bitmapped file such that the total number of
1982 * sectors in the file, including data and all indirect blocks,
1983 * does not exceed 2^48 -1
1984 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1985 * total number of 512 bytes blocks of the file
1988 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1990 * !has_huge_files or CONFIG_LBD is not enabled
1991 * implies the inode i_block represent total blocks in
1992 * 512 bytes 32 == size of vfs inode i_blocks * 8
1994 upper_limit = (1LL << 32) - 1;
1996 /* total blocks in file system block size */
1997 upper_limit >>= (bits - 9);
2001 * We use 48 bit ext4_inode i_blocks
2002 * With EXT4_HUGE_FILE_FL set the i_blocks
2003 * represent total number of blocks in
2004 * file system block size
2006 upper_limit = (1LL << 48) - 1;
2010 /* indirect blocks */
2012 /* double indirect blocks */
2013 meta_blocks += 1 + (1LL << (bits-2));
2014 /* tripple indirect blocks */
2015 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2017 upper_limit -= meta_blocks;
2018 upper_limit <<= bits;
2020 res += 1LL << (bits-2);
2021 res += 1LL << (2*(bits-2));
2022 res += 1LL << (3*(bits-2));
2024 if (res > upper_limit)
2027 if (res > MAX_LFS_FILESIZE)
2028 res = MAX_LFS_FILESIZE;
2033 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2034 ext4_fsblk_t logical_sb_block, int nr)
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2037 ext4_group_t bg, first_meta_bg;
2040 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2042 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2044 return logical_sb_block + nr + 1;
2045 bg = sbi->s_desc_per_block * nr;
2046 if (ext4_bg_has_super(sb, bg))
2048 return (has_super + ext4_group_first_block_no(sb, bg));
2052 * ext4_get_stripe_size: Get the stripe size.
2053 * @sbi: In memory super block info
2055 * If we have specified it via mount option, then
2056 * use the mount option value. If the value specified at mount time is
2057 * greater than the blocks per group use the super block value.
2058 * If the super block value is greater than blocks per group return 0.
2059 * Allocator needs it be less than blocks per group.
2062 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2064 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2065 unsigned long stripe_width =
2066 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2068 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2069 return sbi->s_stripe;
2071 if (stripe_width <= sbi->s_blocks_per_group)
2072 return stripe_width;
2074 if (stride <= sbi->s_blocks_per_group)
2083 struct attribute attr;
2084 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2085 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2086 const char *, size_t);
2090 static int parse_strtoul(const char *buf,
2091 unsigned long max, unsigned long *value)
2095 while (*buf && isspace(*buf))
2097 *value = simple_strtoul(buf, &endp, 0);
2098 while (*endp && isspace(*endp))
2100 if (*endp || *value > max)
2106 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2107 struct ext4_sb_info *sbi,
2110 return snprintf(buf, PAGE_SIZE, "%llu\n",
2111 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2114 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2115 struct ext4_sb_info *sbi, char *buf)
2117 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2119 return snprintf(buf, PAGE_SIZE, "%lu\n",
2120 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2121 sbi->s_sectors_written_start) >> 1);
2124 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2125 struct ext4_sb_info *sbi, char *buf)
2127 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2129 return snprintf(buf, PAGE_SIZE, "%llu\n",
2130 sbi->s_kbytes_written +
2131 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2135 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2136 struct ext4_sb_info *sbi,
2137 const char *buf, size_t count)
2141 if (parse_strtoul(buf, 0x40000000, &t))
2144 if (!is_power_of_2(t))
2147 sbi->s_inode_readahead_blks = t;
2151 static ssize_t sbi_ui_show(struct ext4_attr *a,
2152 struct ext4_sb_info *sbi, char *buf)
2154 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2156 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2159 static ssize_t sbi_ui_store(struct ext4_attr *a,
2160 struct ext4_sb_info *sbi,
2161 const char *buf, size_t count)
2163 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2166 if (parse_strtoul(buf, 0xffffffff, &t))
2172 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2173 static struct ext4_attr ext4_attr_##_name = { \
2174 .attr = {.name = __stringify(_name), .mode = _mode }, \
2177 .offset = offsetof(struct ext4_sb_info, _elname), \
2179 #define EXT4_ATTR(name, mode, show, store) \
2180 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2182 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2183 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2184 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2185 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2186 #define ATTR_LIST(name) &ext4_attr_##name.attr
2188 EXT4_RO_ATTR(delayed_allocation_blocks);
2189 EXT4_RO_ATTR(session_write_kbytes);
2190 EXT4_RO_ATTR(lifetime_write_kbytes);
2191 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2192 inode_readahead_blks_store, s_inode_readahead_blks);
2193 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2194 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2195 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2196 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2197 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2198 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2200 static struct attribute *ext4_attrs[] = {
2201 ATTR_LIST(delayed_allocation_blocks),
2202 ATTR_LIST(session_write_kbytes),
2203 ATTR_LIST(lifetime_write_kbytes),
2204 ATTR_LIST(inode_readahead_blks),
2205 ATTR_LIST(mb_stats),
2206 ATTR_LIST(mb_max_to_scan),
2207 ATTR_LIST(mb_min_to_scan),
2208 ATTR_LIST(mb_order2_req),
2209 ATTR_LIST(mb_stream_req),
2210 ATTR_LIST(mb_group_prealloc),
2214 static ssize_t ext4_attr_show(struct kobject *kobj,
2215 struct attribute *attr, char *buf)
2217 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2219 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2221 return a->show ? a->show(a, sbi, buf) : 0;
2224 static ssize_t ext4_attr_store(struct kobject *kobj,
2225 struct attribute *attr,
2226 const char *buf, size_t len)
2228 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2230 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2232 return a->store ? a->store(a, sbi, buf, len) : 0;
2235 static void ext4_sb_release(struct kobject *kobj)
2237 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2239 complete(&sbi->s_kobj_unregister);
2243 static struct sysfs_ops ext4_attr_ops = {
2244 .show = ext4_attr_show,
2245 .store = ext4_attr_store,
2248 static struct kobj_type ext4_ktype = {
2249 .default_attrs = ext4_attrs,
2250 .sysfs_ops = &ext4_attr_ops,
2251 .release = ext4_sb_release,
2254 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2255 __releases(kernel_lock)
2256 __acquires(kernel_lock)
2259 struct buffer_head *bh;
2260 struct ext4_super_block *es = NULL;
2261 struct ext4_sb_info *sbi;
2263 ext4_fsblk_t sb_block = get_sb_block(&data);
2264 ext4_fsblk_t logical_sb_block;
2265 unsigned long offset = 0;
2266 unsigned long journal_devnum = 0;
2267 unsigned long def_mount_opts;
2273 unsigned int db_count;
2275 int needs_recovery, has_huge_files;
2279 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2281 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2285 sbi->s_blockgroup_lock =
2286 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2287 if (!sbi->s_blockgroup_lock) {
2291 sb->s_fs_info = sbi;
2292 sbi->s_mount_opt = 0;
2293 sbi->s_resuid = EXT4_DEF_RESUID;
2294 sbi->s_resgid = EXT4_DEF_RESGID;
2295 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2296 sbi->s_sb_block = sb_block;
2297 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2302 /* Cleanup superblock name */
2303 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2306 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2308 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
2313 * The ext4 superblock will not be buffer aligned for other than 1kB
