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);
71 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type = {
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #define IS_EXT3_SB(sb) (0)
87 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
95 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
103 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le32_to_cpu(bg->bg_inode_table_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
111 __u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
119 __u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
127 __u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
135 __u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
143 void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
151 void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
159 void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
167 void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
175 void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
183 void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
191 void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
200 /* Just increment the non-pointer handle value */
201 static handle_t *ext4_get_nojournal(void)
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
209 handle = (handle_t *)ref_cnt;
211 current->journal_info = handle;
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t *handle)
219 unsigned long ref_cnt = (unsigned long)handle;
221 BUG_ON(ref_cnt == 0);
224 handle = (handle_t *)ref_cnt;
226 current->journal_info = handle;
230 * Wrappers for jbd2_journal_start/end.
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
237 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
244 /* Special case here: if the journal has aborted behind our
245 * backs (eg. EIO in the commit thread), then we still need to
246 * take the FS itself readonly cleanly. */
247 journal = EXT4_SB(sb)->s_journal;
249 if (is_journal_aborted(journal)) {
250 ext4_abort(sb, __func__, "Detected aborted journal");
251 return ERR_PTR(-EROFS);
253 return jbd2_journal_start(journal, nblocks);
255 return ext4_get_nojournal();
259 * The only special thing we need to do here is to make sure that all
260 * jbd2_journal_stop calls result in the superblock being marked dirty, so
261 * that sync() will call the filesystem's write_super callback if
264 int __ext4_journal_stop(const char *where, handle_t *handle)
266 struct super_block *sb;
270 if (!ext4_handle_valid(handle)) {
271 ext4_put_nojournal(handle);
274 sb = handle->h_transaction->t_journal->j_private;
276 rc = jbd2_journal_stop(handle);
281 __ext4_std_error(sb, where, err);
285 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
286 struct buffer_head *bh, handle_t *handle, int err)
289 const char *errstr = ext4_decode_error(NULL, err, nbuf);
291 BUG_ON(!ext4_handle_valid(handle));
294 BUFFER_TRACE(bh, "abort");
299 if (is_handle_aborted(handle))
302 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
303 caller, errstr, err_fn);
305 jbd2_journal_abort_handle(handle);
308 /* Deal with the reporting of failure conditions on a filesystem such as
309 * inconsistencies detected or read IO failures.
311 * On ext2, we can store the error state of the filesystem in the
312 * superblock. That is not possible on ext4, because we may have other
313 * write ordering constraints on the superblock which prevent us from
314 * writing it out straight away; and given that the journal is about to
315 * be aborted, we can't rely on the current, or future, transactions to
316 * write out the superblock safely.
318 * We'll just use the jbd2_journal_abort() error code to record an error in
319 * the journal instead. On recovery, the journal will complain about
320 * that error until we've noted it down and cleared it.
323 static void ext4_handle_error(struct super_block *sb)
325 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
327 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
328 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
330 if (sb->s_flags & MS_RDONLY)
333 if (!test_opt(sb, ERRORS_CONT)) {
334 journal_t *journal = EXT4_SB(sb)->s_journal;
336 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
338 jbd2_journal_abort(journal, -EIO);
340 if (test_opt(sb, ERRORS_RO)) {
341 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
342 sb->s_flags |= MS_RDONLY;
344 ext4_commit_super(sb, 1);
345 if (test_opt(sb, ERRORS_PANIC))
346 panic("EXT4-fs (device %s): panic forced after error\n",
350 void __ext4_error(struct super_block *sb, const char *function,
351 const char *fmt, ...)
356 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
361 ext4_handle_error(sb);
364 void ext4_error_inode(const char *function, struct inode *inode,
365 const char *fmt, ...)
370 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
371 inode->i_sb->s_id, function, inode->i_ino, current->comm);
376 ext4_handle_error(inode->i_sb);
379 void ext4_error_file(const char *function, struct file *file,
380 const char *fmt, ...)
383 struct inode *inode = file->f_dentry->d_inode;
384 char pathname[80], *path;
387 path = d_path(&(file->f_path), pathname, sizeof(pathname));
391 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
392 inode->i_sb->s_id, function, inode->i_ino, current->comm, path);
397 ext4_handle_error(inode->i_sb);
400 static const char *ext4_decode_error(struct super_block *sb, int errno,
407 errstr = "IO failure";
410 errstr = "Out of memory";
413 if (!sb || (EXT4_SB(sb)->s_journal &&
414 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
415 errstr = "Journal has aborted";
417 errstr = "Readonly filesystem";
420 /* If the caller passed in an extra buffer for unknown
421 * errors, textualise them now. Else we just return
424 /* Check for truncated error codes... */
425 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
434 /* __ext4_std_error decodes expected errors from journaling functions
435 * automatically and invokes the appropriate error response. */
437 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
442 /* Special case: if the error is EROFS, and we're not already
443 * inside a transaction, then there's really no point in logging
445 if (errno == -EROFS && journal_current_handle() == NULL &&
446 (sb->s_flags & MS_RDONLY))
449 errstr = ext4_decode_error(sb, errno, nbuf);
450 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
451 sb->s_id, function, errstr);
453 ext4_handle_error(sb);
457 * ext4_abort is a much stronger failure handler than ext4_error. The
458 * abort function may be used to deal with unrecoverable failures such
459 * as journal IO errors or ENOMEM at a critical moment in log management.
461 * We unconditionally force the filesystem into an ABORT|READONLY state,
462 * unless the error response on the fs has been set to panic in which
463 * case we take the easy way out and panic immediately.
466 void ext4_abort(struct super_block *sb, const char *function,
467 const char *fmt, ...)
472 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
477 if (test_opt(sb, ERRORS_PANIC))
478 panic("EXT4-fs panic from previous error\n");
480 if (sb->s_flags & MS_RDONLY)
483 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
484 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
485 sb->s_flags |= MS_RDONLY;
486 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
487 if (EXT4_SB(sb)->s_journal)
488 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
491 void ext4_msg (struct super_block * sb, const char *prefix,
492 const char *fmt, ...)
497 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
503 void __ext4_warning(struct super_block *sb, const char *function,
504 const char *fmt, ...)
509 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
516 void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
517 const char *function, const char *fmt, ...)
522 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
525 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
530 if (test_opt(sb, ERRORS_CONT)) {
531 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
532 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
533 ext4_commit_super(sb, 0);
536 ext4_unlock_group(sb, grp);
537 ext4_handle_error(sb);
539 * We only get here in the ERRORS_RO case; relocking the group
540 * may be dangerous, but nothing bad will happen since the
541 * filesystem will have already been marked read/only and the
542 * journal has been aborted. We return 1 as a hint to callers
543 * who might what to use the return value from
544 * ext4_grp_locked_error() to distinguish beween the
545 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
546 * aggressively from the ext4 function in question, with a
547 * more appropriate error code.
549 ext4_lock_group(sb, grp);
553 void ext4_update_dynamic_rev(struct super_block *sb)
555 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
557 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
561 "updating to rev %d because of new feature flag, "
562 "running e2fsck is recommended",
565 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
566 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
567 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
568 /* leave es->s_feature_*compat flags alone */
569 /* es->s_uuid will be set by e2fsck if empty */
572 * The rest of the superblock fields should be zero, and if not it
573 * means they are likely already in use, so leave them alone. We
574 * can leave it up to e2fsck to clean up any inconsistencies there.
579 * Open the external journal device
581 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
583 struct block_device *bdev;
584 char b[BDEVNAME_SIZE];
586 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
592 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
593 __bdevname(dev, b), PTR_ERR(bdev));
598 * Release the journal device
600 static int ext4_blkdev_put(struct block_device *bdev)
603 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
606 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
608 struct block_device *bdev;
611 bdev = sbi->journal_bdev;
613 ret = ext4_blkdev_put(bdev);
614 sbi->journal_bdev = NULL;
619 static inline struct inode *orphan_list_entry(struct list_head *l)
621 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
624 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
628 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
629 le32_to_cpu(sbi->s_es->s_last_orphan));
631 printk(KERN_ERR "sb_info orphan list:\n");
632 list_for_each(l, &sbi->s_orphan) {
633 struct inode *inode = orphan_list_entry(l);
635 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
636 inode->i_sb->s_id, inode->i_ino, inode,
637 inode->i_mode, inode->i_nlink,
642 static void ext4_put_super(struct super_block *sb)
644 struct ext4_sb_info *sbi = EXT4_SB(sb);
645 struct ext4_super_block *es = sbi->s_es;
648 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
650 flush_workqueue(sbi->dio_unwritten_wq);
651 destroy_workqueue(sbi->dio_unwritten_wq);
656 ext4_commit_super(sb, 1);
658 if (sbi->s_journal) {
659 err = jbd2_journal_destroy(sbi->s_journal);
660 sbi->s_journal = NULL;
662 ext4_abort(sb, __func__,
663 "Couldn't clean up the journal");
666 ext4_release_system_zone(sb);
668 ext4_ext_release(sb);
669 ext4_xattr_put_super(sb);
671 if (!(sb->s_flags & MS_RDONLY)) {
672 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
673 es->s_state = cpu_to_le16(sbi->s_mount_state);
674 ext4_commit_super(sb, 1);
677 remove_proc_entry(sb->s_id, ext4_proc_root);
679 kobject_del(&sbi->s_kobj);
681 for (i = 0; i < sbi->s_gdb_count; i++)
682 brelse(sbi->s_group_desc[i]);
683 kfree(sbi->s_group_desc);
684 if (is_vmalloc_addr(sbi->s_flex_groups))
685 vfree(sbi->s_flex_groups);
687 kfree(sbi->s_flex_groups);
688 percpu_counter_destroy(&sbi->s_freeblocks_counter);
689 percpu_counter_destroy(&sbi->s_freeinodes_counter);
690 percpu_counter_destroy(&sbi->s_dirs_counter);
691 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
694 for (i = 0; i < MAXQUOTAS; i++)
695 kfree(sbi->s_qf_names[i]);
698 /* Debugging code just in case the in-memory inode orphan list
699 * isn't empty. The on-disk one can be non-empty if we've
700 * detected an error and taken the fs readonly, but the
701 * in-memory list had better be clean by this point. */
702 if (!list_empty(&sbi->s_orphan))
703 dump_orphan_list(sb, sbi);
704 J_ASSERT(list_empty(&sbi->s_orphan));
706 invalidate_bdev(sb->s_bdev);
707 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
709 * Invalidate the journal device's buffers. We don't want them
710 * floating about in memory - the physical journal device may
711 * hotswapped, and it breaks the `ro-after' testing code.
713 sync_blockdev(sbi->journal_bdev);
714 invalidate_bdev(sbi->journal_bdev);
715 ext4_blkdev_remove(sbi);
717 sb->s_fs_info = NULL;
719 * Now that we are completely done shutting down the
720 * superblock, we need to actually destroy the kobject.
