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 flush_workqueue(sbi->dio_unwritten_wq);
649 destroy_workqueue(sbi->dio_unwritten_wq);
654 ext4_commit_super(sb, 1);
656 if (sbi->s_journal) {
657 err = jbd2_journal_destroy(sbi->s_journal);
658 sbi->s_journal = NULL;
660 ext4_abort(sb, __func__,
661 "Couldn't clean up the journal");
664 ext4_release_system_zone(sb);
666 ext4_ext_release(sb);
667 ext4_xattr_put_super(sb);
669 if (!(sb->s_flags & MS_RDONLY)) {
670 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
671 es->s_state = cpu_to_le16(sbi->s_mount_state);
672 ext4_commit_super(sb, 1);
675 remove_proc_entry(sb->s_id, ext4_proc_root);
677 kobject_del(&sbi->s_kobj);
679 for (i = 0; i < sbi->s_gdb_count; i++)
680 brelse(sbi->s_group_desc[i]);
681 kfree(sbi->s_group_desc);
682 if (is_vmalloc_addr(sbi->s_flex_groups))
683 vfree(sbi->s_flex_groups);
685 kfree(sbi->s_flex_groups);
686 percpu_counter_destroy(&sbi->s_freeblocks_counter);
687 percpu_counter_destroy(&sbi->s_freeinodes_counter);
688 percpu_counter_destroy(&sbi->s_dirs_counter);
689 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
692 for (i = 0; i < MAXQUOTAS; i++)
693 kfree(sbi->s_qf_names[i]);
696 /* Debugging code just in case the in-memory inode orphan list
697 * isn't empty. The on-disk one can be non-empty if we've
698 * detected an error and taken the fs readonly, but the
699 * in-memory list had better be clean by this point. */
700 if (!list_empty(&sbi->s_orphan))
701 dump_orphan_list(sb, sbi);
702 J_ASSERT(list_empty(&sbi->s_orphan));
704 invalidate_bdev(sb->s_bdev);
705 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
707 * Invalidate the journal device's buffers. We don't want them
708 * floating about in memory - the physical journal device may
709 * hotswapped, and it breaks the `ro-after' testing code.
711 sync_blockdev(sbi->journal_bdev);
712 invalidate_bdev(sbi->journal_bdev);
713 ext4_blkdev_remove(sbi);
715 sb->s_fs_info = NULL;
717 * Now that we are completely done shutting down the
718 * superblock, we need to actually destroy the kobject.
722 kobject_put(&sbi->s_kobj);
723 wait_for_completion(&sbi->s_kobj_unregister);
724 kfree(sbi->s_blockgroup_lock);
728 static struct kmem_cache *ext4_inode_cachep;
731 * Called inside transaction, so use GFP_NOFS
733 static struct inode *ext4_alloc_inode(struct super_block *sb)
735 struct ext4_inode_info *ei;
737 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
741 ei->vfs_inode.i_version = 1;
742 ei->vfs_inode.i_data.writeback_index = 0;
743 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
744 INIT_LIST_HEAD(&ei->i_prealloc_list);
745 spin_lock_init(&ei->i_prealloc_lock);
747 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
748 * therefore it can be null here. Don't check it, just initialize
751 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
752 ei->i_reserved_data_blocks = 0;
753 ei->i_reserved_meta_blocks = 0;
754 ei->i_allocated_meta_blocks = 0;
755 ei->i_da_metadata_calc_len = 0;
756 ei->i_delalloc_reserved_flag = 0;
757 spin_lock_init(&(ei->i_block_reservation_lock));
759 ei->i_reserved_quota = 0;
761 INIT_LIST_HEAD(&ei->i_completed_io_list);
762 spin_lock_init(&ei->i_completed_io_lock);
763 ei->cur_aio_dio = NULL;
765 ei->i_datasync_tid = 0;
767 return &ei->vfs_inode;
770 static void ext4_destroy_inode(struct inode *inode)
772 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
773 ext4_msg(inode->i_sb, KERN_ERR,
774 "Inode %lu (%p): orphan list check failed!",
775 inode->i_ino, EXT4_I(inode));
776 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
777 EXT4_I(inode), sizeof(struct ext4_inode_info),
781 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
784 static void init_once(void *foo)
786 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
788 INIT_LIST_HEAD(&ei->i_orphan);
789 #ifdef CONFIG_EXT4_FS_XATTR
790 init_rwsem(&ei->xattr_sem);
792 init_rwsem(&ei->i_data_sem);
793 inode_init_once(&ei->vfs_inode);
796 static int init_inodecache(void)
798 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
799 sizeof(struct ext4_inode_info),
800 0, (SLAB_RECLAIM_ACCOUNT|
803 if (ext4_inode_cachep == NULL)
808 static void destroy_inodecache(void)
810 kmem_cache_destroy(ext4_inode_cachep);
813 static void ext4_clear_inode(struct inode *inode)
816 ext4_discard_preallocations(inode);
817 if (EXT4_JOURNAL(inode))
818 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
819 &EXT4_I(inode)->jinode);
822 static inline void ext4_show_quota_options(struct seq_file *seq,
823 struct super_block *sb)
825 #if defined(CONFIG_QUOTA)
826 struct ext4_sb_info *sbi = EXT4_SB(sb);
828 if (sbi->s_jquota_fmt) {
831 switch (sbi->s_jquota_fmt) {
842 seq_printf(seq, ",jqfmt=%s", fmtname);