2314 * block sizes. We need to calculate the offset from buffer start.
2316 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2317 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2318 offset = do_div(logical_sb_block, blocksize);
2320 logical_sb_block = sb_block;
2323 if (!(bh = sb_bread(sb, logical_sb_block))) {
2324 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
2328 * Note: s_es must be initialized as soon as possible because
2329 * some ext4 macro-instructions depend on its value
2331 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2333 sb->s_magic = le16_to_cpu(es->s_magic);
2334 if (sb->s_magic != EXT4_SUPER_MAGIC)
2336 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2338 /* Set defaults before we parse the mount options */
2339 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2340 if (def_mount_opts & EXT4_DEFM_DEBUG)
2341 set_opt(sbi->s_mount_opt, DEBUG);
2342 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2343 set_opt(sbi->s_mount_opt, GRPID);
2344 if (def_mount_opts & EXT4_DEFM_UID16)
2345 set_opt(sbi->s_mount_opt, NO_UID32);
2346 #ifdef CONFIG_EXT4_FS_XATTR
2347 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2348 set_opt(sbi->s_mount_opt, XATTR_USER);
2350 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2351 if (def_mount_opts & EXT4_DEFM_ACL)
2352 set_opt(sbi->s_mount_opt, POSIX_ACL);
2354 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2355 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2356 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2357 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2358 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2359 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2361 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2362 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2363 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2364 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2366 set_opt(sbi->s_mount_opt, ERRORS_RO);
2368 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2369 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2370 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2371 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2372 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2373 sbi->s_mb_history_max = default_mb_history_length;
2375 set_opt(sbi->s_mount_opt, BARRIER);
2378 * enable delayed allocation by default
2379 * Use -o nodelalloc to turn it off
2381 set_opt(sbi->s_mount_opt, DELALLOC);
2384 if (!parse_options((char *) data, sb, &journal_devnum,
2385 &journal_ioprio, NULL, 0))
2388 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2389 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2391 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2392 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2393 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2394 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2396 "EXT4-fs warning: feature flags set on rev 0 fs, "
2397 "running e2fsck is recommended\n");
2400 * Check feature flags regardless of the revision level, since we
2401 * previously didn't change the revision level when setting the flags,
2402 * so there is a chance incompat flags are set on a rev 0 filesystem.
2404 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2406 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2407 "unsupported optional features (%x).\n", sb->s_id,
2408 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2409 ~EXT4_FEATURE_INCOMPAT_SUPP));
2412 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2413 if (!(sb->s_flags & MS_RDONLY) && features) {
2414 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2415 "unsupported optional features (%x).\n", sb->s_id,
2416 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2417 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2420 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2421 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2422 if (has_huge_files) {
2424 * Large file size enabled file system can only be
2425 * mount if kernel is build with CONFIG_LBD
2427 if (sizeof(root->i_blocks) < sizeof(u64) &&
2428 !(sb->s_flags & MS_RDONLY)) {
2429 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2430 "files cannot be mounted read-write "
2431 "without CONFIG_LBD.\n", sb->s_id);
2435 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2437 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2438 blocksize > EXT4_MAX_BLOCK_SIZE) {
2440 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2441 blocksize, sb->s_id);
2445 if (sb->s_blocksize != blocksize) {
2447 /* Validate the filesystem blocksize */
2448 if (!sb_set_blocksize(sb, blocksize)) {
2449 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2455 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2456 offset = do_div(logical_sb_block, blocksize);
2457 bh = sb_bread(sb, logical_sb_block);
2460 "EXT4-fs: Can't read superblock on 2nd try.\n");
2463 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2465 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2467 "EXT4-fs: Magic mismatch, very weird !\n");
2472 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2474 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2477 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2478 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2480 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2481 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2482 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2483 (!is_power_of_2(sbi->s_inode_size)) ||
2484 (sbi->s_inode_size > blocksize)) {
2486 "EXT4-fs: unsupported inode size: %d\n",
2490 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2491 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2493 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2494 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2495 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2496 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2497 !is_power_of_2(sbi->s_desc_size)) {
2499 "EXT4-fs: unsupported descriptor size %lu\n",
2504 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2505 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2506 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2507 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2509 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2510 if (sbi->s_inodes_per_block == 0)
2512 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2513 sbi->s_inodes_per_block;
2514 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2516 sbi->s_mount_state = le16_to_cpu(es->s_state);
2517 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2518 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2519 for (i = 0; i < 4; i++)
2520 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2521 sbi->s_def_hash_version = es->s_def_hash_version;
2522 i = le32_to_cpu(es->s_flags);
2523 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2524 sbi->s_hash_unsigned = 3;
2525 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2526 #ifdef __CHAR_UNSIGNED__
2527 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2528 sbi->s_hash_unsigned = 3;
2530 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2535 if (sbi->s_blocks_per_group > blocksize * 8) {
2537 "EXT4-fs: #blocks per group too big: %lu\n",
2538 sbi->s_blocks_per_group);
2541 if (sbi->s_inodes_per_group > blocksize * 8) {
2543 "EXT4-fs: #inodes per group too big: %lu\n",
2544 sbi->s_inodes_per_group);
2548 if (ext4_blocks_count(es) >
2549 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2550 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2551 " too large to mount safely\n", sb->s_id);
2552 if (sizeof(sector_t) < 8)
2553 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2558 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2561 /* check blocks count against device size */
2562 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2563 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2564 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu "
2565 "exceeds size of device (%llu blocks)\n",
2566 ext4_blocks_count(es), blocks_count);
2571 * It makes no sense for the first data block to be beyond the end
2572 * of the filesystem.
