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"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
73 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
81 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
89 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
92 return le32_to_cpu(bg->bg_inode_table_lo) |
93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
97 __u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
105 __u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
113 __u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
121 __u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
129 void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
137 void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
145 void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
153 void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
161 void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
169 void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
177 void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
186 /* Just increment the non-pointer handle value */
187 static handle_t *ext4_get_nojournal(void)
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
195 handle = (handle_t *)ref_cnt;
197 current->journal_info = handle;
202 /* Decrement the non-pointer handle value */
203 static void ext4_put_nojournal(handle_t *handle)
205 unsigned long ref_cnt = (unsigned long)handle;
207 BUG_ON(ref_cnt == 0);
210 handle = (handle_t *)ref_cnt;
212 current->journal_info = handle;
216 * Wrappers for jbd2_journal_start/end.
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
223 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
233 journal = EXT4_SB(sb)->s_journal;
235 if (is_journal_aborted(journal)) {
236 ext4_abort(sb, __func__, "Detected aborted journal");
237 return ERR_PTR(-EROFS);
239 return jbd2_journal_start(journal, nblocks);
241 return ext4_get_nojournal();
245 * The only special thing we need to do here is to make sure that all
246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
247 * that sync() will call the filesystem's write_super callback if
250 int __ext4_journal_stop(const char *where, handle_t *handle)
252 struct super_block *sb;
256 if (!ext4_handle_valid(handle)) {
257 ext4_put_nojournal(handle);
260 sb = handle->h_transaction->t_journal->j_private;
262 rc = jbd2_journal_stop(handle);
267 __ext4_std_error(sb, where, err);
271 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
272 struct buffer_head *bh, handle_t *handle, int err)
275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
277 BUG_ON(!ext4_handle_valid(handle));
280 BUFFER_TRACE(bh, "abort");
285 if (is_handle_aborted(handle))
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
291 jbd2_journal_abort_handle(handle);
294 /* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
297 * On ext2, we can store the error state of the filesystem in the
298 * superblock. That is not possible on ext4, because we may have other
299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
304 * We'll just use the jbd2_journal_abort() error code to record an error in
305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
309 static void ext4_handle_error(struct super_block *sb)
311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316 if (sb->s_flags & MS_RDONLY)
319 if (!test_opt(sb, ERRORS_CONT)) {
320 journal_t *journal = EXT4_SB(sb)->s_journal;
322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
324 jbd2_journal_abort(journal, -EIO);
326 if (test_opt(sb, ERRORS_RO)) {
327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
328 sb->s_flags |= MS_RDONLY;
330 ext4_commit_super(sb, 1);
331 if (test_opt(sb, ERRORS_PANIC))
332 panic("EXT4-fs (device %s): panic forced after error\n",
336 void __ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
347 ext4_handle_error(sb);
350 void ext4_error_inode(const char *function, struct inode *inode,
351 const char *fmt, ...)
356 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
357 inode->i_sb->s_id, function, inode->i_ino, current->comm);
362 ext4_handle_error(inode->i_sb);
365 void ext4_error_file(const char *function, struct file *file,
366 const char *fmt, ...)
369 struct inode *inode = file->f_dentry->d_inode;
370 char pathname[80], *path;
373 path = d_path(&(file->f_path), pathname, sizeof(pathname));
377 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
378 inode->i_sb->s_id, function, inode->i_ino, current->comm, path);
383 ext4_handle_error(inode->i_sb);
386 static const char *ext4_decode_error(struct super_block *sb, int errno,
393 errstr = "IO failure";
396 errstr = "Out of memory";
399 if (!sb || (EXT4_SB(sb)->s_journal &&
400 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
401 errstr = "Journal has aborted";
403 errstr = "Readonly filesystem";
406 /* If the caller passed in an extra buffer for unknown
407 * errors, textualise them now. Else we just return
410 /* Check for truncated error codes... */
411 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
420 /* __ext4_std_error decodes expected errors from journaling functions
421 * automatically and invokes the appropriate error response. */
423 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
428 /* Special case: if the error is EROFS, and we're not already
429 * inside a transaction, then there's really no point in logging
431 if (errno == -EROFS && journal_current_handle() == NULL &&
432 (sb->s_flags & MS_RDONLY))
435 errstr = ext4_decode_error(sb, errno, nbuf);
436 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
437 sb->s_id, function, errstr);
439 ext4_handle_error(sb);
443 * ext4_abort is a much stronger failure handler than ext4_error. The
444 * abort function may be used to deal with unrecoverable failures such
445 * as journal IO errors or ENOMEM at a critical moment in log management.
447 * We unconditionally force the filesystem into an ABORT|READONLY state,
448 * unless the error response on the fs has been set to panic in which
449 * case we take the easy way out and panic immediately.
452 void ext4_abort(struct super_block *sb, const char *function,
453 const char *fmt, ...)
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
463 if (test_opt(sb, ERRORS_PANIC))
464 panic("EXT4-fs panic from previous error\n");
466 if (sb->s_flags & MS_RDONLY)
469 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
470 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
471 sb->s_flags |= MS_RDONLY;
472 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
473 if (EXT4_SB(sb)->s_journal)
474 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
477 void ext4_msg (struct super_block * sb, const char *prefix,
478 const char *fmt, ...)
483 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
489 void __ext4_warning(struct super_block *sb, const char *function,
490 const char *fmt, ...)
495 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
502 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
503 const char *function, const char *fmt, ...)
508 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
511 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
516 if (test_opt(sb, ERRORS_CONT)) {
517 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
518 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
519 ext4_commit_super(sb, 0);
522 ext4_unlock_group(sb, grp);
523 ext4_handle_error(sb);
525 * We only get here in the ERRORS_RO case; relocking the group
526 * may be dangerous, but nothing bad will happen since the
527 * filesystem will have already been marked read/only and the
528 * journal has been aborted. We return 1 as a hint to callers
529 * who might what to use the return value from
530 * ext4_grp_locked_error() to distinguish beween the
531 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
532 * aggressively from the ext4 function in question, with a
533 * more appropriate error code.
535 ext4_lock_group(sb, grp);
539 void ext4_update_dynamic_rev(struct super_block *sb)
541 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
543 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
547 "updating to rev %d because of new feature flag, "
548 "running e2fsck is recommended",
551 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
552 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
553 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
554 /* leave es->s_feature_*compat flags alone */
555 /* es->s_uuid will be set by e2fsck if empty */
558 * The rest of the superblock fields should be zero, and if not it
559 * means they are likely already in use, so leave them alone. We
560 * can leave it up to e2fsck to clean up any inconsistencies there.
565 * Open the external journal device
567 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
569 struct block_device *bdev;
570 char b[BDEVNAME_SIZE];
572 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
578 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
579 __bdevname(dev, b), PTR_ERR(bdev));
584 * Release the journal device
586 static int ext4_blkdev_put(struct block_device *bdev)
589 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
592 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
594 struct block_device *bdev;
597 bdev = sbi->journal_bdev;
599 ret = ext4_blkdev_put(bdev);
600 sbi->journal_bdev = NULL;
605 static inline struct inode *orphan_list_entry(struct list_head *l)
607 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
610 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
614 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
615 le32_to_cpu(sbi->s_es->s_last_orphan));
617 printk(KERN_ERR "sb_info orphan list:\n");
618 list_for_each(l, &sbi->s_orphan) {
619 struct inode *inode = orphan_list_entry(l);
621 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
622 inode->i_sb->s_id, inode->i_ino, inode,
623 inode->i_mode, inode->i_nlink,
628 static void ext4_put_super(struct super_block *sb)
630 struct ext4_sb_info *sbi = EXT4_SB(sb);
631 struct ext4_super_block *es = sbi->s_es;
634 flush_workqueue(sbi->dio_unwritten_wq);
635 destroy_workqueue(sbi->dio_unwritten_wq);
640 ext4_commit_super(sb, 1);
642 if (sbi->s_journal) {
643 err = jbd2_journal_destroy(sbi->s_journal);
644 sbi->s_journal = NULL;
646 ext4_abort(sb, __func__,
647 "Couldn't clean up the journal");
650 ext4_release_system_zone(sb);
652 ext4_ext_release(sb);
653 ext4_xattr_put_super(sb);
655 if (!(sb->s_flags & MS_RDONLY)) {
656 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
657 es->s_state = cpu_to_le16(sbi->s_mount_state);
658 ext4_commit_super(sb, 1);
661 remove_proc_entry(sb->s_id, ext4_proc_root);
663 kobject_del(&sbi->s_kobj);
665 for (i = 0; i < sbi->s_gdb_count; i++)
666 brelse(sbi->s_group_desc[i]);
667 kfree(sbi->s_group_desc);
668 if (is_vmalloc_addr(sbi->s_flex_groups))
669 vfree(sbi->s_flex_groups);
671 kfree(sbi->s_flex_groups);
672 percpu_counter_destroy(&sbi->s_freeblocks_counter);
673 percpu_counter_destroy(&sbi->s_freeinodes_counter);
674 percpu_counter_destroy(&sbi->s_dirs_counter);
675 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
678 for (i = 0; i < MAXQUOTAS; i++)
679 kfree(sbi->s_qf_names[i]);
682 /* Debugging code just in case the in-memory inode orphan list
683 * isn't empty. The on-disk one can be non-empty if we've
684 * detected an error and taken the fs readonly, but the
685 * in-memory list had better be clean by this point. */
686 if (!list_empty(&sbi->s_orphan))
687 dump_orphan_list(sb, sbi);
688 J_ASSERT(list_empty(&sbi->s_orphan));
690 invalidate_bdev(sb->s_bdev);
691 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
693 * Invalidate the journal device's buffers. We don't want them
694 * floating about in memory - the physical journal device may
695 * hotswapped, and it breaks the `ro-after' testing code.
697 sync_blockdev(sbi->journal_bdev);
698 invalidate_bdev(sbi->journal_bdev);
699 ext4_blkdev_remove(sbi);
701 sb->s_fs_info = NULL;
703 * Now that we are completely done shutting down the
704 * superblock, we need to actually destroy the kobject.