845 if (sbi->s_qf_names[USRQUOTA])
846 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
848 if (sbi->s_qf_names[GRPQUOTA])
849 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
851 if (test_opt(sb, USRQUOTA))
852 seq_puts(seq, ",usrquota");
854 if (test_opt(sb, GRPQUOTA))
855 seq_puts(seq, ",grpquota");
861 * - it's set to a non-default value OR
862 * - if the per-sb default is different from the global default
864 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
867 unsigned long def_mount_opts;
868 struct super_block *sb = vfs->mnt_sb;
869 struct ext4_sb_info *sbi = EXT4_SB(sb);
870 struct ext4_super_block *es = sbi->s_es;
872 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
873 def_errors = le16_to_cpu(es->s_errors);
875 if (sbi->s_sb_block != 1)
876 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
877 if (test_opt(sb, MINIX_DF))
878 seq_puts(seq, ",minixdf");
879 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
880 seq_puts(seq, ",grpid");
881 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
882 seq_puts(seq, ",nogrpid");
883 if (sbi->s_resuid != EXT4_DEF_RESUID ||
884 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
885 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
887 if (sbi->s_resgid != EXT4_DEF_RESGID ||
888 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
889 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
891 if (test_opt(sb, ERRORS_RO)) {
892 if (def_errors == EXT4_ERRORS_PANIC ||
893 def_errors == EXT4_ERRORS_CONTINUE) {
894 seq_puts(seq, ",errors=remount-ro");
897 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
898 seq_puts(seq, ",errors=continue");
899 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
900 seq_puts(seq, ",errors=panic");
901 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
902 seq_puts(seq, ",nouid32");
903 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
904 seq_puts(seq, ",debug");
905 if (test_opt(sb, OLDALLOC))
906 seq_puts(seq, ",oldalloc");
907 #ifdef CONFIG_EXT4_FS_XATTR
908 if (test_opt(sb, XATTR_USER) &&
909 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
910 seq_puts(seq, ",user_xattr");
911 if (!test_opt(sb, XATTR_USER) &&
912 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
913 seq_puts(seq, ",nouser_xattr");
916 #ifdef CONFIG_EXT4_FS_POSIX_ACL
917 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
918 seq_puts(seq, ",acl");
919 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
920 seq_puts(seq, ",noacl");
922 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
923 seq_printf(seq, ",commit=%u",
924 (unsigned) (sbi->s_commit_interval / HZ));
926 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
927 seq_printf(seq, ",min_batch_time=%u",
928 (unsigned) sbi->s_min_batch_time);
930 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
931 seq_printf(seq, ",max_batch_time=%u",
932 (unsigned) sbi->s_min_batch_time);
936 * We're changing the default of barrier mount option, so
937 * let's always display its mount state so it's clear what its
940 seq_puts(seq, ",barrier=");
941 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
942 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
943 seq_puts(seq, ",journal_async_commit");
944 if (test_opt(sb, NOBH))
945 seq_puts(seq, ",nobh");
946 if (test_opt(sb, I_VERSION))
947 seq_puts(seq, ",i_version");
948 if (!test_opt(sb, DELALLOC))
949 seq_puts(seq, ",nodelalloc");
953 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
955 * journal mode get enabled in different ways
956 * So just print the value even if we didn't specify it
958 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
959 seq_puts(seq, ",data=journal");
960 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
961 seq_puts(seq, ",data=ordered");
962 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
963 seq_puts(seq, ",data=writeback");
965 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
966 seq_printf(seq, ",inode_readahead_blks=%u",
967 sbi->s_inode_readahead_blks);
969 if (test_opt(sb, DATA_ERR_ABORT))
970 seq_puts(seq, ",data_err=abort");
972 if (test_opt(sb, NO_AUTO_DA_ALLOC))
973 seq_puts(seq, ",noauto_da_alloc");
975 if (test_opt(sb, DISCARD))
976 seq_puts(seq, ",discard");
978 if (test_opt(sb, NOLOAD))
979 seq_puts(seq, ",norecovery");
981 if (test_opt(sb, DIOREAD_NOLOCK))
982 seq_puts(seq, ",dioread_nolock");
984 ext4_show_quota_options(seq, sb);
989 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
990 u64 ino, u32 generation)
994 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
995 return ERR_PTR(-ESTALE);
996 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
997 return ERR_PTR(-ESTALE);
999 /* iget isn't really right if the inode is currently unallocated!!