2574 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2575 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2576 "block %u is beyond end of filesystem (%llu)\n",
2577 le32_to_cpu(es->s_first_data_block),
2578 ext4_blocks_count(es));
2581 blocks_count = (ext4_blocks_count(es) -
2582 le32_to_cpu(es->s_first_data_block) +
2583 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2584 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2585 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2586 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2587 "(block count %llu, first data block %u, "
2588 "blocks per group %lu)\n", sbi->s_groups_count,
2589 ext4_blocks_count(es),
2590 le32_to_cpu(es->s_first_data_block),
2591 EXT4_BLOCKS_PER_GROUP(sb));
2594 sbi->s_groups_count = blocks_count;
2595 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2596 EXT4_DESC_PER_BLOCK(sb);
2597 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2599 if (sbi->s_group_desc == NULL) {
2600 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2604 #ifdef CONFIG_PROC_FS
2606 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2609 bgl_lock_init(sbi->s_blockgroup_lock);
2611 for (i = 0; i < db_count; i++) {
2612 block = descriptor_loc(sb, logical_sb_block, i);
2613 sbi->s_group_desc[i] = sb_bread(sb, block);
2614 if (!sbi->s_group_desc[i]) {
2615 printk(KERN_ERR "EXT4-fs: "
2616 "can't read group descriptor %d\n", i);
2621 if (!ext4_check_descriptors(sb)) {
2622 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2625 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2626 if (!ext4_fill_flex_info(sb)) {
2628 "EXT4-fs: unable to initialize "
2629 "flex_bg meta info!\n");
2633 sbi->s_gdb_count = db_count;
2634 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2635 spin_lock_init(&sbi->s_next_gen_lock);
2637 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2638 ext4_count_free_blocks(sb));
2640 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2641 ext4_count_free_inodes(sb));
2644 err = percpu_counter_init(&sbi->s_dirs_counter,
2645 ext4_count_dirs(sb));
2648 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2651 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2655 sbi->s_stripe = ext4_get_stripe_size(sbi);
2658 * set up enough so that it can read an inode
2660 if (!test_opt(sb, NOLOAD) &&
2661 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2662 sb->s_op = &ext4_sops;
2664 sb->s_op = &ext4_nojournal_sops;
2665 sb->s_export_op = &ext4_export_ops;
2666 sb->s_xattr = ext4_xattr_handlers;
2668 sb->s_qcop = &ext4_qctl_operations;
2669 sb->dq_op = &ext4_quota_operations;
2671 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2672 mutex_init(&sbi->s_orphan_lock);
2673 mutex_init(&sbi->s_resize_lock);
2677 needs_recovery = (es->s_last_orphan != 0 ||
2678 EXT4_HAS_INCOMPAT_FEATURE(sb,
2679 EXT4_FEATURE_INCOMPAT_RECOVER));
2682 * The first inode we look at is the journal inode. Don't try
2683 * root first: it may be modified in the journal!