708 kobject_put(&sbi->s_kobj);
709 wait_for_completion(&sbi->s_kobj_unregister);
710 kfree(sbi->s_blockgroup_lock);
714 static struct kmem_cache *ext4_inode_cachep;
717 * Called inside transaction, so use GFP_NOFS
719 static struct inode *ext4_alloc_inode(struct super_block *sb)
721 struct ext4_inode_info *ei;
723 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
727 ei->vfs_inode.i_version = 1;
728 ei->vfs_inode.i_data.writeback_index = 0;
729 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
730 INIT_LIST_HEAD(&ei->i_prealloc_list);
731 spin_lock_init(&ei->i_prealloc_lock);
733 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
734 * therefore it can be null here. Don't check it, just initialize
737 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
738 ei->i_reserved_data_blocks = 0;
739 ei->i_reserved_meta_blocks = 0;
740 ei->i_allocated_meta_blocks = 0;
741 ei->i_da_metadata_calc_len = 0;
742 ei->i_delalloc_reserved_flag = 0;
743 spin_lock_init(&(ei->i_block_reservation_lock));
745 ei->i_reserved_quota = 0;
747 INIT_LIST_HEAD(&ei->i_completed_io_list);
748 spin_lock_init(&ei->i_completed_io_lock);
749 ei->cur_aio_dio = NULL;
751 ei->i_datasync_tid = 0;
753 return &ei->vfs_inode;
756 static void ext4_destroy_inode(struct inode *inode)
758 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
759 ext4_msg(inode->i_sb, KERN_ERR,
760 "Inode %lu (%p): orphan list check failed!",
761 inode->i_ino, EXT4_I(inode));
762 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
763 EXT4_I(inode), sizeof(struct ext4_inode_info),
767 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
770 static void init_once(void *foo)
772 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
774 INIT_LIST_HEAD(&ei->i_orphan);
775 #ifdef CONFIG_EXT4_FS_XATTR
776 init_rwsem(&ei->xattr_sem);
778 init_rwsem(&ei->i_data_sem);
779 inode_init_once(&ei->vfs_inode);
782 static int init_inodecache(void)
784 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
785 sizeof(struct ext4_inode_info),
786 0, (SLAB_RECLAIM_ACCOUNT|
789 if (ext4_inode_cachep == NULL)
794 static void destroy_inodecache(void)
796 kmem_cache_destroy(ext4_inode_cachep);
799 static void ext4_clear_inode(struct inode *inode)
801 ext4_discard_preallocations(inode);
802 if (EXT4_JOURNAL(inode))
803 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
804 &EXT4_I(inode)->jinode);
807 static inline void ext4_show_quota_options(struct seq_file *seq,
808 struct super_block *sb)
810 #if defined(CONFIG_QUOTA)
811 struct ext4_sb_info *sbi = EXT4_SB(sb);
813 if (sbi->s_jquota_fmt) {
816 switch (sbi->s_jquota_fmt) {
827 seq_printf(seq, ",jqfmt=%s", fmtname);
830 if (sbi->s_qf_names[USRQUOTA])
831 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
833 if (sbi->s_qf_names[GRPQUOTA])
834 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
836 if (test_opt(sb, USRQUOTA))
837 seq_puts(seq, ",usrquota");
839 if (test_opt(sb, GRPQUOTA))
840 seq_puts(seq, ",grpquota");
846 * - it's set to a non-default value OR
847 * - if the per-sb default is different from the global default
849 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
852 unsigned long def_mount_opts;
853 struct super_block *sb = vfs->mnt_sb;
854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
857 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
858 def_errors = le16_to_cpu(es->s_errors);
860 if (sbi->s_sb_block != 1)
861 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
862 if (test_opt(sb, MINIX_DF))
863 seq_puts(seq, ",minixdf");
864 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
865 seq_puts(seq, ",grpid");
866 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
867 seq_puts(seq, ",nogrpid");
868 if (sbi->s_resuid != EXT4_DEF_RESUID ||
869 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
870 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
872 if (sbi->s_resgid != EXT4_DEF_RESGID ||
873 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
874 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
876 if (test_opt(sb, ERRORS_RO)) {
877 if (def_errors == EXT4_ERRORS_PANIC ||
878 def_errors == EXT4_ERRORS_CONTINUE) {
879 seq_puts(seq, ",errors=remount-ro");
882 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
883 seq_puts(seq, ",errors=continue");
884 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
885 seq_puts(seq, ",errors=panic");
886 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
887 seq_puts(seq, ",nouid32");
888 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
889 seq_puts(seq, ",debug");
890 if (test_opt(sb, OLDALLOC))
891 seq_puts(seq, ",oldalloc");
892 #ifdef CONFIG_EXT4_FS_XATTR
893 if (test_opt(sb, XATTR_USER) &&
894 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
895 seq_puts(seq, ",user_xattr");
896 if (!test_opt(sb, XATTR_USER) &&
897 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
898 seq_puts(seq, ",nouser_xattr");
901 #ifdef CONFIG_EXT4_FS_POSIX_ACL
902 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
903 seq_puts(seq, ",acl");
904 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
905 seq_puts(seq, ",noacl");
907 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
908 seq_printf(seq, ",commit=%u",
909 (unsigned) (sbi->s_commit_interval / HZ));
911 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
912 seq_printf(seq, ",min_batch_time=%u",
913 (unsigned) sbi->s_min_batch_time);
915 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
916 seq_printf(seq, ",max_batch_time=%u",
917 (unsigned) sbi->s_min_batch_time);
921 * We're changing the default of barrier mount option, so
922 * let's always display its mount state so it's clear what its
925 seq_puts(seq, ",barrier=");
926 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
927 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
928 seq_puts(seq, ",journal_async_commit");
929 if (test_opt(sb, NOBH))
930 seq_puts(seq, ",nobh");
931 if (test_opt(sb, I_VERSION))
932 seq_puts(seq, ",i_version");
933 if (!test_opt(sb, DELALLOC))
934 seq_puts(seq, ",nodelalloc");
938 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
940 * journal mode get enabled in different ways
941 * So just print the value even if we didn't specify it
943 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
944 seq_puts(seq, ",data=journal");
945 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
946 seq_puts(seq, ",data=ordered");
947 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
948 seq_puts(seq, ",data=writeback");
950 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
951 seq_printf(seq, ",inode_readahead_blks=%u",
952 sbi->s_inode_readahead_blks);
954 if (test_opt(sb, DATA_ERR_ABORT))
955 seq_puts(seq, ",data_err=abort");
957 if (test_opt(sb, NO_AUTO_DA_ALLOC))
958 seq_puts(seq, ",noauto_da_alloc");
960 if (test_opt(sb, DISCARD))
961 seq_puts(seq, ",discard");
963 if (test_opt(sb, NOLOAD))
964 seq_puts(seq, ",norecovery");
966 if (test_opt(sb, DIOREAD_NOLOCK))
967 seq_puts(seq, ",dioread_nolock");
969 ext4_show_quota_options(seq, sb);
974 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
975 u64 ino, u32 generation)
979 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
980 return ERR_PTR(-ESTALE);
981 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
982 return ERR_PTR(-ESTALE);
984 /* iget isn't really right if the inode is currently unallocated!!
986 * ext4_read_inode will return a bad_inode if the inode had been
987 * deleted, so we should be safe.
989 * Currently we don't know the generation for parent directory, so
990 * a generation of 0 means "accept any"
992 inode = ext4_iget(sb, ino);
994 return ERR_CAST(inode);
995 if (generation && inode->i_generation != generation) {
997 return ERR_PTR(-ESTALE);
1003 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1004 int fh_len, int fh_type)
1006 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1007 ext4_nfs_get_inode);
1010 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1011 int fh_len, int fh_type)
1013 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1014 ext4_nfs_get_inode);
1018 * Try to release metadata pages (indirect blocks, directories) which are
1019 * mapped via the block device. Since these pages could have journal heads
1020 * which would prevent try_to_free_buffers() from freeing them, we must use
1021 * jbd2 layer's try_to_free_buffers() function to release them.
1023 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1026 journal_t *journal = EXT4_SB(sb)->s_journal;
1028 WARN_ON(PageChecked(page));
1029 if (!page_has_buffers(page))
1032 return jbd2_journal_try_to_free_buffers(journal, page,
1033 wait & ~__GFP_WAIT);
1034 return try_to_free_buffers(page);
1038 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1039 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1041 static int ext4_write_dquot(struct dquot *dquot);
1042 static int ext4_acquire_dquot(struct dquot *dquot);
1043 static int ext4_release_dquot(struct dquot *dquot);
1044 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1045 static int ext4_write_info(struct super_block *sb, int type);
1046 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1047 char *path, int remount);
1048 static int ext4_quota_on_mount(struct super_block *sb, int type);
1049 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1050 size_t len, loff_t off);
1051 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1052 const char *data, size_t len, loff_t off);
1054 static const struct dquot_operations ext4_quota_operations = {
1055 .initialize = dquot_initialize,
1057 .alloc_space = dquot_alloc_space,
1058 .reserve_space = dquot_reserve_space,
1059 .claim_space = dquot_claim_space,
1060 .release_rsv = dquot_release_reserved_space,
1062 .get_reserved_space = ext4_get_reserved_space,
1064 .alloc_inode = dquot_alloc_inode,
1065 .free_space = dquot_free_space,
1066 .free_inode = dquot_free_inode,
1067 .transfer = dquot_transfer,
1068 .write_dquot = ext4_write_dquot,
1069 .acquire_dquot = ext4_acquire_dquot,
1070 .release_dquot = ext4_release_dquot,
1071 .mark_dirty = ext4_mark_dquot_dirty,
1072 .write_info = ext4_write_info,
1073 .alloc_dquot = dquot_alloc,
1074 .destroy_dquot = dquot_destroy,
1077 static const struct quotactl_ops ext4_qctl_operations = {
1078 .quota_on = ext4_quota_on,
1079 .quota_off = vfs_quota_off,
1080 .quota_sync = vfs_quota_sync,
1081 .get_info = vfs_get_dqinfo,
1082 .set_info = vfs_set_dqinfo,
1083 .get_dqblk = vfs_get_dqblk,
1084 .set_dqblk = vfs_set_dqblk
1088 static const struct super_operations ext4_sops = {
1089 .alloc_inode = ext4_alloc_inode,
1090 .destroy_inode = ext4_destroy_inode,
1091 .write_inode = ext4_write_inode,
1092 .dirty_inode = ext4_dirty_inode,
1093 .delete_inode = ext4_delete_inode,
1094 .put_super = ext4_put_super,
1095 .sync_fs = ext4_sync_fs,
1096 .freeze_fs = ext4_freeze,
1097 .unfreeze_fs = ext4_unfreeze,
1098 .statfs = ext4_statfs,
1099 .remount_fs = ext4_remount,
1100 .clear_inode = ext4_clear_inode,
1101 .show_options = ext4_show_options,
1103 .quota_read = ext4_quota_read,
1104 .quota_write = ext4_quota_write,
1106 .bdev_try_to_free_page = bdev_try_to_free_page,
1109 static const struct super_operations ext4_nojournal_sops = {
1110 .alloc_inode = ext4_alloc_inode,
1111 .destroy_inode = ext4_destroy_inode,
1112 .write_inode = ext4_write_inode,
1113 .dirty_inode = ext4_dirty_inode,
1114 .delete_inode = ext4_delete_inode,
1115 .write_super = ext4_write_super,
1116 .put_super = ext4_put_super,
1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
1119 .clear_inode = ext4_clear_inode,
1120 .show_options = ext4_show_options,
1122 .quota_read = ext4_quota_read,
1123 .quota_write = ext4_quota_write,
1125 .bdev_try_to_free_page = bdev_try_to_free_page,
1128 static const struct export_operations ext4_export_ops = {
1129 .fh_to_dentry = ext4_fh_to_dentry,
1130 .fh_to_parent = ext4_fh_to_parent,
1131 .get_parent = ext4_get_parent,
1135 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1136 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1137 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1138 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1139 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1140 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1141 Opt_journal_update, Opt_journal_dev,