1001 * ext4_read_inode will return a bad_inode if the inode had been
1002 * deleted, so we should be safe.
1004 * Currently we don't know the generation for parent directory, so
1005 * a generation of 0 means "accept any"
1007 inode = ext4_iget(sb, ino);
1009 return ERR_CAST(inode);
1010 if (generation && inode->i_generation != generation) {
1012 return ERR_PTR(-ESTALE);
1018 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1019 int fh_len, int fh_type)
1021 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1022 ext4_nfs_get_inode);
1025 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1026 int fh_len, int fh_type)
1028 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1029 ext4_nfs_get_inode);
1033 * Try to release metadata pages (indirect blocks, directories) which are
1034 * mapped via the block device. Since these pages could have journal heads
1035 * which would prevent try_to_free_buffers() from freeing them, we must use
1036 * jbd2 layer's try_to_free_buffers() function to release them.
1038 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1041 journal_t *journal = EXT4_SB(sb)->s_journal;
1043 WARN_ON(PageChecked(page));
1044 if (!page_has_buffers(page))
1047 return jbd2_journal_try_to_free_buffers(journal, page,
1048 wait & ~__GFP_WAIT);
1049 return try_to_free_buffers(page);
1053 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1054 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1056 static int ext4_write_dquot(struct dquot *dquot);
1057 static int ext4_acquire_dquot(struct dquot *dquot);
1058 static int ext4_release_dquot(struct dquot *dquot);
1059 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1060 static int ext4_write_info(struct super_block *sb, int type);
1061 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1062 char *path, int remount);
1063 static int ext4_quota_on_mount(struct super_block *sb, int type);
1064 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1065 size_t len, loff_t off);
1066 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1067 const char *data, size_t len, loff_t off);
1069 static const struct dquot_operations ext4_quota_operations = {
1071 .get_reserved_space = ext4_get_reserved_space,
1073 .write_dquot = ext4_write_dquot,
1074 .acquire_dquot = ext4_acquire_dquot,
1075 .release_dquot = ext4_release_dquot,
1076 .mark_dirty = ext4_mark_dquot_dirty,
1077 .write_info = ext4_write_info,
1078 .alloc_dquot = dquot_alloc,
1079 .destroy_dquot = dquot_destroy,
1082 static const struct quotactl_ops ext4_qctl_operations = {
1083 .quota_on = ext4_quota_on,
1084 .quota_off = vfs_quota_off,
1085 .quota_sync = vfs_quota_sync,
1086 .get_info = vfs_get_dqinfo,
1087 .set_info = vfs_set_dqinfo,
1088 .get_dqblk = vfs_get_dqblk,
1089 .set_dqblk = vfs_set_dqblk
1093 static const struct super_operations ext4_sops = {
1094 .alloc_inode = ext4_alloc_inode,
1095 .destroy_inode = ext4_destroy_inode,
1096 .write_inode = ext4_write_inode,
1097 .dirty_inode = ext4_dirty_inode,
1098 .delete_inode = ext4_delete_inode,
1099 .put_super = ext4_put_super,
1100 .sync_fs = ext4_sync_fs,
1101 .freeze_fs = ext4_freeze,
1102 .unfreeze_fs = ext4_unfreeze,
1103 .statfs = ext4_statfs,
1104 .remount_fs = ext4_remount,
1105 .clear_inode = ext4_clear_inode,
1106 .show_options = ext4_show_options,
1108 .quota_read = ext4_quota_read,
1109 .quota_write = ext4_quota_write,
1111 .bdev_try_to_free_page = bdev_try_to_free_page,
1114 static const struct super_operations ext4_nojournal_sops = {
1115 .alloc_inode = ext4_alloc_inode,
1116 .destroy_inode = ext4_destroy_inode,
1117 .write_inode = ext4_write_inode,
1118 .dirty_inode = ext4_dirty_inode,
1119 .delete_inode = ext4_delete_inode,
1120 .write_super = ext4_write_super,
1121 .put_super = ext4_put_super,
1122 .statfs = ext4_statfs,
1123 .remount_fs = ext4_remount,
1124 .clear_inode = ext4_clear_inode,
1125 .show_options = ext4_show_options,
1127 .quota_read = ext4_quota_read,
1128 .quota_write = ext4_quota_write,
1130 .bdev_try_to_free_page = bdev_try_to_free_page,
1133 static const struct export_operations ext4_export_ops = {
1134 .fh_to_dentry = ext4_fh_to_dentry,
1135 .fh_to_parent = ext4_fh_to_parent,
1136 .get_parent = ext4_get_parent,
1140 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1141 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1142 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1143 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1144 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1145 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1146 Opt_journal_update, Opt_journal_dev,
