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/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/ext4.h>
52 static int default_mb_history_length = 1000;
54 module_param_named(default_mb_history_length, default_mb_history_length,
56 MODULE_PARM_DESC(default_mb_history_length,
57 "Default number of entries saved for mb_history");
59 struct proc_dir_entry *ext4_proc_root;
60 static struct kset *ext4_kset;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static void ext4_write_super(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
79 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
80 struct ext4_group_desc *bg)
82 return le32_to_cpu(bg->bg_block_bitmap_lo) |
83 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
84 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
87 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
95 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_inode_table_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
103 __u32 ext4_free_blks_count(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
111 __u32 ext4_free_inodes_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
119 __u32 ext4_used_dirs_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
127 __u32 ext4_itable_unused_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_itable_unused_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
135 void ext4_block_bitmap_set(struct super_block *sb,
136 struct ext4_group_desc *bg, ext4_fsblk_t blk)
138 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
139 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
140 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
143 void ext4_inode_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
151 void ext4_inode_table_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
159 void ext4_free_blks_set(struct super_block *sb,
160 struct ext4_group_desc *bg, __u32 count)
162 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
167 void ext4_free_inodes_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
170 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
175 void ext4_used_dirs_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
183 void ext4_itable_unused_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
192 * Wrappers for jbd2_journal_start/end.
194 * The only special thing we need to do here is to make sure that all
195 * journal_end calls result in the superblock being marked dirty, so
196 * that sync() will call the filesystem's write_super callback if
199 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
203 if (sb->s_flags & MS_RDONLY)
204 return ERR_PTR(-EROFS);
206 /* Special case here: if the journal has aborted behind our
207 * backs (eg. EIO in the commit thread), then we still need to
208 * take the FS itself readonly cleanly. */
209 journal = EXT4_SB(sb)->s_journal;
211 if (is_journal_aborted(journal)) {
212 ext4_abort(sb, __func__, "Detected aborted journal");
213 return ERR_PTR(-EROFS);
215 return jbd2_journal_start(journal, nblocks);
218 * We're not journaling, return the appropriate indication.
220 current->journal_info = EXT4_NOJOURNAL_HANDLE;
221 return current->journal_info;
225 * The only special thing we need to do here is to make sure that all
226 * jbd2_journal_stop calls result in the superblock being marked dirty, so
227 * that sync() will call the filesystem's write_super callback if
230 int __ext4_journal_stop(const char *where, handle_t *handle)
232 struct super_block *sb;
236 if (!ext4_handle_valid(handle)) {
238 * Do this here since we don't call jbd2_journal_stop() in
241 current->journal_info = NULL;
244 sb = handle->h_transaction->t_journal->j_private;
246 rc = jbd2_journal_stop(handle);
251 __ext4_std_error(sb, where, err);
255 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
256 struct buffer_head *bh, handle_t *handle, int err)
259 const char *errstr = ext4_decode_error(NULL, err, nbuf);
261 BUG_ON(!ext4_handle_valid(handle));
264 BUFFER_TRACE(bh, "abort");
269 if (is_handle_aborted(handle))
272 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
273 caller, errstr, err_fn);
275 jbd2_journal_abort_handle(handle);
278 /* Deal with the reporting of failure conditions on a filesystem such as
279 * inconsistencies detected or read IO failures.
281 * On ext2, we can store the error state of the filesystem in the
282 * superblock. That is not possible on ext4, because we may have other
283 * write ordering constraints on the superblock which prevent us from
284 * writing it out straight away; and given that the journal is about to
285 * be aborted, we can't rely on the current, or future, transactions to
286 * write out the superblock safely.
288 * We'll just use the jbd2_journal_abort() error code to record an error in
289 * the journal instead. On recovery, the journal will compain about
290 * that error until we've noted it down and cleared it.
293 static void ext4_handle_error(struct super_block *sb)
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 if (sb->s_flags & MS_RDONLY)
303 if (!test_opt(sb, ERRORS_CONT)) {
304 journal_t *journal = EXT4_SB(sb)->s_journal;
306 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
308 jbd2_journal_abort(journal, -EIO);
310 if (test_opt(sb, ERRORS_RO)) {
311 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
312 sb->s_flags |= MS_RDONLY;
314 ext4_commit_super(sb, 1);
315 if (test_opt(sb, ERRORS_PANIC))
316 panic("EXT4-fs (device %s): panic forced after error\n",
320 void ext4_error(struct super_block *sb, const char *function,
321 const char *fmt, ...)
326 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
331 ext4_handle_error(sb);
334 static const char *ext4_decode_error(struct super_block *sb, int errno,
341 errstr = "IO failure";
344 errstr = "Out of memory";
347 if (!sb || (EXT4_SB(sb)->s_journal &&
348 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
349 errstr = "Journal has aborted";
351 errstr = "Readonly filesystem";
354 /* If the caller passed in an extra buffer for unknown
355 * errors, textualise them now. Else we just return
358 /* Check for truncated error codes... */
359 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
368 /* __ext4_std_error decodes expected errors from journaling functions
369 * automatically and invokes the appropriate error response. */
371 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
376 /* Special case: if the error is EROFS, and we're not already
377 * inside a transaction, then there's really no point in logging
379 if (errno == -EROFS && journal_current_handle() == NULL &&
380 (sb->s_flags & MS_RDONLY))
383 errstr = ext4_decode_error(sb, errno, nbuf);
384 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
385 sb->s_id, function, errstr);
387 ext4_handle_error(sb);
391 * ext4_abort is a much stronger failure handler than ext4_error. The
392 * abort function may be used to deal with unrecoverable failures such
393 * as journal IO errors or ENOMEM at a critical moment in log management.
395 * We unconditionally force the filesystem into an ABORT|READONLY state,
396 * unless the error response on the fs has been set to panic in which
397 * case we take the easy way out and panic immediately.
400 void ext4_abort(struct super_block *sb, const char *function,
401 const char *fmt, ...)
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 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
418 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
419 sb->s_flags |= MS_RDONLY;
420 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
421 if (EXT4_SB(sb)->s_journal)
422 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
425 void ext4_msg (struct super_block * sb, const char *prefix,
426 const char *fmt, ...)
431 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
437 void ext4_warning(struct super_block *sb, const char *function,
438 const char *fmt, ...)
443 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
450 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
451 const char *function, const char *fmt, ...)
456 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
459 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
464 if (test_opt(sb, ERRORS_CONT)) {
465 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
466 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
467 ext4_commit_super(sb, 0);
470 ext4_unlock_group(sb, grp);
471 ext4_handle_error(sb);
473 * We only get here in the ERRORS_RO case; relocking the group
474 * may be dangerous, but nothing bad will happen since the
475 * filesystem will have already been marked read/only and the
476 * journal has been aborted. We return 1 as a hint to callers
477 * who might what to use the return value from
478 * ext4_grp_locked_error() to distinguish beween the
479 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
480 * aggressively from the ext4 function in question, with a
481 * more appropriate error code.
483 ext4_lock_group(sb, grp);
487 void ext4_update_dynamic_rev(struct super_block *sb)
489 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
491 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
494 ext4_warning(sb, __func__,
495 "updating to rev %d because of new feature flag, "
496 "running e2fsck is recommended",
499 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
500 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
501 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
502 /* leave es->s_feature_*compat flags alone */
503 /* es->s_uuid will be set by e2fsck if empty */
506 * The rest of the superblock fields should be zero, and if not it
507 * means they are likely already in use, so leave them alone. We
508 * can leave it up to e2fsck to clean up any inconsistencies there.
513 * Open the external journal device
515 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
517 struct block_device *bdev;
518 char b[BDEVNAME_SIZE];
520 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
526 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
527 __bdevname(dev, b), PTR_ERR(bdev));
532 * Release the journal device
534 static int ext4_blkdev_put(struct block_device *bdev)
537 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
540 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
542 struct block_device *bdev;
545 bdev = sbi->journal_bdev;
547 ret = ext4_blkdev_put(bdev);
548 sbi->journal_bdev = NULL;
553 static inline struct inode *orphan_list_entry(struct list_head *l)
555 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
558 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
562 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
563 le32_to_cpu(sbi->s_es->s_last_orphan));
565 printk(KERN_ERR "sb_info orphan list:\n");
566 list_for_each(l, &sbi->s_orphan) {
567 struct inode *inode = orphan_list_entry(l);
569 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
570 inode->i_sb->s_id, inode->i_ino, inode,
571 inode->i_mode, inode->i_nlink,
576 static void ext4_put_super(struct super_block *sb)
578 struct ext4_sb_info *sbi = EXT4_SB(sb);
579 struct ext4_super_block *es = sbi->s_es;
585 ext4_commit_super(sb, 1);
587 ext4_release_system_zone(sb);
589 ext4_ext_release(sb);
590 ext4_xattr_put_super(sb);
591 if (sbi->s_journal) {
592 err = jbd2_journal_destroy(sbi->s_journal);
593 sbi->s_journal = NULL;
595 ext4_abort(sb, __func__,
596 "Couldn't clean up the journal");
598 if (!(sb->s_flags & MS_RDONLY)) {
599 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
600 es->s_state = cpu_to_le16(sbi->s_mount_state);
601 ext4_commit_super(sb, 1);
604 remove_proc_entry(sb->s_id, ext4_proc_root);
606 kobject_del(&sbi->s_kobj);
608 for (i = 0; i < sbi->s_gdb_count; i++)
609 brelse(sbi->s_group_desc[i]);
610 kfree(sbi->s_group_desc);
611 if (is_vmalloc_addr(sbi->s_flex_groups))
612 vfree(sbi->s_flex_groups);
614 kfree(sbi->s_flex_groups);
615 percpu_counter_destroy(&sbi->s_freeblocks_counter);
616 percpu_counter_destroy(&sbi->s_freeinodes_counter);
617 percpu_counter_destroy(&sbi->s_dirs_counter);
618 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
621 for (i = 0; i < MAXQUOTAS; i++)
622 kfree(sbi->s_qf_names[i]);
625 /* Debugging code just in case the in-memory inode orphan list
626 * isn't empty. The on-disk one can be non-empty if we've
627 * detected an error and taken the fs readonly, but the
628 * in-memory list had better be clean by this point. */
629 if (!list_empty(&sbi->s_orphan))
630 dump_orphan_list(sb, sbi);
631 J_ASSERT(list_empty(&sbi->s_orphan));
633 invalidate_bdev(sb->s_bdev);
634 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
636 * Invalidate the journal device's buffers. We don't want them
637 * floating about in memory - the physical journal device may
638 * hotswapped, and it breaks the `ro-after' testing code.