2685 if (!test_opt(sb, NOLOAD) &&
2686 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2687 if (ext4_load_journal(sb, es, journal_devnum))
2689 if (!(sb->s_flags & MS_RDONLY) &&
2690 EXT4_SB(sb)->s_journal->j_failed_commit) {
2691 printk(KERN_CRIT "EXT4-fs error (device %s): "
2692 "ext4_fill_super: Journal transaction "
2693 "%u is corrupt\n", sb->s_id,
2694 EXT4_SB(sb)->s_journal->j_failed_commit);
2695 if (test_opt(sb, ERRORS_RO)) {
2697 "Mounting filesystem read-only\n");
2698 sb->s_flags |= MS_RDONLY;
2699 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2700 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2702 if (test_opt(sb, ERRORS_PANIC)) {
2703 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2704 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2705 ext4_commit_super(sb, 1);
2709 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2710 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2711 printk(KERN_ERR "EXT4-fs: required journal recovery "
2712 "suppressed and not mounted read-only\n");
2715 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2716 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2717 sbi->s_journal = NULL;
2722 if (ext4_blocks_count(es) > 0xffffffffULL &&
2723 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2724 JBD2_FEATURE_INCOMPAT_64BIT)) {
2725 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
2729 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2730 jbd2_journal_set_features(sbi->s_journal,
2731 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2732 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2733 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2734 jbd2_journal_set_features(sbi->s_journal,
2735 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2736 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2737 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2739 jbd2_journal_clear_features(sbi->s_journal,
2740 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2741 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2744 /* We have now updated the journal if required, so we can
2745 * validate the data journaling mode. */
2746 switch (test_opt(sb, DATA_FLAGS)) {
2748 /* No mode set, assume a default based on the journal
2749 * capabilities: ORDERED_DATA if the journal can
2750 * cope, else JOURNAL_DATA
2752 if (jbd2_journal_check_available_features
2753 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2754 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2756 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2759 case EXT4_MOUNT_ORDERED_DATA:
2760 case EXT4_MOUNT_WRITEBACK_DATA:
2761 if (!jbd2_journal_check_available_features
2762 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2763 printk(KERN_ERR "EXT4-fs: Journal does not support "
2764 "requested data journaling mode\n");
2770 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2774 if (test_opt(sb, NOBH)) {
2775 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2776 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2777 "its supported only with writeback mode\n");
2778 clear_opt(sbi->s_mount_opt, NOBH);
2782 * The jbd2_journal_load will have done any necessary log recovery,
2783 * so we can safely mount the rest of the filesystem now.
2786 root = ext4_iget(sb, EXT4_ROOT_INO);
2788 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2789 ret = PTR_ERR(root);
2792 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2794 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2797 sb->s_root = d_alloc_root(root);
2799 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2805 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2807 /* determine the minimum size of new large inodes, if present */
2808 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2809 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2810 EXT4_GOOD_OLD_INODE_SIZE;
2811 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2812 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2813 if (sbi->s_want_extra_isize <
2814 le16_to_cpu(es->s_want_extra_isize))
2815 sbi->s_want_extra_isize =
2816 le16_to_cpu(es->s_want_extra_isize);
2817 if (sbi->s_want_extra_isize <
2818 le16_to_cpu(es->s_min_extra_isize))
2819 sbi->s_want_extra_isize =
2820 le16_to_cpu(es->s_min_extra_isize);
2823 /* Check if enough inode space is available */
2824 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2825 sbi->s_inode_size) {
2826 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2827 EXT4_GOOD_OLD_INODE_SIZE;
2828 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2832 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2833 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2834 "requested data journaling mode\n");
2835 clear_opt(sbi->s_mount_opt, DELALLOC);
2836 } else if (test_opt(sb, DELALLOC))
2837 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2839 err = ext4_setup_system_zone(sb);
2841 printk(KERN_ERR "EXT4-fs: failed to initialize system "
2842 "zone (%d)\n", err);
2847 err = ext4_mb_init(sb, needs_recovery);
2849 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2854 sbi->s_kobj.kset = ext4_kset;
2855 init_completion(&sbi->s_kobj_unregister);
2856 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2859 ext4_mb_release(sb);
2860 ext4_ext_release(sb);
2864 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2865 ext4_orphan_cleanup(sb, es);
2866 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2867 if (needs_recovery) {
2868 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2869 ext4_mark_recovery_complete(sb, es);
2871 if (EXT4_SB(sb)->s_journal) {
2872 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2873 descr = " journalled data mode";
2874 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2875 descr = " ordered data mode";
2877 descr = " writeback data mode";
2879 descr = "out journal";
2881 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2889 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2894 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2895 ext4_release_system_zone(sb);
2896 if (sbi->s_journal) {
2897 jbd2_journal_destroy(sbi->s_journal);
2898 sbi->s_journal = NULL;
2901 if (sbi->s_flex_groups) {
2902 if (is_vmalloc_addr(sbi->s_flex_groups))
2903 vfree(sbi->s_flex_groups);
2905 kfree(sbi->s_flex_groups);
2907 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2908 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2909 percpu_counter_destroy(&sbi->s_dirs_counter);
2910 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2912 for (i = 0; i < db_count; i++)
2913 brelse(sbi->s_group_desc[i]);
2914 kfree(sbi->s_group_desc);
2917 remove_proc_entry(sb->s_id, ext4_proc_root);
2920 for (i = 0; i < MAXQUOTAS; i++)
2921 kfree(sbi->s_qf_names[i]);
2923 ext4_blkdev_remove(sbi);
2926 sb->s_fs_info = NULL;
2933 * Setup any per-fs journal parameters now. We'll do this both on
2934 * initial mount, once the journal has been initialised but before we've
2935 * done any recovery; and again on any subsequent remount.