1147 Opt_journal_checksum, Opt_journal_async_commit,
1148 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1149 Opt_data_err_abort, Opt_data_err_ignore,
1150 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1151 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1152 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1153 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1154 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1155 Opt_block_validity, Opt_noblock_validity,
1156 Opt_inode_readahead_blks, Opt_journal_ioprio,
1157 Opt_dioread_nolock, Opt_dioread_lock,
1158 Opt_discard, Opt_nodiscard,
1161 static const match_table_t tokens = {
1162 {Opt_bsd_df, "bsddf"},
1163 {Opt_minix_df, "minixdf"},
1164 {Opt_grpid, "grpid"},
1165 {Opt_grpid, "bsdgroups"},
1166 {Opt_nogrpid, "nogrpid"},
1167 {Opt_nogrpid, "sysvgroups"},
1168 {Opt_resgid, "resgid=%u"},
1169 {Opt_resuid, "resuid=%u"},
1171 {Opt_err_cont, "errors=continue"},
1172 {Opt_err_panic, "errors=panic"},
1173 {Opt_err_ro, "errors=remount-ro"},
1174 {Opt_nouid32, "nouid32"},
1175 {Opt_debug, "debug"},
1176 {Opt_oldalloc, "oldalloc"},
1177 {Opt_orlov, "orlov"},
1178 {Opt_user_xattr, "user_xattr"},
1179 {Opt_nouser_xattr, "nouser_xattr"},
1181 {Opt_noacl, "noacl"},
1182 {Opt_noload, "noload"},
1183 {Opt_noload, "norecovery"},
1186 {Opt_commit, "commit=%u"},
1187 {Opt_min_batch_time, "min_batch_time=%u"},
1188 {Opt_max_batch_time, "max_batch_time=%u"},
1189 {Opt_journal_update, "journal=update"},
1190 {Opt_journal_dev, "journal_dev=%u"},
1191 {Opt_journal_checksum, "journal_checksum"},
1192 {Opt_journal_async_commit, "journal_async_commit"},
1193 {Opt_abort, "abort"},
1194 {Opt_data_journal, "data=journal"},
1195 {Opt_data_ordered, "data=ordered"},
1196 {Opt_data_writeback, "data=writeback"},
1197 {Opt_data_err_abort, "data_err=abort"},
1198 {Opt_data_err_ignore, "data_err=ignore"},
1199 {Opt_offusrjquota, "usrjquota="},
1200 {Opt_usrjquota, "usrjquota=%s"},
1201 {Opt_offgrpjquota, "grpjquota="},
1202 {Opt_grpjquota, "grpjquota=%s"},
1203 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1204 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1205 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1206 {Opt_grpquota, "grpquota"},
1207 {Opt_noquota, "noquota"},
1208 {Opt_quota, "quota"},
1209 {Opt_usrquota, "usrquota"},
1210 {Opt_barrier, "barrier=%u"},
1211 {Opt_barrier, "barrier"},
1212 {Opt_nobarrier, "nobarrier"},
1213 {Opt_i_version, "i_version"},
1214 {Opt_stripe, "stripe=%u"},
1215 {Opt_resize, "resize"},
1216 {Opt_delalloc, "delalloc"},
1217 {Opt_nodelalloc, "nodelalloc"},
1218 {Opt_block_validity, "block_validity"},
1219 {Opt_noblock_validity, "noblock_validity"},
1220 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1221 {Opt_journal_ioprio, "journal_ioprio=%u"},
1222 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1223 {Opt_auto_da_alloc, "auto_da_alloc"},
1224 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1225 {Opt_dioread_nolock, "dioread_nolock"},
1226 {Opt_dioread_lock, "dioread_lock"},
1227 {Opt_discard, "discard"},
1228 {Opt_nodiscard, "nodiscard"},
1232 static ext4_fsblk_t get_sb_block(void **data)
1234 ext4_fsblk_t sb_block;
1235 char *options = (char *) *data;
1237 if (!options || strncmp(options, "sb=", 3) != 0)
1238 return 1; /* Default location */
1241 /* TODO: use simple_strtoll with >32bit ext4 */
1242 sb_block = simple_strtoul(options, &options, 0);
1243 if (*options && *options != ',') {
1244 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1248 if (*options == ',')
1250 *data = (void *) options;
1255 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1256 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1257 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1260 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1262 struct ext4_sb_info *sbi = EXT4_SB(sb);
1265 if (sb_any_quota_loaded(sb) &&
1266 !sbi->s_qf_names[qtype]) {
1267 ext4_msg(sb, KERN_ERR,
1268 "Cannot change journaled "
1269 "quota options when quota turned on");
1272 qname = match_strdup(args);
1274 ext4_msg(sb, KERN_ERR,
1275 "Not enough memory for storing quotafile name");
1278 if (sbi->s_qf_names[qtype] &&
1279 strcmp(sbi->s_qf_names[qtype], qname)) {
1280 ext4_msg(sb, KERN_ERR,
1281 "%s quota file already specified", QTYPE2NAME(qtype));
1285 sbi->s_qf_names[qtype] = qname;