724 kobject_put(&sbi->s_kobj);
725 wait_for_completion(&sbi->s_kobj_unregister);
726 kfree(sbi->s_blockgroup_lock);
730 static struct kmem_cache *ext4_inode_cachep;
733 * Called inside transaction, so use GFP_NOFS
735 static struct inode *ext4_alloc_inode(struct super_block *sb)
737 struct ext4_inode_info *ei;
739 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
743 ei->vfs_inode.i_version = 1;
744 ei->vfs_inode.i_data.writeback_index = 0;
745 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
746 INIT_LIST_HEAD(&ei->i_prealloc_list);
747 spin_lock_init(&ei->i_prealloc_lock);
749 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
750 * therefore it can be null here. Don't check it, just initialize
753 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
754 ei->i_reserved_data_blocks = 0;
755 ei->i_reserved_meta_blocks = 0;
756 ei->i_allocated_meta_blocks = 0;
757 ei->i_da_metadata_calc_len = 0;
758 ei->i_delalloc_reserved_flag = 0;
759 spin_lock_init(&(ei->i_block_reservation_lock));
761 ei->i_reserved_quota = 0;
763 INIT_LIST_HEAD(&ei->i_completed_io_list);
764 spin_lock_init(&ei->i_completed_io_lock);
765 ei->cur_aio_dio = NULL;
767 ei->i_datasync_tid = 0;
769 return &ei->vfs_inode;
772 static void ext4_destroy_inode(struct inode *inode)
774 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
775 ext4_msg(inode->i_sb, KERN_ERR,
776 "Inode %lu (%p): orphan list check failed!",
777 inode->i_ino, EXT4_I(inode));
778 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
779 EXT4_I(inode), sizeof(struct ext4_inode_info),
783 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
786 static void init_once(void *foo)
788 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
790 INIT_LIST_HEAD(&ei->i_orphan);
791 #ifdef CONFIG_EXT4_FS_XATTR
792 init_rwsem(&ei->xattr_sem);
794 init_rwsem(&ei->i_data_sem);
795 inode_init_once(&ei->vfs_inode);
798 static int init_inodecache(void)
800 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
801 sizeof(struct ext4_inode_info),
802 0, (SLAB_RECLAIM_ACCOUNT|
805 if (ext4_inode_cachep == NULL)
810 static void destroy_inodecache(void)
812 kmem_cache_destroy(ext4_inode_cachep);
815 static void ext4_clear_inode(struct inode *inode)
818 ext4_discard_preallocations(inode);
819 if (EXT4_JOURNAL(inode))
820 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
821 &EXT4_I(inode)->jinode);
824 static inline void ext4_show_quota_options(struct seq_file *seq,
825 struct super_block *sb)
827 #if defined(CONFIG_QUOTA)
828 struct ext4_sb_info *sbi = EXT4_SB(sb);
830 if (sbi->s_jquota_fmt) {
833 switch (sbi->s_jquota_fmt) {
844 seq_printf(seq, ",jqfmt=%s", fmtname);
847 if (sbi->s_qf_names[USRQUOTA])
848 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
850 if (sbi->s_qf_names[GRPQUOTA])
851 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
853 if (test_opt(sb, USRQUOTA))
854 seq_puts(seq, ",usrquota");
856 if (test_opt(sb, GRPQUOTA))
857 seq_puts(seq, ",grpquota");
863 * - it's set to a non-default value OR
864 * - if the per-sb default is different from the global default
866 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
869 unsigned long def_mount_opts;
870 struct super_block *sb = vfs->mnt_sb;
871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
874 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
875 def_errors = le16_to_cpu(es->s_errors);
877 if (sbi->s_sb_block != 1)
878 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
879 if (test_opt(sb, MINIX_DF))
880 seq_puts(seq, ",minixdf");
881 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
882 seq_puts(seq, ",grpid");
883 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
884 seq_puts(seq, ",nogrpid");
885 if (sbi->s_resuid != EXT4_DEF_RESUID ||
886 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
887 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
889 if (sbi->s_resgid != EXT4_DEF_RESGID ||
890 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
891 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
893 if (test_opt(sb, ERRORS_RO)) {
894 if (def_errors == EXT4_ERRORS_PANIC ||
895 def_errors == EXT4_ERRORS_CONTINUE) {
896 seq_puts(seq, ",errors=remount-ro");
899 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
900 seq_puts(seq, ",errors=continue");
901 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
902 seq_puts(seq, ",errors=panic");
903 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
904 seq_puts(seq, ",nouid32");
905 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
906 seq_puts(seq, ",debug");
907 if (test_opt(sb, OLDALLOC))
908 seq_puts(seq, ",oldalloc");
909 #ifdef CONFIG_EXT4_FS_XATTR
910 if (test_opt(sb, XATTR_USER) &&
911 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
912 seq_puts(seq, ",user_xattr");
913 if (!test_opt(sb, XATTR_USER) &&
914 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
915 seq_puts(seq, ",nouser_xattr");
918 #ifdef CONFIG_EXT4_FS_POSIX_ACL
919 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
920 seq_puts(seq, ",acl");
921 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
922 seq_puts(seq, ",noacl");
924 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
925 seq_printf(seq, ",commit=%u",
926 (unsigned) (sbi->s_commit_interval / HZ));
928 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
929 seq_printf(seq, ",min_batch_time=%u",
930 (unsigned) sbi->s_min_batch_time);
932 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
933 seq_printf(seq, ",max_batch_time=%u",
934 (unsigned) sbi->s_min_batch_time);
938 * We're changing the default of barrier mount option, so
939 * let's always display its mount state so it's clear what its
942 seq_puts(seq, ",barrier=");
943 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
944 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
945 seq_puts(seq, ",journal_async_commit");
946 if (test_opt(sb, NOBH))
947 seq_puts(seq, ",nobh");
948 if (test_opt(sb, I_VERSION))
949 seq_puts(seq, ",i_version");
950 if (!test_opt(sb, DELALLOC))
951 seq_puts(seq, ",nodelalloc");
955 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
957 * journal mode get enabled in different ways
958 * So just print the value even if we didn't specify it
960 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
961 seq_puts(seq, ",data=journal");
962 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
963 seq_puts(seq, ",data=ordered");
964 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
965 seq_puts(seq, ",data=writeback");
967 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
968 seq_printf(seq, ",inode_readahead_blks=%u",
969 sbi->s_inode_readahead_blks);
971 if (test_opt(sb, DATA_ERR_ABORT))
972 seq_puts(seq, ",data_err=abort");
974 if (test_opt(sb, NO_AUTO_DA_ALLOC))
975 seq_puts(seq, ",noauto_da_alloc");
977 if (test_opt(sb, DISCARD))
978 seq_puts(seq, ",discard");
980 if (test_opt(sb, NOLOAD))
981 seq_puts(seq, ",norecovery");
983 if (test_opt(sb, DIOREAD_NOLOCK))
984 seq_puts(seq, ",dioread_nolock");
986 ext4_show_quota_options(seq, sb);
991 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
992 u64 ino, u32 generation)
996 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
997 return ERR_PTR(-ESTALE);
998 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
999 return ERR_PTR(-ESTALE);
1001 /* iget isn't really right if the inode is currently unallocated!!
1003 * ext4_read_inode will return a bad_inode if the inode had been
1004 * deleted, so we should be safe.
1006 * Currently we don't know the generation for parent directory, so
1007 * a generation of 0 means "accept any"
1009 inode = ext4_iget(sb, ino);
1011 return ERR_CAST(inode);
1012 if (generation && inode->i_generation != generation) {
1014 return ERR_PTR(-ESTALE);
1020 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1021 int fh_len, int fh_type)
1023 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1024 ext4_nfs_get_inode);
1027 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1028 int fh_len, int fh_type)
1030 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1031 ext4_nfs_get_inode);
1035 * Try to release metadata pages (indirect blocks, directories) which are
1036 * mapped via the block device. Since these pages could have journal heads
1037 * which would prevent try_to_free_buffers() from freeing them, we must use
1038 * jbd2 layer's try_to_free_buffers() function to release them.
1040 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1043 journal_t *journal = EXT4_SB(sb)->s_journal;
1045 WARN_ON(PageChecked(page));
1046 if (!page_has_buffers(page))
1049 return jbd2_journal_try_to_free_buffers(journal, page,
1050 wait & ~__GFP_WAIT);
1051 return try_to_free_buffers(page);
1055 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1056 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1058 static int ext4_write_dquot(struct dquot *dquot);
1059 static int ext4_acquire_dquot(struct dquot *dquot);
1060 static int ext4_release_dquot(struct dquot *dquot);
1061 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1062 static int ext4_write_info(struct super_block *sb, int type);
1063 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1064 char *path, int remount);
1065 static int ext4_quota_on_mount(struct super_block *sb, int type);
1066 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1067 size_t len, loff_t off);
1068 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1069 const char *data, size_t len, loff_t off);
1071 static const struct dquot_operations ext4_quota_operations = {
1073 .get_reserved_space = ext4_get_reserved_space,
1075 .write_dquot = ext4_write_dquot,
1076 .acquire_dquot = ext4_acquire_dquot,
1077 .release_dquot = ext4_release_dquot,
1078 .mark_dirty = ext4_mark_dquot_dirty,
1079 .write_info = ext4_write_info,
1080 .alloc_dquot = dquot_alloc,
1081 .destroy_dquot = dquot_destroy,
1084 static const struct quotactl_ops ext4_qctl_operations = {
1085 .quota_on = ext4_quota_on,
1086 .quota_off = vfs_quota_off,
1087 .quota_sync = vfs_quota_sync,
1088 .get_info = vfs_get_dqinfo,
1089 .set_info = vfs_set_dqinfo,
1090 .get_dqblk = vfs_get_dqblk,
1091 .set_dqblk = vfs_set_dqblk
1095 static const struct super_operations ext4_sops = {
1096 .alloc_inode = ext4_alloc_inode,
1097 .destroy_inode = ext4_destroy_inode,
1098 .write_inode = ext4_write_inode,
1099 .dirty_inode = ext4_dirty_inode,
1100 .delete_inode = ext4_delete_inode,
1101 .put_super = ext4_put_super,
1102 .sync_fs = ext4_sync_fs,
1103 .freeze_fs = ext4_freeze,
1104 .unfreeze_fs = ext4_unfreeze,
1105 .statfs = ext4_statfs,
1106 .remount_fs = ext4_remount,
1107 .clear_inode = ext4_clear_inode,
1108 .show_options = ext4_show_options,
1110 .quota_read = ext4_quota_read,
1111 .quota_write = ext4_quota_write,
1113 .bdev_try_to_free_page = bdev_try_to_free_page,
1116 static const struct super_operations ext4_nojournal_sops = {
1117 .alloc_inode = ext4_alloc_inode,
1118 .destroy_inode = ext4_destroy_inode,
1119 .write_inode = ext4_write_inode,
1120 .dirty_inode = ext4_dirty_inode,
1121 .delete_inode = ext4_delete_inode,
1122 .write_super = ext4_write_super,
1123 .put_super = ext4_put_super,
1124 .statfs = ext4_statfs,
1125 .remount_fs = ext4_remount,
1126 .clear_inode = ext4_clear_inode,
1127 .show_options = ext4_show_options,
1129 .quota_read = ext4_quota_read,
1130 .quota_write = ext4_quota_write,
1132 .bdev_try_to_free_page = bdev_try_to_free_page,
1135 static const struct export_operations ext4_export_ops = {
1136 .fh_to_dentry = ext4_fh_to_dentry,
1137 .fh_to_parent = ext4_fh_to_parent,
1138 .get_parent = ext4_get_parent,
1142 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1143 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1144 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1145 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1146 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1147 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1148 Opt_journal_update, Opt_journal_dev,
1149 Opt_journal_checksum, Opt_journal_async_commit,
1150 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1151 Opt_data_err_abort, Opt_data_err_ignore,