640 sync_blockdev(sbi->journal_bdev);
641 invalidate_bdev(sbi->journal_bdev);
642 ext4_blkdev_remove(sbi);
644 sb->s_fs_info = NULL;
646 * Now that we are completely done shutting down the
647 * superblock, we need to actually destroy the kobject.
651 kobject_put(&sbi->s_kobj);
652 wait_for_completion(&sbi->s_kobj_unregister);
653 kfree(sbi->s_blockgroup_lock);
657 static struct kmem_cache *ext4_inode_cachep;
660 * Called inside transaction, so use GFP_NOFS
662 static struct inode *ext4_alloc_inode(struct super_block *sb)
664 struct ext4_inode_info *ei;
666 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
670 ei->vfs_inode.i_version = 1;
671 ei->vfs_inode.i_data.writeback_index = 0;
672 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
673 INIT_LIST_HEAD(&ei->i_prealloc_list);
674 spin_lock_init(&ei->i_prealloc_lock);
676 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
677 * therefore it can be null here. Don't check it, just initialize
680 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
681 ei->i_reserved_data_blocks = 0;
682 ei->i_reserved_meta_blocks = 0;
683 ei->i_allocated_meta_blocks = 0;
684 ei->i_delalloc_reserved_flag = 0;
685 spin_lock_init(&(ei->i_block_reservation_lock));
687 return &ei->vfs_inode;
690 static void ext4_destroy_inode(struct inode *inode)
692 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
693 ext4_msg(inode->i_sb, KERN_ERR,
694 "Inode %lu (%p): orphan list check failed!",
695 inode->i_ino, EXT4_I(inode));
696 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
697 EXT4_I(inode), sizeof(struct ext4_inode_info),
701 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
704 static void init_once(void *foo)
706 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
708 INIT_LIST_HEAD(&ei->i_orphan);
709 #ifdef CONFIG_EXT4_FS_XATTR
710 init_rwsem(&ei->xattr_sem);
712 init_rwsem(&ei->i_data_sem);
713 inode_init_once(&ei->vfs_inode);
716 static int init_inodecache(void)
718 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
719 sizeof(struct ext4_inode_info),
720 0, (SLAB_RECLAIM_ACCOUNT|
723 if (ext4_inode_cachep == NULL)
728 static void destroy_inodecache(void)
730 kmem_cache_destroy(ext4_inode_cachep);
733 static void ext4_clear_inode(struct inode *inode)
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,
938 journal_t *journal = EXT4_SB(sb)->s_journal;
940 WARN_ON(PageChecked(page));
941 if (!page_has_buffers(page))
944 return jbd2_journal_try_to_free_buffers(journal, page,
946 return try_to_free_buffers(page);
950 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
951 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
953 static int ext4_write_dquot(struct dquot *dquot);
954 static int ext4_acquire_dquot(struct dquot *dquot);
955 static int ext4_release_dquot(struct dquot *dquot);
956 static int ext4_mark_dquot_dirty(struct dquot *dquot);
957 static int ext4_write_info(struct super_block *sb, int type);
958 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
959 char *path, int remount);
960 static int ext4_quota_on_mount(struct super_block *sb, int type);
961 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
962 size_t len, loff_t off);
963 static ssize_t ext4_quota_write(struct super_block *sb, int type,
964 const char *data, size_t len, loff_t off);
966 static struct dquot_operations ext4_quota_operations = {
967 .initialize = dquot_initialize,
969 .alloc_space = dquot_alloc_space,
970 .reserve_space = dquot_reserve_space,
971 .claim_space = dquot_claim_space,
972 .release_rsv = dquot_release_reserved_space,
973 .get_reserved_space = ext4_get_reserved_space,
974 .alloc_inode = dquot_alloc_inode,
975 .free_space = dquot_free_space,
976 .free_inode = dquot_free_inode,
977 .transfer = dquot_transfer,
978 .write_dquot = ext4_write_dquot,
979 .acquire_dquot = ext4_acquire_dquot,
980 .release_dquot = ext4_release_dquot,
981 .mark_dirty = ext4_mark_dquot_dirty,
982 .write_info = ext4_write_info,
983 .alloc_dquot = dquot_alloc,
984 .destroy_dquot = dquot_destroy,
987 static struct quotactl_ops ext4_qctl_operations = {
988 .quota_on = ext4_quota_on,
989 .quota_off = vfs_quota_off,
990 .quota_sync = vfs_quota_sync,
991 .get_info = vfs_get_dqinfo,
992 .set_info = vfs_set_dqinfo,
993 .get_dqblk = vfs_get_dqblk,
994 .set_dqblk = vfs_set_dqblk
998 static const struct super_operations ext4_sops = {
999 .alloc_inode = ext4_alloc_inode,
1000 .destroy_inode = ext4_destroy_inode,
1001 .write_inode = ext4_write_inode,
1002 .dirty_inode = ext4_dirty_inode,
1003 .delete_inode = ext4_delete_inode,
1004 .put_super = ext4_put_super,
1005 .sync_fs = ext4_sync_fs,
1006 .freeze_fs = ext4_freeze,
1007 .unfreeze_fs = ext4_unfreeze,
1008 .statfs = ext4_statfs,
1009 .remount_fs = ext4_remount,
1010 .clear_inode = ext4_clear_inode,
1011 .show_options = ext4_show_options,
1013 .quota_read = ext4_quota_read,
1014 .quota_write = ext4_quota_write,
1016 .bdev_try_to_free_page = bdev_try_to_free_page,
1019 static const struct super_operations ext4_nojournal_sops = {
1020 .alloc_inode = ext4_alloc_inode,
1021 .destroy_inode = ext4_destroy_inode,
1022 .write_inode = ext4_write_inode,
1023 .dirty_inode = ext4_dirty_inode,
1024 .delete_inode = ext4_delete_inode,
1025 .write_super = ext4_write_super,
1026 .put_super = ext4_put_super,
1027 .statfs = ext4_statfs,
1028 .remount_fs = ext4_remount,
1029 .clear_inode = ext4_clear_inode,
1030 .show_options = ext4_show_options,
1032 .quota_read = ext4_quota_read,
1033 .quota_write = ext4_quota_write,
1035 .bdev_try_to_free_page = bdev_try_to_free_page,
1038 static const struct export_operations ext4_export_ops = {
1039 .fh_to_dentry = ext4_fh_to_dentry,
1040 .fh_to_parent = ext4_fh_to_parent,
1041 .get_parent = ext4_get_parent,
1045 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1046 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1047 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1048 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1049 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1050 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1051 Opt_journal_update, Opt_journal_dev,
1052 Opt_journal_checksum, Opt_journal_async_commit,
1053 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1054 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
1055 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1056 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1057 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1058 Opt_usrquota, Opt_grpquota, Opt_i_version,
1059 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1060 Opt_block_validity, Opt_noblock_validity,
1061 Opt_inode_readahead_blks, Opt_journal_ioprio
1064 static const match_table_t tokens = {
1065 {Opt_bsd_df, "bsddf"},
1066 {Opt_minix_df, "minixdf"},
1067 {Opt_grpid, "grpid"},
1068 {Opt_grpid, "bsdgroups"},
1069 {Opt_nogrpid, "nogrpid"},
1070 {Opt_nogrpid, "sysvgroups"},
1071 {Opt_resgid, "resgid=%u"},
1072 {Opt_resuid, "resuid=%u"},
1074 {Opt_err_cont, "errors=continue"},
1075 {Opt_err_panic, "errors=panic"},
1076 {Opt_err_ro, "errors=remount-ro"},
1077 {Opt_nouid32, "nouid32"},
1078 {Opt_debug, "debug"},
1079 {Opt_oldalloc, "oldalloc"},
1080 {Opt_orlov, "orlov"},
1081 {Opt_user_xattr, "user_xattr"},
1082 {Opt_nouser_xattr, "nouser_xattr"},
1084 {Opt_noacl, "noacl"},
1085 {Opt_noload, "noload"},
1088 {Opt_commit, "commit=%u"},
1089 {Opt_min_batch_time, "min_batch_time=%u"},
1090 {Opt_max_batch_time, "max_batch_time=%u"},
1091 {Opt_journal_update, "journal=update"},
1092 {Opt_journal_dev, "journal_dev=%u"},
1093 {Opt_journal_checksum, "journal_checksum"},
1094 {Opt_journal_async_commit, "journal_async_commit"},
1095 {Opt_abort, "abort"},
1096 {Opt_data_journal, "data=journal"},
1097 {Opt_data_ordered, "data=ordered"},
1098 {Opt_data_writeback, "data=writeback"},
1099 {Opt_data_err_abort, "data_err=abort"},
1100 {Opt_data_err_ignore, "data_err=ignore"},
1101 {Opt_mb_history_length, "mb_history_length=%u"},
1102 {Opt_offusrjquota, "usrjquota="},
1103 {Opt_usrjquota, "usrjquota=%s"},
1104 {Opt_offgrpjquota, "grpjquota="},
1105 {Opt_grpjquota, "grpjquota=%s"},
1106 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1107 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1108 {Opt_grpquota, "grpquota"},
1109 {Opt_noquota, "noquota"},
1110 {Opt_quota, "quota"},
1111 {Opt_usrquota, "usrquota"},
1112 {Opt_barrier, "barrier=%u"},
1113 {Opt_barrier, "barrier"},
1114 {Opt_nobarrier, "nobarrier"},
1115 {Opt_i_version, "i_version"},
1116 {Opt_stripe, "stripe=%u"},
1117 {Opt_resize, "resize"},
1118 {Opt_delalloc, "delalloc"},
1119 {Opt_nodelalloc, "nodelalloc"},
1120 {Opt_block_validity, "block_validity"},
1121 {Opt_noblock_validity, "noblock_validity"},
1122 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1123 {Opt_journal_ioprio, "journal_ioprio=%u"},
1124 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1125 {Opt_auto_da_alloc, "auto_da_alloc"},
1126 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1130 static ext4_fsblk_t get_sb_block(void **data)
1132 ext4_fsblk_t sb_block;
1133 char *options = (char *) *data;
1135 if (!options || strncmp(options, "sb=", 3) != 0)
1136 return 1; /* Default location */
1139 /* TODO: use simple_strtoll with >32bit ext4 */
1140 sb_block = simple_strtoul(options, &options, 0);
1141 if (*options && *options != ',') {
1142 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1146 if (*options == ',')
1148 *data = (void *) options;
1153 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1155 static int parse_options(char *options, struct super_block *sb,
1156 unsigned long *journal_devnum,
1157 unsigned int *journal_ioprio,
1158 ext4_fsblk_t *n_blocks_count, int is_remount)
1160 struct ext4_sb_info *sbi = EXT4_SB(sb);
1162 substring_t args[MAX_OPT_ARGS];
1173 while ((p = strsep(&options, ",")) != NULL) {
1178 token = match_token(p, tokens, args);
1181 clear_opt(sbi->s_mount_opt, MINIX_DF);
1184 set_opt(sbi->s_mount_opt, MINIX_DF);
1187 set_opt(sbi->s_mount_opt, GRPID);
1190 clear_opt(sbi->s_mount_opt, GRPID);
1193 if (match_int(&args[0], &option))
1195 sbi->s_resuid = option;
1198 if (match_int(&args[0], &option))
1200 sbi->s_resgid = option;
1203 /* handled by get_sb_block() instead of here */
1204 /* *sb_block = match_int(&args[0]); */
1207 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1208 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1209 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1212 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1213 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1214 set_opt(sbi->s_mount_opt, ERRORS_RO);