1142 Opt_journal_checksum, Opt_journal_async_commit,
1143 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1144 Opt_data_err_abort, Opt_data_err_ignore,
1145 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1146 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1147 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1148 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1149 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1150 Opt_block_validity, Opt_noblock_validity,
1151 Opt_inode_readahead_blks, Opt_journal_ioprio,
1152 Opt_dioread_nolock, Opt_dioread_lock,
1153 Opt_discard, Opt_nodiscard,
1156 static const match_table_t tokens = {
1157 {Opt_bsd_df, "bsddf"},
1158 {Opt_minix_df, "minixdf"},
1159 {Opt_grpid, "grpid"},
1160 {Opt_grpid, "bsdgroups"},
1161 {Opt_nogrpid, "nogrpid"},
1162 {Opt_nogrpid, "sysvgroups"},
1163 {Opt_resgid, "resgid=%u"},
1164 {Opt_resuid, "resuid=%u"},
1166 {Opt_err_cont, "errors=continue"},
1167 {Opt_err_panic, "errors=panic"},
1168 {Opt_err_ro, "errors=remount-ro"},
1169 {Opt_nouid32, "nouid32"},
1170 {Opt_debug, "debug"},
1171 {Opt_oldalloc, "oldalloc"},
1172 {Opt_orlov, "orlov"},
1173 {Opt_user_xattr, "user_xattr"},
1174 {Opt_nouser_xattr, "nouser_xattr"},
1176 {Opt_noacl, "noacl"},
1177 {Opt_noload, "noload"},
1178 {Opt_noload, "norecovery"},
1181 {Opt_commit, "commit=%u"},
1182 {Opt_min_batch_time, "min_batch_time=%u"},
1183 {Opt_max_batch_time, "max_batch_time=%u"},
1184 {Opt_journal_update, "journal=update"},
1185 {Opt_journal_dev, "journal_dev=%u"},
1186 {Opt_journal_checksum, "journal_checksum"},
1187 {Opt_journal_async_commit, "journal_async_commit"},
1188 {Opt_abort, "abort"},
1189 {Opt_data_journal, "data=journal"},
1190 {Opt_data_ordered, "data=ordered"},
1191 {Opt_data_writeback, "data=writeback"},
1192 {Opt_data_err_abort, "data_err=abort"},
1193 {Opt_data_err_ignore, "data_err=ignore"},
1194 {Opt_offusrjquota, "usrjquota="},
1195 {Opt_usrjquota, "usrjquota=%s"},
1196 {Opt_offgrpjquota, "grpjquota="},
1197 {Opt_grpjquota, "grpjquota=%s"},
1198 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1199 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1200 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1201 {Opt_grpquota, "grpquota"},
1202 {Opt_noquota, "noquota"},
1203 {Opt_quota, "quota"},
1204 {Opt_usrquota, "usrquota"},
1205 {Opt_barrier, "barrier=%u"},
1206 {Opt_barrier, "barrier"},
1207 {Opt_nobarrier, "nobarrier"},
1208 {Opt_i_version, "i_version"},
1209 {Opt_stripe, "stripe=%u"},
1210 {Opt_resize, "resize"},
1211 {Opt_delalloc, "delalloc"},
1212 {Opt_nodelalloc, "nodelalloc"},
1213 {Opt_block_validity, "block_validity"},
1214 {Opt_noblock_validity, "noblock_validity"},
1215 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1216 {Opt_journal_ioprio, "journal_ioprio=%u"},
1217 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1218 {Opt_auto_da_alloc, "auto_da_alloc"},
1219 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1220 {Opt_dioread_nolock, "dioread_nolock"},
1221 {Opt_dioread_lock, "dioread_lock"},
1222 {Opt_discard, "discard"},
1223 {Opt_nodiscard, "nodiscard"},
1227 static ext4_fsblk_t get_sb_block(void **data)
1229 ext4_fsblk_t sb_block;
1230 char *options = (char *) *data;
1232 if (!options || strncmp(options, "sb=", 3) != 0)
1233 return 1; /* Default location */
1236 /* TODO: use simple_strtoll with >32bit ext4 */
1237 sb_block = simple_strtoul(options, &options, 0);
1238 if (*options && *options != ',') {
1239 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1243 if (*options == ',')
1245 *data = (void *) options;
1250 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1251 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1252 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1255 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1257 struct ext4_sb_info *sbi = EXT4_SB(sb);
1260 if (sb_any_quota_loaded(sb) &&
1261 !sbi->s_qf_names[qtype]) {
1262 ext4_msg(sb, KERN_ERR,
1263 "Cannot change journaled "
1264 "quota options when quota turned on");
1267 qname = match_strdup(args);
1269 ext4_msg(sb, KERN_ERR,
1270 "Not enough memory for storing quotafile name");
1273 if (sbi->s_qf_names[qtype] &&
1274 strcmp(sbi->s_qf_names[qtype], qname)) {
1275 ext4_msg(sb, KERN_ERR,
1276 "%s quota file already specified", QTYPE2NAME(qtype));
1280 sbi->s_qf_names[qtype] = qname;
1281 if (strchr(sbi->s_qf_names[qtype], '/')) {
1282 ext4_msg(sb, KERN_ERR,
1283 "quotafile must be on filesystem root");
1284 kfree(sbi->s_qf_names[qtype]);
1285 sbi->s_qf_names[qtype] = NULL;
1288 set_opt(sbi->s_mount_opt, QUOTA);
1292 static int clear_qf_name(struct super_block *sb, int qtype)
1295 struct ext4_sb_info *sbi = EXT4_SB(sb);
1297 if (sb_any_quota_loaded(sb) &&
1298 sbi->s_qf_names[qtype]) {
1299 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1300 " when quota turned on");
1304 * The space will be released later when all options are confirmed
1307 sbi->s_qf_names[qtype] = NULL;
1312 static int parse_options(char *options, struct super_block *sb,
1313 unsigned long *journal_devnum,
1314 unsigned int *journal_ioprio,
1315 ext4_fsblk_t *n_blocks_count, int is_remount)
1317 struct ext4_sb_info *sbi = EXT4_SB(sb);
1319 substring_t args[MAX_OPT_ARGS];
1329 while ((p = strsep(&options, ",")) != NULL) {
1335 * Initialize args struct so we know whether arg was
1336 * found; some options take optional arguments.
1338 args[0].to = args[0].from = 0;
1339 token = match_token(p, tokens, args);
1342 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1343 clear_opt(sbi->s_mount_opt, MINIX_DF);
1346 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1347 set_opt(sbi->s_mount_opt, MINIX_DF);
1351 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1352 set_opt(sbi->s_mount_opt, GRPID);
1356 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1357 clear_opt(sbi->s_mount_opt, GRPID);
1361 if (match_int(&args[0], &option))
1363 sbi->s_resuid = option;
1366 if (match_int(&args[0], &option))
1368 sbi->s_resgid = option;
1371 /* handled by get_sb_block() instead of here */
1372 /* *sb_block = match_int(&args[0]); */
1375 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1376 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1377 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1380 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1381 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1382 set_opt(sbi->s_mount_opt, ERRORS_RO);
1385 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1386 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1387 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1390 set_opt(sbi->s_mount_opt, NO_UID32);
1393 set_opt(sbi->s_mount_opt, DEBUG);
1396 set_opt(sbi->s_mount_opt, OLDALLOC);
1399 clear_opt(sbi->s_mount_opt, OLDALLOC);
1401 #ifdef CONFIG_EXT4_FS_XATTR
1402 case Opt_user_xattr:
1403 set_opt(sbi->s_mount_opt, XATTR_USER);
1405 case Opt_nouser_xattr:
1406 clear_opt(sbi->s_mount_opt, XATTR_USER);
1409 case Opt_user_xattr:
1410 case Opt_nouser_xattr:
1411 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1414 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1416 set_opt(sbi->s_mount_opt, POSIX_ACL);
1419 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1424 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1427 case Opt_journal_update:
1429 /* Eventually we will want to be able to create
1430 a journal file here. For now, only allow the
1431 user to specify an existing inode to be the
1434 ext4_msg(sb, KERN_ERR,
1435 "Cannot specify journal on remount");
1438 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1440 case Opt_journal_dev:
1442 ext4_msg(sb, KERN_ERR,
1443 "Cannot specify journal on remount");
1446 if (match_int(&args[0], &option))
1448 *journal_devnum = option;
1450 case Opt_journal_checksum:
1451 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1453 case Opt_journal_async_commit:
1454 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1455 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1458 set_opt(sbi->s_mount_opt, NOLOAD);
1461 if (match_int(&args[0], &option))
1466 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1467 sbi->s_commit_interval = HZ * option;
1469 case Opt_max_batch_time:
1470 if (match_int(&args[0], &option))
1475 option = EXT4_DEF_MAX_BATCH_TIME;
1476 sbi->s_max_batch_time = option;
1478 case Opt_min_batch_time:
1479 if (match_int(&args[0], &option))
1483 sbi->s_min_batch_time = option;
1485 case Opt_data_journal:
1486 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1488 case Opt_data_ordered:
1489 data_opt = EXT4_MOUNT_ORDERED_DATA;
1491 case Opt_data_writeback:
1492 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1495 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1496 ext4_msg(sb, KERN_ERR,
1497 "Cannot change data mode on remount");
1501 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1502 sbi->s_mount_opt |= data_opt;
1505 case Opt_data_err_abort:
1506 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1508 case Opt_data_err_ignore:
1509 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1513 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1517 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1520 case Opt_offusrjquota:
1521 if (!clear_qf_name(sb, USRQUOTA))
1524 case Opt_offgrpjquota:
1525 if (!clear_qf_name(sb, GRPQUOTA))
1529 case Opt_jqfmt_vfsold:
1530 qfmt = QFMT_VFS_OLD;
1532 case Opt_jqfmt_vfsv0:
1535 case Opt_jqfmt_vfsv1:
1538 if (sb_any_quota_loaded(sb) &&
1539 sbi->s_jquota_fmt != qfmt) {
1540 ext4_msg(sb, KERN_ERR, "Cannot change "
1541 "journaled quota options when "
1545 sbi->s_jquota_fmt = qfmt;
1549 set_opt(sbi->s_mount_opt, QUOTA);
1550 set_opt(sbi->s_mount_opt, USRQUOTA);
1553 set_opt(sbi->s_mount_opt, QUOTA);
1554 set_opt(sbi->s_mount_opt, GRPQUOTA);
1557 if (sb_any_quota_loaded(sb)) {
1558 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1559 "options when quota turned on");
1562 clear_opt(sbi->s_mount_opt, QUOTA);
1563 clear_opt(sbi->s_mount_opt, USRQUOTA);
1564 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1570 ext4_msg(sb, KERN_ERR,
1571 "quota options not supported");
1575 case Opt_offusrjquota:
1576 case Opt_offgrpjquota:
1577 case Opt_jqfmt_vfsold:
1578 case Opt_jqfmt_vfsv0:
1579 case Opt_jqfmt_vfsv1:
1580 ext4_msg(sb, KERN_ERR,
1581 "journaled quota options not supported");
1587 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1590 clear_opt(sbi->s_mount_opt, BARRIER);
1594 if (match_int(&args[0], &option))
1597 option = 1; /* No argument, default to 1 */
1599 set_opt(sbi->s_mount_opt, BARRIER);
1601 clear_opt(sbi->s_mount_opt, BARRIER);
1607 ext4_msg(sb, KERN_ERR,
1608 "resize option only available "
1612 if (match_int(&args[0], &option) != 0)
1614 *n_blocks_count = option;
1617 set_opt(sbi->s_mount_opt, NOBH);
1620 clear_opt(sbi->s_mount_opt, NOBH);
1623 set_opt(sbi->s_mount_opt, I_VERSION);
1624 sb->s_flags |= MS_I_VERSION;
1626 case Opt_nodelalloc:
1627 clear_opt(sbi->s_mount_opt, DELALLOC);
1630 if (match_int(&args[0], &option))
1634 sbi->s_stripe = option;
1637 set_opt(sbi->s_mount_opt, DELALLOC);
1639 case Opt_block_validity:
1640 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1642 case Opt_noblock_validity:
1643 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1645 case Opt_inode_readahead_blks:
1646 if (match_int(&args[0], &option))
1648 if (option < 0 || option > (1 << 30))
1650 if (!is_power_of_2(option)) {
1651 ext4_msg(sb, KERN_ERR,
1652 "EXT4-fs: inode_readahead_blks"
1653 " must be a power of 2");
1656 sbi->s_inode_readahead_blks = option;
1658 case Opt_journal_ioprio:
1659 if (match_int(&args[0], &option))
1661 if (option < 0 || option > 7)
1663 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1666 case Opt_noauto_da_alloc:
1667 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1669 case Opt_auto_da_alloc:
1671 if (match_int(&args[0], &option))
1674 option = 1; /* No argument, default to 1 */
1676 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1678 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1681 set_opt(sbi->s_mount_opt, DISCARD);
1684 clear_opt(sbi->s_mount_opt, DISCARD);