1152 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1153 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1154 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1155 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1156 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1157 Opt_block_validity, Opt_noblock_validity,
1158 Opt_inode_readahead_blks, Opt_journal_ioprio,
1159 Opt_dioread_nolock, Opt_dioread_lock,
1160 Opt_discard, Opt_nodiscard,
1163 static const match_table_t tokens = {
1164 {Opt_bsd_df, "bsddf"},
1165 {Opt_minix_df, "minixdf"},
1166 {Opt_grpid, "grpid"},
1167 {Opt_grpid, "bsdgroups"},
1168 {Opt_nogrpid, "nogrpid"},
1169 {Opt_nogrpid, "sysvgroups"},
1170 {Opt_resgid, "resgid=%u"},
1171 {Opt_resuid, "resuid=%u"},
1173 {Opt_err_cont, "errors=continue"},
1174 {Opt_err_panic, "errors=panic"},
1175 {Opt_err_ro, "errors=remount-ro"},
1176 {Opt_nouid32, "nouid32"},
1177 {Opt_debug, "debug"},
1178 {Opt_oldalloc, "oldalloc"},
1179 {Opt_orlov, "orlov"},
1180 {Opt_user_xattr, "user_xattr"},
1181 {Opt_nouser_xattr, "nouser_xattr"},
1183 {Opt_noacl, "noacl"},
1184 {Opt_noload, "noload"},
1185 {Opt_noload, "norecovery"},
1188 {Opt_commit, "commit=%u"},
1189 {Opt_min_batch_time, "min_batch_time=%u"},
1190 {Opt_max_batch_time, "max_batch_time=%u"},
1191 {Opt_journal_update, "journal=update"},
1192 {Opt_journal_dev, "journal_dev=%u"},
1193 {Opt_journal_checksum, "journal_checksum"},
1194 {Opt_journal_async_commit, "journal_async_commit"},
1195 {Opt_abort, "abort"},
1196 {Opt_data_journal, "data=journal"},
1197 {Opt_data_ordered, "data=ordered"},
1198 {Opt_data_writeback, "data=writeback"},
1199 {Opt_data_err_abort, "data_err=abort"},
1200 {Opt_data_err_ignore, "data_err=ignore"},
1201 {Opt_offusrjquota, "usrjquota="},
1202 {Opt_usrjquota, "usrjquota=%s"},
1203 {Opt_offgrpjquota, "grpjquota="},
1204 {Opt_grpjquota, "grpjquota=%s"},
1205 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1206 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1207 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1208 {Opt_grpquota, "grpquota"},
1209 {Opt_noquota, "noquota"},
1210 {Opt_quota, "quota"},
1211 {Opt_usrquota, "usrquota"},
1212 {Opt_barrier, "barrier=%u"},
1213 {Opt_barrier, "barrier"},
1214 {Opt_nobarrier, "nobarrier"},
1215 {Opt_i_version, "i_version"},
1216 {Opt_stripe, "stripe=%u"},
1217 {Opt_resize, "resize"},
1218 {Opt_delalloc, "delalloc"},
1219 {Opt_nodelalloc, "nodelalloc"},
1220 {Opt_block_validity, "block_validity"},
1221 {Opt_noblock_validity, "noblock_validity"},
1222 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1223 {Opt_journal_ioprio, "journal_ioprio=%u"},
1224 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1225 {Opt_auto_da_alloc, "auto_da_alloc"},
1226 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1227 {Opt_dioread_nolock, "dioread_nolock"},
1228 {Opt_dioread_lock, "dioread_lock"},
1229 {Opt_discard, "discard"},
1230 {Opt_nodiscard, "nodiscard"},
1234 static ext4_fsblk_t get_sb_block(void **data)
1236 ext4_fsblk_t sb_block;
1237 char *options = (char *) *data;
1239 if (!options || strncmp(options, "sb=", 3) != 0)
1240 return 1; /* Default location */
1243 /* TODO: use simple_strtoll with >32bit ext4 */
1244 sb_block = simple_strtoul(options, &options, 0);
1245 if (*options && *options != ',') {
1246 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1250 if (*options == ',')
1252 *data = (void *) options;
1257 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1258 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1259 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1262 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1264 struct ext4_sb_info *sbi = EXT4_SB(sb);
1267 if (sb_any_quota_loaded(sb) &&
1268 !sbi->s_qf_names[qtype]) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Cannot change journaled "
1271 "quota options when quota turned on");
1274 qname = match_strdup(args);
1276 ext4_msg(sb, KERN_ERR,
1277 "Not enough memory for storing quotafile name");
1280 if (sbi->s_qf_names[qtype] &&
1281 strcmp(sbi->s_qf_names[qtype], qname)) {
1282 ext4_msg(sb, KERN_ERR,
1283 "%s quota file already specified", QTYPE2NAME(qtype));
1287 sbi->s_qf_names[qtype] = qname;
1288 if (strchr(sbi->s_qf_names[qtype], '/')) {
1289 ext4_msg(sb, KERN_ERR,
1290 "quotafile must be on filesystem root");
1291 kfree(sbi->s_qf_names[qtype]);
1292 sbi->s_qf_names[qtype] = NULL;
1295 set_opt(sbi->s_mount_opt, QUOTA);
1299 static int clear_qf_name(struct super_block *sb, int qtype)
1302 struct ext4_sb_info *sbi = EXT4_SB(sb);
1304 if (sb_any_quota_loaded(sb) &&
1305 sbi->s_qf_names[qtype]) {
1306 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1307 " when quota turned on");
1311 * The space will be released later when all options are confirmed
1314 sbi->s_qf_names[qtype] = NULL;
1319 static int parse_options(char *options, struct super_block *sb,
1320 unsigned long *journal_devnum,
1321 unsigned int *journal_ioprio,
1322 ext4_fsblk_t *n_blocks_count, int is_remount)
1324 struct ext4_sb_info *sbi = EXT4_SB(sb);
1326 substring_t args[MAX_OPT_ARGS];
1336 while ((p = strsep(&options, ",")) != NULL) {
1342 * Initialize args struct so we know whether arg was
1343 * found; some options take optional arguments.
1345 args[0].to = args[0].from = 0;
1346 token = match_token(p, tokens, args);
1349 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1350 clear_opt(sbi->s_mount_opt, MINIX_DF);
1353 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1354 set_opt(sbi->s_mount_opt, MINIX_DF);
1358 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1359 set_opt(sbi->s_mount_opt, GRPID);
1363 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1364 clear_opt(sbi->s_mount_opt, GRPID);
1368 if (match_int(&args[0], &option))
1370 sbi->s_resuid = option;
1373 if (match_int(&args[0], &option))
1375 sbi->s_resgid = option;
1378 /* handled by get_sb_block() instead of here */
1379 /* *sb_block = match_int(&args[0]); */
1382 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1383 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1384 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1387 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1388 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1389 set_opt(sbi->s_mount_opt, ERRORS_RO);
1392 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1393 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1394 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1397 set_opt(sbi->s_mount_opt, NO_UID32);
1400 set_opt(sbi->s_mount_opt, DEBUG);
1403 set_opt(sbi->s_mount_opt, OLDALLOC);
1406 clear_opt(sbi->s_mount_opt, OLDALLOC);
1408 #ifdef CONFIG_EXT4_FS_XATTR
1409 case Opt_user_xattr:
1410 set_opt(sbi->s_mount_opt, XATTR_USER);
1412 case Opt_nouser_xattr:
1413 clear_opt(sbi->s_mount_opt, XATTR_USER);
1416 case Opt_user_xattr:
1417 case Opt_nouser_xattr:
1418 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1421 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1423 set_opt(sbi->s_mount_opt, POSIX_ACL);
1426 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1431 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1434 case Opt_journal_update:
1436 /* Eventually we will want to be able to create
1437 a journal file here. For now, only allow the
1438 user to specify an existing inode to be the
1441 ext4_msg(sb, KERN_ERR,
1442 "Cannot specify journal on remount");
1445 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1447 case Opt_journal_dev:
1449 ext4_msg(sb, KERN_ERR,
1450 "Cannot specify journal on remount");
1453 if (match_int(&args[0], &option))
1455 *journal_devnum = option;
1457 case Opt_journal_checksum:
1458 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1460 case Opt_journal_async_commit:
1461 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1462 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1465 set_opt(sbi->s_mount_opt, NOLOAD);
1468 if (match_int(&args[0], &option))
1473 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1474 sbi->s_commit_interval = HZ * option;
1476 case Opt_max_batch_time:
1477 if (match_int(&args[0], &option))
1482 option = EXT4_DEF_MAX_BATCH_TIME;
1483 sbi->s_max_batch_time = option;
1485 case Opt_min_batch_time:
1486 if (match_int(&args[0], &option))
1490 sbi->s_min_batch_time = option;
1492 case Opt_data_journal:
1493 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1495 case Opt_data_ordered:
1496 data_opt = EXT4_MOUNT_ORDERED_DATA;
1498 case Opt_data_writeback:
1499 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1502 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1503 ext4_msg(sb, KERN_ERR,
1504 "Cannot change data mode on remount");
1508 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1509 sbi->s_mount_opt |= data_opt;
1512 case Opt_data_err_abort:
1513 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1515 case Opt_data_err_ignore:
1516 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1520 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1524 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1527 case Opt_offusrjquota:
1528 if (!clear_qf_name(sb, USRQUOTA))
1531 case Opt_offgrpjquota:
1532 if (!clear_qf_name(sb, GRPQUOTA))
1536 case Opt_jqfmt_vfsold:
1537 qfmt = QFMT_VFS_OLD;
1539 case Opt_jqfmt_vfsv0:
1542 case Opt_jqfmt_vfsv1:
1545 if (sb_any_quota_loaded(sb) &&
1546 sbi->s_jquota_fmt != qfmt) {
1547 ext4_msg(sb, KERN_ERR, "Cannot change "
1548 "journaled quota options when "
1552 sbi->s_jquota_fmt = qfmt;
1556 set_opt(sbi->s_mount_opt, QUOTA);
1557 set_opt(sbi->s_mount_opt, USRQUOTA);
1560 set_opt(sbi->s_mount_opt, QUOTA);
1561 set_opt(sbi->s_mount_opt, GRPQUOTA);
1564 if (sb_any_quota_loaded(sb)) {
1565 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1566 "options when quota turned on");
1569 clear_opt(sbi->s_mount_opt, QUOTA);
1570 clear_opt(sbi->s_mount_opt, USRQUOTA);
1571 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1577 ext4_msg(sb, KERN_ERR,
1578 "quota options not supported");
1582 case Opt_offusrjquota:
1583 case Opt_offgrpjquota:
1584 case Opt_jqfmt_vfsold:
1585 case Opt_jqfmt_vfsv0:
1586 case Opt_jqfmt_vfsv1:
1587 ext4_msg(sb, KERN_ERR,
1588 "journaled quota options not supported");
1594 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1597 clear_opt(sbi->s_mount_opt, BARRIER);
1601 if (match_int(&args[0], &option))
1604 option = 1; /* No argument, default to 1 */
1606 set_opt(sbi->s_mount_opt, BARRIER);
1608 clear_opt(sbi->s_mount_opt, BARRIER);
1614 ext4_msg(sb, KERN_ERR,
1615 "resize option only available "
1619 if (match_int(&args[0], &option) != 0)
1621 *n_blocks_count = option;
1624 set_opt(sbi->s_mount_opt, NOBH);
1627 clear_opt(sbi->s_mount_opt, NOBH);
1630 set_opt(sbi->s_mount_opt, I_VERSION);
1631 sb->s_flags |= MS_I_VERSION;
1633 case Opt_nodelalloc:
1634 clear_opt(sbi->s_mount_opt, DELALLOC);
1637 if (match_int(&args[0], &option))
1641 sbi->s_stripe = option;
1644 set_opt(sbi->s_mount_opt, DELALLOC);
1646 case Opt_block_validity:
1647 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1649 case Opt_noblock_validity:
1650 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1652 case Opt_inode_readahead_blks:
1653 if (match_int(&args[0], &option))
1655 if (option < 0 || option > (1 << 30))
1657 if (!is_power_of_2(option)) {
1658 ext4_msg(sb, KERN_ERR,
1659 "EXT4-fs: inode_readahead_blks"
1660 " must be a power of 2");
1663 sbi->s_inode_readahead_blks = option;
1665 case Opt_journal_ioprio:
1666 if (match_int(&args[0], &option))
1668 if (option < 0 || option > 7)
1670 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1673 case Opt_noauto_da_alloc:
1674 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1676 case Opt_auto_da_alloc:
1678 if (match_int(&args[0], &option))
1681 option = 1; /* No argument, default to 1 */
1683 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1685 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1688 set_opt(sbi->s_mount_opt, DISCARD);
1691 clear_opt(sbi->s_mount_opt, DISCARD);
1693 case Opt_dioread_nolock:
1694 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1696 case Opt_dioread_lock:
1697 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1700 ext4_msg(sb, KERN_ERR,
1701 "Unrecognized mount option \"%s\" "
1702 "or missing value", p);
1707 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1708 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1709 clear_opt(sbi->s_mount_opt, USRQUOTA);
1711 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1712 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1714 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1715 ext4_msg(sb, KERN_ERR, "old and new quota "
1720 if (!sbi->s_jquota_fmt) {
1721 ext4_msg(sb, KERN_ERR, "journaled quota format "
1726 if (sbi->s_jquota_fmt) {
1727 ext4_msg(sb, KERN_ERR, "journaled quota format "
1728 "specified with no journaling "
1737 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1740 struct ext4_sb_info *sbi = EXT4_SB(sb);
1743 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1744 ext4_msg(sb, KERN_ERR, "revision level too high, "
1745 "forcing read-only mode");
1750 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1751 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1752 "running e2fsck is recommended");
1753 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1754 ext4_msg(sb, KERN_WARNING,
1755 "warning: mounting fs with errors, "
1756 "running e2fsck is recommended");
1757 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1758 le16_to_cpu(es->s_mnt_count) >=
1759 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1760 ext4_msg(sb, KERN_WARNING,
1761 "warning: maximal mount count reached, "
1762 "running e2fsck is recommended");
1763 else if (le32_to_cpu(es->s_checkinterval) &&
1764 (le32_to_cpu(es->s_lastcheck) +
1765 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1766 ext4_msg(sb, KERN_WARNING,
1767 "warning: checktime reached, "
1768 "running e2fsck is recommended");
1769 if (!sbi->s_journal)
1770 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1771 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1772 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1773 le16_add_cpu(&es->s_mnt_count, 1);
1774 es->s_mtime = cpu_to_le32(get_seconds());
1775 ext4_update_dynamic_rev(sb);
1777 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1779 ext4_commit_super(sb, 1);
1780 if (test_opt(sb, DEBUG))
1781 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1782 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1784 sbi->s_groups_count,
1785 EXT4_BLOCKS_PER_GROUP(sb),
1786 EXT4_INODES_PER_GROUP(sb),
1792 static int ext4_fill_flex_info(struct super_block *sb)
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 struct ext4_group_desc *gdp = NULL;
1796 ext4_group_t flex_group_count;
1797 ext4_group_t flex_group;
1798 int groups_per_flex = 0;
1802 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1803 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1805 if (groups_per_flex < 2) {
1806 sbi->s_log_groups_per_flex = 0;
1810 /* We allocate both existing and potentially added groups */
1811 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1812 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1813 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1814 size = flex_group_count * sizeof(struct flex_groups);
1815 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1816 if (sbi->s_flex_groups == NULL) {
1817 sbi->s_flex_groups = vmalloc(size);
1818 if (sbi->s_flex_groups)
1819 memset(sbi->s_flex_groups, 0, size);
1821 if (sbi->s_flex_groups == NULL) {
1822 ext4_msg(sb, KERN_ERR, "not enough memory for "
1823 "%u flex groups", flex_group_count);
1827 for (i = 0; i < sbi->s_groups_count; i++) {
1828 gdp = ext4_get_group_desc(sb, i, NULL);
1830 flex_group = ext4_flex_group(sbi, i);
1831 atomic_add(ext4_free_inodes_count(sb, gdp),
1832 &sbi->s_flex_groups[flex_group].free_inodes);
1833 atomic_add(ext4_free_blks_count(sb, gdp),
1834 &sbi->s_flex_groups[flex_group].free_blocks);
1835 atomic_add(ext4_used_dirs_count(sb, gdp),
1836 &sbi->s_flex_groups[flex_group].used_dirs);
1844 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1845 struct ext4_group_desc *gdp)
1849 if (sbi->s_es->s_feature_ro_compat &
1850 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1851 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1852 __le32 le_group = cpu_to_le32(block_group);
1854 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1855 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1856 crc = crc16(crc, (__u8 *)gdp, offset);
1857 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1858 /* for checksum of struct ext4_group_desc do the rest...*/
1859 if ((sbi->s_es->s_feature_incompat &
1860 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1861 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1862 crc = crc16(crc, (__u8 *)gdp + offset,
1863 le16_to_cpu(sbi->s_es->s_desc_size) -
1867 return cpu_to_le16(crc);
1870 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1871 struct ext4_group_desc *gdp)
1873 if ((sbi->s_es->s_feature_ro_compat &
1874 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1875 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1881 /* Called at mount-time, super-block is locked */
1882 static int ext4_check_descriptors(struct super_block *sb)
1884 struct ext4_sb_info *sbi = EXT4_SB(sb);
1885 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1886 ext4_fsblk_t last_block;
1887 ext4_fsblk_t block_bitmap;
1888 ext4_fsblk_t inode_bitmap;
1889 ext4_fsblk_t inode_table;
1890 int flexbg_flag = 0;
1893 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1896 ext4_debug("Checking group descriptors");
1898 for (i = 0; i < sbi->s_groups_count; i++) {
1899 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1901 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1902 last_block = ext4_blocks_count(sbi->s_es) - 1;
1904 last_block = first_block +
1905 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1907 block_bitmap = ext4_block_bitmap(sb, gdp);
1908 if (block_bitmap < first_block || block_bitmap > last_block) {
1909 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1910 "Block bitmap for group %u not in group "
1911 "(block %llu)!", i, block_bitmap);
1914 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1915 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1916 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1917 "Inode bitmap for group %u not in group "
1918 "(block %llu)!", i, inode_bitmap);
1921 inode_table = ext4_inode_table(sb, gdp);
1922 if (inode_table < first_block ||
1923 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1925 "Inode table for group %u not in group "
1926 "(block %llu)!", i, inode_table);
1929 ext4_lock_group(sb, i);
1930 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1931 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1932 "Checksum for group %u failed (%u!=%u)",
1933 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1934 gdp)), le16_to_cpu(gdp->bg_checksum));
1935 if (!(sb->s_flags & MS_RDONLY)) {
1936 ext4_unlock_group(sb, i);
1940 ext4_unlock_group(sb, i);
1942 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1945 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1946 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1950 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1951 * the superblock) which were deleted from all directories, but held open by
1952 * a process at the time of a crash. We walk the list and try to delete these
1953 * inodes at recovery time (only with a read-write filesystem).
1955 * In order to keep the orphan inode chain consistent during traversal (in
1956 * case of crash during recovery), we link each inode into the superblock
1957 * orphan list_head and handle it the same way as an inode deletion during
1958 * normal operation (which journals the operations for us).
1960 * We only do an iget() and an iput() on each inode, which is very safe if we
1961 * accidentally point at an in-use or already deleted inode. The worst that
1962 * can happen in this case is that we get a "bit already cleared" message from
1963 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1964 * e2fsck was run on this filesystem, and it must have already done the orphan
1965 * inode cleanup for us, so we can safely abort without any further action.
1967 static void ext4_orphan_cleanup(struct super_block *sb,
1968 struct ext4_super_block *es)
1970 unsigned int s_flags = sb->s_flags;
1971 int nr_orphans = 0, nr_truncates = 0;
1975 if (!es->s_last_orphan) {
1976 jbd_debug(4, "no orphan inodes to clean up\n");
1980 if (bdev_read_only(sb->s_bdev)) {
1981 ext4_msg(sb, KERN_ERR, "write access "
1982 "unavailable, skipping orphan cleanup");
1986 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1987 if (es->s_last_orphan)
1988 jbd_debug(1, "Errors on filesystem, "
1989 "clearing orphan list.\n");
1990 es->s_last_orphan = 0;
1991 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1995 if (s_flags & MS_RDONLY) {
1996 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1997 sb->s_flags &= ~MS_RDONLY;
2000 /* Needed for iput() to work correctly and not trash data */
2001 sb->s_flags |= MS_ACTIVE;
2002 /* Turn on quotas so that they are updated correctly */
2003 for (i = 0; i < MAXQUOTAS; i++) {
2004 if (EXT4_SB(sb)->s_qf_names[i]) {
2005 int ret = ext4_quota_on_mount(sb, i);
2007 ext4_msg(sb, KERN_ERR,
2008 "Cannot turn on journaled "
2009 "quota: error %d", ret);
2014 while (es->s_last_orphan) {
2015 struct inode *inode;
2017 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2018 if (IS_ERR(inode)) {
2019 es->s_last_orphan = 0;
2023 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2024 dquot_initialize(inode);
2025 if (inode->i_nlink) {
2026 ext4_msg(sb, KERN_DEBUG,
2027 "%s: truncating inode %lu to %lld bytes",
2028 __func__, inode->i_ino, inode->i_size);
2029 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2030 inode->i_ino, inode->i_size);
2031 ext4_truncate(inode);
2034 ext4_msg(sb, KERN_DEBUG,
2035 "%s: deleting unreferenced inode %lu",
2036 __func__, inode->i_ino);
2037 jbd_debug(2, "deleting unreferenced inode %lu\n",
2041 iput(inode); /* The delete magic happens here! */
2044 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2047 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2048 PLURAL(nr_orphans));
2050 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2051 PLURAL(nr_truncates));
2053 /* Turn quotas off */
2054 for (i = 0; i < MAXQUOTAS; i++) {
2055 if (sb_dqopt(sb)->files[i])
2056 vfs_quota_off(sb, i, 0);
2059 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2063 * Maximal extent format file size.
2064 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2065 * extent format containers, within a sector_t, and within i_blocks
2066 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2067 * so that won't be a limiting factor.
2069 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2071 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2074 loff_t upper_limit = MAX_LFS_FILESIZE;
2076 /* small i_blocks in vfs inode? */
2077 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2079 * CONFIG_LBDAF is not enabled implies the inode
2080 * i_block represent total blocks in 512 bytes
2081 * 32 == size of vfs inode i_blocks * 8
2083 upper_limit = (1LL << 32) - 1;
2085 /* total blocks in file system block size */
2086 upper_limit >>= (blkbits - 9);
2087 upper_limit <<= blkbits;
2090 /* 32-bit extent-start container, ee_block */
2095 /* Sanity check against vm- & vfs- imposed limits */
2096 if (res > upper_limit)
2103 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2104 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2105 * We need to be 1 filesystem block less than the 2^48 sector limit.
2107 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2109 loff_t res = EXT4_NDIR_BLOCKS;
2112 /* This is calculated to be the largest file size for a dense, block
2113 * mapped file such that the file's total number of 512-byte sectors,
2114 * including data and all indirect blocks, does not exceed (2^48 - 1).
2116 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2117 * number of 512-byte sectors of the file.