1217 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1218 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1219 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1222 set_opt(sbi->s_mount_opt, NO_UID32);
1225 set_opt(sbi->s_mount_opt, DEBUG);
1228 set_opt(sbi->s_mount_opt, OLDALLOC);
1231 clear_opt(sbi->s_mount_opt, OLDALLOC);
1233 #ifdef CONFIG_EXT4_FS_XATTR
1234 case Opt_user_xattr:
1235 set_opt(sbi->s_mount_opt, XATTR_USER);
1237 case Opt_nouser_xattr:
1238 clear_opt(sbi->s_mount_opt, XATTR_USER);
1241 case Opt_user_xattr:
1242 case Opt_nouser_xattr:
1243 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1246 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1248 set_opt(sbi->s_mount_opt, POSIX_ACL);
1251 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1256 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1259 case Opt_journal_update:
1261 /* Eventually we will want to be able to create
1262 a journal file here. For now, only allow the
1263 user to specify an existing inode to be the
1266 ext4_msg(sb, KERN_ERR,
1267 "Cannot specify journal on remount");
1270 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1272 case Opt_journal_dev:
1274 ext4_msg(sb, KERN_ERR,
1275 "Cannot specify journal on remount");
1278 if (match_int(&args[0], &option))
1280 *journal_devnum = option;
1282 case Opt_journal_checksum:
1283 break; /* Kept for backwards compatibility */
1284 case Opt_journal_async_commit:
1285 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1288 set_opt(sbi->s_mount_opt, NOLOAD);
1291 if (match_int(&args[0], &option))
1296 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1297 sbi->s_commit_interval = HZ * option;
1299 case Opt_max_batch_time:
1300 if (match_int(&args[0], &option))
1305 option = EXT4_DEF_MAX_BATCH_TIME;
1306 sbi->s_max_batch_time = option;
1308 case Opt_min_batch_time:
1309 if (match_int(&args[0], &option))
1313 sbi->s_min_batch_time = option;
1315 case Opt_data_journal:
1316 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1318 case Opt_data_ordered:
1319 data_opt = EXT4_MOUNT_ORDERED_DATA;
1321 case Opt_data_writeback:
1322 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1325 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1327 ext4_msg(sb, KERN_ERR,
1328 "Cannot change data mode on remount");
1332 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1333 sbi->s_mount_opt |= data_opt;
1336 case Opt_data_err_abort:
1337 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1339 case Opt_data_err_ignore:
1340 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342 case Opt_mb_history_length:
1343 if (match_int(&args[0], &option))
1347 sbi->s_mb_history_max = option;
1356 if (sb_any_quota_loaded(sb) &&
1357 !sbi->s_qf_names[qtype]) {
1358 ext4_msg(sb, KERN_ERR,
1359 "Cannot change journaled "
1360 "quota options when quota turned on");
1363 qname = match_strdup(&args[0]);
1365 ext4_msg(sb, KERN_ERR,
1366 "Not enough memory for "
1367 "storing quotafile name");
1370 if (sbi->s_qf_names[qtype] &&
1371 strcmp(sbi->s_qf_names[qtype], qname)) {
1372 ext4_msg(sb, KERN_ERR,
1373 "%s quota file already "
1374 "specified", QTYPE2NAME(qtype));
1378 sbi->s_qf_names[qtype] = qname;
1379 if (strchr(sbi->s_qf_names[qtype], '/')) {
1380 ext4_msg(sb, KERN_ERR,
1381 "quotafile must be on "
1383 kfree(sbi->s_qf_names[qtype]);
1384 sbi->s_qf_names[qtype] = NULL;
1387 set_opt(sbi->s_mount_opt, QUOTA);
1389 case Opt_offusrjquota:
1392 case Opt_offgrpjquota:
1395 if (sb_any_quota_loaded(sb) &&
1396 sbi->s_qf_names[qtype]) {
1397 ext4_msg(sb, KERN_ERR, "Cannot change "
1398 "journaled quota options when "
1403 * The space will be released later when all options
1404 * are confirmed to be correct
1406 sbi->s_qf_names[qtype] = NULL;
1408 case Opt_jqfmt_vfsold:
1409 qfmt = QFMT_VFS_OLD;
1411 case Opt_jqfmt_vfsv0:
1414 if (sb_any_quota_loaded(sb) &&
1415 sbi->s_jquota_fmt != qfmt) {
1416 ext4_msg(sb, KERN_ERR, "Cannot change "
1417 "journaled quota options when "
1421 sbi->s_jquota_fmt = qfmt;
1425 set_opt(sbi->s_mount_opt, QUOTA);
1426 set_opt(sbi->s_mount_opt, USRQUOTA);
1429 set_opt(sbi->s_mount_opt, QUOTA);
1430 set_opt(sbi->s_mount_opt, GRPQUOTA);
1433 if (sb_any_quota_loaded(sb)) {
1434 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1435 "options when quota turned on");
1438 clear_opt(sbi->s_mount_opt, QUOTA);
1439 clear_opt(sbi->s_mount_opt, USRQUOTA);
1440 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1446 ext4_msg(sb, KERN_ERR,
1447 "quota options not supported");
1451 case Opt_offusrjquota:
1452 case Opt_offgrpjquota:
1453 case Opt_jqfmt_vfsold:
1454 case Opt_jqfmt_vfsv0:
1455 ext4_msg(sb, KERN_ERR,
1456 "journaled quota options not supported");
1462 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1465 clear_opt(sbi->s_mount_opt, BARRIER);
1468 if (match_int(&args[0], &option)) {
1469 set_opt(sbi->s_mount_opt, BARRIER);
1473 set_opt(sbi->s_mount_opt, BARRIER);
1475 clear_opt(sbi->s_mount_opt, BARRIER);
1481 ext4_msg(sb, KERN_ERR,
1482 "resize option only available "
1486 if (match_int(&args[0], &option) != 0)
1488 *n_blocks_count = option;
1491 set_opt(sbi->s_mount_opt, NOBH);
1494 clear_opt(sbi->s_mount_opt, NOBH);
1497 set_opt(sbi->s_mount_opt, I_VERSION);
1498 sb->s_flags |= MS_I_VERSION;
1500 case Opt_nodelalloc:
1501 clear_opt(sbi->s_mount_opt, DELALLOC);
1504 if (match_int(&args[0], &option))
1508 sbi->s_stripe = option;
1511 set_opt(sbi->s_mount_opt, DELALLOC);
1513 case Opt_block_validity:
1514 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1516 case Opt_noblock_validity:
1517 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519 case Opt_inode_readahead_blks:
1520 if (match_int(&args[0], &option))
1522 if (option < 0 || option > (1 << 30))
1524 if (!is_power_of_2(option)) {
1525 ext4_msg(sb, KERN_ERR,
1526 "EXT4-fs: inode_readahead_blks"
1527 " must be a power of 2");
1530 sbi->s_inode_readahead_blks = option;
1532 case Opt_journal_ioprio:
1533 if (match_int(&args[0], &option))
1535 if (option < 0 || option > 7)
1537 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1540 case Opt_noauto_da_alloc:
1541 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1543 case Opt_auto_da_alloc:
1544 if (match_int(&args[0], &option)) {
1545 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1549 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1551 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1554 ext4_msg(sb, KERN_ERR,
1555 "Unrecognized mount option \"%s\" "
1556 "or missing value", p);
1561 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1562 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1563 sbi->s_qf_names[USRQUOTA])
1564 clear_opt(sbi->s_mount_opt, USRQUOTA);
1566 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1567 sbi->s_qf_names[GRPQUOTA])
1568 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1570 if ((sbi->s_qf_names[USRQUOTA] &&
1571 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1572 (sbi->s_qf_names[GRPQUOTA] &&
1573 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1574 ext4_msg(sb, KERN_ERR, "old and new quota "
1579 if (!sbi->s_jquota_fmt) {
1580 ext4_msg(sb, KERN_ERR, "journaled quota format "
1585 if (sbi->s_jquota_fmt) {
1586 ext4_msg(sb, KERN_ERR, "journaled quota format "
1587 "specified with no journaling "
1596 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1599 struct ext4_sb_info *sbi = EXT4_SB(sb);
1602 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1603 ext4_msg(sb, KERN_ERR, "revision level too high, "
1604 "forcing read-only mode");
1609 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1610 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1611 "running e2fsck is recommended");
1612 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1613 ext4_msg(sb, KERN_WARNING,
1614 "warning: mounting fs with errors, "
1615 "running e2fsck is recommended");
1616 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1617 le16_to_cpu(es->s_mnt_count) >=
1618 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1619 ext4_msg(sb, KERN_WARNING,
1620 "warning: maximal mount count reached, "
1621 "running e2fsck is recommended");
1622 else if (le32_to_cpu(es->s_checkinterval) &&
1623 (le32_to_cpu(es->s_lastcheck) +
1624 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1625 ext4_msg(sb, KERN_WARNING,
1626 "warning: checktime reached, "
1627 "running e2fsck is recommended");
1628 if (!sbi->s_journal)
1629 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1630 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1631 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1632 le16_add_cpu(&es->s_mnt_count, 1);
1633 es->s_mtime = cpu_to_le32(get_seconds());
1634 ext4_update_dynamic_rev(sb);
1636 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1638 ext4_commit_super(sb, 1);
1639 if (test_opt(sb, DEBUG))
1640 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1641 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1643 sbi->s_groups_count,
1644 EXT4_BLOCKS_PER_GROUP(sb),
1645 EXT4_INODES_PER_GROUP(sb),
1648 if (EXT4_SB(sb)->s_journal) {
1649 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1650 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1651 "external", EXT4_SB(sb)->s_journal->j_devname);
1653 ext4_msg(sb, KERN_INFO, "no journal");
1658 static int ext4_fill_flex_info(struct super_block *sb)
1660 struct ext4_sb_info *sbi = EXT4_SB(sb);
1661 struct ext4_group_desc *gdp = NULL;
1662 ext4_group_t flex_group_count;
1663 ext4_group_t flex_group;
1664 int groups_per_flex = 0;
1668 if (!sbi->s_es->s_log_groups_per_flex) {
1669 sbi->s_log_groups_per_flex = 0;
1673 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1674 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1676 /* We allocate both existing and potentially added groups */
1677 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1678 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1679 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1680 size = flex_group_count * sizeof(struct flex_groups);
1681 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1682 if (sbi->s_flex_groups == NULL) {
1683 sbi->s_flex_groups = vmalloc(size);
1684 if (sbi->s_flex_groups)
1685 memset(sbi->s_flex_groups, 0, size);
1687 if (sbi->s_flex_groups == NULL) {
1688 ext4_msg(sb, KERN_ERR, "not enough memory for "
1689 "%u flex groups", flex_group_count);