2937 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2939 struct ext4_sb_info *sbi = EXT4_SB(sb);
2941 journal->j_commit_interval = sbi->s_commit_interval;
2942 journal->j_min_batch_time = sbi->s_min_batch_time;
2943 journal->j_max_batch_time = sbi->s_max_batch_time;
2945 spin_lock(&journal->j_state_lock);
2946 if (test_opt(sb, BARRIER))
2947 journal->j_flags |= JBD2_BARRIER;
2949 journal->j_flags &= ~JBD2_BARRIER;
2950 if (test_opt(sb, DATA_ERR_ABORT))
2951 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2953 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2954 spin_unlock(&journal->j_state_lock);
2957 static journal_t *ext4_get_journal(struct super_block *sb,
2958 unsigned int journal_inum)
2960 struct inode *journal_inode;
2963 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2965 /* First, test for the existence of a valid inode on disk. Bad
2966 * things happen if we iget() an unused inode, as the subsequent
2967 * iput() will try to delete it. */
2969 journal_inode = ext4_iget(sb, journal_inum);
2970 if (IS_ERR(journal_inode)) {
2971 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2974 if (!journal_inode->i_nlink) {
2975 make_bad_inode(journal_inode);
2976 iput(journal_inode);
2977 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2981 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2982 journal_inode, journal_inode->i_size);
2983 if (!S_ISREG(journal_inode->i_mode)) {
2984 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2985 iput(journal_inode);
2989 journal = jbd2_journal_init_inode(journal_inode);
2991 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2992 iput(journal_inode);
2995 journal->j_private = sb;
2996 ext4_init_journal_params(sb, journal);
3000 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3003 struct buffer_head *bh;
3007 int hblock, blocksize;
3008 ext4_fsblk_t sb_block;
3009 unsigned long offset;
3010 struct ext4_super_block *es;
3011 struct block_device *bdev;
3013 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3015 bdev = ext4_blkdev_get(j_dev);
3019 if (bd_claim(bdev, sb)) {
3021 "EXT4-fs: failed to claim external journal device.\n");
3022 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3026 blocksize = sb->s_blocksize;
3027 hblock = bdev_hardsect_size(bdev);
3028 if (blocksize < hblock) {
3030 "EXT4-fs: blocksize too small for journal device.\n");
3034 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3035 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3036 set_blocksize(bdev, blocksize);
3037 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3038 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
3039 "external journal\n");
3043 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3044 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3045 !(le32_to_cpu(es->s_feature_incompat) &
3046 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3047 printk(KERN_ERR "EXT4-fs: external journal has "
3048 "bad superblock\n");
3053 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3054 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
3059 len = ext4_blocks_count(es);
3060 start = sb_block + 1;
3061 brelse(bh); /* we're done with the superblock */
3063 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3064 start, len, blocksize);
3066 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
3069 journal->j_private = sb;
3070 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3071 wait_on_buffer(journal->j_sb_buffer);
3072 if (!buffer_uptodate(journal->j_sb_buffer)) {
3073 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
3076 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3077 printk(KERN_ERR "EXT4-fs: External journal has more than one "
3078 "user (unsupported) - %d\n",
3079 be32_to_cpu(journal->j_superblock->s_nr_users));
3082 EXT4_SB(sb)->journal_bdev = bdev;
3083 ext4_init_journal_params(sb, journal);
3086 jbd2_journal_destroy(journal);
3088 ext4_blkdev_put(bdev);
3092 static int ext4_load_journal(struct super_block *sb,
3093 struct ext4_super_block *es,
3094 unsigned long journal_devnum)
3097 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3100 int really_read_only;
3102 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3104 if (journal_devnum &&
3105 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3106 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
3107 "numbers have changed\n");
3108 journal_dev = new_decode_dev(journal_devnum);
3110 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3112 really_read_only = bdev_read_only(sb->s_bdev);
3115 * Are we loading a blank journal or performing recovery after a
3116 * crash? For recovery, we need to check in advance whether we
3117 * can get read-write access to the device.
3120 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3121 if (sb->s_flags & MS_RDONLY) {
3122 printk(KERN_INFO "EXT4-fs: INFO: recovery "
3123 "required on readonly filesystem.\n");
3124 if (really_read_only) {
3125 printk(KERN_ERR "EXT4-fs: write access "
3126 "unavailable, cannot proceed.\n");
3129 printk(KERN_INFO "EXT4-fs: write access will "
3130 "be enabled during recovery.\n");
3134 if (journal_inum && journal_dev) {
3135 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
3136 "and inode journals!\n");
3141 if (!(journal = ext4_get_journal(sb, journal_inum)))
3144 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3148 if (journal->j_flags & JBD2_BARRIER)
3149 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
3151 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
3153 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3154 err = jbd2_journal_update_format(journal);
3156 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
3157 jbd2_journal_destroy(journal);
3162 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3163 err = jbd2_journal_wipe(journal, !really_read_only);
3165 err = jbd2_journal_load(journal);
3168 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
3169 jbd2_journal_destroy(journal);
3173 EXT4_SB(sb)->s_journal = journal;
3174 ext4_clear_journal_err(sb, es);
3176 if (journal_devnum &&
3177 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3178 es->s_journal_dev = cpu_to_le32(journal_devnum);
3180 /* Make sure we flush the recovery flag to disk. */
3181 ext4_commit_super(sb, 1);
3187 static int ext4_commit_super(struct super_block *sb, int sync)
3189 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3190 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3195 if (buffer_write_io_error(sbh)) {
3197 * Oh, dear. A previous attempt to write the
3198 * superblock failed. This could happen because the
3199 * USB device was yanked out. Or it could happen to
3200 * be a transient write error and maybe the block will
3201 * be remapped. Nothing we can do but to retry the
3202 * write and hope for the best.
3204 printk(KERN_ERR "EXT4-fs: previous I/O error to "
3205 "superblock detected for %s.\n", sb->s_id);
3206 clear_buffer_write_io_error(sbh);
3207 set_buffer_uptodate(sbh);
3209 es->s_wtime = cpu_to_le32(get_seconds());
3210 es->s_kbytes_written =
3211 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3212 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3213 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3214 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3215 &EXT4_SB(sb)->s_freeblocks_counter));
3216 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3217 &EXT4_SB(sb)->s_freeinodes_counter));
3219 BUFFER_TRACE(sbh, "marking dirty");
3220 mark_buffer_dirty(sbh);
3222 error = sync_dirty_buffer(sbh);
3226 error = buffer_write_io_error(sbh);
3228 printk(KERN_ERR "EXT4-fs: I/O error while writing "
3229 "superblock for %s.\n", sb->s_id);
3230 clear_buffer_write_io_error(sbh);
3231 set_buffer_uptodate(sbh);
3239 * Have we just finished recovery? If so, and if we are mounting (or
3240 * remounting) the filesystem readonly, then we will end up with a
3241 * consistent fs on disk. Record that fact.