1686 case Opt_dioread_nolock:
1687 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1689 case Opt_dioread_lock:
1690 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1693 ext4_msg(sb, KERN_ERR,
1694 "Unrecognized mount option \"%s\" "
1695 "or missing value", p);
1700 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1701 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1702 clear_opt(sbi->s_mount_opt, USRQUOTA);
1704 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1705 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1707 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1708 ext4_msg(sb, KERN_ERR, "old and new quota "
1713 if (!sbi->s_jquota_fmt) {
1714 ext4_msg(sb, KERN_ERR, "journaled quota format "
1719 if (sbi->s_jquota_fmt) {
1720 ext4_msg(sb, KERN_ERR, "journaled quota format "
1721 "specified with no journaling "
1730 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1733 struct ext4_sb_info *sbi = EXT4_SB(sb);
1736 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1737 ext4_msg(sb, KERN_ERR, "revision level too high, "
1738 "forcing read-only mode");
1743 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1744 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1745 "running e2fsck is recommended");
1746 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1747 ext4_msg(sb, KERN_WARNING,
1748 "warning: mounting fs with errors, "
1749 "running e2fsck is recommended");
1750 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1751 le16_to_cpu(es->s_mnt_count) >=
1752 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1753 ext4_msg(sb, KERN_WARNING,
1754 "warning: maximal mount count reached, "
1755 "running e2fsck is recommended");
1756 else if (le32_to_cpu(es->s_checkinterval) &&
1757 (le32_to_cpu(es->s_lastcheck) +
1758 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1759 ext4_msg(sb, KERN_WARNING,
1760 "warning: checktime reached, "
1761 "running e2fsck is recommended");
1762 if (!sbi->s_journal)
1763 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1764 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1765 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1766 le16_add_cpu(&es->s_mnt_count, 1);
1767 es->s_mtime = cpu_to_le32(get_seconds());
1768 ext4_update_dynamic_rev(sb);
1770 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1772 ext4_commit_super(sb, 1);
1773 if (test_opt(sb, DEBUG))
1774 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1775 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1777 sbi->s_groups_count,
1778 EXT4_BLOCKS_PER_GROUP(sb),
1779 EXT4_INODES_PER_GROUP(sb),
1785 static int ext4_fill_flex_info(struct super_block *sb)
1787 struct ext4_sb_info *sbi = EXT4_SB(sb);
1788 struct ext4_group_desc *gdp = NULL;
1789 ext4_group_t flex_group_count;
1790 ext4_group_t flex_group;
1791 int groups_per_flex = 0;
1795 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1796 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1798 if (groups_per_flex < 2) {
1799 sbi->s_log_groups_per_flex = 0;
1803 /* We allocate both existing and potentially added groups */
1804 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1805 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1806 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1807 size = flex_group_count * sizeof(struct flex_groups);
1808 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1809 if (sbi->s_flex_groups == NULL) {
1810 sbi->s_flex_groups = vmalloc(size);
1811 if (sbi->s_flex_groups)
1812 memset(sbi->s_flex_groups, 0, size);
1814 if (sbi->s_flex_groups == NULL) {
1815 ext4_msg(sb, KERN_ERR, "not enough memory for "
1816 "%u flex groups", flex_group_count);
1820 for (i = 0; i < sbi->s_groups_count; i++) {
1821 gdp = ext4_get_group_desc(sb, i, NULL);
1823 flex_group = ext4_flex_group(sbi, i);
1824 atomic_add(ext4_free_inodes_count(sb, gdp),
1825 &sbi->s_flex_groups[flex_group].free_inodes);
1826 atomic_add(ext4_free_blks_count(sb, gdp),
1827 &sbi->s_flex_groups[flex_group].free_blocks);
1828 atomic_add(ext4_used_dirs_count(sb, gdp),
1829 &sbi->s_flex_groups[flex_group].used_dirs);
1837 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1838 struct ext4_group_desc *gdp)
1842 if (sbi->s_es->s_feature_ro_compat &
1843 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1844 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1845 __le32 le_group = cpu_to_le32(block_group);
1847 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1848 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1849 crc = crc16(crc, (__u8 *)gdp, offset);
1850 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1851 /* for checksum of struct ext4_group_desc do the rest...*/
1852 if ((sbi->s_es->s_feature_incompat &
1853 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1854 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1855 crc = crc16(crc, (__u8 *)gdp + offset,
1856 le16_to_cpu(sbi->s_es->s_desc_size) -
1860 return cpu_to_le16(crc);
1863 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1864 struct ext4_group_desc *gdp)
1866 if ((sbi->s_es->s_feature_ro_compat &
1867 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1868 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1874 /* Called at mount-time, super-block is locked */
1875 static int ext4_check_descriptors(struct super_block *sb)
1877 struct ext4_sb_info *sbi = EXT4_SB(sb);
1878 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1879 ext4_fsblk_t last_block;
1880 ext4_fsblk_t block_bitmap;
1881 ext4_fsblk_t inode_bitmap;
1882 ext4_fsblk_t inode_table;
1883 int flexbg_flag = 0;
1886 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1889 ext4_debug("Checking group descriptors");
1891 for (i = 0; i < sbi->s_groups_count; i++) {
1892 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1894 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1895 last_block = ext4_blocks_count(sbi->s_es) - 1;
1897 last_block = first_block +
1898 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1900 block_bitmap = ext4_block_bitmap(sb, gdp);
1901 if (block_bitmap < first_block || block_bitmap > last_block) {
1902 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1903 "Block bitmap for group %u not in group "
1904 "(block %llu)!", i, block_bitmap);
1907 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1908 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1909 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1910 "Inode bitmap for group %u not in group "
1911 "(block %llu)!", i, inode_bitmap);
1914 inode_table = ext4_inode_table(sb, gdp);
1915 if (inode_table < first_block ||
1916 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1917 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1918 "Inode table for group %u not in group "
1919 "(block %llu)!", i, inode_table);
1922 ext4_lock_group(sb, i);
1923 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1925 "Checksum for group %u failed (%u!=%u)",
1926 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1927 gdp)), le16_to_cpu(gdp->bg_checksum));
1928 if (!(sb->s_flags & MS_RDONLY)) {
1929 ext4_unlock_group(sb, i);
1933 ext4_unlock_group(sb, i);
1935 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1938 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1939 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1943 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1944 * the superblock) which were deleted from all directories, but held open by
1945 * a process at the time of a crash. We walk the list and try to delete these
1946 * inodes at recovery time (only with a read-write filesystem).
1948 * In order to keep the orphan inode chain consistent during traversal (in
1949 * case of crash during recovery), we link each inode into the superblock
1950 * orphan list_head and handle it the same way as an inode deletion during
1951 * normal operation (which journals the operations for us).
1953 * We only do an iget() and an iput() on each inode, which is very safe if we
1954 * accidentally point at an in-use or already deleted inode. The worst that
1955 * can happen in this case is that we get a "bit already cleared" message from
1956 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1957 * e2fsck was run on this filesystem, and it must have already done the orphan
1958 * inode cleanup for us, so we can safely abort without any further action.
1960 static void ext4_orphan_cleanup(struct super_block *sb,
1961 struct ext4_super_block *es)
1963 unsigned int s_flags = sb->s_flags;
1964 int nr_orphans = 0, nr_truncates = 0;
1968 if (!es->s_last_orphan) {
1969 jbd_debug(4, "no orphan inodes to clean up\n");
1973 if (bdev_read_only(sb->s_bdev)) {
1974 ext4_msg(sb, KERN_ERR, "write access "
1975 "unavailable, skipping orphan cleanup");
1979 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1980 if (es->s_last_orphan)
1981 jbd_debug(1, "Errors on filesystem, "
1982 "clearing orphan list.\n");
1983 es->s_last_orphan = 0;
1984 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1988 if (s_flags & MS_RDONLY) {
1989 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1990 sb->s_flags &= ~MS_RDONLY;
1993 /* Needed for iput() to work correctly and not trash data */
1994 sb->s_flags |= MS_ACTIVE;
1995 /* Turn on quotas so that they are updated correctly */
1996 for (i = 0; i < MAXQUOTAS; i++) {
1997 if (EXT4_SB(sb)->s_qf_names[i]) {
1998 int ret = ext4_quota_on_mount(sb, i);
2000 ext4_msg(sb, KERN_ERR,
2001 "Cannot turn on journaled "
2002 "quota: error %d", ret);
2007 while (es->s_last_orphan) {
2008 struct inode *inode;
2010 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2011 if (IS_ERR(inode)) {
2012 es->s_last_orphan = 0;
2016 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2018 if (inode->i_nlink) {
2019 ext4_msg(sb, KERN_DEBUG,
2020 "%s: truncating inode %lu to %lld bytes",
2021 __func__, inode->i_ino, inode->i_size);
2022 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2023 inode->i_ino, inode->i_size);
2024 ext4_truncate(inode);
2027 ext4_msg(sb, KERN_DEBUG,
2028 "%s: deleting unreferenced inode %lu",
2029 __func__, inode->i_ino);
2030 jbd_debug(2, "deleting unreferenced inode %lu\n",
2034 iput(inode); /* The delete magic happens here! */
2037 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2040 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2041 PLURAL(nr_orphans));
2043 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2044 PLURAL(nr_truncates));
2046 /* Turn quotas off */
2047 for (i = 0; i < MAXQUOTAS; i++) {
2048 if (sb_dqopt(sb)->files[i])
2049 vfs_quota_off(sb, i, 0);
2052 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2056 * Maximal extent format file size.
2057 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2058 * extent format containers, within a sector_t, and within i_blocks
2059 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2060 * so that won't be a limiting factor.
2062 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2064 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2067 loff_t upper_limit = MAX_LFS_FILESIZE;
2069 /* small i_blocks in vfs inode? */
2070 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2072 * CONFIG_LBDAF is not enabled implies the inode
2073 * i_block represent total blocks in 512 bytes
2074 * 32 == size of vfs inode i_blocks * 8
2076 upper_limit = (1LL << 32) - 1;
2078 /* total blocks in file system block size */
2079 upper_limit >>= (blkbits - 9);
2080 upper_limit <<= blkbits;
2083 /* 32-bit extent-start container, ee_block */
2088 /* Sanity check against vm- & vfs- imposed limits */
2089 if (res > upper_limit)
2096 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2097 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2098 * We need to be 1 filesystem block less than the 2^48 sector limit.
2100 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2102 loff_t res = EXT4_NDIR_BLOCKS;
2105 /* This is calculated to be the largest file size for a dense, block
2106 * mapped file such that the file's total number of 512-byte sectors,
2107 * including data and all indirect blocks, does not exceed (2^48 - 1).
2109 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2110 * number of 512-byte sectors of the file.