1693 for (i = 0; i < sbi->s_groups_count; i++) {
1694 gdp = ext4_get_group_desc(sb, i, NULL);
1696 flex_group = ext4_flex_group(sbi, i);
1697 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1698 ext4_free_inodes_count(sb, gdp));
1699 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1700 ext4_free_blks_count(sb, gdp));
1701 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1702 ext4_used_dirs_count(sb, gdp));
1710 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1711 struct ext4_group_desc *gdp)
1715 if (sbi->s_es->s_feature_ro_compat &
1716 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1717 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1718 __le32 le_group = cpu_to_le32(block_group);
1720 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1721 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1722 crc = crc16(crc, (__u8 *)gdp, offset);
1723 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1724 /* for checksum of struct ext4_group_desc do the rest...*/
1725 if ((sbi->s_es->s_feature_incompat &
1726 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1727 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1728 crc = crc16(crc, (__u8 *)gdp + offset,
1729 le16_to_cpu(sbi->s_es->s_desc_size) -
1733 return cpu_to_le16(crc);
1736 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1737 struct ext4_group_desc *gdp)
1739 if ((sbi->s_es->s_feature_ro_compat &
1740 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1741 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1747 /* Called at mount-time, super-block is locked */
1748 static int ext4_check_descriptors(struct super_block *sb)
1750 struct ext4_sb_info *sbi = EXT4_SB(sb);
1751 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1752 ext4_fsblk_t last_block;
1753 ext4_fsblk_t block_bitmap;
1754 ext4_fsblk_t inode_bitmap;
1755 ext4_fsblk_t inode_table;
1756 int flexbg_flag = 0;
1759 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1762 ext4_debug("Checking group descriptors");
1764 for (i = 0; i < sbi->s_groups_count; i++) {
1765 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1767 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1768 last_block = ext4_blocks_count(sbi->s_es) - 1;
1770 last_block = first_block +
1771 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1773 block_bitmap = ext4_block_bitmap(sb, gdp);
1774 if (block_bitmap < first_block || block_bitmap > last_block) {
1775 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1776 "Block bitmap for group %u not in group "
1777 "(block %llu)!", i, block_bitmap);
1780 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1781 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1782 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1783 "Inode bitmap for group %u not in group "
1784 "(block %llu)!", i, inode_bitmap);
1787 inode_table = ext4_inode_table(sb, gdp);
1788 if (inode_table < first_block ||
1789 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1790 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1791 "Inode table for group %u not in group "
1792 "(block %llu)!", i, inode_table);
1795 ext4_lock_group(sb, i);
1796 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1797 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1798 "Checksum for group %u failed (%u!=%u)",
1799 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1800 gdp)), le16_to_cpu(gdp->bg_checksum));
1801 if (!(sb->s_flags & MS_RDONLY)) {
1802 ext4_unlock_group(sb, i);
1806 ext4_unlock_group(sb, i);
1808 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1811 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1812 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1816 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1817 * the superblock) which were deleted from all directories, but held open by
1818 * a process at the time of a crash. We walk the list and try to delete these
1819 * inodes at recovery time (only with a read-write filesystem).
1821 * In order to keep the orphan inode chain consistent during traversal (in
1822 * case of crash during recovery), we link each inode into the superblock
1823 * orphan list_head and handle it the same way as an inode deletion during
1824 * normal operation (which journals the operations for us).
1826 * We only do an iget() and an iput() on each inode, which is very safe if we
1827 * accidentally point at an in-use or already deleted inode. The worst that
1828 * can happen in this case is that we get a "bit already cleared" message from
1829 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1830 * e2fsck was run on this filesystem, and it must have already done the orphan
1831 * inode cleanup for us, so we can safely abort without any further action.
1833 static void ext4_orphan_cleanup(struct super_block *sb,
1834 struct ext4_super_block *es)
1836 unsigned int s_flags = sb->s_flags;
1837 int nr_orphans = 0, nr_truncates = 0;
1841 if (!es->s_last_orphan) {
1842 jbd_debug(4, "no orphan inodes to clean up\n");
1846 if (bdev_read_only(sb->s_bdev)) {
1847 ext4_msg(sb, KERN_ERR, "write access "
1848 "unavailable, skipping orphan cleanup");
1852 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1853 if (es->s_last_orphan)
1854 jbd_debug(1, "Errors on filesystem, "
1855 "clearing orphan list.\n");
1856 es->s_last_orphan = 0;
1857 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1861 if (s_flags & MS_RDONLY) {
1862 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1863 sb->s_flags &= ~MS_RDONLY;
1866 /* Needed for iput() to work correctly and not trash data */
1867 sb->s_flags |= MS_ACTIVE;
1868 /* Turn on quotas so that they are updated correctly */
1869 for (i = 0; i < MAXQUOTAS; i++) {
1870 if (EXT4_SB(sb)->s_qf_names[i]) {
1871 int ret = ext4_quota_on_mount(sb, i);
1873 ext4_msg(sb, KERN_ERR,
1874 "Cannot turn on journaled "
1875 "quota: error %d", ret);
1880 while (es->s_last_orphan) {
1881 struct inode *inode;
1883 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1884 if (IS_ERR(inode)) {
1885 es->s_last_orphan = 0;
1889 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1891 if (inode->i_nlink) {
1892 ext4_msg(sb, KERN_DEBUG,
1893 "%s: truncating inode %lu to %lld bytes",
1894 __func__, inode->i_ino, inode->i_size);
1895 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1896 inode->i_ino, inode->i_size);
1897 ext4_truncate(inode);
1900 ext4_msg(sb, KERN_DEBUG,
1901 "%s: deleting unreferenced inode %lu",
1902 __func__, inode->i_ino);
1903 jbd_debug(2, "deleting unreferenced inode %lu\n",
1907 iput(inode); /* The delete magic happens here! */
1910 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1913 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1914 PLURAL(nr_orphans));
1916 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1917 PLURAL(nr_truncates));
1919 /* Turn quotas off */
1920 for (i = 0; i < MAXQUOTAS; i++) {
1921 if (sb_dqopt(sb)->files[i])
1922 vfs_quota_off(sb, i, 0);
1925 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1929 * Maximal extent format file size.
1930 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1931 * extent format containers, within a sector_t, and within i_blocks
1932 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1933 * so that won't be a limiting factor.
1935 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1937 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1940 loff_t upper_limit = MAX_LFS_FILESIZE;
1942 /* small i_blocks in vfs inode? */
1943 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1945 * CONFIG_LBDAF is not enabled implies the inode
1946 * i_block represent total blocks in 512 bytes
1947 * 32 == size of vfs inode i_blocks * 8
1949 upper_limit = (1LL << 32) - 1;
1951 /* total blocks in file system block size */
1952 upper_limit >>= (blkbits - 9);
1953 upper_limit <<= blkbits;
1956 /* 32-bit extent-start container, ee_block */
1961 /* Sanity check against vm- & vfs- imposed limits */
1962 if (res > upper_limit)
1969 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1970 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1971 * We need to be 1 filesystem block less than the 2^48 sector limit.
1973 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1975 loff_t res = EXT4_NDIR_BLOCKS;
1978 /* This is calculated to be the largest file size for a dense, block
1979 * mapped file such that the file's total number of 512-byte sectors,
1980 * including data and all indirect blocks, does not exceed (2^48 - 1).
1982 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1983 * number of 512-byte sectors of the file.
1986 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1988 * !has_huge_files or CONFIG_LBDAF not enabled implies that
1989 * the inode i_block field represents total file blocks in
1990 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
1992 upper_limit = (1LL << 32) - 1;
1994 /* total blocks in file system block size */
1995 upper_limit >>= (bits - 9);
1999 * We use 48 bit ext4_inode i_blocks
2000 * With EXT4_HUGE_FILE_FL set the i_blocks
2001 * represent total number of blocks in
2002 * file system block size
2004 upper_limit = (1LL << 48) - 1;
2008 /* indirect blocks */
2010 /* double indirect blocks */
2011 meta_blocks += 1 + (1LL << (bits-2));
2012 /* tripple indirect blocks */
2013 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2015 upper_limit -= meta_blocks;
2016 upper_limit <<= bits;
2018 res += 1LL << (bits-2);
2019 res += 1LL << (2*(bits-2));
2020 res += 1LL << (3*(bits-2));
2022 if (res > upper_limit)
2025 if (res > MAX_LFS_FILESIZE)
2026 res = MAX_LFS_FILESIZE;
2031 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2032 ext4_fsblk_t logical_sb_block, int nr)
2034 struct ext4_sb_info *sbi = EXT4_SB(sb);
2035 ext4_group_t bg, first_meta_bg;
2038 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2040 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2042 return logical_sb_block + nr + 1;
2043 bg = sbi->s_desc_per_block * nr;
2044 if (ext4_bg_has_super(sb, bg))
2047 return (has_super + ext4_group_first_block_no(sb, bg));
2051 * ext4_get_stripe_size: Get the stripe size.
2052 * @sbi: In memory super block info
2054 * If we have specified it via mount option, then
2055 * use the mount option value. If the value specified at mount time is
2056 * greater than the blocks per group use the super block value.
2057 * If the super block value is greater than blocks per group return 0.
2058 * Allocator needs it be less than blocks per group.