3243 static void ext4_mark_recovery_complete(struct super_block *sb,
3244 struct ext4_super_block *es)
3246 journal_t *journal = EXT4_SB(sb)->s_journal;
3248 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3249 BUG_ON(journal != NULL);
3252 jbd2_journal_lock_updates(journal);
3253 if (jbd2_journal_flush(journal) < 0)
3256 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3257 sb->s_flags & MS_RDONLY) {
3258 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3259 ext4_commit_super(sb, 1);
3263 jbd2_journal_unlock_updates(journal);
3267 * If we are mounting (or read-write remounting) a filesystem whose journal
3268 * has recorded an error from a previous lifetime, move that error to the
3269 * main filesystem now.
3271 static void ext4_clear_journal_err(struct super_block *sb,
3272 struct ext4_super_block *es)
3278 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3280 journal = EXT4_SB(sb)->s_journal;
3283 * Now check for any error status which may have been recorded in the
3284 * journal by a prior ext4_error() or ext4_abort()
3287 j_errno = jbd2_journal_errno(journal);
3291 errstr = ext4_decode_error(sb, j_errno, nbuf);
3292 ext4_warning(sb, __func__, "Filesystem error recorded "
3293 "from previous mount: %s", errstr);
3294 ext4_warning(sb, __func__, "Marking fs in need of "
3295 "filesystem check.");
3297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3299 ext4_commit_super(sb, 1);
3301 jbd2_journal_clear_err(journal);
3306 * Force the running and committing transactions to commit,
3307 * and wait on the commit.
3309 int ext4_force_commit(struct super_block *sb)
3314 if (sb->s_flags & MS_RDONLY)
3317 journal = EXT4_SB(sb)->s_journal;
3319 ret = ext4_journal_force_commit(journal);
3324 static void ext4_write_super(struct super_block *sb)
3326 ext4_commit_super(sb, 1);
3329 static int ext4_sync_fs(struct super_block *sb, int wait)
3334 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
3335 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3337 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3343 * LVM calls this function before a (read-only) snapshot is created. This
3344 * gives us a chance to flush the journal completely and mark the fs clean.
3346 static int ext4_freeze(struct super_block *sb)
3351 if (sb->s_flags & MS_RDONLY)
3354 journal = EXT4_SB(sb)->s_journal;
3356 /* Now we set up the journal barrier. */
3357 jbd2_journal_lock_updates(journal);
3360 * Don't clear the needs_recovery flag if we failed to flush
3363 error = jbd2_journal_flush(journal);
3366 jbd2_journal_unlock_updates(journal);
3370 /* Journal blocked and flushed, clear needs_recovery flag. */
3371 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3372 error = ext4_commit_super(sb, 1);
3379 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3380 * flag here, even though the filesystem is not technically dirty yet.
3382 static int ext4_unfreeze(struct super_block *sb)
3384 if (sb->s_flags & MS_RDONLY)
3388 /* Reset the needs_recovery flag before the fs is unlocked. */
3389 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3390 ext4_commit_super(sb, 1);
3392 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3396 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3398 struct ext4_super_block *es;
3399 struct ext4_sb_info *sbi = EXT4_SB(sb);
3400 ext4_fsblk_t n_blocks_count = 0;
3401 unsigned long old_sb_flags;
3402 struct ext4_mount_options old_opts;
3404 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3410 /* Store the original options */
3411 old_sb_flags = sb->s_flags;
3412 old_opts.s_mount_opt = sbi->s_mount_opt;
3413 old_opts.s_resuid = sbi->s_resuid;
3414 old_opts.s_resgid = sbi->s_resgid;
3415 old_opts.s_commit_interval = sbi->s_commit_interval;
3416 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3417 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3419 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3420 for (i = 0; i < MAXQUOTAS; i++)
3421 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3423 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3424 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3427 * Allow the "check" option to be passed as a remount option.
3429 if (!parse_options(data, sb, NULL, &journal_ioprio,
3430 &n_blocks_count, 1)) {
3435 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3436 ext4_abort(sb, __func__, "Abort forced by user");
3438 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3439 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3443 if (sbi->s_journal) {
3444 ext4_init_journal_params(sb, sbi->s_journal);
3445 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3448 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3449 n_blocks_count > ext4_blocks_count(es)) {
3450 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3455 if (*flags & MS_RDONLY) {
3457 * First of all, the unconditional stuff we have to do
3458 * to disable replay of the journal when we next remount
3460 sb->s_flags |= MS_RDONLY;
3463 * OK, test if we are remounting a valid rw partition
3464 * readonly, and if so set the rdonly flag and then
3465 * mark the partition as valid again.
3467 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3468 (sbi->s_mount_state & EXT4_VALID_FS))
3469 es->s_state = cpu_to_le16(sbi->s_mount_state);
3472 ext4_mark_recovery_complete(sb, es);
3475 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3476 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3477 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3478 "remount RDWR because of unsupported "
3479 "optional features (%x).\n", sb->s_id,
3480 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3481 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3487 * Make sure the group descriptor checksums
3488 * are sane. If they aren't, refuse to
3491 for (g = 0; g < sbi->s_groups_count; g++) {
3492 struct ext4_group_desc *gdp =
3493 ext4_get_group_desc(sb, g, NULL);
3495 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3497 "EXT4-fs: ext4_remount: "
3498 "Checksum for group %u failed (%u!=%u)\n",
3499 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3500 le16_to_cpu(gdp->bg_checksum));
3507 * If we have an unprocessed orphan list hanging
3508 * around from a previously readonly bdev mount,
3509 * require a full umount/remount for now.