2120 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2122 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2123 * the inode i_block field represents total file blocks in
2124 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2126 upper_limit = (1LL << 32) - 1;
2128 /* total blocks in file system block size */
2129 upper_limit >>= (bits - 9);
2133 * We use 48 bit ext4_inode i_blocks
2134 * With EXT4_HUGE_FILE_FL set the i_blocks
2135 * represent total number of blocks in
2136 * file system block size
2138 upper_limit = (1LL << 48) - 1;
2142 /* indirect blocks */
2144 /* double indirect blocks */
2145 meta_blocks += 1 + (1LL << (bits-2));
2146 /* tripple indirect blocks */
2147 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2149 upper_limit -= meta_blocks;
2150 upper_limit <<= bits;
2152 res += 1LL << (bits-2);
2153 res += 1LL << (2*(bits-2));
2154 res += 1LL << (3*(bits-2));
2156 if (res > upper_limit)
2159 if (res > MAX_LFS_FILESIZE)
2160 res = MAX_LFS_FILESIZE;
2165 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2166 ext4_fsblk_t logical_sb_block, int nr)
2168 struct ext4_sb_info *sbi = EXT4_SB(sb);
2169 ext4_group_t bg, first_meta_bg;
2172 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2174 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2176 return logical_sb_block + nr + 1;
2177 bg = sbi->s_desc_per_block * nr;
2178 if (ext4_bg_has_super(sb, bg))
2181 return (has_super + ext4_group_first_block_no(sb, bg));
2185 * ext4_get_stripe_size: Get the stripe size.
2186 * @sbi: In memory super block info
2188 * If we have specified it via mount option, then
2189 * use the mount option value. If the value specified at mount time is
2190 * greater than the blocks per group use the super block value.
2191 * If the super block value is greater than blocks per group return 0.
2192 * Allocator needs it be less than blocks per group.
2195 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2197 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2198 unsigned long stripe_width =
2199 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2201 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2202 return sbi->s_stripe;
2204 if (stripe_width <= sbi->s_blocks_per_group)
2205 return stripe_width;
2207 if (stride <= sbi->s_blocks_per_group)
2216 struct attribute attr;
2217 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2218 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2219 const char *, size_t);
2223 static int parse_strtoul(const char *buf,
2224 unsigned long max, unsigned long *value)
2228 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2229 endp = skip_spaces(endp);
2230 if (*endp || *value > max)
2236 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2237 struct ext4_sb_info *sbi,
2240 return snprintf(buf, PAGE_SIZE, "%llu\n",
2241 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2244 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2245 struct ext4_sb_info *sbi, char *buf)
2247 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2249 return snprintf(buf, PAGE_SIZE, "%lu\n",
2250 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2251 sbi->s_sectors_written_start) >> 1);
2254 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2255 struct ext4_sb_info *sbi, char *buf)
2257 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2259 return snprintf(buf, PAGE_SIZE, "%llu\n",
2260 (unsigned long long)(sbi->s_kbytes_written +
2261 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2262 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2265 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2266 struct ext4_sb_info *sbi,
2267 const char *buf, size_t count)
2271 if (parse_strtoul(buf, 0x40000000, &t))
2274 if (!is_power_of_2(t))
2277 sbi->s_inode_readahead_blks = t;
2281 static ssize_t sbi_ui_show(struct ext4_attr *a,
2282 struct ext4_sb_info *sbi, char *buf)
2284 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2286 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2289 static ssize_t sbi_ui_store(struct ext4_attr *a,
2290 struct ext4_sb_info *sbi,
2291 const char *buf, size_t count)
2293 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2296 if (parse_strtoul(buf, 0xffffffff, &t))
2302 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2303 static struct ext4_attr ext4_attr_##_name = { \
2304 .attr = {.name = __stringify(_name), .mode = _mode }, \
2307 .offset = offsetof(struct ext4_sb_info, _elname), \
2309 #define EXT4_ATTR(name, mode, show, store) \
2310 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2312 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2313 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2314 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2315 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2316 #define ATTR_LIST(name) &ext4_attr_##name.attr
2318 EXT4_RO_ATTR(delayed_allocation_blocks);
2319 EXT4_RO_ATTR(session_write_kbytes);
2320 EXT4_RO_ATTR(lifetime_write_kbytes);
2321 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2322 inode_readahead_blks_store, s_inode_readahead_blks);
2323 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2324 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2325 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2326 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2327 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2328 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2329 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2330 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2332 static struct attribute *ext4_attrs[] = {
2333 ATTR_LIST(delayed_allocation_blocks),
2334 ATTR_LIST(session_write_kbytes),
2335 ATTR_LIST(lifetime_write_kbytes),
2336 ATTR_LIST(inode_readahead_blks),
2337 ATTR_LIST(inode_goal),
2338 ATTR_LIST(mb_stats),
2339 ATTR_LIST(mb_max_to_scan),
2340 ATTR_LIST(mb_min_to_scan),
2341 ATTR_LIST(mb_order2_req),
2342 ATTR_LIST(mb_stream_req),
2343 ATTR_LIST(mb_group_prealloc),
2344 ATTR_LIST(max_writeback_mb_bump),
2348 static ssize_t ext4_attr_show(struct kobject *kobj,
2349 struct attribute *attr, char *buf)
2351 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2353 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2355 return a->show ? a->show(a, sbi, buf) : 0;
2358 static ssize_t ext4_attr_store(struct kobject *kobj,
2359 struct attribute *attr,
2360 const char *buf, size_t len)
2362 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2364 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2366 return a->store ? a->store(a, sbi, buf, len) : 0;
2369 static void ext4_sb_release(struct kobject *kobj)
2371 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2373 complete(&sbi->s_kobj_unregister);
2377 static const struct sysfs_ops ext4_attr_ops = {
2378 .show = ext4_attr_show,
2379 .store = ext4_attr_store,
2382 static struct kobj_type ext4_ktype = {
2383 .default_attrs = ext4_attrs,
2384 .sysfs_ops = &ext4_attr_ops,
2385 .release = ext4_sb_release,
2389 * Check whether this filesystem can be mounted based on
2390 * the features present and the RDONLY/RDWR mount requested.
2391 * Returns 1 if this filesystem can be mounted as requested,
2392 * 0 if it cannot be.
2394 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2396 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2397 ext4_msg(sb, KERN_ERR,
2398 "Couldn't mount because of "
2399 "unsupported optional features (%x)",
2400 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2401 ~EXT4_FEATURE_INCOMPAT_SUPP));
2408 /* Check that feature set is OK for a read-write mount */
2409 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2410 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2411 "unsupported optional features (%x)",
2412 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2413 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2417 * Large file size enabled file system can only be mounted
2418 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2420 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2421 if (sizeof(blkcnt_t) < sizeof(u64)) {
2422 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2423 "cannot be mounted RDWR without "
2431 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2432 __releases(kernel_lock)
2433 __acquires(kernel_lock)
2435 struct buffer_head *bh;
2436 struct ext4_super_block *es = NULL;
2437 struct ext4_sb_info *sbi;
2439 ext4_fsblk_t sb_block = get_sb_block(&data);
2440 ext4_fsblk_t logical_sb_block;
2441 unsigned long offset = 0;
2442 unsigned long journal_devnum = 0;
2443 unsigned long def_mount_opts;
2449 unsigned int db_count;
2451 int needs_recovery, has_huge_files;
2454 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2456 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2460 sbi->s_blockgroup_lock =
2461 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2462 if (!sbi->s_blockgroup_lock) {
2466 sb->s_fs_info = sbi;
2467 sbi->s_mount_opt = 0;
2468 sbi->s_resuid = EXT4_DEF_RESUID;
2469 sbi->s_resgid = EXT4_DEF_RESGID;
2470 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2471 sbi->s_sb_block = sb_block;
2472 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2477 /* Cleanup superblock name */
2478 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2481 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2483 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2488 * The ext4 superblock will not be buffer aligned for other than 1kB
2489 * block sizes. We need to calculate the offset from buffer start.
2491 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2492 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2493 offset = do_div(logical_sb_block, blocksize);
2495 logical_sb_block = sb_block;
2498 if (!(bh = sb_bread(sb, logical_sb_block))) {
2499 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2503 * Note: s_es must be initialized as soon as possible because
2504 * some ext4 macro-instructions depend on its value
2506 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2508 sb->s_magic = le16_to_cpu(es->s_magic);
2509 if (sb->s_magic != EXT4_SUPER_MAGIC)
2511 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2513 /* Set defaults before we parse the mount options */
2514 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2515 if (def_mount_opts & EXT4_DEFM_DEBUG)
2516 set_opt(sbi->s_mount_opt, DEBUG);
2517 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2518 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2520 set_opt(sbi->s_mount_opt, GRPID);
2522 if (def_mount_opts & EXT4_DEFM_UID16)
2523 set_opt(sbi->s_mount_opt, NO_UID32);
2524 #ifdef CONFIG_EXT4_FS_XATTR
2525 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2526 set_opt(sbi->s_mount_opt, XATTR_USER);
2528 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2529 if (def_mount_opts & EXT4_DEFM_ACL)
2530 set_opt(sbi->s_mount_opt, POSIX_ACL);
2532 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2533 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2534 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2535 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2536 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2537 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2539 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2540 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2541 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2542 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2544 set_opt(sbi->s_mount_opt, ERRORS_RO);
2546 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2547 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2548 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2549 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2550 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2552 set_opt(sbi->s_mount_opt, BARRIER);
2555 * enable delayed allocation by default
2556 * Use -o nodelalloc to turn it off
2558 if (!IS_EXT3_SB(sb))
2559 set_opt(sbi->s_mount_opt, DELALLOC);
2561 if (!parse_options((char *) data, sb, &journal_devnum,
2562 &journal_ioprio, NULL, 0))
2565 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2566 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2568 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2569 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2570 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2571 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2572 ext4_msg(sb, KERN_WARNING,
2573 "feature flags set on rev 0 fs, "
2574 "running e2fsck is recommended");
2577 * Check feature flags regardless of the revision level, since we
2578 * previously didn't change the revision level when setting the flags,
2579 * so there is a chance incompat flags are set on a rev 0 filesystem.
2581 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2584 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2586 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2587 blocksize > EXT4_MAX_BLOCK_SIZE) {
2588 ext4_msg(sb, KERN_ERR,
2589 "Unsupported filesystem blocksize %d", blocksize);
2593 if (sb->s_blocksize != blocksize) {
2594 /* Validate the filesystem blocksize */
2595 if (!sb_set_blocksize(sb, blocksize)) {
2596 ext4_msg(sb, KERN_ERR, "bad block size %d",
2602 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2603 offset = do_div(logical_sb_block, blocksize);
2604 bh = sb_bread(sb, logical_sb_block);
2606 ext4_msg(sb, KERN_ERR,
2607 "Can't read superblock on 2nd try");
2610 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2612 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2613 ext4_msg(sb, KERN_ERR,
2614 "Magic mismatch, very weird!");
2619 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2620 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2621 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2623 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2625 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2626 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2627 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2629 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2630 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2631 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2632 (!is_power_of_2(sbi->s_inode_size)) ||
2633 (sbi->s_inode_size > blocksize)) {
2634 ext4_msg(sb, KERN_ERR,
2635 "unsupported inode size: %d",
2639 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2640 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2643 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2644 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2645 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2646 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2647 !is_power_of_2(sbi->s_desc_size)) {
2648 ext4_msg(sb, KERN_ERR,
2649 "unsupported descriptor size %lu",
2654 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2656 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2657 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2658 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2661 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2662 if (sbi->s_inodes_per_block == 0)
2664 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2665 sbi->s_inodes_per_block;
2666 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2668 sbi->s_mount_state = le16_to_cpu(es->s_state);
2669 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2670 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2672 for (i = 0; i < 4; i++)
2673 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2674 sbi->s_def_hash_version = es->s_def_hash_version;
2675 i = le32_to_cpu(es->s_flags);
2676 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2677 sbi->s_hash_unsigned = 3;
2678 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2679 #ifdef __CHAR_UNSIGNED__
2680 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2681 sbi->s_hash_unsigned = 3;
2683 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2688 if (sbi->s_blocks_per_group > blocksize * 8) {
2689 ext4_msg(sb, KERN_ERR,
2690 "#blocks per group too big: %lu",
2691 sbi->s_blocks_per_group);
2694 if (sbi->s_inodes_per_group > blocksize * 8) {
2695 ext4_msg(sb, KERN_ERR,
2696 "#inodes per group too big: %lu",
2697 sbi->s_inodes_per_group);
2702 * Test whether we have more sectors than will fit in sector_t,
2703 * and whether the max offset is addressable by the page cache.