1286 if (strchr(sbi->s_qf_names[qtype], '/')) {
1287 ext4_msg(sb, KERN_ERR,
1288 "quotafile must be on filesystem root");
1289 kfree(sbi->s_qf_names[qtype]);
1290 sbi->s_qf_names[qtype] = NULL;
1293 set_opt(sbi->s_mount_opt, QUOTA);
1297 static int clear_qf_name(struct super_block *sb, int qtype)
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1302 if (sb_any_quota_loaded(sb) &&
1303 sbi->s_qf_names[qtype]) {
1304 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1305 " when quota turned on");
1309 * The space will be released later when all options are confirmed
1312 sbi->s_qf_names[qtype] = NULL;
1317 static int parse_options(char *options, struct super_block *sb,
1318 unsigned long *journal_devnum,
1319 unsigned int *journal_ioprio,
1320 ext4_fsblk_t *n_blocks_count, int is_remount)
1322 struct ext4_sb_info *sbi = EXT4_SB(sb);
1324 substring_t args[MAX_OPT_ARGS];
1334 while ((p = strsep(&options, ",")) != NULL) {
1340 * Initialize args struct so we know whether arg was
1341 * found; some options take optional arguments.
1343 args[0].to = args[0].from = 0;
1344 token = match_token(p, tokens, args);
1347 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1348 clear_opt(sbi->s_mount_opt, MINIX_DF);
1351 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1352 set_opt(sbi->s_mount_opt, MINIX_DF);
1356 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1357 set_opt(sbi->s_mount_opt, GRPID);
1361 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1362 clear_opt(sbi->s_mount_opt, GRPID);
1366 if (match_int(&args[0], &option))
1368 sbi->s_resuid = option;
1371 if (match_int(&args[0], &option))
1373 sbi->s_resgid = option;
1376 /* handled by get_sb_block() instead of here */
1377 /* *sb_block = match_int(&args[0]); */
1380 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1381 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1382 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1385 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1386 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1387 set_opt(sbi->s_mount_opt, ERRORS_RO);
1390 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1391 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1392 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1395 set_opt(sbi->s_mount_opt, NO_UID32);
1398 set_opt(sbi->s_mount_opt, DEBUG);
1401 set_opt(sbi->s_mount_opt, OLDALLOC);
1404 clear_opt(sbi->s_mount_opt, OLDALLOC);
1406 #ifdef CONFIG_EXT4_FS_XATTR
1407 case Opt_user_xattr:
1408 set_opt(sbi->s_mount_opt, XATTR_USER);
1410 case Opt_nouser_xattr:
1411 clear_opt(sbi->s_mount_opt, XATTR_USER);
1414 case Opt_user_xattr:
1415 case Opt_nouser_xattr:
1416 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1419 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1421 set_opt(sbi->s_mount_opt, POSIX_ACL);
1424 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1429 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1432 case Opt_journal_update:
1434 /* Eventually we will want to be able to create
1435 a journal file here. For now, only allow the
1436 user to specify an existing inode to be the
1439 ext4_msg(sb, KERN_ERR,
1440 "Cannot specify journal on remount");
1443 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1445 case Opt_journal_dev:
1447 ext4_msg(sb, KERN_ERR,
1448 "Cannot specify journal on remount");
1451 if (match_int(&args[0], &option))
1453 *journal_devnum = option;
1455 case Opt_journal_checksum:
1456 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1458 case Opt_journal_async_commit:
1459 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1460 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1463 set_opt(sbi->s_mount_opt, NOLOAD);
1466 if (match_int(&args[0], &option))
1471 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1472 sbi->s_commit_interval = HZ * option;
1474 case Opt_max_batch_time:
1475 if (match_int(&args[0], &option))
1480 option = EXT4_DEF_MAX_BATCH_TIME;
1481 sbi->s_max_batch_time = option;
1483 case Opt_min_batch_time:
1484 if (match_int(&args[0], &option))
1488 sbi->s_min_batch_time = option;
1490 case Opt_data_journal:
1491 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1493 case Opt_data_ordered:
1494 data_opt = EXT4_MOUNT_ORDERED_DATA;
1496 case Opt_data_writeback:
1497 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1500 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1501 ext4_msg(sb, KERN_ERR,
1502 "Cannot change data mode on remount");