2113 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2115 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2116 * the inode i_block field represents total file blocks in
2117 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2119 upper_limit = (1LL << 32) - 1;
2121 /* total blocks in file system block size */
2122 upper_limit >>= (bits - 9);
2126 * We use 48 bit ext4_inode i_blocks
2127 * With EXT4_HUGE_FILE_FL set the i_blocks
2128 * represent total number of blocks in
2129 * file system block size
2131 upper_limit = (1LL << 48) - 1;
2135 /* indirect blocks */
2137 /* double indirect blocks */
2138 meta_blocks += 1 + (1LL << (bits-2));
2139 /* tripple indirect blocks */
2140 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2142 upper_limit -= meta_blocks;
2143 upper_limit <<= bits;
2145 res += 1LL << (bits-2);
2146 res += 1LL << (2*(bits-2));
2147 res += 1LL << (3*(bits-2));
2149 if (res > upper_limit)
2152 if (res > MAX_LFS_FILESIZE)
2153 res = MAX_LFS_FILESIZE;
2158 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2159 ext4_fsblk_t logical_sb_block, int nr)
2161 struct ext4_sb_info *sbi = EXT4_SB(sb);
2162 ext4_group_t bg, first_meta_bg;
2165 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2167 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2169 return logical_sb_block + nr + 1;
2170 bg = sbi->s_desc_per_block * nr;
2171 if (ext4_bg_has_super(sb, bg))
2174 return (has_super + ext4_group_first_block_no(sb, bg));
2178 * ext4_get_stripe_size: Get the stripe size.
2179 * @sbi: In memory super block info
2181 * If we have specified it via mount option, then
2182 * use the mount option value. If the value specified at mount time is
2183 * greater than the blocks per group use the super block value.
2184 * If the super block value is greater than blocks per group return 0.
2185 * Allocator needs it be less than blocks per group.
2188 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2190 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2191 unsigned long stripe_width =
2192 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2194 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2195 return sbi->s_stripe;
2197 if (stripe_width <= sbi->s_blocks_per_group)
2198 return stripe_width;
2200 if (stride <= sbi->s_blocks_per_group)
2209 struct attribute attr;
2210 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2211 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2212 const char *, size_t);
2216 static int parse_strtoul(const char *buf,
2217 unsigned long max, unsigned long *value)
2221 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2222 endp = skip_spaces(endp);
2223 if (*endp || *value > max)
2229 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2230 struct ext4_sb_info *sbi,
2233 return snprintf(buf, PAGE_SIZE, "%llu\n",
2234 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2237 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2238 struct ext4_sb_info *sbi, char *buf)
2240 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2242 return snprintf(buf, PAGE_SIZE, "%lu\n",
2243 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2244 sbi->s_sectors_written_start) >> 1);
2247 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2248 struct ext4_sb_info *sbi, char *buf)
2250 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2252 return snprintf(buf, PAGE_SIZE, "%llu\n",
2253 (unsigned long long)(sbi->s_kbytes_written +
2254 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2255 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2258 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2259 struct ext4_sb_info *sbi,
2260 const char *buf, size_t count)
2264 if (parse_strtoul(buf, 0x40000000, &t))
2267 if (!is_power_of_2(t))
2270 sbi->s_inode_readahead_blks = t;
2274 static ssize_t sbi_ui_show(struct ext4_attr *a,
2275 struct ext4_sb_info *sbi, char *buf)
2277 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2279 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2282 static ssize_t sbi_ui_store(struct ext4_attr *a,
2283 struct ext4_sb_info *sbi,
2284 const char *buf, size_t count)
2286 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2289 if (parse_strtoul(buf, 0xffffffff, &t))
2295 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2296 static struct ext4_attr ext4_attr_##_name = { \
2297 .attr = {.name = __stringify(_name), .mode = _mode }, \
2300 .offset = offsetof(struct ext4_sb_info, _elname), \
2302 #define EXT4_ATTR(name, mode, show, store) \
2303 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2305 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2306 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2307 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2308 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2309 #define ATTR_LIST(name) &ext4_attr_##name.attr
2311 EXT4_RO_ATTR(delayed_allocation_blocks);
2312 EXT4_RO_ATTR(session_write_kbytes);
2313 EXT4_RO_ATTR(lifetime_write_kbytes);
2314 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2315 inode_readahead_blks_store, s_inode_readahead_blks);
2316 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2317 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2318 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2319 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2320 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2321 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2322 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2323 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2325 static struct attribute *ext4_attrs[] = {
2326 ATTR_LIST(delayed_allocation_blocks),
2327 ATTR_LIST(session_write_kbytes),
2328 ATTR_LIST(lifetime_write_kbytes),
2329 ATTR_LIST(inode_readahead_blks),
2330 ATTR_LIST(inode_goal),
2331 ATTR_LIST(mb_stats),
2332 ATTR_LIST(mb_max_to_scan),
2333 ATTR_LIST(mb_min_to_scan),
2334 ATTR_LIST(mb_order2_req),
2335 ATTR_LIST(mb_stream_req),
2336 ATTR_LIST(mb_group_prealloc),
2337 ATTR_LIST(max_writeback_mb_bump),
2341 static ssize_t ext4_attr_show(struct kobject *kobj,
2342 struct attribute *attr, char *buf)
2344 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2346 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2348 return a->show ? a->show(a, sbi, buf) : 0;
2351 static ssize_t ext4_attr_store(struct kobject *kobj,
2352 struct attribute *attr,
2353 const char *buf, size_t len)
2355 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2357 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2359 return a->store ? a->store(a, sbi, buf, len) : 0;
2362 static void ext4_sb_release(struct kobject *kobj)
2364 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2366 complete(&sbi->s_kobj_unregister);
2370 static struct sysfs_ops ext4_attr_ops = {
2371 .show = ext4_attr_show,
2372 .store = ext4_attr_store,
2375 static struct kobj_type ext4_ktype = {
2376 .default_attrs = ext4_attrs,
2377 .sysfs_ops = &ext4_attr_ops,
2378 .release = ext4_sb_release,
2382 * Check whether this filesystem can be mounted based on
2383 * the features present and the RDONLY/RDWR mount requested.
2384 * Returns 1 if this filesystem can be mounted as requested,
2385 * 0 if it cannot be.
2387 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2389 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2390 ext4_msg(sb, KERN_ERR,
2391 "Couldn't mount because of "
2392 "unsupported optional features (%x)",
2393 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2394 ~EXT4_FEATURE_INCOMPAT_SUPP));
2401 /* Check that feature set is OK for a read-write mount */
2402 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2403 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2404 "unsupported optional features (%x)",
2405 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2406 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2410 * Large file size enabled file system can only be mounted
2411 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2413 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2414 if (sizeof(blkcnt_t) < sizeof(u64)) {
2415 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2416 "cannot be mounted RDWR without "
2424 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2425 __releases(kernel_lock)
2426 __acquires(kernel_lock)
2428 struct buffer_head *bh;
2429 struct ext4_super_block *es = NULL;
2430 struct ext4_sb_info *sbi;
2432 ext4_fsblk_t sb_block = get_sb_block(&data);
2433 ext4_fsblk_t logical_sb_block;
2434 unsigned long offset = 0;
2435 unsigned long journal_devnum = 0;
2436 unsigned long def_mount_opts;
2442 unsigned int db_count;
2444 int needs_recovery, has_huge_files;
2447 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2449 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2453 sbi->s_blockgroup_lock =
2454 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2455 if (!sbi->s_blockgroup_lock) {
2459 sb->s_fs_info = sbi;
2460 sbi->s_mount_opt = 0;
2461 sbi->s_resuid = EXT4_DEF_RESUID;
2462 sbi->s_resgid = EXT4_DEF_RESGID;
2463 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2464 sbi->s_sb_block = sb_block;
2465 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2470 /* Cleanup superblock name */
2471 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2474 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2476 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2481 * The ext4 superblock will not be buffer aligned for other than 1kB
2482 * block sizes. We need to calculate the offset from buffer start.
2484 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2485 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2486 offset = do_div(logical_sb_block, blocksize);
2488 logical_sb_block = sb_block;
2491 if (!(bh = sb_bread(sb, logical_sb_block))) {
2492 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2496 * Note: s_es must be initialized as soon as possible because
2497 * some ext4 macro-instructions depend on its value
2499 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2501 sb->s_magic = le16_to_cpu(es->s_magic);
2502 if (sb->s_magic != EXT4_SUPER_MAGIC)
2504 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2506 /* Set defaults before we parse the mount options */
2507 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2508 if (def_mount_opts & EXT4_DEFM_DEBUG)
2509 set_opt(sbi->s_mount_opt, DEBUG);
2510 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2511 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2513 set_opt(sbi->s_mount_opt, GRPID);
2515 if (def_mount_opts & EXT4_DEFM_UID16)
2516 set_opt(sbi->s_mount_opt, NO_UID32);
2517 #ifdef CONFIG_EXT4_FS_XATTR
2518 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2519 set_opt(sbi->s_mount_opt, XATTR_USER);
2521 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2522 if (def_mount_opts & EXT4_DEFM_ACL)
2523 set_opt(sbi->s_mount_opt, POSIX_ACL);
2525 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2526 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2527 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2528 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2529 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2530 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2532 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2533 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2534 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2535 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2537 set_opt(sbi->s_mount_opt, ERRORS_RO);
2539 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2540 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2541 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2542 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2543 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2545 set_opt(sbi->s_mount_opt, BARRIER);
2548 * enable delayed allocation by default
2549 * Use -o nodelalloc to turn it off
2551 set_opt(sbi->s_mount_opt, DELALLOC);
2553 if (!parse_options((char *) data, sb, &journal_devnum,
2554 &journal_ioprio, NULL, 0))
2557 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2558 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2560 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2561 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2562 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2563 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2564 ext4_msg(sb, KERN_WARNING,
2565 "feature flags set on rev 0 fs, "
2566 "running e2fsck is recommended");
2569 * Check feature flags regardless of the revision level, since we
2570 * previously didn't change the revision level when setting the flags,
2571 * so there is a chance incompat flags are set on a rev 0 filesystem.
2573 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2576 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2578 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2579 blocksize > EXT4_MAX_BLOCK_SIZE) {
2580 ext4_msg(sb, KERN_ERR,
2581 "Unsupported filesystem blocksize %d", blocksize);
2585 if (sb->s_blocksize != blocksize) {
2586 /* Validate the filesystem blocksize */
2587 if (!sb_set_blocksize(sb, blocksize)) {
2588 ext4_msg(sb, KERN_ERR, "bad block size %d",
2594 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2595 offset = do_div(logical_sb_block, blocksize);
2596 bh = sb_bread(sb, logical_sb_block);
2598 ext4_msg(sb, KERN_ERR,
2599 "Can't read superblock on 2nd try");
2602 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2604 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2605 ext4_msg(sb, KERN_ERR,
2606 "Magic mismatch, very weird!");
2611 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2612 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2613 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2615 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2617 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2618 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2619 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2621 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2622 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2623 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2624 (!is_power_of_2(sbi->s_inode_size)) ||
2625 (sbi->s_inode_size > blocksize)) {
2626 ext4_msg(sb, KERN_ERR,
2627 "unsupported inode size: %d",
2631 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2632 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2635 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2636 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2637 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2638 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2639 !is_power_of_2(sbi->s_desc_size)) {
2640 ext4_msg(sb, KERN_ERR,
2641 "unsupported descriptor size %lu",
2646 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2648 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2649 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2650 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2653 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2654 if (sbi->s_inodes_per_block == 0)
2656 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2657 sbi->s_inodes_per_block;
2658 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2660 sbi->s_mount_state = le16_to_cpu(es->s_state);
2661 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2662 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2664 for (i = 0; i < 4; i++)
2665 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2666 sbi->s_def_hash_version = es->s_def_hash_version;
2667 i = le32_to_cpu(es->s_flags);
2668 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2669 sbi->s_hash_unsigned = 3;
2670 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2671 #ifdef __CHAR_UNSIGNED__
2672 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2673 sbi->s_hash_unsigned = 3;
2675 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2680 if (sbi->s_blocks_per_group > blocksize * 8) {
2681 ext4_msg(sb, KERN_ERR,
2682 "#blocks per group too big: %lu",
2683 sbi->s_blocks_per_group);
2686 if (sbi->s_inodes_per_group > blocksize * 8) {
2687 ext4_msg(sb, KERN_ERR,
2688 "#inodes per group too big: %lu",
2689 sbi->s_inodes_per_group);
2694 * Test whether we have more sectors than will fit in sector_t,
2695 * and whether the max offset is addressable by the page cache.