2061 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2063 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2064 unsigned long stripe_width =
2065 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2067 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2068 return sbi->s_stripe;
2070 if (stripe_width <= sbi->s_blocks_per_group)
2071 return stripe_width;
2073 if (stride <= sbi->s_blocks_per_group)
2082 struct attribute attr;
2083 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2084 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2085 const char *, size_t);
2089 static int parse_strtoul(const char *buf,
2090 unsigned long max, unsigned long *value)
2094 while (*buf && isspace(*buf))
2096 *value = simple_strtoul(buf, &endp, 0);
2097 while (*endp && isspace(*endp))
2099 if (*endp || *value > max)
2105 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2106 struct ext4_sb_info *sbi,
2109 return snprintf(buf, PAGE_SIZE, "%llu\n",
2110 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2113 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2114 struct ext4_sb_info *sbi, char *buf)
2116 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2118 return snprintf(buf, PAGE_SIZE, "%lu\n",
2119 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2120 sbi->s_sectors_written_start) >> 1);
2123 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2124 struct ext4_sb_info *sbi, char *buf)
2126 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2128 return snprintf(buf, PAGE_SIZE, "%llu\n",
2129 sbi->s_kbytes_written +
2130 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2131 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2134 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2135 struct ext4_sb_info *sbi,
2136 const char *buf, size_t count)
2140 if (parse_strtoul(buf, 0x40000000, &t))
2143 if (!is_power_of_2(t))
2146 sbi->s_inode_readahead_blks = t;
2150 static ssize_t sbi_ui_show(struct ext4_attr *a,
2151 struct ext4_sb_info *sbi, char *buf)
2153 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2155 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2158 static ssize_t sbi_ui_store(struct ext4_attr *a,
2159 struct ext4_sb_info *sbi,
2160 const char *buf, size_t count)
2162 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2165 if (parse_strtoul(buf, 0xffffffff, &t))
2171 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2172 static struct ext4_attr ext4_attr_##_name = { \
2173 .attr = {.name = __stringify(_name), .mode = _mode }, \
2176 .offset = offsetof(struct ext4_sb_info, _elname), \
2178 #define EXT4_ATTR(name, mode, show, store) \
2179 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2181 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2182 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2183 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2184 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2185 #define ATTR_LIST(name) &ext4_attr_##name.attr
2187 EXT4_RO_ATTR(delayed_allocation_blocks);
2188 EXT4_RO_ATTR(session_write_kbytes);
2189 EXT4_RO_ATTR(lifetime_write_kbytes);
2190 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2191 inode_readahead_blks_store, s_inode_readahead_blks);
2192 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
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(inode_goal),
2206 ATTR_LIST(mb_stats),
2207 ATTR_LIST(mb_max_to_scan),
2208 ATTR_LIST(mb_min_to_scan),
2209 ATTR_LIST(mb_order2_req),
2210 ATTR_LIST(mb_stream_req),
2211 ATTR_LIST(mb_group_prealloc),
2215 static ssize_t ext4_attr_show(struct kobject *kobj,
2216 struct attribute *attr, char *buf)
2218 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2220 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2222 return a->show ? a->show(a, sbi, buf) : 0;
2225 static ssize_t ext4_attr_store(struct kobject *kobj,
2226 struct attribute *attr,
2227 const char *buf, size_t len)
2229 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2231 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2233 return a->store ? a->store(a, sbi, buf, len) : 0;
2236 static void ext4_sb_release(struct kobject *kobj)
2238 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 complete(&sbi->s_kobj_unregister);
2244 static struct sysfs_ops ext4_attr_ops = {
2245 .show = ext4_attr_show,
2246 .store = ext4_attr_store,
2249 static struct kobj_type ext4_ktype = {
2250 .default_attrs = ext4_attrs,
2251 .sysfs_ops = &ext4_attr_ops,
2252 .release = ext4_sb_release,
2256 * Check whether this filesystem can be mounted based on
2257 * the features present and the RDONLY/RDWR mount requested.
2258 * Returns 1 if this filesystem can be mounted as requested,
2259 * 0 if it cannot be.
2261 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2263 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2264 ext4_msg(sb, KERN_ERR,
2265 "Couldn't mount because of "
2266 "unsupported optional features (%x)",
2267 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2268 ~EXT4_FEATURE_INCOMPAT_SUPP));
2275 /* Check that feature set is OK for a read-write mount */
2276 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2277 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2278 "unsupported optional features (%x)",
2279 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2280 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2284 * Large file size enabled file system can only be mounted
2285 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2287 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2288 if (sizeof(blkcnt_t) < sizeof(u64)) {
2289 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2290 "cannot be mounted RDWR without "
2298 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2299 __releases(kernel_lock)
2300 __acquires(kernel_lock)
2302 struct buffer_head *bh;
2303 struct ext4_super_block *es = NULL;
2304 struct ext4_sb_info *sbi;
2306 ext4_fsblk_t sb_block = get_sb_block(&data);
2307 ext4_fsblk_t logical_sb_block;
2308 unsigned long offset = 0;
2309 unsigned long journal_devnum = 0;
2310 unsigned long def_mount_opts;
2316 unsigned int db_count;
2318 int needs_recovery, has_huge_files;
2321 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2323 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2327 sbi->s_blockgroup_lock =
2328 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2329 if (!sbi->s_blockgroup_lock) {
2333 sb->s_fs_info = sbi;
2334 sbi->s_mount_opt = 0;
2335 sbi->s_resuid = EXT4_DEF_RESUID;
2336 sbi->s_resgid = EXT4_DEF_RESGID;
2337 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2338 sbi->s_sb_block = sb_block;
2339 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2344 /* Cleanup superblock name */
2345 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2348 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2350 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2355 * The ext4 superblock will not be buffer aligned for other than 1kB
2356 * block sizes. We need to calculate the offset from buffer start.
2358 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2359 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2360 offset = do_div(logical_sb_block, blocksize);
2362 logical_sb_block = sb_block;
2365 if (!(bh = sb_bread(sb, logical_sb_block))) {
2366 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2370 * Note: s_es must be initialized as soon as possible because
2371 * some ext4 macro-instructions depend on its value
2373 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2375 sb->s_magic = le16_to_cpu(es->s_magic);
2376 if (sb->s_magic != EXT4_SUPER_MAGIC)
2378 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2380 /* Set defaults before we parse the mount options */
2381 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2382 if (def_mount_opts & EXT4_DEFM_DEBUG)
2383 set_opt(sbi->s_mount_opt, DEBUG);
2384 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
2385 set_opt(sbi->s_mount_opt, GRPID);
2386 if (def_mount_opts & EXT4_DEFM_UID16)
2387 set_opt(sbi->s_mount_opt, NO_UID32);
2388 #ifdef CONFIG_EXT4_FS_XATTR
2389 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2390 set_opt(sbi->s_mount_opt, XATTR_USER);
2392 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2393 if (def_mount_opts & EXT4_DEFM_ACL)
2394 set_opt(sbi->s_mount_opt, POSIX_ACL);
2396 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2397 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2398 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2399 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2400 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2401 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2403 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2404 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2405 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2406 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2408 set_opt(sbi->s_mount_opt, ERRORS_RO);
2410 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2411 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2412 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2413 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2414 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2415 sbi->s_mb_history_max = default_mb_history_length;
2417 set_opt(sbi->s_mount_opt, BARRIER);
2420 * enable delayed allocation by default
2421 * Use -o nodelalloc to turn it off
2423 set_opt(sbi->s_mount_opt, DELALLOC);
2425 if (!parse_options((char *) data, sb, &journal_devnum,
2426 &journal_ioprio, NULL, 0))
2429 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2430 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2432 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2433 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2434 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2435 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2436 ext4_msg(sb, KERN_WARNING,
2437 "feature flags set on rev 0 fs, "
2438 "running e2fsck is recommended");
2441 * Check feature flags regardless of the revision level, since we
2442 * previously didn't change the revision level when setting the flags,
2443 * so there is a chance incompat flags are set on a rev 0 filesystem.
2445 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2448 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2450 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2451 blocksize > EXT4_MAX_BLOCK_SIZE) {
2452 ext4_msg(sb, KERN_ERR,
2453 "Unsupported filesystem blocksize %d", blocksize);
2457 if (sb->s_blocksize != blocksize) {
2458 /* Validate the filesystem blocksize */
2459 if (!sb_set_blocksize(sb, blocksize)) {
2460 ext4_msg(sb, KERN_ERR, "bad block size %d",
2466 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2467 offset = do_div(logical_sb_block, blocksize);
2468 bh = sb_bread(sb, logical_sb_block);
2470 ext4_msg(sb, KERN_ERR,
2471 "Can't read superblock on 2nd try");
2474 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2476 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2477 ext4_msg(sb, KERN_ERR,
2478 "Magic mismatch, very weird!");
2483 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2484 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2485 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2487 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2489 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2490 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2491 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2493 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2494 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2495 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2496 (!is_power_of_2(sbi->s_inode_size)) ||
2497 (sbi->s_inode_size > blocksize)) {
2498 ext4_msg(sb, KERN_ERR,
2499 "unsupported inode size: %d",
2503 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2504 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2507 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2508 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2509 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2510 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2511 !is_power_of_2(sbi->s_desc_size)) {
2512 ext4_msg(sb, KERN_ERR,
2513 "unsupported descriptor size %lu",
2518 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2520 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2521 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2522 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2525 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2526 if (sbi->s_inodes_per_block == 0)
2528 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2529 sbi->s_inodes_per_block;
2530 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2532 sbi->s_mount_state = le16_to_cpu(es->s_state);
2533 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2534 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2536 for (i = 0; i < 4; i++)
2537 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2538 sbi->s_def_hash_version = es->s_def_hash_version;
2539 i = le32_to_cpu(es->s_flags);
2540 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2541 sbi->s_hash_unsigned = 3;
2542 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2543 #ifdef __CHAR_UNSIGNED__
2544 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2545 sbi->s_hash_unsigned = 3;
2547 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2552 if (sbi->s_blocks_per_group > blocksize * 8) {
2553 ext4_msg(sb, KERN_ERR,
2554 "#blocks per group too big: %lu",
2555 sbi->s_blocks_per_group);
2558 if (sbi->s_inodes_per_group > blocksize * 8) {
2559 ext4_msg(sb, KERN_ERR,
2560 "#inodes per group too big: %lu",
2561 sbi->s_inodes_per_group);
2566 * Test whether we have more sectors than will fit in sector_t,
2567 * and whether the max offset is addressable by the page cache.
2569 if ((ext4_blocks_count(es) >
2570 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2571 (ext4_blocks_count(es) >
2572 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2573 ext4_msg(sb, KERN_ERR, "filesystem"
2574 " too large to mount safely on this system");
2575 if (sizeof(sector_t) < 8)
2576 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2581 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2584 /* check blocks count against device size */
2585 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2586 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2587 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2588 "exceeds size of device (%llu blocks)",
2589 ext4_blocks_count(es), blocks_count);
2594 * It makes no sense for the first data block to be beyond the end
2595 * of the filesystem.