1506 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1507 sbi->s_mount_opt |= data_opt;
1510 case Opt_data_err_abort:
1511 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1513 case Opt_data_err_ignore:
1514 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1518 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1522 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1525 case Opt_offusrjquota:
1526 if (!clear_qf_name(sb, USRQUOTA))
1529 case Opt_offgrpjquota:
1530 if (!clear_qf_name(sb, GRPQUOTA))
1534 case Opt_jqfmt_vfsold:
1535 qfmt = QFMT_VFS_OLD;
1537 case Opt_jqfmt_vfsv0:
1540 case Opt_jqfmt_vfsv1:
1543 if (sb_any_quota_loaded(sb) &&
1544 sbi->s_jquota_fmt != qfmt) {
1545 ext4_msg(sb, KERN_ERR, "Cannot change "
1546 "journaled quota options when "
1550 sbi->s_jquota_fmt = qfmt;
1554 set_opt(sbi->s_mount_opt, QUOTA);
1555 set_opt(sbi->s_mount_opt, USRQUOTA);
1558 set_opt(sbi->s_mount_opt, QUOTA);
1559 set_opt(sbi->s_mount_opt, GRPQUOTA);
1562 if (sb_any_quota_loaded(sb)) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564 "options when quota turned on");
1567 clear_opt(sbi->s_mount_opt, QUOTA);
1568 clear_opt(sbi->s_mount_opt, USRQUOTA);
1569 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1575 ext4_msg(sb, KERN_ERR,
1576 "quota options not supported");
1580 case Opt_offusrjquota:
1581 case Opt_offgrpjquota:
1582 case Opt_jqfmt_vfsold:
1583 case Opt_jqfmt_vfsv0:
1584 case Opt_jqfmt_vfsv1:
1585 ext4_msg(sb, KERN_ERR,
1586 "journaled quota options not supported");
1592 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1595 clear_opt(sbi->s_mount_opt, BARRIER);
1599 if (match_int(&args[0], &option))
1602 option = 1; /* No argument, default to 1 */
1604 set_opt(sbi->s_mount_opt, BARRIER);
1606 clear_opt(sbi->s_mount_opt, BARRIER);
1612 ext4_msg(sb, KERN_ERR,
1613 "resize option only available "
1617 if (match_int(&args[0], &option) != 0)
1619 *n_blocks_count = option;
1622 set_opt(sbi->s_mount_opt, NOBH);
1625 clear_opt(sbi->s_mount_opt, NOBH);
1628 set_opt(sbi->s_mount_opt, I_VERSION);
1629 sb->s_flags |= MS_I_VERSION;
1631 case Opt_nodelalloc:
1632 clear_opt(sbi->s_mount_opt, DELALLOC);
1635 if (match_int(&args[0], &option))
1639 sbi->s_stripe = option;
1642 set_opt(sbi->s_mount_opt, DELALLOC);
1644 case Opt_block_validity:
1645 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1647 case Opt_noblock_validity:
1648 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1650 case Opt_inode_readahead_blks:
1651 if (match_int(&args[0], &option))
1653 if (option < 0 || option > (1 << 30))
1655 if (!is_power_of_2(option)) {
1656 ext4_msg(sb, KERN_ERR,
1657 "EXT4-fs: inode_readahead_blks"
1658 " must be a power of 2");
1661 sbi->s_inode_readahead_blks = option;
1663 case Opt_journal_ioprio:
1664 if (match_int(&args[0], &option))
1666 if (option < 0 || option > 7)
1668 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1671 case Opt_noauto_da_alloc:
1672 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1674 case Opt_auto_da_alloc:
1676 if (match_int(&args[0], &option))
1679 option = 1; /* No argument, default to 1 */
1681 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1683 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1686 set_opt(sbi->s_mount_opt, DISCARD);
1689 clear_opt(sbi->s_mount_opt, DISCARD);
1691 case Opt_dioread_nolock:
1692 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1694 case Opt_dioread_lock:
1695 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1698 ext4_msg(sb, KERN_ERR,
1699 "Unrecognized mount option \"%s\" "
1700 "or missing value", p);
1705 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1706 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1707 clear_opt(sbi->s_mount_opt, USRQUOTA);
1709 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1710 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1712 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1713 ext4_msg(sb, KERN_ERR, "old and new quota "
1718 if (!sbi->s_jquota_fmt) {
1719 ext4_msg(sb, KERN_ERR, "journaled quota format "
1724 if (sbi->s_jquota_fmt) {
1725 ext4_msg(sb, KERN_ERR, "journaled quota format "
1726 "specified with no journaling "
1735 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1738 struct ext4_sb_info *sbi = EXT4_SB(sb);
1741 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1742 ext4_msg(sb, KERN_ERR, "revision level too high, "
1743 "forcing read-only mode");