3511 if (es->s_last_orphan) {
3512 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3513 "remount RDWR because of unprocessed "
3514 "orphan inode list. Please "
3515 "umount/remount instead.\n",
3522 * Mounting a RDONLY partition read-write, so reread
3523 * and store the current valid flag. (It may have
3524 * been changed by e2fsck since we originally mounted
3528 ext4_clear_journal_err(sb, es);
3529 sbi->s_mount_state = le16_to_cpu(es->s_state);
3530 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3532 if (!ext4_setup_super(sb, es, 0))
3533 sb->s_flags &= ~MS_RDONLY;
3536 ext4_setup_system_zone(sb);
3537 if (sbi->s_journal == NULL)
3538 ext4_commit_super(sb, 1);
3541 /* Release old quota file names */
3542 for (i = 0; i < MAXQUOTAS; i++)
3543 if (old_opts.s_qf_names[i] &&
3544 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3545 kfree(old_opts.s_qf_names[i]);
3549 sb->s_flags = old_sb_flags;
3550 sbi->s_mount_opt = old_opts.s_mount_opt;
3551 sbi->s_resuid = old_opts.s_resuid;
3552 sbi->s_resgid = old_opts.s_resgid;
3553 sbi->s_commit_interval = old_opts.s_commit_interval;
3554 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3555 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3557 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3558 for (i = 0; i < MAXQUOTAS; i++) {
3559 if (sbi->s_qf_names[i] &&
3560 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3561 kfree(sbi->s_qf_names[i]);
3562 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3568 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3570 struct super_block *sb = dentry->d_sb;
3571 struct ext4_sb_info *sbi = EXT4_SB(sb);
3572 struct ext4_super_block *es = sbi->s_es;
3575 if (test_opt(sb, MINIX_DF)) {
3576 sbi->s_overhead_last = 0;
3577 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3578 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3579 ext4_fsblk_t overhead = 0;
3582 * Compute the overhead (FS structures). This is constant
3583 * for a given filesystem unless the number of block groups
3584 * changes so we cache the previous value until it does.
3588 * All of the blocks before first_data_block are
3591 overhead = le32_to_cpu(es->s_first_data_block);
3594 * Add the overhead attributed to the superblock and
3595 * block group descriptors. If the sparse superblocks
3596 * feature is turned on, then not all groups have this.
3598 for (i = 0; i < ngroups; i++) {
3599 overhead += ext4_bg_has_super(sb, i) +
3600 ext4_bg_num_gdb(sb, i);
3605 * Every block group has an inode bitmap, a block
3606 * bitmap, and an inode table.
3608 overhead += ngroups * (2 + sbi->s_itb_per_group);
3609 sbi->s_overhead_last = overhead;
3611 sbi->s_blocks_last = ext4_blocks_count(es);
3614 buf->f_type = EXT4_SUPER_MAGIC;
3615 buf->f_bsize = sb->s_blocksize;
3616 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3617 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3618 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3619 ext4_free_blocks_count_set(es, buf->f_bfree);
3620 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3621 if (buf->f_bfree < ext4_r_blocks_count(es))
3623 buf->f_files = le32_to_cpu(es->s_inodes_count);
3624 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3625 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3626 buf->f_namelen = EXT4_NAME_LEN;
3627 fsid = le64_to_cpup((void *)es->s_uuid) ^
3628 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3629 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3630 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3634 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3635 * is locked for write. Otherwise the are possible deadlocks:
3636 * Process 1 Process 2
3637 * ext4_create() quota_sync()
3638 * jbd2_journal_start() write_dquot()
3639 * vfs_dq_init() down(dqio_mutex)
3640 * down(dqio_mutex) jbd2_journal_start()
3646 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3648 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3651 static int ext4_write_dquot(struct dquot *dquot)
3655 struct inode *inode;
3657 inode = dquot_to_inode(dquot);
3658 handle = ext4_journal_start(inode,
3659 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3661 return PTR_ERR(handle);
3662 ret = dquot_commit(dquot);
3663 err = ext4_journal_stop(handle);
3669 static int ext4_acquire_dquot(struct dquot *dquot)
3674 handle = ext4_journal_start(dquot_to_inode(dquot),
3675 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3677 return PTR_ERR(handle);
3678 ret = dquot_acquire(dquot);
3679 err = ext4_journal_stop(handle);
3685 static int ext4_release_dquot(struct dquot *dquot)
3690 handle = ext4_journal_start(dquot_to_inode(dquot),
3691 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3692 if (IS_ERR(handle)) {
3693 /* Release dquot anyway to avoid endless cycle in dqput() */
3694 dquot_release(dquot);
3695 return PTR_ERR(handle);
3697 ret = dquot_release(dquot);
3698 err = ext4_journal_stop(handle);
3704 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3706 /* Are we journaling quotas? */
3707 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3708 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3709 dquot_mark_dquot_dirty(dquot);
3710 return ext4_write_dquot(dquot);
3712 return dquot_mark_dquot_dirty(dquot);
3716 static int ext4_write_info(struct super_block *sb, int type)
3721 /* Data block + inode block */
3722 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3724 return PTR_ERR(handle);
3725 ret = dquot_commit_info(sb, type);
3726 err = ext4_journal_stop(handle);
3733 * Turn on quotas during mount time - we need to find
3734 * the quota file and such...