2705 if ((ext4_blocks_count(es) >
2706 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2707 (ext4_blocks_count(es) >
2708 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2709 ext4_msg(sb, KERN_ERR, "filesystem"
2710 " too large to mount safely on this system");
2711 if (sizeof(sector_t) < 8)
2712 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2717 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2720 /* check blocks count against device size */
2721 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2722 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2723 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2724 "exceeds size of device (%llu blocks)",
2725 ext4_blocks_count(es), blocks_count);
2730 * It makes no sense for the first data block to be beyond the end
2731 * of the filesystem.
2733 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2734 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2735 "block %u is beyond end of filesystem (%llu)",
2736 le32_to_cpu(es->s_first_data_block),
2737 ext4_blocks_count(es));
2740 blocks_count = (ext4_blocks_count(es) -
2741 le32_to_cpu(es->s_first_data_block) +
2742 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2743 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2744 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2745 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2746 "(block count %llu, first data block %u, "
2747 "blocks per group %lu)", sbi->s_groups_count,
2748 ext4_blocks_count(es),
2749 le32_to_cpu(es->s_first_data_block),
2750 EXT4_BLOCKS_PER_GROUP(sb));
2753 sbi->s_groups_count = blocks_count;
2754 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2755 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2756 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2757 EXT4_DESC_PER_BLOCK(sb);
2758 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2760 if (sbi->s_group_desc == NULL) {
2761 ext4_msg(sb, KERN_ERR, "not enough memory");
2765 #ifdef CONFIG_PROC_FS
2767 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2770 bgl_lock_init(sbi->s_blockgroup_lock);
2772 for (i = 0; i < db_count; i++) {
2773 block = descriptor_loc(sb, logical_sb_block, i);
2774 sbi->s_group_desc[i] = sb_bread(sb, block);
2775 if (!sbi->s_group_desc[i]) {
2776 ext4_msg(sb, KERN_ERR,
2777 "can't read group descriptor %d", i);
2782 if (!ext4_check_descriptors(sb)) {
2783 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2786 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2787 if (!ext4_fill_flex_info(sb)) {
2788 ext4_msg(sb, KERN_ERR,
2789 "unable to initialize "
2790 "flex_bg meta info!");
2794 sbi->s_gdb_count = db_count;
2795 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2796 spin_lock_init(&sbi->s_next_gen_lock);
2798 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2799 ext4_count_free_blocks(sb));
2801 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2802 ext4_count_free_inodes(sb));
2805 err = percpu_counter_init(&sbi->s_dirs_counter,
2806 ext4_count_dirs(sb));
2809 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2812 ext4_msg(sb, KERN_ERR, "insufficient memory");
2816 sbi->s_stripe = ext4_get_stripe_size(sbi);
2817 sbi->s_max_writeback_mb_bump = 128;
2820 * set up enough so that it can read an inode
2822 if (!test_opt(sb, NOLOAD) &&
2823 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2824 sb->s_op = &ext4_sops;
2826 sb->s_op = &ext4_nojournal_sops;
2827 sb->s_export_op = &ext4_export_ops;
2828 sb->s_xattr = ext4_xattr_handlers;
2830 sb->s_qcop = &ext4_qctl_operations;
2831 sb->dq_op = &ext4_quota_operations;
2833 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2834 mutex_init(&sbi->s_orphan_lock);
2835 mutex_init(&sbi->s_resize_lock);
2839 needs_recovery = (es->s_last_orphan != 0 ||
2840 EXT4_HAS_INCOMPAT_FEATURE(sb,
2841 EXT4_FEATURE_INCOMPAT_RECOVER));
2844 * The first inode we look at is the journal inode. Don't try
2845 * root first: it may be modified in the journal!
2847 if (!test_opt(sb, NOLOAD) &&
2848 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2849 if (ext4_load_journal(sb, es, journal_devnum))
2851 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2852 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2853 ext4_msg(sb, KERN_ERR, "required journal recovery "
2854 "suppressed and not mounted read-only");
2855 goto failed_mount_wq;
2857 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2858 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2859 sbi->s_journal = NULL;
2864 if (ext4_blocks_count(es) > 0xffffffffULL &&
2865 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2866 JBD2_FEATURE_INCOMPAT_64BIT)) {
2867 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2868 goto failed_mount_wq;
2871 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2872 jbd2_journal_set_features(sbi->s_journal,
2873 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2874 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2875 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2876 jbd2_journal_set_features(sbi->s_journal,
2877 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2878 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2879 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2881 jbd2_journal_clear_features(sbi->s_journal,
2882 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2883 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2886 /* We have now updated the journal if required, so we can
2887 * validate the data journaling mode. */
2888 switch (test_opt(sb, DATA_FLAGS)) {
2890 /* No mode set, assume a default based on the journal
2891 * capabilities: ORDERED_DATA if the journal can
2892 * cope, else JOURNAL_DATA
2894 if (jbd2_journal_check_available_features
2895 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2896 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2898 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2901 case EXT4_MOUNT_ORDERED_DATA:
2902 case EXT4_MOUNT_WRITEBACK_DATA:
2903 if (!jbd2_journal_check_available_features
2904 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2905 ext4_msg(sb, KERN_ERR, "Journal does not support "
2906 "requested data journaling mode");
2907 goto failed_mount_wq;
2912 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2915 if (test_opt(sb, NOBH)) {
2916 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2917 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2918 "its supported only with writeback mode");
2919 clear_opt(sbi->s_mount_opt, NOBH);
2921 if (test_opt(sb, DIOREAD_NOLOCK)) {
2922 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2923 "not supported with nobh mode");
2924 goto failed_mount_wq;
2927 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2928 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2929 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2930 goto failed_mount_wq;
2934 * The jbd2_journal_load will have done any necessary log recovery,
2935 * so we can safely mount the rest of the filesystem now.
2938 root = ext4_iget(sb, EXT4_ROOT_INO);
2940 ext4_msg(sb, KERN_ERR, "get root inode failed");
2941 ret = PTR_ERR(root);
2944 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2946 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2949 sb->s_root = d_alloc_root(root);
2951 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2957 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2959 /* determine the minimum size of new large inodes, if present */
2960 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2961 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2962 EXT4_GOOD_OLD_INODE_SIZE;
2963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2964 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2965 if (sbi->s_want_extra_isize <
2966 le16_to_cpu(es->s_want_extra_isize))
2967 sbi->s_want_extra_isize =
2968 le16_to_cpu(es->s_want_extra_isize);
2969 if (sbi->s_want_extra_isize <
2970 le16_to_cpu(es->s_min_extra_isize))
2971 sbi->s_want_extra_isize =
2972 le16_to_cpu(es->s_min_extra_isize);
2975 /* Check if enough inode space is available */
2976 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2977 sbi->s_inode_size) {
2978 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2979 EXT4_GOOD_OLD_INODE_SIZE;
2980 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2984 if (test_opt(sb, DELALLOC) &&
2985 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2986 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2987 "requested data journaling mode");
2988 clear_opt(sbi->s_mount_opt, DELALLOC);
2990 if (test_opt(sb, DIOREAD_NOLOCK)) {
2991 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2992 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2993 "option - requested data journaling mode");
2994 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2996 if (sb->s_blocksize < PAGE_SIZE) {
2997 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2998 "option - block size is too small");
2999 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3003 err = ext4_setup_system_zone(sb);
3005 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3006 "zone (%d)\n", err);
3011 err = ext4_mb_init(sb, needs_recovery);
3013 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3018 sbi->s_kobj.kset = ext4_kset;
3019 init_completion(&sbi->s_kobj_unregister);
3020 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3023 ext4_mb_release(sb);
3024 ext4_ext_release(sb);
3028 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3029 ext4_orphan_cleanup(sb, es);
3030 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3031 if (needs_recovery) {
3032 ext4_msg(sb, KERN_INFO, "recovery complete");
3033 ext4_mark_recovery_complete(sb, es);
3035 if (EXT4_SB(sb)->s_journal) {
3036 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3037 descr = " journalled data mode";
3038 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3039 descr = " ordered data mode";
3041 descr = " writeback data mode";
3043 descr = "out journal";
3045 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
3052 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3056 ext4_msg(sb, KERN_ERR, "mount failed");
3057 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3059 ext4_release_system_zone(sb);
3060 if (sbi->s_journal) {
3061 jbd2_journal_destroy(sbi->s_journal);
3062 sbi->s_journal = NULL;
3065 if (sbi->s_flex_groups) {
3066 if (is_vmalloc_addr(sbi->s_flex_groups))
3067 vfree(sbi->s_flex_groups);
3069 kfree(sbi->s_flex_groups);
3071 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3072 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3073 percpu_counter_destroy(&sbi->s_dirs_counter);
3074 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3076 for (i = 0; i < db_count; i++)
3077 brelse(sbi->s_group_desc[i]);
3078 kfree(sbi->s_group_desc);
3081 remove_proc_entry(sb->s_id, ext4_proc_root);
3084 for (i = 0; i < MAXQUOTAS; i++)
3085 kfree(sbi->s_qf_names[i]);
3087 ext4_blkdev_remove(sbi);
3090 sb->s_fs_info = NULL;
3091 kfree(sbi->s_blockgroup_lock);
3098 * Setup any per-fs journal parameters now. We'll do this both on
3099 * initial mount, once the journal has been initialised but before we've
3100 * done any recovery; and again on any subsequent remount.