1748 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1749 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1750 "running e2fsck is recommended");
1751 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1752 ext4_msg(sb, KERN_WARNING,
1753 "warning: mounting fs with errors, "
1754 "running e2fsck is recommended");
1755 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1756 le16_to_cpu(es->s_mnt_count) >=
1757 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1758 ext4_msg(sb, KERN_WARNING,
1759 "warning: maximal mount count reached, "
1760 "running e2fsck is recommended");
1761 else if (le32_to_cpu(es->s_checkinterval) &&
1762 (le32_to_cpu(es->s_lastcheck) +
1763 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1764 ext4_msg(sb, KERN_WARNING,
1765 "warning: checktime reached, "
1766 "running e2fsck is recommended");
1767 if (!sbi->s_journal)
1768 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1769 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1770 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1771 le16_add_cpu(&es->s_mnt_count, 1);
1772 es->s_mtime = cpu_to_le32(get_seconds());
1773 ext4_update_dynamic_rev(sb);
1775 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1777 ext4_commit_super(sb, 1);
1778 if (test_opt(sb, DEBUG))
1779 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1780 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1782 sbi->s_groups_count,
1783 EXT4_BLOCKS_PER_GROUP(sb),
1784 EXT4_INODES_PER_GROUP(sb),
1790 static int ext4_fill_flex_info(struct super_block *sb)
1792 struct ext4_sb_info *sbi = EXT4_SB(sb);
1793 struct ext4_group_desc *gdp = NULL;
1794 ext4_group_t flex_group_count;
1795 ext4_group_t flex_group;
1796 int groups_per_flex = 0;
1800 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1801 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1803 if (groups_per_flex < 2) {
1804 sbi->s_log_groups_per_flex = 0;
1808 /* We allocate both existing and potentially added groups */
1809 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1810 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1811 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1812 size = flex_group_count * sizeof(struct flex_groups);
1813 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1814 if (sbi->s_flex_groups == NULL) {
1815 sbi->s_flex_groups = vmalloc(size);
1816 if (sbi->s_flex_groups)
1817 memset(sbi->s_flex_groups, 0, size);
1819 if (sbi->s_flex_groups == NULL) {
1820 ext4_msg(sb, KERN_ERR, "not enough memory for "
1821 "%u flex groups", flex_group_count);
1825 for (i = 0; i < sbi->s_groups_count; i++) {
1826 gdp = ext4_get_group_desc(sb, i, NULL);
1828 flex_group = ext4_flex_group(sbi, i);
1829 atomic_add(ext4_free_inodes_count(sb, gdp),
1830 &sbi->s_flex_groups[flex_group].free_inodes);
1831 atomic_add(ext4_free_blks_count(sb, gdp),
1832 &sbi->s_flex_groups[flex_group].free_blocks);
1833 atomic_add(ext4_used_dirs_count(sb, gdp),
1834 &sbi->s_flex_groups[flex_group].used_dirs);
1842 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1843 struct ext4_group_desc *gdp)
1847 if (sbi->s_es->s_feature_ro_compat &
1848 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1849 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1850 __le32 le_group = cpu_to_le32(block_group);
1852 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1853 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1854 crc = crc16(crc, (__u8 *)gdp, offset);
1855 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1856 /* for checksum of struct ext4_group_desc do the rest...*/
1857 if ((sbi->s_es->s_feature_incompat &
1858 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1859 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1860 crc = crc16(crc, (__u8 *)gdp + offset,
1861 le16_to_cpu(sbi->s_es->s_desc_size) -
1865 return cpu_to_le16(crc);
1868 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1869 struct ext4_group_desc *gdp)
1871 if ((sbi->s_es->s_feature_ro_compat &
1872 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1873 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1879 /* Called at mount-time, super-block is locked */
1880 static int ext4_check_descriptors(struct super_block *sb)
1882 struct ext4_sb_info *sbi = EXT4_SB(sb);
1883 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1884 ext4_fsblk_t last_block;
1885 ext4_fsblk_t block_bitmap;
1886 ext4_fsblk_t inode_bitmap;
1887 ext4_fsblk_t inode_table;
1888 int flexbg_flag = 0;
1891 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1894 ext4_debug("Checking group descriptors");