2697 if ((ext4_blocks_count(es) >
2698 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2699 (ext4_blocks_count(es) >
2700 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2701 ext4_msg(sb, KERN_ERR, "filesystem"
2702 " too large to mount safely on this system");
2703 if (sizeof(sector_t) < 8)
2704 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2709 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2712 /* check blocks count against device size */
2713 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2714 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2715 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2716 "exceeds size of device (%llu blocks)",
2717 ext4_blocks_count(es), blocks_count);
2722 * It makes no sense for the first data block to be beyond the end
2723 * of the filesystem.
2725 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2726 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2727 "block %u is beyond end of filesystem (%llu)",
2728 le32_to_cpu(es->s_first_data_block),
2729 ext4_blocks_count(es));
2732 blocks_count = (ext4_blocks_count(es) -
2733 le32_to_cpu(es->s_first_data_block) +
2734 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2735 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2736 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2737 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2738 "(block count %llu, first data block %u, "
2739 "blocks per group %lu)", sbi->s_groups_count,
2740 ext4_blocks_count(es),
2741 le32_to_cpu(es->s_first_data_block),
2742 EXT4_BLOCKS_PER_GROUP(sb));
2745 sbi->s_groups_count = blocks_count;
2746 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2747 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2748 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2749 EXT4_DESC_PER_BLOCK(sb);
2750 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2752 if (sbi->s_group_desc == NULL) {
2753 ext4_msg(sb, KERN_ERR, "not enough memory");
2757 #ifdef CONFIG_PROC_FS
2759 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2762 bgl_lock_init(sbi->s_blockgroup_lock);
2764 for (i = 0; i < db_count; i++) {
2765 block = descriptor_loc(sb, logical_sb_block, i);
2766 sbi->s_group_desc[i] = sb_bread(sb, block);
2767 if (!sbi->s_group_desc[i]) {
2768 ext4_msg(sb, KERN_ERR,
2769 "can't read group descriptor %d", i);
2774 if (!ext4_check_descriptors(sb)) {
2775 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2778 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2779 if (!ext4_fill_flex_info(sb)) {
2780 ext4_msg(sb, KERN_ERR,
2781 "unable to initialize "
2782 "flex_bg meta info!");
2786 sbi->s_gdb_count = db_count;
2787 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2788 spin_lock_init(&sbi->s_next_gen_lock);
2790 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2791 ext4_count_free_blocks(sb));
2793 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2794 ext4_count_free_inodes(sb));
2797 err = percpu_counter_init(&sbi->s_dirs_counter,
2798 ext4_count_dirs(sb));
2801 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2804 ext4_msg(sb, KERN_ERR, "insufficient memory");
2808 sbi->s_stripe = ext4_get_stripe_size(sbi);
2809 sbi->s_max_writeback_mb_bump = 128;
2812 * set up enough so that it can read an inode
2814 if (!test_opt(sb, NOLOAD) &&
2815 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2816 sb->s_op = &ext4_sops;
2818 sb->s_op = &ext4_nojournal_sops;
2819 sb->s_export_op = &ext4_export_ops;
2820 sb->s_xattr = ext4_xattr_handlers;
2822 sb->s_qcop = &ext4_qctl_operations;
2823 sb->dq_op = &ext4_quota_operations;
2825 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2826 mutex_init(&sbi->s_orphan_lock);
2827 mutex_init(&sbi->s_resize_lock);
2831 needs_recovery = (es->s_last_orphan != 0 ||
2832 EXT4_HAS_INCOMPAT_FEATURE(sb,
2833 EXT4_FEATURE_INCOMPAT_RECOVER));
2836 * The first inode we look at is the journal inode. Don't try
2837 * root first: it may be modified in the journal!
2839 if (!test_opt(sb, NOLOAD) &&
2840 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2841 if (ext4_load_journal(sb, es, journal_devnum))
2843 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2844 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2845 ext4_msg(sb, KERN_ERR, "required journal recovery "
2846 "suppressed and not mounted read-only");
2847 goto failed_mount_wq;
2849 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2850 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2851 sbi->s_journal = NULL;
2856 if (ext4_blocks_count(es) > 0xffffffffULL &&
2857 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2858 JBD2_FEATURE_INCOMPAT_64BIT)) {
2859 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2860 goto failed_mount_wq;
2863 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2864 jbd2_journal_set_features(sbi->s_journal,
2865 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2866 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2867 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2868 jbd2_journal_set_features(sbi->s_journal,
2869 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2870 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2871 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2873 jbd2_journal_clear_features(sbi->s_journal,
2874 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2875 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2878 /* We have now updated the journal if required, so we can
2879 * validate the data journaling mode. */
2880 switch (test_opt(sb, DATA_FLAGS)) {
2882 /* No mode set, assume a default based on the journal
2883 * capabilities: ORDERED_DATA if the journal can
2884 * cope, else JOURNAL_DATA
2886 if (jbd2_journal_check_available_features
2887 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2888 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2890 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2893 case EXT4_MOUNT_ORDERED_DATA:
2894 case EXT4_MOUNT_WRITEBACK_DATA:
2895 if (!jbd2_journal_check_available_features
2896 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2897 ext4_msg(sb, KERN_ERR, "Journal does not support "
2898 "requested data journaling mode");
2899 goto failed_mount_wq;
2904 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2907 if (test_opt(sb, NOBH)) {
2908 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2909 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2910 "its supported only with writeback mode");
2911 clear_opt(sbi->s_mount_opt, NOBH);
2913 if (test_opt(sb, DIOREAD_NOLOCK)) {
2914 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2915 "not supported with nobh mode");
2916 goto failed_mount_wq;
2919 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2920 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2921 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2922 goto failed_mount_wq;
2926 * The jbd2_journal_load will have done any necessary log recovery,
2927 * so we can safely mount the rest of the filesystem now.
2930 root = ext4_iget(sb, EXT4_ROOT_INO);
2932 ext4_msg(sb, KERN_ERR, "get root inode failed");
2933 ret = PTR_ERR(root);
2936 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2938 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2941 sb->s_root = d_alloc_root(root);
2943 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2949 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2951 /* determine the minimum size of new large inodes, if present */
2952 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2953 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2954 EXT4_GOOD_OLD_INODE_SIZE;
2955 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2956 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2957 if (sbi->s_want_extra_isize <
2958 le16_to_cpu(es->s_want_extra_isize))
2959 sbi->s_want_extra_isize =
2960 le16_to_cpu(es->s_want_extra_isize);
2961 if (sbi->s_want_extra_isize <
2962 le16_to_cpu(es->s_min_extra_isize))
2963 sbi->s_want_extra_isize =
2964 le16_to_cpu(es->s_min_extra_isize);
2967 /* Check if enough inode space is available */
2968 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2969 sbi->s_inode_size) {
2970 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2971 EXT4_GOOD_OLD_INODE_SIZE;
2972 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2976 if (test_opt(sb, DELALLOC) &&
2977 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2978 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2979 "requested data journaling mode");
2980 clear_opt(sbi->s_mount_opt, DELALLOC);
2982 if (test_opt(sb, DIOREAD_NOLOCK)) {
2983 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2984 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2985 "option - requested data journaling mode");
2986 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2988 if (sb->s_blocksize < PAGE_SIZE) {
2989 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2990 "option - block size is too small");
2991 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2995 err = ext4_setup_system_zone(sb);
2997 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2998 "zone (%d)\n", err);
3003 err = ext4_mb_init(sb, needs_recovery);
3005 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3010 sbi->s_kobj.kset = ext4_kset;
3011 init_completion(&sbi->s_kobj_unregister);
3012 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3015 ext4_mb_release(sb);
3016 ext4_ext_release(sb);
3020 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3021 ext4_orphan_cleanup(sb, es);
3022 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3023 if (needs_recovery) {
3024 ext4_msg(sb, KERN_INFO, "recovery complete");
3025 ext4_mark_recovery_complete(sb, es);
3027 if (EXT4_SB(sb)->s_journal) {
3028 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3029 descr = " journalled data mode";
3030 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3031 descr = " ordered data mode";
3033 descr = " writeback data mode";
3035 descr = "out journal";
3037 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
3044 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3048 ext4_msg(sb, KERN_ERR, "mount failed");
3049 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3051 ext4_release_system_zone(sb);
3052 if (sbi->s_journal) {
3053 jbd2_journal_destroy(sbi->s_journal);
3054 sbi->s_journal = NULL;
3057 if (sbi->s_flex_groups) {
3058 if (is_vmalloc_addr(sbi->s_flex_groups))
3059 vfree(sbi->s_flex_groups);
3061 kfree(sbi->s_flex_groups);
3063 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3065 percpu_counter_destroy(&sbi->s_dirs_counter);
3066 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3068 for (i = 0; i < db_count; i++)
3069 brelse(sbi->s_group_desc[i]);
3070 kfree(sbi->s_group_desc);
3073 remove_proc_entry(sb->s_id, ext4_proc_root);
3076 for (i = 0; i < MAXQUOTAS; i++)
3077 kfree(sbi->s_qf_names[i]);
3079 ext4_blkdev_remove(sbi);
3082 sb->s_fs_info = NULL;
3083 kfree(sbi->s_blockgroup_lock);
3090 * Setup any per-fs journal parameters now. We'll do this both on
3091 * initial mount, once the journal has been initialised but before we've
3092 * done any recovery; and again on any subsequent remount.