2597 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2598 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2599 "block %u is beyond end of filesystem (%llu)",
2600 le32_to_cpu(es->s_first_data_block),
2601 ext4_blocks_count(es));
2604 blocks_count = (ext4_blocks_count(es) -
2605 le32_to_cpu(es->s_first_data_block) +
2606 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2607 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2608 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2609 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2610 "(block count %llu, first data block %u, "
2611 "blocks per group %lu)", sbi->s_groups_count,
2612 ext4_blocks_count(es),
2613 le32_to_cpu(es->s_first_data_block),
2614 EXT4_BLOCKS_PER_GROUP(sb));
2617 sbi->s_groups_count = blocks_count;
2618 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2619 EXT4_DESC_PER_BLOCK(sb);
2620 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2622 if (sbi->s_group_desc == NULL) {
2623 ext4_msg(sb, KERN_ERR, "not enough memory");
2627 #ifdef CONFIG_PROC_FS
2629 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2632 bgl_lock_init(sbi->s_blockgroup_lock);
2634 for (i = 0; i < db_count; i++) {
2635 block = descriptor_loc(sb, logical_sb_block, i);
2636 sbi->s_group_desc[i] = sb_bread(sb, block);
2637 if (!sbi->s_group_desc[i]) {
2638 ext4_msg(sb, KERN_ERR,
2639 "can't read group descriptor %d", i);
2644 if (!ext4_check_descriptors(sb)) {
2645 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2648 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2649 if (!ext4_fill_flex_info(sb)) {
2650 ext4_msg(sb, KERN_ERR,
2651 "unable to initialize "
2652 "flex_bg meta info!");
2656 sbi->s_gdb_count = db_count;
2657 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2658 spin_lock_init(&sbi->s_next_gen_lock);
2660 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2661 ext4_count_free_blocks(sb));
2663 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2664 ext4_count_free_inodes(sb));
2667 err = percpu_counter_init(&sbi->s_dirs_counter,
2668 ext4_count_dirs(sb));
2671 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2674 ext4_msg(sb, KERN_ERR, "insufficient memory");
2678 sbi->s_stripe = ext4_get_stripe_size(sbi);
2681 * set up enough so that it can read an inode
2683 if (!test_opt(sb, NOLOAD) &&
2684 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2685 sb->s_op = &ext4_sops;
2687 sb->s_op = &ext4_nojournal_sops;
2688 sb->s_export_op = &ext4_export_ops;
2689 sb->s_xattr = ext4_xattr_handlers;
2691 sb->s_qcop = &ext4_qctl_operations;
2692 sb->dq_op = &ext4_quota_operations;
2694 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2695 mutex_init(&sbi->s_orphan_lock);
2696 mutex_init(&sbi->s_resize_lock);
2700 needs_recovery = (es->s_last_orphan != 0 ||
2701 EXT4_HAS_INCOMPAT_FEATURE(sb,
2702 EXT4_FEATURE_INCOMPAT_RECOVER));
2705 * The first inode we look at is the journal inode. Don't try
2706 * root first: it may be modified in the journal!
2708 if (!test_opt(sb, NOLOAD) &&
2709 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2710 if (ext4_load_journal(sb, es, journal_devnum))
2712 if (!(sb->s_flags & MS_RDONLY) &&
2713 EXT4_SB(sb)->s_journal->j_failed_commit) {
2714 ext4_msg(sb, KERN_CRIT, "error: "
2715 "ext4_fill_super: Journal transaction "
2717 EXT4_SB(sb)->s_journal->j_failed_commit);
2718 if (test_opt(sb, ERRORS_RO)) {
2719 ext4_msg(sb, KERN_CRIT,
2720 "Mounting filesystem read-only");
2721 sb->s_flags |= MS_RDONLY;
2722 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2723 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2725 if (test_opt(sb, ERRORS_PANIC)) {
2726 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2727 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2728 ext4_commit_super(sb, 1);
2732 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2733 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2734 ext4_msg(sb, KERN_ERR, "required journal recovery "
2735 "suppressed and not mounted read-only");
2738 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2739 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2740 sbi->s_journal = NULL;
2745 if (ext4_blocks_count(es) > 0xffffffffULL &&
2746 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2747 JBD2_FEATURE_INCOMPAT_64BIT)) {
2748 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2752 jbd2_journal_set_features(sbi->s_journal,
2753 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2754 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2755 jbd2_journal_set_features(sbi->s_journal, 0, 0,
2756 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2758 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2759 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2761 /* We have now updated the journal if required, so we can
2762 * validate the data journaling mode. */
2763 switch (test_opt(sb, DATA_FLAGS)) {
2765 /* No mode set, assume a default based on the journal
2766 * capabilities: ORDERED_DATA if the journal can
2767 * cope, else JOURNAL_DATA
2769 if (jbd2_journal_check_available_features
2770 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2771 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2773 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2776 case EXT4_MOUNT_ORDERED_DATA:
2777 case EXT4_MOUNT_WRITEBACK_DATA:
2778 if (!jbd2_journal_check_available_features
2779 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2780 ext4_msg(sb, KERN_ERR, "Journal does not support "
2781 "requested data journaling mode");
2787 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2791 if (test_opt(sb, NOBH)) {
2792 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2793 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2794 "its supported only with writeback mode");
2795 clear_opt(sbi->s_mount_opt, NOBH);
2799 * The jbd2_journal_load will have done any necessary log recovery,
2800 * so we can safely mount the rest of the filesystem now.
2803 root = ext4_iget(sb, EXT4_ROOT_INO);
2805 ext4_msg(sb, KERN_ERR, "get root inode failed");
2806 ret = PTR_ERR(root);
2809 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2811 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2814 sb->s_root = d_alloc_root(root);
2816 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2822 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2824 /* determine the minimum size of new large inodes, if present */
2825 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2826 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2827 EXT4_GOOD_OLD_INODE_SIZE;
2828 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2829 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2830 if (sbi->s_want_extra_isize <
2831 le16_to_cpu(es->s_want_extra_isize))
2832 sbi->s_want_extra_isize =
2833 le16_to_cpu(es->s_want_extra_isize);
2834 if (sbi->s_want_extra_isize <
2835 le16_to_cpu(es->s_min_extra_isize))
2836 sbi->s_want_extra_isize =
2837 le16_to_cpu(es->s_min_extra_isize);
2840 /* Check if enough inode space is available */
2841 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2842 sbi->s_inode_size) {
2843 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2844 EXT4_GOOD_OLD_INODE_SIZE;
2845 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2849 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2850 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2851 "requested data journaling mode");
2852 clear_opt(sbi->s_mount_opt, DELALLOC);
2853 } else if (test_opt(sb, DELALLOC))
2854 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
2856 err = ext4_setup_system_zone(sb);
2858 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2859 "zone (%d)\n", err);
2864 err = ext4_mb_init(sb, needs_recovery);
2866 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2871 sbi->s_kobj.kset = ext4_kset;
2872 init_completion(&sbi->s_kobj_unregister);
2873 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2876 ext4_mb_release(sb);
2877 ext4_ext_release(sb);
2881 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2882 ext4_orphan_cleanup(sb, es);
2883 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2884 if (needs_recovery) {
2885 ext4_msg(sb, KERN_INFO, "recovery complete");
2886 ext4_mark_recovery_complete(sb, es);
2888 if (EXT4_SB(sb)->s_journal) {
2889 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2890 descr = " journalled data mode";
2891 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2892 descr = " ordered data mode";
2894 descr = " writeback data mode";
2896 descr = "out journal";
2898 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
2905 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
2909 ext4_msg(sb, KERN_ERR, "mount failed");
2910 ext4_release_system_zone(sb);
2911 if (sbi->s_journal) {
2912 jbd2_journal_destroy(sbi->s_journal);
2913 sbi->s_journal = NULL;
2916 if (sbi->s_flex_groups) {
2917 if (is_vmalloc_addr(sbi->s_flex_groups))
2918 vfree(sbi->s_flex_groups);
2920 kfree(sbi->s_flex_groups);
2922 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2923 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2924 percpu_counter_destroy(&sbi->s_dirs_counter);
2925 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2927 for (i = 0; i < db_count; i++)
2928 brelse(sbi->s_group_desc[i]);
2929 kfree(sbi->s_group_desc);
2932 remove_proc_entry(sb->s_id, ext4_proc_root);
2935 for (i = 0; i < MAXQUOTAS; i++)
2936 kfree(sbi->s_qf_names[i]);
2938 ext4_blkdev_remove(sbi);
2941 sb->s_fs_info = NULL;
2942 kfree(sbi->s_blockgroup_lock);
2949 * Setup any per-fs journal parameters now. We'll do this both on
2950 * initial mount, once the journal has been initialised but before we've
2951 * done any recovery; and again on any subsequent remount.
2953 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2955 struct ext4_sb_info *sbi = EXT4_SB(sb);
2957 journal->j_commit_interval = sbi->s_commit_interval;
2958 journal->j_min_batch_time = sbi->s_min_batch_time;
2959 journal->j_max_batch_time = sbi->s_max_batch_time;
2961 spin_lock(&journal->j_state_lock);
2962 if (test_opt(sb, BARRIER))
2963 journal->j_flags |= JBD2_BARRIER;
2965 journal->j_flags &= ~JBD2_BARRIER;
2966 if (test_opt(sb, DATA_ERR_ABORT))
2967 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2969 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2970 spin_unlock(&journal->j_state_lock);
2973 static journal_t *ext4_get_journal(struct super_block *sb,
2974 unsigned int journal_inum)
2976 struct inode *journal_inode;
2979 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2981 /* First, test for the existence of a valid inode on disk. Bad
2982 * things happen if we iget() an unused inode, as the subsequent
2983 * iput() will try to delete it. */
2985 journal_inode = ext4_iget(sb, journal_inum);
2986 if (IS_ERR(journal_inode)) {
2987 ext4_msg(sb, KERN_ERR, "no journal found");
2990 if (!journal_inode->i_nlink) {
2991 make_bad_inode(journal_inode);
2992 iput(journal_inode);
2993 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
2997 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2998 journal_inode, journal_inode->i_size);
2999 if (!S_ISREG(journal_inode->i_mode)) {
3000 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3001 iput(journal_inode);
3005 journal = jbd2_journal_init_inode(journal_inode);
3007 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3008 iput(journal_inode);
3011 journal->j_private = sb;
3012 ext4_init_journal_params(sb, journal);
3016 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3019 struct buffer_head *bh;
3023 int hblock, blocksize;
3024 ext4_fsblk_t sb_block;
3025 unsigned long offset;
3026 struct ext4_super_block *es;
3027 struct block_device *bdev;
3029 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3031 bdev = ext4_blkdev_get(j_dev, sb);
3035 if (bd_claim(bdev, sb)) {
3036 ext4_msg(sb, KERN_ERR,
3037 "failed to claim external journal device");
3038 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3042 blocksize = sb->s_blocksize;
3043 hblock = bdev_logical_block_size(bdev);
3044 if (blocksize < hblock) {
3045 ext4_msg(sb, KERN_ERR,
3046 "blocksize too small for journal device");
3050 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3051 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3052 set_blocksize(bdev, blocksize);
3053 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3054 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3055 "external journal");
3059 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3060 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3061 !(le32_to_cpu(es->s_feature_incompat) &
3062 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3063 ext4_msg(sb, KERN_ERR, "external journal has "
3069 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3070 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3075 len = ext4_blocks_count(es);
3076 start = sb_block + 1;
3077 brelse(bh); /* we're done with the superblock */
3079 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3080 start, len, blocksize);
3082 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3085 journal->j_private = sb;
3086 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3087 wait_on_buffer(journal->j_sb_buffer);
3088 if (!buffer_uptodate(journal->j_sb_buffer)) {
3089 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3092 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3093 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3094 "user (unsupported) - %d",
3095 be32_to_cpu(journal->j_superblock->s_nr_users));
3098 EXT4_SB(sb)->journal_bdev = bdev;
3099 ext4_init_journal_params(sb, journal);
3103 jbd2_journal_destroy(journal);
3105 ext4_blkdev_put(bdev);
3109 static int ext4_load_journal(struct super_block *sb,
3110 struct ext4_super_block *es,
3111 unsigned long journal_devnum)
3114 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3117 int really_read_only;
3119 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3121 if (journal_devnum &&
3122 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3123 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3124 "numbers have changed");
3125 journal_dev = new_decode_dev(journal_devnum);
3127 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3129 really_read_only = bdev_read_only(sb->s_bdev);
3132 * Are we loading a blank journal or performing recovery after a
3133 * crash? For recovery, we need to check in advance whether we
3134 * can get read-write access to the device.