3102 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3106 journal->j_commit_interval = sbi->s_commit_interval;
3107 journal->j_min_batch_time = sbi->s_min_batch_time;
3108 journal->j_max_batch_time = sbi->s_max_batch_time;
3110 spin_lock(&journal->j_state_lock);
3111 if (test_opt(sb, BARRIER))
3112 journal->j_flags |= JBD2_BARRIER;
3114 journal->j_flags &= ~JBD2_BARRIER;
3115 if (test_opt(sb, DATA_ERR_ABORT))
3116 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3118 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3119 spin_unlock(&journal->j_state_lock);
3122 static journal_t *ext4_get_journal(struct super_block *sb,
3123 unsigned int journal_inum)
3125 struct inode *journal_inode;
3128 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3130 /* First, test for the existence of a valid inode on disk. Bad
3131 * things happen if we iget() an unused inode, as the subsequent
3132 * iput() will try to delete it. */
3134 journal_inode = ext4_iget(sb, journal_inum);
3135 if (IS_ERR(journal_inode)) {
3136 ext4_msg(sb, KERN_ERR, "no journal found");
3139 if (!journal_inode->i_nlink) {
3140 make_bad_inode(journal_inode);
3141 iput(journal_inode);
3142 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3146 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3147 journal_inode, journal_inode->i_size);
3148 if (!S_ISREG(journal_inode->i_mode)) {
3149 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3150 iput(journal_inode);
3154 journal = jbd2_journal_init_inode(journal_inode);
3156 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3157 iput(journal_inode);
3160 journal->j_private = sb;
3161 ext4_init_journal_params(sb, journal);
3165 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3168 struct buffer_head *bh;
3172 int hblock, blocksize;
3173 ext4_fsblk_t sb_block;
3174 unsigned long offset;
3175 struct ext4_super_block *es;
3176 struct block_device *bdev;
3178 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3180 bdev = ext4_blkdev_get(j_dev, sb);
3184 if (bd_claim(bdev, sb)) {
3185 ext4_msg(sb, KERN_ERR,
3186 "failed to claim external journal device");
3187 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3191 blocksize = sb->s_blocksize;
3192 hblock = bdev_logical_block_size(bdev);
3193 if (blocksize < hblock) {
3194 ext4_msg(sb, KERN_ERR,
3195 "blocksize too small for journal device");
3199 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3200 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3201 set_blocksize(bdev, blocksize);
3202 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3203 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3204 "external journal");
3208 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3209 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3210 !(le32_to_cpu(es->s_feature_incompat) &
3211 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3212 ext4_msg(sb, KERN_ERR, "external journal has "
3218 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3219 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3224 len = ext4_blocks_count(es);
3225 start = sb_block + 1;
3226 brelse(bh); /* we're done with the superblock */
3228 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3229 start, len, blocksize);
3231 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3234 journal->j_private = sb;
3235 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3236 wait_on_buffer(journal->j_sb_buffer);
3237 if (!buffer_uptodate(journal->j_sb_buffer)) {
3238 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3241 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3242 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3243 "user (unsupported) - %d",
3244 be32_to_cpu(journal->j_superblock->s_nr_users));
3247 EXT4_SB(sb)->journal_bdev = bdev;
3248 ext4_init_journal_params(sb, journal);
3252 jbd2_journal_destroy(journal);
3254 ext4_blkdev_put(bdev);
3258 static int ext4_load_journal(struct super_block *sb,
3259 struct ext4_super_block *es,
3260 unsigned long journal_devnum)
3263 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3266 int really_read_only;
3268 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3270 if (journal_devnum &&
3271 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3272 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3273 "numbers have changed");
3274 journal_dev = new_decode_dev(journal_devnum);
3276 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3278 really_read_only = bdev_read_only(sb->s_bdev);
3281 * Are we loading a blank journal or performing recovery after a
3282 * crash? For recovery, we need to check in advance whether we
3283 * can get read-write access to the device.
3285 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3286 if (sb->s_flags & MS_RDONLY) {
3287 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3288 "required on readonly filesystem");
3289 if (really_read_only) {
3290 ext4_msg(sb, KERN_ERR, "write access "
3291 "unavailable, cannot proceed");
3294 ext4_msg(sb, KERN_INFO, "write access will "
3295 "be enabled during recovery");
3299 if (journal_inum && journal_dev) {
3300 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3301 "and inode journals!");
3306 if (!(journal = ext4_get_journal(sb, journal_inum)))
3309 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3313 if (!(journal->j_flags & JBD2_BARRIER))
3314 ext4_msg(sb, KERN_INFO, "barriers disabled");
3316 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3317 err = jbd2_journal_update_format(journal);
3319 ext4_msg(sb, KERN_ERR, "error updating journal");
3320 jbd2_journal_destroy(journal);
3325 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3326 err = jbd2_journal_wipe(journal, !really_read_only);
3328 err = jbd2_journal_load(journal);
3331 ext4_msg(sb, KERN_ERR, "error loading journal");
3332 jbd2_journal_destroy(journal);
3336 EXT4_SB(sb)->s_journal = journal;
3337 ext4_clear_journal_err(sb, es);
3339 if (journal_devnum &&
3340 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3341 es->s_journal_dev = cpu_to_le32(journal_devnum);
3343 /* Make sure we flush the recovery flag to disk. */
3344 ext4_commit_super(sb, 1);
3350 static int ext4_commit_super(struct super_block *sb, int sync)
3352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3353 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3358 if (buffer_write_io_error(sbh)) {
3360 * Oh, dear. A previous attempt to write the
3361 * superblock failed. This could happen because the
3362 * USB device was yanked out. Or it could happen to
3363 * be a transient write error and maybe the block will
3364 * be remapped. Nothing we can do but to retry the
3365 * write and hope for the best.
3367 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3368 "superblock detected");
3369 clear_buffer_write_io_error(sbh);
3370 set_buffer_uptodate(sbh);
3373 * If the file system is mounted read-only, don't update the
3374 * superblock write time. This avoids updating the superblock
3375 * write time when we are mounting the root file system
3376 * read/only but we need to replay the journal; at that point,
3377 * for people who are east of GMT and who make their clock
3378 * tick in localtime for Windows bug-for-bug compatibility,
3379 * the clock is set in the future, and this will cause e2fsck
3380 * to complain and force a full file system check.
3382 if (!(sb->s_flags & MS_RDONLY))
3383 es->s_wtime = cpu_to_le32(get_seconds());
3384 es->s_kbytes_written =
3385 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3386 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3387 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3388 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3389 &EXT4_SB(sb)->s_freeblocks_counter));
3390 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3391 &EXT4_SB(sb)->s_freeinodes_counter));
3393 BUFFER_TRACE(sbh, "marking dirty");
3394 mark_buffer_dirty(sbh);
3396 error = sync_dirty_buffer(sbh);
3400 error = buffer_write_io_error(sbh);
3402 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3404 clear_buffer_write_io_error(sbh);
3405 set_buffer_uptodate(sbh);
3412 * Have we just finished recovery? If so, and if we are mounting (or
3413 * remounting) the filesystem readonly, then we will end up with a
3414 * consistent fs on disk. Record that fact.
3416 static void ext4_mark_recovery_complete(struct super_block *sb,
3417 struct ext4_super_block *es)
3419 journal_t *journal = EXT4_SB(sb)->s_journal;
3421 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3422 BUG_ON(journal != NULL);
3425 jbd2_journal_lock_updates(journal);
3426 if (jbd2_journal_flush(journal) < 0)
3429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3430 sb->s_flags & MS_RDONLY) {
3431 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3432 ext4_commit_super(sb, 1);
3436 jbd2_journal_unlock_updates(journal);
3440 * If we are mounting (or read-write remounting) a filesystem whose journal
3441 * has recorded an error from a previous lifetime, move that error to the
3442 * main filesystem now.
3444 static void ext4_clear_journal_err(struct super_block *sb,
3445 struct ext4_super_block *es)
3451 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3453 journal = EXT4_SB(sb)->s_journal;
3456 * Now check for any error status which may have been recorded in the
3457 * journal by a prior ext4_error() or ext4_abort()
3460 j_errno = jbd2_journal_errno(journal);
3464 errstr = ext4_decode_error(sb, j_errno, nbuf);
3465 ext4_warning(sb, "Filesystem error recorded "
3466 "from previous mount: %s", errstr);
3467 ext4_warning(sb, "Marking fs in need of filesystem check.");
3469 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3470 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3471 ext4_commit_super(sb, 1);
3473 jbd2_journal_clear_err(journal);
3478 * Force the running and committing transactions to commit,
3479 * and wait on the commit.
3481 int ext4_force_commit(struct super_block *sb)
3486 if (sb->s_flags & MS_RDONLY)
3489 journal = EXT4_SB(sb)->s_journal;
3491 ret = ext4_journal_force_commit(journal);
3496 static void ext4_write_super(struct super_block *sb)
3499 ext4_commit_super(sb, 1);
3503 static int ext4_sync_fs(struct super_block *sb, int wait)
3507 struct ext4_sb_info *sbi = EXT4_SB(sb);
3509 trace_ext4_sync_fs(sb, wait);
3510 flush_workqueue(sbi->dio_unwritten_wq);
3511 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3513 jbd2_log_wait_commit(sbi->s_journal, target);
3519 * LVM calls this function before a (read-only) snapshot is created. This
3520 * gives us a chance to flush the journal completely and mark the fs clean.
3522 static int ext4_freeze(struct super_block *sb)
3527 if (sb->s_flags & MS_RDONLY)
3530 journal = EXT4_SB(sb)->s_journal;
3532 /* Now we set up the journal barrier. */
3533 jbd2_journal_lock_updates(journal);
3536 * Don't clear the needs_recovery flag if we failed to flush
3539 error = jbd2_journal_flush(journal);
3542 jbd2_journal_unlock_updates(journal);
3546 /* Journal blocked and flushed, clear needs_recovery flag. */
3547 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3548 error = ext4_commit_super(sb, 1);
3555 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3556 * flag here, even though the filesystem is not technically dirty yet.
3558 static int ext4_unfreeze(struct super_block *sb)
3560 if (sb->s_flags & MS_RDONLY)
3564 /* Reset the needs_recovery flag before the fs is unlocked. */
3565 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3566 ext4_commit_super(sb, 1);
3568 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3572 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3574 struct ext4_super_block *es;
3575 struct ext4_sb_info *sbi = EXT4_SB(sb);
3576 ext4_fsblk_t n_blocks_count = 0;
3577 unsigned long old_sb_flags;
3578 struct ext4_mount_options old_opts;
3579 int enable_quota = 0;
3581 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3589 /* Store the original options */
3591 old_sb_flags = sb->s_flags;
3592 old_opts.s_mount_opt = sbi->s_mount_opt;
3593 old_opts.s_resuid = sbi->s_resuid;
3594 old_opts.s_resgid = sbi->s_resgid;
3595 old_opts.s_commit_interval = sbi->s_commit_interval;
3596 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3597 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3599 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3600 for (i = 0; i < MAXQUOTAS; i++)
3601 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3603 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3604 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3607 * Allow the "check" option to be passed as a remount option.
3609 if (!parse_options(data, sb, NULL, &journal_ioprio,
3610 &n_blocks_count, 1)) {
3615 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3616 ext4_abort(sb, __func__, "Abort forced by user");
3618 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3619 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3623 if (sbi->s_journal) {
3624 ext4_init_journal_params(sb, sbi->s_journal);
3625 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3628 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3629 n_blocks_count > ext4_blocks_count(es)) {
3630 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3635 if (*flags & MS_RDONLY) {
3636 err = dquot_suspend(sb, -1);
3641 * First of all, the unconditional stuff we have to do
3642 * to disable replay of the journal when we next remount
3644 sb->s_flags |= MS_RDONLY;
3647 * OK, test if we are remounting a valid rw partition
3648 * readonly, and if so set the rdonly flag and then
3649 * mark the partition as valid again.
3651 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3652 (sbi->s_mount_state & EXT4_VALID_FS))
3653 es->s_state = cpu_to_le16(sbi->s_mount_state);
3656 ext4_mark_recovery_complete(sb, es);
3658 /* Make sure we can mount this feature set readwrite */
3659 if (!ext4_feature_set_ok(sb, 0)) {
3664 * Make sure the group descriptor checksums
3665 * are sane. If they aren't, refuse to remount r/w.
3667 for (g = 0; g < sbi->s_groups_count; g++) {
3668 struct ext4_group_desc *gdp =
3669 ext4_get_group_desc(sb, g, NULL);
3671 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3672 ext4_msg(sb, KERN_ERR,
3673 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3674 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3675 le16_to_cpu(gdp->bg_checksum));
3682 * If we have an unprocessed orphan list hanging
3683 * around from a previously readonly bdev mount,
3684 * require a full umount/remount for now.