3094 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3096 struct ext4_sb_info *sbi = EXT4_SB(sb);
3098 journal->j_commit_interval = sbi->s_commit_interval;
3099 journal->j_min_batch_time = sbi->s_min_batch_time;
3100 journal->j_max_batch_time = sbi->s_max_batch_time;
3102 spin_lock(&journal->j_state_lock);
3103 if (test_opt(sb, BARRIER))
3104 journal->j_flags |= JBD2_BARRIER;
3106 journal->j_flags &= ~JBD2_BARRIER;
3107 if (test_opt(sb, DATA_ERR_ABORT))
3108 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3110 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3111 spin_unlock(&journal->j_state_lock);
3114 static journal_t *ext4_get_journal(struct super_block *sb,
3115 unsigned int journal_inum)
3117 struct inode *journal_inode;
3120 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3122 /* First, test for the existence of a valid inode on disk. Bad
3123 * things happen if we iget() an unused inode, as the subsequent
3124 * iput() will try to delete it. */
3126 journal_inode = ext4_iget(sb, journal_inum);
3127 if (IS_ERR(journal_inode)) {
3128 ext4_msg(sb, KERN_ERR, "no journal found");
3131 if (!journal_inode->i_nlink) {
3132 make_bad_inode(journal_inode);
3133 iput(journal_inode);
3134 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3138 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3139 journal_inode, journal_inode->i_size);
3140 if (!S_ISREG(journal_inode->i_mode)) {
3141 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3142 iput(journal_inode);
3146 journal = jbd2_journal_init_inode(journal_inode);
3148 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3149 iput(journal_inode);
3152 journal->j_private = sb;
3153 ext4_init_journal_params(sb, journal);
3157 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3160 struct buffer_head *bh;
3164 int hblock, blocksize;
3165 ext4_fsblk_t sb_block;
3166 unsigned long offset;
3167 struct ext4_super_block *es;
3168 struct block_device *bdev;
3170 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3172 bdev = ext4_blkdev_get(j_dev, sb);
3176 if (bd_claim(bdev, sb)) {
3177 ext4_msg(sb, KERN_ERR,
3178 "failed to claim external journal device");
3179 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3183 blocksize = sb->s_blocksize;
3184 hblock = bdev_logical_block_size(bdev);
3185 if (blocksize < hblock) {
3186 ext4_msg(sb, KERN_ERR,
3187 "blocksize too small for journal device");
3191 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3192 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3193 set_blocksize(bdev, blocksize);
3194 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3195 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3196 "external journal");
3200 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3201 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3202 !(le32_to_cpu(es->s_feature_incompat) &
3203 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3204 ext4_msg(sb, KERN_ERR, "external journal has "
3210 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3211 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3216 len = ext4_blocks_count(es);
3217 start = sb_block + 1;
3218 brelse(bh); /* we're done with the superblock */
3220 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3221 start, len, blocksize);
3223 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3226 journal->j_private = sb;
3227 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3228 wait_on_buffer(journal->j_sb_buffer);
3229 if (!buffer_uptodate(journal->j_sb_buffer)) {
3230 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3233 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3234 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3235 "user (unsupported) - %d",
3236 be32_to_cpu(journal->j_superblock->s_nr_users));
3239 EXT4_SB(sb)->journal_bdev = bdev;
3240 ext4_init_journal_params(sb, journal);
3244 jbd2_journal_destroy(journal);
3246 ext4_blkdev_put(bdev);
3250 static int ext4_load_journal(struct super_block *sb,
3251 struct ext4_super_block *es,
3252 unsigned long journal_devnum)
3255 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3258 int really_read_only;
3260 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3262 if (journal_devnum &&
3263 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3264 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3265 "numbers have changed");
3266 journal_dev = new_decode_dev(journal_devnum);
3268 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3270 really_read_only = bdev_read_only(sb->s_bdev);
3273 * Are we loading a blank journal or performing recovery after a
3274 * crash? For recovery, we need to check in advance whether we
3275 * can get read-write access to the device.
3277 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3278 if (sb->s_flags & MS_RDONLY) {
3279 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3280 "required on readonly filesystem");
3281 if (really_read_only) {
3282 ext4_msg(sb, KERN_ERR, "write access "
3283 "unavailable, cannot proceed");
3286 ext4_msg(sb, KERN_INFO, "write access will "
3287 "be enabled during recovery");
3291 if (journal_inum && journal_dev) {
3292 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3293 "and inode journals!");
3298 if (!(journal = ext4_get_journal(sb, journal_inum)))
3301 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3305 if (!(journal->j_flags & JBD2_BARRIER))
3306 ext4_msg(sb, KERN_INFO, "barriers disabled");
3308 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3309 err = jbd2_journal_update_format(journal);
3311 ext4_msg(sb, KERN_ERR, "error updating journal");
3312 jbd2_journal_destroy(journal);
3317 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3318 err = jbd2_journal_wipe(journal, !really_read_only);
3320 err = jbd2_journal_load(journal);
3323 ext4_msg(sb, KERN_ERR, "error loading journal");
3324 jbd2_journal_destroy(journal);
3328 EXT4_SB(sb)->s_journal = journal;
3329 ext4_clear_journal_err(sb, es);
3331 if (journal_devnum &&
3332 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3333 es->s_journal_dev = cpu_to_le32(journal_devnum);
3335 /* Make sure we flush the recovery flag to disk. */
3336 ext4_commit_super(sb, 1);
3342 static int ext4_commit_super(struct super_block *sb, int sync)
3344 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3345 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3350 if (buffer_write_io_error(sbh)) {
3352 * Oh, dear. A previous attempt to write the
3353 * superblock failed. This could happen because the
3354 * USB device was yanked out. Or it could happen to
3355 * be a transient write error and maybe the block will
3356 * be remapped. Nothing we can do but to retry the
3357 * write and hope for the best.
3359 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3360 "superblock detected");
3361 clear_buffer_write_io_error(sbh);
3362 set_buffer_uptodate(sbh);
3365 * If the file system is mounted read-only, don't update the
3366 * superblock write time. This avoids updating the superblock
3367 * write time when we are mounting the root file system
3368 * read/only but we need to replay the journal; at that point,
3369 * for people who are east of GMT and who make their clock
3370 * tick in localtime for Windows bug-for-bug compatibility,
3371 * the clock is set in the future, and this will cause e2fsck
3372 * to complain and force a full file system check.
3374 if (!(sb->s_flags & MS_RDONLY))
3375 es->s_wtime = cpu_to_le32(get_seconds());
3376 es->s_kbytes_written =
3377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3378 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3379 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3380 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3381 &EXT4_SB(sb)->s_freeblocks_counter));
3382 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3383 &EXT4_SB(sb)->s_freeinodes_counter));
3385 BUFFER_TRACE(sbh, "marking dirty");
3386 mark_buffer_dirty(sbh);
3388 error = sync_dirty_buffer(sbh);
3392 error = buffer_write_io_error(sbh);
3394 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3396 clear_buffer_write_io_error(sbh);
3397 set_buffer_uptodate(sbh);
3404 * Have we just finished recovery? If so, and if we are mounting (or
3405 * remounting) the filesystem readonly, then we will end up with a
3406 * consistent fs on disk. Record that fact.
3408 static void ext4_mark_recovery_complete(struct super_block *sb,
3409 struct ext4_super_block *es)
3411 journal_t *journal = EXT4_SB(sb)->s_journal;
3413 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3414 BUG_ON(journal != NULL);
3417 jbd2_journal_lock_updates(journal);
3418 if (jbd2_journal_flush(journal) < 0)
3421 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3422 sb->s_flags & MS_RDONLY) {
3423 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3424 ext4_commit_super(sb, 1);
3428 jbd2_journal_unlock_updates(journal);
3432 * If we are mounting (or read-write remounting) a filesystem whose journal
3433 * has recorded an error from a previous lifetime, move that error to the
3434 * main filesystem now.
3436 static void ext4_clear_journal_err(struct super_block *sb,
3437 struct ext4_super_block *es)
3443 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3445 journal = EXT4_SB(sb)->s_journal;
3448 * Now check for any error status which may have been recorded in the
3449 * journal by a prior ext4_error() or ext4_abort()
3452 j_errno = jbd2_journal_errno(journal);
3456 errstr = ext4_decode_error(sb, j_errno, nbuf);
3457 ext4_warning(sb, "Filesystem error recorded "
3458 "from previous mount: %s", errstr);
3459 ext4_warning(sb, "Marking fs in need of filesystem check.");
3461 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3462 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3463 ext4_commit_super(sb, 1);
3465 jbd2_journal_clear_err(journal);
3470 * Force the running and committing transactions to commit,
3471 * and wait on the commit.
3473 int ext4_force_commit(struct super_block *sb)
3478 if (sb->s_flags & MS_RDONLY)
3481 journal = EXT4_SB(sb)->s_journal;
3483 ret = ext4_journal_force_commit(journal);
3488 static void ext4_write_super(struct super_block *sb)
3491 ext4_commit_super(sb, 1);
3495 static int ext4_sync_fs(struct super_block *sb, int wait)
3499 struct ext4_sb_info *sbi = EXT4_SB(sb);
3501 trace_ext4_sync_fs(sb, wait);
3502 flush_workqueue(sbi->dio_unwritten_wq);
3503 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3505 jbd2_log_wait_commit(sbi->s_journal, target);
3511 * LVM calls this function before a (read-only) snapshot is created. This
3512 * gives us a chance to flush the journal completely and mark the fs clean.
3514 static int ext4_freeze(struct super_block *sb)
3519 if (sb->s_flags & MS_RDONLY)
3522 journal = EXT4_SB(sb)->s_journal;
3524 /* Now we set up the journal barrier. */
3525 jbd2_journal_lock_updates(journal);
3528 * Don't clear the needs_recovery flag if we failed to flush
3531 error = jbd2_journal_flush(journal);
3534 jbd2_journal_unlock_updates(journal);
3538 /* Journal blocked and flushed, clear needs_recovery flag. */
3539 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3540 error = ext4_commit_super(sb, 1);
3547 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3548 * flag here, even though the filesystem is not technically dirty yet.
3550 static int ext4_unfreeze(struct super_block *sb)
3552 if (sb->s_flags & MS_RDONLY)
3556 /* Reset the needs_recovery flag before the fs is unlocked. */
3557 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3558 ext4_commit_super(sb, 1);
3560 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3564 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3566 struct ext4_super_block *es;
3567 struct ext4_sb_info *sbi = EXT4_SB(sb);
3568 ext4_fsblk_t n_blocks_count = 0;
3569 unsigned long old_sb_flags;
3570 struct ext4_mount_options old_opts;
3572 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3580 /* Store the original options */
3582 old_sb_flags = sb->s_flags;
3583 old_opts.s_mount_opt = sbi->s_mount_opt;
3584 old_opts.s_resuid = sbi->s_resuid;
3585 old_opts.s_resgid = sbi->s_resgid;
3586 old_opts.s_commit_interval = sbi->s_commit_interval;
3587 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3588 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3590 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3591 for (i = 0; i < MAXQUOTAS; i++)
3592 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3594 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3595 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3598 * Allow the "check" option to be passed as a remount option.
3600 if (!parse_options(data, sb, NULL, &journal_ioprio,
3601 &n_blocks_count, 1)) {
3606 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3607 ext4_abort(sb, __func__, "Abort forced by user");
3609 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3610 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3614 if (sbi->s_journal) {
3615 ext4_init_journal_params(sb, sbi->s_journal);
3616 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3619 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3620 n_blocks_count > ext4_blocks_count(es)) {
3621 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3626 if (*flags & MS_RDONLY) {
3628 * First of all, the unconditional stuff we have to do
3629 * to disable replay of the journal when we next remount
3631 sb->s_flags |= MS_RDONLY;
3634 * OK, test if we are remounting a valid rw partition
3635 * readonly, and if so set the rdonly flag and then
3636 * mark the partition as valid again.
3638 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3639 (sbi->s_mount_state & EXT4_VALID_FS))
3640 es->s_state = cpu_to_le16(sbi->s_mount_state);
3643 ext4_mark_recovery_complete(sb, es);
3645 /* Make sure we can mount this feature set readwrite */
3646 if (!ext4_feature_set_ok(sb, 0)) {
3651 * Make sure the group descriptor checksums
3652 * are sane. If they aren't, refuse to remount r/w.
3654 for (g = 0; g < sbi->s_groups_count; g++) {
3655 struct ext4_group_desc *gdp =
3656 ext4_get_group_desc(sb, g, NULL);
3658 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3659 ext4_msg(sb, KERN_ERR,
3660 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3661 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3662 le16_to_cpu(gdp->bg_checksum));
3669 * If we have an unprocessed orphan list hanging
3670 * around from a previously readonly bdev mount,
3671 * require a full umount/remount for now.