3136 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3137 if (sb->s_flags & MS_RDONLY) {
3138 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3139 "required on readonly filesystem");
3140 if (really_read_only) {
3141 ext4_msg(sb, KERN_ERR, "write access "
3142 "unavailable, cannot proceed");
3145 ext4_msg(sb, KERN_INFO, "write access will "
3146 "be enabled during recovery");
3150 if (journal_inum && journal_dev) {
3151 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3152 "and inode journals!");
3157 if (!(journal = ext4_get_journal(sb, journal_inum)))
3160 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3164 if (journal->j_flags & JBD2_BARRIER)
3165 ext4_msg(sb, KERN_INFO, "barriers enabled");
3167 ext4_msg(sb, KERN_INFO, "barriers disabled");
3169 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3170 err = jbd2_journal_update_format(journal);
3172 ext4_msg(sb, KERN_ERR, "error updating journal");
3173 jbd2_journal_destroy(journal);
3178 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3179 err = jbd2_journal_wipe(journal, !really_read_only);
3181 err = jbd2_journal_load(journal);
3184 ext4_msg(sb, KERN_ERR, "error loading journal");
3185 jbd2_journal_destroy(journal);
3189 EXT4_SB(sb)->s_journal = journal;
3190 ext4_clear_journal_err(sb, es);
3192 if (journal_devnum &&
3193 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3194 es->s_journal_dev = cpu_to_le32(journal_devnum);
3196 /* Make sure we flush the recovery flag to disk. */
3197 ext4_commit_super(sb, 1);
3203 static int ext4_commit_super(struct super_block *sb, int sync)
3205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3206 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3211 if (buffer_write_io_error(sbh)) {
3213 * Oh, dear. A previous attempt to write the
3214 * superblock failed. This could happen because the
3215 * USB device was yanked out. Or it could happen to
3216 * be a transient write error and maybe the block will
3217 * be remapped. Nothing we can do but to retry the
3218 * write and hope for the best.
3220 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3221 "superblock detected");
3222 clear_buffer_write_io_error(sbh);
3223 set_buffer_uptodate(sbh);
3225 es->s_wtime = cpu_to_le32(get_seconds());
3226 es->s_kbytes_written =
3227 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3228 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3229 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3230 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3231 &EXT4_SB(sb)->s_freeblocks_counter));
3232 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3233 &EXT4_SB(sb)->s_freeinodes_counter));
3235 BUFFER_TRACE(sbh, "marking dirty");
3236 mark_buffer_dirty(sbh);
3238 error = sync_dirty_buffer(sbh);
3242 error = buffer_write_io_error(sbh);
3244 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3246 clear_buffer_write_io_error(sbh);
3247 set_buffer_uptodate(sbh);
3254 * Have we just finished recovery? If so, and if we are mounting (or
3255 * remounting) the filesystem readonly, then we will end up with a
3256 * consistent fs on disk. Record that fact.
3258 static void ext4_mark_recovery_complete(struct super_block *sb,
3259 struct ext4_super_block *es)
3261 journal_t *journal = EXT4_SB(sb)->s_journal;
3263 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3264 BUG_ON(journal != NULL);
3267 jbd2_journal_lock_updates(journal);
3268 if (jbd2_journal_flush(journal) < 0)
3271 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3272 sb->s_flags & MS_RDONLY) {
3273 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3274 ext4_commit_super(sb, 1);
3278 jbd2_journal_unlock_updates(journal);
3282 * If we are mounting (or read-write remounting) a filesystem whose journal
3283 * has recorded an error from a previous lifetime, move that error to the
3284 * main filesystem now.
3286 static void ext4_clear_journal_err(struct super_block *sb,
3287 struct ext4_super_block *es)
3293 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3295 journal = EXT4_SB(sb)->s_journal;
3298 * Now check for any error status which may have been recorded in the
3299 * journal by a prior ext4_error() or ext4_abort()
3302 j_errno = jbd2_journal_errno(journal);
3306 errstr = ext4_decode_error(sb, j_errno, nbuf);
3307 ext4_warning(sb, __func__, "Filesystem error recorded "
3308 "from previous mount: %s", errstr);
3309 ext4_warning(sb, __func__, "Marking fs in need of "
3310 "filesystem check.");
3312 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3313 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3314 ext4_commit_super(sb, 1);
3316 jbd2_journal_clear_err(journal);
3321 * Force the running and committing transactions to commit,
3322 * and wait on the commit.
3324 int ext4_force_commit(struct super_block *sb)
3329 if (sb->s_flags & MS_RDONLY)
3332 journal = EXT4_SB(sb)->s_journal;
3334 ret = ext4_journal_force_commit(journal);
3339 static void ext4_write_super(struct super_block *sb)
3342 ext4_commit_super(sb, 1);
3346 static int ext4_sync_fs(struct super_block *sb, int wait)
3351 trace_ext4_sync_fs(sb, wait);
3352 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3354 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
3360 * LVM calls this function before a (read-only) snapshot is created. This
3361 * gives us a chance to flush the journal completely and mark the fs clean.
3363 static int ext4_freeze(struct super_block *sb)
3368 if (sb->s_flags & MS_RDONLY)
3371 journal = EXT4_SB(sb)->s_journal;
3373 /* Now we set up the journal barrier. */
3374 jbd2_journal_lock_updates(journal);
3377 * Don't clear the needs_recovery flag if we failed to flush
3380 error = jbd2_journal_flush(journal);
3383 jbd2_journal_unlock_updates(journal);
3387 /* Journal blocked and flushed, clear needs_recovery flag. */
3388 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3389 error = ext4_commit_super(sb, 1);
3396 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3397 * flag here, even though the filesystem is not technically dirty yet.
3399 static int ext4_unfreeze(struct super_block *sb)
3401 if (sb->s_flags & MS_RDONLY)
3405 /* Reset the needs_recovery flag before the fs is unlocked. */
3406 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3407 ext4_commit_super(sb, 1);
3409 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3413 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3415 struct ext4_super_block *es;
3416 struct ext4_sb_info *sbi = EXT4_SB(sb);
3417 ext4_fsblk_t n_blocks_count = 0;
3418 unsigned long old_sb_flags;
3419 struct ext4_mount_options old_opts;
3421 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3429 /* Store the original options */
3431 old_sb_flags = sb->s_flags;
3432 old_opts.s_mount_opt = sbi->s_mount_opt;
3433 old_opts.s_resuid = sbi->s_resuid;
3434 old_opts.s_resgid = sbi->s_resgid;
3435 old_opts.s_commit_interval = sbi->s_commit_interval;
3436 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3437 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3439 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3440 for (i = 0; i < MAXQUOTAS; i++)
3441 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3443 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3444 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3447 * Allow the "check" option to be passed as a remount option.
3449 if (!parse_options(data, sb, NULL, &journal_ioprio,
3450 &n_blocks_count, 1)) {
3455 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3456 ext4_abort(sb, __func__, "Abort forced by user");
3458 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3459 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3463 if (sbi->s_journal) {
3464 ext4_init_journal_params(sb, sbi->s_journal);
3465 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3468 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3469 n_blocks_count > ext4_blocks_count(es)) {
3470 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3475 if (*flags & MS_RDONLY) {
3477 * First of all, the unconditional stuff we have to do
3478 * to disable replay of the journal when we next remount
3480 sb->s_flags |= MS_RDONLY;
3483 * OK, test if we are remounting a valid rw partition
3484 * readonly, and if so set the rdonly flag and then
3485 * mark the partition as valid again.
3487 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3488 (sbi->s_mount_state & EXT4_VALID_FS))
3489 es->s_state = cpu_to_le16(sbi->s_mount_state);
3492 ext4_mark_recovery_complete(sb, es);
3494 /* Make sure we can mount this feature set readwrite */
3495 if (!ext4_feature_set_ok(sb, 0)) {
3500 * Make sure the group descriptor checksums
3501 * are sane. If they aren't, refuse to remount r/w.
3503 for (g = 0; g < sbi->s_groups_count; g++) {
3504 struct ext4_group_desc *gdp =
3505 ext4_get_group_desc(sb, g, NULL);
3507 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3508 ext4_msg(sb, KERN_ERR,
3509 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3510 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3511 le16_to_cpu(gdp->bg_checksum));
3518 * If we have an unprocessed orphan list hanging
3519 * around from a previously readonly bdev mount,
3520 * require a full umount/remount for now.