3736 static int ext4_quota_on_mount(struct super_block *sb, int type)
3738 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3739 EXT4_SB(sb)->s_jquota_fmt, type);
3743 * Standard function to be called on quota_on
3745 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3746 char *name, int remount)
3751 if (!test_opt(sb, QUOTA))
3753 /* When remounting, no checks are needed and in fact, name is NULL */
3755 return vfs_quota_on(sb, type, format_id, name, remount);
3757 err = kern_path(name, LOOKUP_FOLLOW, &path);
3761 /* Quotafile not on the same filesystem? */
3762 if (path.mnt->mnt_sb != sb) {
3766 /* Journaling quota? */
3767 if (EXT4_SB(sb)->s_qf_names[type]) {
3768 /* Quotafile not in fs root? */
3769 if (path.dentry->d_parent != sb->s_root)
3771 "EXT4-fs: Quota file not on filesystem root. "
3772 "Journaled quota will not work.\n");
3776 * When we journal data on quota file, we have to flush journal to see
3777 * all updates to the file when we bypass pagecache...
3779 if (EXT4_SB(sb)->s_journal &&
3780 ext4_should_journal_data(path.dentry->d_inode)) {
3782 * We don't need to lock updates but journal_flush() could
3783 * otherwise be livelocked...
3785 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3786 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3787 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3794 err = vfs_quota_on_path(sb, type, format_id, &path);
3799 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3800 * acquiring the locks... As quota files are never truncated and quota code
3801 * itself serializes the operations (and noone else should touch the files)
3802 * we don't have to be afraid of races */
3803 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3804 size_t len, loff_t off)
3806 struct inode *inode = sb_dqopt(sb)->files[type];
3807 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3809 int offset = off & (sb->s_blocksize - 1);
3812 struct buffer_head *bh;
3813 loff_t i_size = i_size_read(inode);
3817 if (off+len > i_size)
3820 while (toread > 0) {
3821 tocopy = sb->s_blocksize - offset < toread ?
3822 sb->s_blocksize - offset : toread;
3823 bh = ext4_bread(NULL, inode, blk, 0, &err);
3826 if (!bh) /* A hole? */
3827 memset(data, 0, tocopy);
3829 memcpy(data, bh->b_data+offset, tocopy);
3839 /* Write to quotafile (we know the transaction is already started and has
3840 * enough credits) */
3841 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3842 const char *data, size_t len, loff_t off)
3844 struct inode *inode = sb_dqopt(sb)->files[type];
3845 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3847 int offset = off & (sb->s_blocksize - 1);
3849 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3850 size_t towrite = len;
3851 struct buffer_head *bh;
3852 handle_t *handle = journal_current_handle();
3854 if (EXT4_SB(sb)->s_journal && !handle) {
3855 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3856 " cancelled because transaction is not started.\n",
3857 (unsigned long long)off, (unsigned long long)len);
3860 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3861 while (towrite > 0) {
3862 tocopy = sb->s_blocksize - offset < towrite ?
3863 sb->s_blocksize - offset : towrite;
3864 bh = ext4_bread(handle, inode, blk, 1, &err);
3867 if (journal_quota) {
3868 err = ext4_journal_get_write_access(handle, bh);
3875 memcpy(bh->b_data+offset, data, tocopy);
3876 flush_dcache_page(bh->b_page);
3879 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3881 /* Always do at least ordered writes for quotas */
3882 err = ext4_jbd2_file_inode(handle, inode);
3883 mark_buffer_dirty(bh);
3894 if (len == towrite) {
3895 mutex_unlock(&inode->i_mutex);
3898 if (inode->i_size < off+len-towrite) {
3899 i_size_write(inode, off+len-towrite);
3900 EXT4_I(inode)->i_disksize = inode->i_size;
3902 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3903 ext4_mark_inode_dirty(handle, inode);
3904 mutex_unlock(&inode->i_mutex);
3905 return len - towrite;
3910 static int ext4_get_sb(struct file_system_type *fs_type,
3911 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3913 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3916 static struct file_system_type ext4_fs_type = {
3917 .owner = THIS_MODULE,
3919 .get_sb = ext4_get_sb,
3920 .kill_sb = kill_block_super,
3921 .fs_flags = FS_REQUIRES_DEV,
3924 #ifdef CONFIG_EXT4DEV_COMPAT
3925 static int ext4dev_get_sb(struct file_system_type *fs_type,
3926 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3928 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3929 "to mount using ext4\n");
3930 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3931 "will go away by 2.6.31\n");
3932 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3935 static struct file_system_type ext4dev_fs_type = {
3936 .owner = THIS_MODULE,
3938 .get_sb = ext4dev_get_sb,
3939 .kill_sb = kill_block_super,
3940 .fs_flags = FS_REQUIRES_DEV,
3942 MODULE_ALIAS("ext4dev");
3945 static int __init init_ext4_fs(void)
3949 err = init_ext4_system_zone();
3952 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3955 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3956 err = init_ext4_mballoc();
3960 err = init_ext4_xattr();
3963 err = init_inodecache();
3966 err = register_filesystem(&ext4_fs_type);
3969 #ifdef CONFIG_EXT4DEV_COMPAT
3970 err = register_filesystem(&ext4dev_fs_type);
3972 unregister_filesystem(&ext4_fs_type);
3978 destroy_inodecache();
3982 exit_ext4_mballoc();
3984 remove_proc_entry("fs/ext4", NULL);
3985 kset_unregister(ext4_kset);
3987 exit_ext4_system_zone();
3991 static void __exit exit_ext4_fs(void)
3993 unregister_filesystem(&ext4_fs_type);
3994 #ifdef CONFIG_EXT4DEV_COMPAT
3995 unregister_filesystem(&ext4dev_fs_type);
3997 destroy_inodecache();
3999 exit_ext4_mballoc();
4000 remove_proc_entry("fs/ext4", NULL);
4001 kset_unregister(ext4_kset);
4002 exit_ext4_system_zone();
4005 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4006 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4007 MODULE_LICENSE("GPL");
4008 module_init(init_ext4_fs)
4009 module_exit(exit_ext4_fs)