3686 if (es->s_last_orphan) {
3687 ext4_msg(sb, KERN_WARNING, "Couldn't "
3688 "remount RDWR because of unprocessed "
3689 "orphan inode list. Please "
3690 "umount/remount instead");
3696 * Mounting a RDONLY partition read-write, so reread
3697 * and store the current valid flag. (It may have
3698 * been changed by e2fsck since we originally mounted
3702 ext4_clear_journal_err(sb, es);
3703 sbi->s_mount_state = le16_to_cpu(es->s_state);
3704 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3706 if (!ext4_setup_super(sb, es, 0))
3707 sb->s_flags &= ~MS_RDONLY;
3711 ext4_setup_system_zone(sb);
3712 if (sbi->s_journal == NULL)
3713 ext4_commit_super(sb, 1);
3716 /* Release old quota file names */
3717 for (i = 0; i < MAXQUOTAS; i++)
3718 if (old_opts.s_qf_names[i] &&
3719 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3720 kfree(old_opts.s_qf_names[i]);
3725 dquot_resume(sb, -1);
3729 sb->s_flags = old_sb_flags;
3730 sbi->s_mount_opt = old_opts.s_mount_opt;
3731 sbi->s_resuid = old_opts.s_resuid;
3732 sbi->s_resgid = old_opts.s_resgid;
3733 sbi->s_commit_interval = old_opts.s_commit_interval;
3734 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3735 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3737 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3738 for (i = 0; i < MAXQUOTAS; i++) {
3739 if (sbi->s_qf_names[i] &&
3740 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3741 kfree(sbi->s_qf_names[i]);
3742 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3750 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3752 struct super_block *sb = dentry->d_sb;
3753 struct ext4_sb_info *sbi = EXT4_SB(sb);
3754 struct ext4_super_block *es = sbi->s_es;
3757 if (test_opt(sb, MINIX_DF)) {
3758 sbi->s_overhead_last = 0;
3759 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3760 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3761 ext4_fsblk_t overhead = 0;
3764 * Compute the overhead (FS structures). This is constant
3765 * for a given filesystem unless the number of block groups
3766 * changes so we cache the previous value until it does.
3770 * All of the blocks before first_data_block are
3773 overhead = le32_to_cpu(es->s_first_data_block);
3776 * Add the overhead attributed to the superblock and
3777 * block group descriptors. If the sparse superblocks
3778 * feature is turned on, then not all groups have this.
3780 for (i = 0; i < ngroups; i++) {
3781 overhead += ext4_bg_has_super(sb, i) +
3782 ext4_bg_num_gdb(sb, i);
3787 * Every block group has an inode bitmap, a block
3788 * bitmap, and an inode table.
3790 overhead += ngroups * (2 + sbi->s_itb_per_group);
3791 sbi->s_overhead_last = overhead;
3793 sbi->s_blocks_last = ext4_blocks_count(es);
3796 buf->f_type = EXT4_SUPER_MAGIC;
3797 buf->f_bsize = sb->s_blocksize;
3798 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3799 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3800 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3801 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3802 if (buf->f_bfree < ext4_r_blocks_count(es))
3804 buf->f_files = le32_to_cpu(es->s_inodes_count);
3805 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3806 buf->f_namelen = EXT4_NAME_LEN;
3807 fsid = le64_to_cpup((void *)es->s_uuid) ^
3808 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3809 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3810 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3815 /* Helper function for writing quotas on sync - we need to start transaction
3816 * before quota file is locked for write. Otherwise the are possible deadlocks:
3817 * Process 1 Process 2
3818 * ext4_create() quota_sync()
3819 * jbd2_journal_start() write_dquot()
3820 * dquot_initialize() down(dqio_mutex)
3821 * down(dqio_mutex) jbd2_journal_start()
3827 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3829 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3832 static int ext4_write_dquot(struct dquot *dquot)
3836 struct inode *inode;
3838 inode = dquot_to_inode(dquot);
3839 handle = ext4_journal_start(inode,
3840 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3842 return PTR_ERR(handle);
3843 ret = dquot_commit(dquot);
3844 err = ext4_journal_stop(handle);
3850 static int ext4_acquire_dquot(struct dquot *dquot)
3855 handle = ext4_journal_start(dquot_to_inode(dquot),
3856 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3858 return PTR_ERR(handle);
3859 ret = dquot_acquire(dquot);
3860 err = ext4_journal_stop(handle);
3866 static int ext4_release_dquot(struct dquot *dquot)
3871 handle = ext4_journal_start(dquot_to_inode(dquot),
3872 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3873 if (IS_ERR(handle)) {
3874 /* Release dquot anyway to avoid endless cycle in dqput() */
3875 dquot_release(dquot);
3876 return PTR_ERR(handle);
3878 ret = dquot_release(dquot);
3879 err = ext4_journal_stop(handle);
3885 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3887 /* Are we journaling quotas? */
3888 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3889 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3890 dquot_mark_dquot_dirty(dquot);
3891 return ext4_write_dquot(dquot);
3893 return dquot_mark_dquot_dirty(dquot);
3897 static int ext4_write_info(struct super_block *sb, int type)
3902 /* Data block + inode block */
3903 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3905 return PTR_ERR(handle);
3906 ret = dquot_commit_info(sb, type);
3907 err = ext4_journal_stop(handle);
3914 * Turn on quotas during mount time - we need to find
3915 * the quota file and such...
3917 static int ext4_quota_on_mount(struct super_block *sb, int type)
3919 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3920 EXT4_SB(sb)->s_jquota_fmt, type);
3924 * Standard function to be called on quota_on
3926 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3927 char *name, int remount)
3932 if (!test_opt(sb, QUOTA))
3934 /* When remounting, no checks are needed and in fact, name is NULL */
3936 return vfs_quota_on(sb, type, format_id, name, remount);
3938 err = kern_path(name, LOOKUP_FOLLOW, &path);
3942 /* Quotafile not on the same filesystem? */
3943 if (path.mnt->mnt_sb != sb) {
3947 /* Journaling quota? */
3948 if (EXT4_SB(sb)->s_qf_names[type]) {
3949 /* Quotafile not in fs root? */
3950 if (path.dentry->d_parent != sb->s_root)
3951 ext4_msg(sb, KERN_WARNING,
3952 "Quota file not on filesystem root. "
3953 "Journaled quota will not work");
3957 * When we journal data on quota file, we have to flush journal to see
3958 * all updates to the file when we bypass pagecache...
3960 if (EXT4_SB(sb)->s_journal &&
3961 ext4_should_journal_data(path.dentry->d_inode)) {
3963 * We don't need to lock updates but journal_flush() could
3964 * otherwise be livelocked...
3966 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3967 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3968 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3975 err = vfs_quota_on_path(sb, type, format_id, &path);
3980 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3981 * acquiring the locks... As quota files are never truncated and quota code
3982 * itself serializes the operations (and noone else should touch the files)
3983 * we don't have to be afraid of races */
3984 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3985 size_t len, loff_t off)
3987 struct inode *inode = sb_dqopt(sb)->files[type];
3988 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3990 int offset = off & (sb->s_blocksize - 1);
3993 struct buffer_head *bh;
3994 loff_t i_size = i_size_read(inode);
3998 if (off+len > i_size)
4001 while (toread > 0) {
4002 tocopy = sb->s_blocksize - offset < toread ?
4003 sb->s_blocksize - offset : toread;
4004 bh = ext4_bread(NULL, inode, blk, 0, &err);
4007 if (!bh) /* A hole? */
4008 memset(data, 0, tocopy);
4010 memcpy(data, bh->b_data+offset, tocopy);
4020 /* Write to quotafile (we know the transaction is already started and has
4021 * enough credits) */
4022 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4023 const char *data, size_t len, loff_t off)
4025 struct inode *inode = sb_dqopt(sb)->files[type];
4026 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4028 int offset = off & (sb->s_blocksize - 1);
4029 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4030 struct buffer_head *bh;
4031 handle_t *handle = journal_current_handle();
4033 if (EXT4_SB(sb)->s_journal && !handle) {
4034 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4035 " cancelled because transaction is not started",
4036 (unsigned long long)off, (unsigned long long)len);
4040 * Since we account only one data block in transaction credits,
4041 * then it is impossible to cross a block boundary.
4043 if (sb->s_blocksize - offset < len) {
4044 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4045 " cancelled because not block aligned",
4046 (unsigned long long)off, (unsigned long long)len);
4050 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4051 bh = ext4_bread(handle, inode, blk, 1, &err);
4054 if (journal_quota) {
4055 err = ext4_journal_get_write_access(handle, bh);
4062 memcpy(bh->b_data+offset, data, len);
4063 flush_dcache_page(bh->b_page);
4066 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4068 /* Always do at least ordered writes for quotas */
4069 err = ext4_jbd2_file_inode(handle, inode);
4070 mark_buffer_dirty(bh);
4075 mutex_unlock(&inode->i_mutex);
4078 if (inode->i_size < off + len) {
4079 i_size_write(inode, off + len);
4080 EXT4_I(inode)->i_disksize = inode->i_size;
4082 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4083 ext4_mark_inode_dirty(handle, inode);
4084 mutex_unlock(&inode->i_mutex);
4090 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4091 const char *dev_name, void *data, struct vfsmount *mnt)
4093 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4096 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4097 static struct file_system_type ext2_fs_type = {
4098 .owner = THIS_MODULE,
4100 .get_sb = ext4_get_sb,
4101 .kill_sb = kill_block_super,
4102 .fs_flags = FS_REQUIRES_DEV,
4105 static inline void register_as_ext2(void)
4107 int err = register_filesystem(&ext2_fs_type);
4110 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4113 static inline void unregister_as_ext2(void)
4115 unregister_filesystem(&ext2_fs_type);
4117 MODULE_ALIAS("ext2");
4119 static inline void register_as_ext2(void) { }
4120 static inline void unregister_as_ext2(void) { }
4123 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4124 static inline void register_as_ext3(void)
4126 int err = register_filesystem(&ext3_fs_type);
4129 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4132 static inline void unregister_as_ext3(void)
4134 unregister_filesystem(&ext3_fs_type);
4136 MODULE_ALIAS("ext3");
4138 static inline void register_as_ext3(void) { }
4139 static inline void unregister_as_ext3(void) { }
4142 static struct file_system_type ext4_fs_type = {
4143 .owner = THIS_MODULE,
4145 .get_sb = ext4_get_sb,
4146 .kill_sb = kill_block_super,
4147 .fs_flags = FS_REQUIRES_DEV,
4150 static int __init init_ext4_fs(void)
4154 err = init_ext4_system_zone();
4157 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4160 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4161 err = init_ext4_mballoc();
4165 err = init_ext4_xattr();
4168 err = init_inodecache();
4173 err = register_filesystem(&ext4_fs_type);
4178 unregister_as_ext2();
4179 unregister_as_ext3();
4180 destroy_inodecache();
4184 exit_ext4_mballoc();
4186 remove_proc_entry("fs/ext4", NULL);
4187 kset_unregister(ext4_kset);
4189 exit_ext4_system_zone();
4193 static void __exit exit_ext4_fs(void)
4195 unregister_as_ext2();
4196 unregister_as_ext3();
4197 unregister_filesystem(&ext4_fs_type);
4198 destroy_inodecache();
4200 exit_ext4_mballoc();
4201 remove_proc_entry("fs/ext4", NULL);
4202 kset_unregister(ext4_kset);
4203 exit_ext4_system_zone();
4206 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4207 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4208 MODULE_LICENSE("GPL");
4209 module_init(init_ext4_fs)
4210 module_exit(exit_ext4_fs)