3522 if (es->s_last_orphan) {
3523 ext4_msg(sb, KERN_WARNING, "Couldn't "
3524 "remount RDWR because of unprocessed "
3525 "orphan inode list. Please "
3526 "umount/remount instead");
3532 * Mounting a RDONLY partition read-write, so reread
3533 * and store the current valid flag. (It may have
3534 * been changed by e2fsck since we originally mounted
3538 ext4_clear_journal_err(sb, es);
3539 sbi->s_mount_state = le16_to_cpu(es->s_state);
3540 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3542 if (!ext4_setup_super(sb, es, 0))
3543 sb->s_flags &= ~MS_RDONLY;
3546 ext4_setup_system_zone(sb);
3547 if (sbi->s_journal == NULL)
3548 ext4_commit_super(sb, 1);
3551 /* Release old quota file names */
3552 for (i = 0; i < MAXQUOTAS; i++)
3553 if (old_opts.s_qf_names[i] &&
3554 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3555 kfree(old_opts.s_qf_names[i]);
3562 sb->s_flags = old_sb_flags;
3563 sbi->s_mount_opt = old_opts.s_mount_opt;
3564 sbi->s_resuid = old_opts.s_resuid;
3565 sbi->s_resgid = old_opts.s_resgid;
3566 sbi->s_commit_interval = old_opts.s_commit_interval;
3567 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3568 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3570 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3571 for (i = 0; i < MAXQUOTAS; i++) {
3572 if (sbi->s_qf_names[i] &&
3573 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3574 kfree(sbi->s_qf_names[i]);
3575 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3583 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3585 struct super_block *sb = dentry->d_sb;
3586 struct ext4_sb_info *sbi = EXT4_SB(sb);
3587 struct ext4_super_block *es = sbi->s_es;
3590 if (test_opt(sb, MINIX_DF)) {
3591 sbi->s_overhead_last = 0;
3592 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3593 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3594 ext4_fsblk_t overhead = 0;
3597 * Compute the overhead (FS structures). This is constant
3598 * for a given filesystem unless the number of block groups
3599 * changes so we cache the previous value until it does.
3603 * All of the blocks before first_data_block are
3606 overhead = le32_to_cpu(es->s_first_data_block);
3609 * Add the overhead attributed to the superblock and
3610 * block group descriptors. If the sparse superblocks
3611 * feature is turned on, then not all groups have this.
3613 for (i = 0; i < ngroups; i++) {
3614 overhead += ext4_bg_has_super(sb, i) +
3615 ext4_bg_num_gdb(sb, i);
3620 * Every block group has an inode bitmap, a block
3621 * bitmap, and an inode table.
3623 overhead += ngroups * (2 + sbi->s_itb_per_group);
3624 sbi->s_overhead_last = overhead;
3626 sbi->s_blocks_last = ext4_blocks_count(es);
3629 buf->f_type = EXT4_SUPER_MAGIC;
3630 buf->f_bsize = sb->s_blocksize;
3631 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3632 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3633 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3634 ext4_free_blocks_count_set(es, buf->f_bfree);
3635 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3636 if (buf->f_bfree < ext4_r_blocks_count(es))
3638 buf->f_files = le32_to_cpu(es->s_inodes_count);
3639 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3640 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3641 buf->f_namelen = EXT4_NAME_LEN;
3642 fsid = le64_to_cpup((void *)es->s_uuid) ^
3643 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3644 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3645 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3650 /* Helper function for writing quotas on sync - we need to start transaction
3651 * before quota file is locked for write. Otherwise the are possible deadlocks:
3652 * Process 1 Process 2
3653 * ext4_create() quota_sync()
3654 * jbd2_journal_start() write_dquot()
3655 * vfs_dq_init() down(dqio_mutex)
3656 * down(dqio_mutex) jbd2_journal_start()
3662 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3664 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3667 static int ext4_write_dquot(struct dquot *dquot)
3671 struct inode *inode;
3673 inode = dquot_to_inode(dquot);
3674 handle = ext4_journal_start(inode,
3675 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3677 return PTR_ERR(handle);
3678 ret = dquot_commit(dquot);
3679 err = ext4_journal_stop(handle);
3685 static int ext4_acquire_dquot(struct dquot *dquot)
3690 handle = ext4_journal_start(dquot_to_inode(dquot),
3691 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3693 return PTR_ERR(handle);
3694 ret = dquot_acquire(dquot);
3695 err = ext4_journal_stop(handle);
3701 static int ext4_release_dquot(struct dquot *dquot)
3706 handle = ext4_journal_start(dquot_to_inode(dquot),
3707 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3708 if (IS_ERR(handle)) {
3709 /* Release dquot anyway to avoid endless cycle in dqput() */
3710 dquot_release(dquot);
3711 return PTR_ERR(handle);
3713 ret = dquot_release(dquot);
3714 err = ext4_journal_stop(handle);
3720 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3722 /* Are we journaling quotas? */
3723 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3724 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3725 dquot_mark_dquot_dirty(dquot);
3726 return ext4_write_dquot(dquot);
3728 return dquot_mark_dquot_dirty(dquot);
3732 static int ext4_write_info(struct super_block *sb, int type)
3737 /* Data block + inode block */
3738 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3740 return PTR_ERR(handle);
3741 ret = dquot_commit_info(sb, type);
3742 err = ext4_journal_stop(handle);
3749 * Turn on quotas during mount time - we need to find
3750 * the quota file and such...
3752 static int ext4_quota_on_mount(struct super_block *sb, int type)
3754 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3755 EXT4_SB(sb)->s_jquota_fmt, type);
3759 * Standard function to be called on quota_on
3761 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3762 char *name, int remount)
3767 if (!test_opt(sb, QUOTA))
3769 /* When remounting, no checks are needed and in fact, name is NULL */
3771 return vfs_quota_on(sb, type, format_id, name, remount);
3773 err = kern_path(name, LOOKUP_FOLLOW, &path);
3777 /* Quotafile not on the same filesystem? */
3778 if (path.mnt->mnt_sb != sb) {
3782 /* Journaling quota? */
3783 if (EXT4_SB(sb)->s_qf_names[type]) {
3784 /* Quotafile not in fs root? */
3785 if (path.dentry->d_parent != sb->s_root)
3786 ext4_msg(sb, KERN_WARNING,
3787 "Quota file not on filesystem root. "
3788 "Journaled quota will not work");
3792 * When we journal data on quota file, we have to flush journal to see
3793 * all updates to the file when we bypass pagecache...
3795 if (EXT4_SB(sb)->s_journal &&
3796 ext4_should_journal_data(path.dentry->d_inode)) {
3798 * We don't need to lock updates but journal_flush() could
3799 * otherwise be livelocked...
3801 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3802 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3803 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3810 err = vfs_quota_on_path(sb, type, format_id, &path);
3815 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3816 * acquiring the locks... As quota files are never truncated and quota code
3817 * itself serializes the operations (and noone else should touch the files)
3818 * we don't have to be afraid of races */
3819 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3820 size_t len, loff_t off)
3822 struct inode *inode = sb_dqopt(sb)->files[type];
3823 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3825 int offset = off & (sb->s_blocksize - 1);
3828 struct buffer_head *bh;
3829 loff_t i_size = i_size_read(inode);
3833 if (off+len > i_size)
3836 while (toread > 0) {
3837 tocopy = sb->s_blocksize - offset < toread ?
3838 sb->s_blocksize - offset : toread;
3839 bh = ext4_bread(NULL, inode, blk, 0, &err);
3842 if (!bh) /* A hole? */
3843 memset(data, 0, tocopy);
3845 memcpy(data, bh->b_data+offset, tocopy);
3855 /* Write to quotafile (we know the transaction is already started and has
3856 * enough credits) */
3857 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3858 const char *data, size_t len, loff_t off)
3860 struct inode *inode = sb_dqopt(sb)->files[type];
3861 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3863 int offset = off & (sb->s_blocksize - 1);
3865 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3866 size_t towrite = len;
3867 struct buffer_head *bh;
3868 handle_t *handle = journal_current_handle();
3870 if (EXT4_SB(sb)->s_journal && !handle) {
3871 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3872 " cancelled because transaction is not started",
3873 (unsigned long long)off, (unsigned long long)len);
3876 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3877 while (towrite > 0) {
3878 tocopy = sb->s_blocksize - offset < towrite ?
3879 sb->s_blocksize - offset : towrite;
3880 bh = ext4_bread(handle, inode, blk, 1, &err);
3883 if (journal_quota) {
3884 err = ext4_journal_get_write_access(handle, bh);
3891 memcpy(bh->b_data+offset, data, tocopy);
3892 flush_dcache_page(bh->b_page);
3895 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3897 /* Always do at least ordered writes for quotas */
3898 err = ext4_jbd2_file_inode(handle, inode);
3899 mark_buffer_dirty(bh);
3910 if (len == towrite) {
3911 mutex_unlock(&inode->i_mutex);
3914 if (inode->i_size < off+len-towrite) {
3915 i_size_write(inode, off+len-towrite);
3916 EXT4_I(inode)->i_disksize = inode->i_size;
3918 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3919 ext4_mark_inode_dirty(handle, inode);
3920 mutex_unlock(&inode->i_mutex);
3921 return len - towrite;
3926 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3927 const char *dev_name, void *data, struct vfsmount *mnt)
3929 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3932 static struct file_system_type ext4_fs_type = {
3933 .owner = THIS_MODULE,
3935 .get_sb = ext4_get_sb,
3936 .kill_sb = kill_block_super,
3937 .fs_flags = FS_REQUIRES_DEV,
3940 #ifdef CONFIG_EXT4DEV_COMPAT
3941 static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3942 const char *dev_name, void *data,struct vfsmount *mnt)
3944 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3945 "to mount using ext4\n", dev_name);
3946 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3947 "will go away by 2.6.31\n", dev_name);
3948 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
3951 static struct file_system_type ext4dev_fs_type = {
3952 .owner = THIS_MODULE,
3954 .get_sb = ext4dev_get_sb,
3955 .kill_sb = kill_block_super,
3956 .fs_flags = FS_REQUIRES_DEV,
3958 MODULE_ALIAS("ext4dev");
3961 static int __init init_ext4_fs(void)
3965 err = init_ext4_system_zone();
3968 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3971 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3972 err = init_ext4_mballoc();
3976 err = init_ext4_xattr();
3979 err = init_inodecache();
3982 err = register_filesystem(&ext4_fs_type);
3985 #ifdef CONFIG_EXT4DEV_COMPAT
3986 err = register_filesystem(&ext4dev_fs_type);
3988 unregister_filesystem(&ext4_fs_type);
3994 destroy_inodecache();
3998 exit_ext4_mballoc();
4000 remove_proc_entry("fs/ext4", NULL);
4001 kset_unregister(ext4_kset);
4003 exit_ext4_system_zone();
4007 static void __exit exit_ext4_fs(void)
4009 unregister_filesystem(&ext4_fs_type);
4010 #ifdef CONFIG_EXT4DEV_COMPAT
4011 unregister_filesystem(&ext4dev_fs_type);
4013 destroy_inodecache();
4015 exit_ext4_mballoc();
4016 remove_proc_entry("fs/ext4", NULL);
4017 kset_unregister(ext4_kset);
4018 exit_ext4_system_zone();
4021 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4022 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4023 MODULE_LICENSE("GPL");
4024 module_init(init_ext4_fs)
4025 module_exit(exit_ext4_fs)