3673 if (es->s_last_orphan) {
3674 ext4_msg(sb, KERN_WARNING, "Couldn't "
3675 "remount RDWR because of unprocessed "
3676 "orphan inode list. Please "
3677 "umount/remount instead");
3683 * Mounting a RDONLY partition read-write, so reread
3684 * and store the current valid flag. (It may have
3685 * been changed by e2fsck since we originally mounted
3689 ext4_clear_journal_err(sb, es);
3690 sbi->s_mount_state = le16_to_cpu(es->s_state);
3691 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3693 if (!ext4_setup_super(sb, es, 0))
3694 sb->s_flags &= ~MS_RDONLY;
3697 ext4_setup_system_zone(sb);
3698 if (sbi->s_journal == NULL)
3699 ext4_commit_super(sb, 1);
3702 /* Release old quota file names */
3703 for (i = 0; i < MAXQUOTAS; i++)
3704 if (old_opts.s_qf_names[i] &&
3705 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3706 kfree(old_opts.s_qf_names[i]);
3713 sb->s_flags = old_sb_flags;
3714 sbi->s_mount_opt = old_opts.s_mount_opt;
3715 sbi->s_resuid = old_opts.s_resuid;
3716 sbi->s_resgid = old_opts.s_resgid;
3717 sbi->s_commit_interval = old_opts.s_commit_interval;
3718 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3719 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3721 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3722 for (i = 0; i < MAXQUOTAS; i++) {
3723 if (sbi->s_qf_names[i] &&
3724 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3725 kfree(sbi->s_qf_names[i]);
3726 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3734 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3736 struct super_block *sb = dentry->d_sb;
3737 struct ext4_sb_info *sbi = EXT4_SB(sb);
3738 struct ext4_super_block *es = sbi->s_es;
3741 if (test_opt(sb, MINIX_DF)) {
3742 sbi->s_overhead_last = 0;
3743 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3744 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3745 ext4_fsblk_t overhead = 0;
3748 * Compute the overhead (FS structures). This is constant
3749 * for a given filesystem unless the number of block groups
3750 * changes so we cache the previous value until it does.
3754 * All of the blocks before first_data_block are
3757 overhead = le32_to_cpu(es->s_first_data_block);
3760 * Add the overhead attributed to the superblock and
3761 * block group descriptors. If the sparse superblocks
3762 * feature is turned on, then not all groups have this.
3764 for (i = 0; i < ngroups; i++) {
3765 overhead += ext4_bg_has_super(sb, i) +
3766 ext4_bg_num_gdb(sb, i);
3771 * Every block group has an inode bitmap, a block
3772 * bitmap, and an inode table.
3774 overhead += ngroups * (2 + sbi->s_itb_per_group);
3775 sbi->s_overhead_last = overhead;
3777 sbi->s_blocks_last = ext4_blocks_count(es);
3780 buf->f_type = EXT4_SUPER_MAGIC;
3781 buf->f_bsize = sb->s_blocksize;
3782 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3783 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3784 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3785 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3786 if (buf->f_bfree < ext4_r_blocks_count(es))
3788 buf->f_files = le32_to_cpu(es->s_inodes_count);
3789 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3790 buf->f_namelen = EXT4_NAME_LEN;
3791 fsid = le64_to_cpup((void *)es->s_uuid) ^
3792 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3793 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3794 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3799 /* Helper function for writing quotas on sync - we need to start transaction
3800 * before quota file is locked for write. Otherwise the are possible deadlocks:
3801 * Process 1 Process 2
3802 * ext4_create() quota_sync()
3803 * jbd2_journal_start() write_dquot()
3804 * vfs_dq_init() down(dqio_mutex)
3805 * down(dqio_mutex) jbd2_journal_start()
3811 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3813 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3816 static int ext4_write_dquot(struct dquot *dquot)
3820 struct inode *inode;
3822 inode = dquot_to_inode(dquot);
3823 handle = ext4_journal_start(inode,
3824 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3826 return PTR_ERR(handle);
3827 ret = dquot_commit(dquot);
3828 err = ext4_journal_stop(handle);
3834 static int ext4_acquire_dquot(struct dquot *dquot)
3839 handle = ext4_journal_start(dquot_to_inode(dquot),
3840 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3842 return PTR_ERR(handle);
3843 ret = dquot_acquire(dquot);
3844 err = ext4_journal_stop(handle);
3850 static int ext4_release_dquot(struct dquot *dquot)
3855 handle = ext4_journal_start(dquot_to_inode(dquot),
3856 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3857 if (IS_ERR(handle)) {
3858 /* Release dquot anyway to avoid endless cycle in dqput() */
3859 dquot_release(dquot);
3860 return PTR_ERR(handle);
3862 ret = dquot_release(dquot);
3863 err = ext4_journal_stop(handle);
3869 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3871 /* Are we journaling quotas? */
3872 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3873 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3874 dquot_mark_dquot_dirty(dquot);
3875 return ext4_write_dquot(dquot);
3877 return dquot_mark_dquot_dirty(dquot);
3881 static int ext4_write_info(struct super_block *sb, int type)
3886 /* Data block + inode block */
3887 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3889 return PTR_ERR(handle);
3890 ret = dquot_commit_info(sb, type);
3891 err = ext4_journal_stop(handle);
3898 * Turn on quotas during mount time - we need to find
3899 * the quota file and such...
3901 static int ext4_quota_on_mount(struct super_block *sb, int type)
3903 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3904 EXT4_SB(sb)->s_jquota_fmt, type);
3908 * Standard function to be called on quota_on
3910 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3911 char *name, int remount)
3916 if (!test_opt(sb, QUOTA))
3918 /* When remounting, no checks are needed and in fact, name is NULL */
3920 return vfs_quota_on(sb, type, format_id, name, remount);
3922 err = kern_path(name, LOOKUP_FOLLOW, &path);
3926 /* Quotafile not on the same filesystem? */
3927 if (path.mnt->mnt_sb != sb) {
3931 /* Journaling quota? */
3932 if (EXT4_SB(sb)->s_qf_names[type]) {
3933 /* Quotafile not in fs root? */
3934 if (path.dentry->d_parent != sb->s_root)
3935 ext4_msg(sb, KERN_WARNING,
3936 "Quota file not on filesystem root. "
3937 "Journaled quota will not work");
3941 * When we journal data on quota file, we have to flush journal to see
3942 * all updates to the file when we bypass pagecache...
3944 if (EXT4_SB(sb)->s_journal &&
3945 ext4_should_journal_data(path.dentry->d_inode)) {
3947 * We don't need to lock updates but journal_flush() could
3948 * otherwise be livelocked...
3950 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3951 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3952 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3959 err = vfs_quota_on_path(sb, type, format_id, &path);
3964 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3965 * acquiring the locks... As quota files are never truncated and quota code
3966 * itself serializes the operations (and noone else should touch the files)
3967 * we don't have to be afraid of races */
3968 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3969 size_t len, loff_t off)
3971 struct inode *inode = sb_dqopt(sb)->files[type];
3972 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3974 int offset = off & (sb->s_blocksize - 1);
3977 struct buffer_head *bh;
3978 loff_t i_size = i_size_read(inode);
3982 if (off+len > i_size)
3985 while (toread > 0) {
3986 tocopy = sb->s_blocksize - offset < toread ?
3987 sb->s_blocksize - offset : toread;
3988 bh = ext4_bread(NULL, inode, blk, 0, &err);
3991 if (!bh) /* A hole? */
3992 memset(data, 0, tocopy);
3994 memcpy(data, bh->b_data+offset, tocopy);
4004 /* Write to quotafile (we know the transaction is already started and has
4005 * enough credits) */
4006 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4007 const char *data, size_t len, loff_t off)
4009 struct inode *inode = sb_dqopt(sb)->files[type];
4010 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4012 int offset = off & (sb->s_blocksize - 1);
4013 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4014 struct buffer_head *bh;
4015 handle_t *handle = journal_current_handle();
4017 if (EXT4_SB(sb)->s_journal && !handle) {
4018 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4019 " cancelled because transaction is not started",
4020 (unsigned long long)off, (unsigned long long)len);
4024 * Since we account only one data block in transaction credits,
4025 * then it is impossible to cross a block boundary.
4027 if (sb->s_blocksize - offset < len) {
4028 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4029 " cancelled because not block aligned",
4030 (unsigned long long)off, (unsigned long long)len);
4034 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4035 bh = ext4_bread(handle, inode, blk, 1, &err);
4038 if (journal_quota) {
4039 err = ext4_journal_get_write_access(handle, bh);
4046 memcpy(bh->b_data+offset, data, len);
4047 flush_dcache_page(bh->b_page);
4050 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4052 /* Always do at least ordered writes for quotas */
4053 err = ext4_jbd2_file_inode(handle, inode);
4054 mark_buffer_dirty(bh);
4059 mutex_unlock(&inode->i_mutex);
4062 if (inode->i_size < off + len) {
4063 i_size_write(inode, off + len);
4064 EXT4_I(inode)->i_disksize = inode->i_size;
4066 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4067 ext4_mark_inode_dirty(handle, inode);
4068 mutex_unlock(&inode->i_mutex);
4074 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4075 const char *dev_name, void *data, struct vfsmount *mnt)
4077 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4080 #if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4081 static struct file_system_type ext2_fs_type = {
4082 .owner = THIS_MODULE,
4084 .get_sb = ext4_get_sb,
4085 .kill_sb = kill_block_super,
4086 .fs_flags = FS_REQUIRES_DEV,
4089 static inline void register_as_ext2(void)
4091 int err = register_filesystem(&ext2_fs_type);
4094 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4097 static inline void unregister_as_ext2(void)
4099 unregister_filesystem(&ext2_fs_type);
4101 MODULE_ALIAS("ext2");
4103 static inline void register_as_ext2(void) { }
4104 static inline void unregister_as_ext2(void) { }
4107 #if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4108 static struct file_system_type ext3_fs_type = {
4109 .owner = THIS_MODULE,
4111 .get_sb = ext4_get_sb,
4112 .kill_sb = kill_block_super,
4113 .fs_flags = FS_REQUIRES_DEV,
4116 static inline void register_as_ext3(void)
4118 int err = register_filesystem(&ext3_fs_type);
4121 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4124 static inline void unregister_as_ext3(void)
4126 unregister_filesystem(&ext3_fs_type);
4128 MODULE_ALIAS("ext3");
4130 static inline void register_as_ext3(void) { }
4131 static inline void unregister_as_ext3(void) { }
4134 static struct file_system_type ext4_fs_type = {
4135 .owner = THIS_MODULE,
4137 .get_sb = ext4_get_sb,
4138 .kill_sb = kill_block_super,
4139 .fs_flags = FS_REQUIRES_DEV,
4142 static int __init init_ext4_fs(void)
4146 err = init_ext4_system_zone();
4149 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4152 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4153 err = init_ext4_mballoc();
4157 err = init_ext4_xattr();
4160 err = init_inodecache();
4165 err = register_filesystem(&ext4_fs_type);
4170 unregister_as_ext2();
4171 unregister_as_ext3();
4172 destroy_inodecache();
4176 exit_ext4_mballoc();
4178 remove_proc_entry("fs/ext4", NULL);
4179 kset_unregister(ext4_kset);
4181 exit_ext4_system_zone();
4185 static void __exit exit_ext4_fs(void)
4187 unregister_as_ext2();
4188 unregister_as_ext3();
4189 unregister_filesystem(&ext4_fs_type);
4190 destroy_inodecache();
4192 exit_ext4_mballoc();
4193 remove_proc_entry("fs/ext4", NULL);
4194 kset_unregister(ext4_kset);
4195 exit_ext4_system_zone();
4198 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4199 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4200 MODULE_LICENSE("GPL");
4201 module_init(init_ext4_fs)
4202 module_exit(exit_ext4_fs)