1896 for (i = 0; i < sbi->s_groups_count; i++) {
1897 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1899 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1900 last_block = ext4_blocks_count(sbi->s_es) - 1;
1902 last_block = first_block +
1903 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1905 block_bitmap = ext4_block_bitmap(sb, gdp);
1906 if (block_bitmap < first_block || block_bitmap > last_block) {
1907 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1908 "Block bitmap for group %u not in group "
1909 "(block %llu)!", i, block_bitmap);
1912 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1913 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1914 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1915 "Inode bitmap for group %u not in group "
1916 "(block %llu)!", i, inode_bitmap);
1919 inode_table = ext4_inode_table(sb, gdp);
1920 if (inode_table < first_block ||
1921 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1922 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1923 "Inode table for group %u not in group "
1924 "(block %llu)!", i, inode_table);
1927 ext4_lock_group(sb, i);
1928 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1929 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1930 "Checksum for group %u failed (%u!=%u)",
1931 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1932 gdp)), le16_to_cpu(gdp->bg_checksum));
1933 if (!(sb->s_flags & MS_RDONLY)) {
1934 ext4_unlock_group(sb, i);
1938 ext4_unlock_group(sb, i);
1940 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1943 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1944 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
1948 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1949 * the superblock) which were deleted from all directories, but held open by
1950 * a process at the time of a crash. We walk the list and try to delete these
1951 * inodes at recovery time (only with a read-write filesystem).
1953 * In order to keep the orphan inode chain consistent during traversal (in
1954 * case of crash during recovery), we link each inode into the superblock
1955 * orphan list_head and handle it the same way as an inode deletion during
1956 * normal operation (which journals the operations for us).
1958 * We only do an iget() and an iput() on each inode, which is very safe if we
1959 * accidentally point at an in-use or already deleted inode. The worst that
1960 * can happen in this case is that we get a "bit already cleared" message from
1961 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1962 * e2fsck was run on this filesystem, and it must have already done the orphan
1963 * inode cleanup for us, so we can safely abort without any further action.
1965 static void ext4_orphan_cleanup(struct super_block *sb,
1966 struct ext4_super_block *es)
1968 unsigned int s_flags = sb->s_flags;
1969 int nr_orphans = 0, nr_truncates = 0;
1973 if (!es->s_last_orphan) {
1974 jbd_debug(4, "no orphan inodes to clean up\n");
1978 if (bdev_read_only(sb->s_bdev)) {
1979 ext4_msg(sb, KERN_ERR, "write access "
1980 "unavailable, skipping orphan cleanup");
1984 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1985 if (es->s_last_orphan)
1986 jbd_debug(1, "Errors on filesystem, "
1987 "clearing orphan list.\n");
1988 es->s_last_orphan = 0;
1989 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1993 if (s_flags & MS_RDONLY) {
1994 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1995 sb->s_flags &= ~MS_RDONLY;
1998 /* Needed for iput() to work correctly and not trash data */
1999 sb->s_flags |= MS_ACTIVE;
2000 /* Turn on quotas so that they are updated correctly */
2001 for (i = 0; i < MAXQUOTAS; i++) {
2002 if (EXT4_SB(sb)->s_qf_names[i]) {
2003 int ret = ext4_quota_on_mount(sb, i);
2005 ext4_msg(sb, KERN_ERR,
2006 "Cannot turn on journaled "
2007 "quota: error %d", ret);
2012 while (es->s_last_orphan) {
2013 struct inode *inode;
2015 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2016 if (IS_ERR(inode)) {
2017 es->s_last_orphan = 0;
2021 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2022 dquot_initialize(inode);
2023 if (inode->i_nlink) {
2024 ext4_msg(sb, KERN_DEBUG,
2025 "%s: truncating inode %lu to %lld bytes",
2026 __func__, inode->i_ino, inode->i_size);
2027 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2028 inode->i_ino, inode->i_size);
2029 ext4_truncate(inode);
2032 ext4_msg(sb, KERN_DEBUG,
2033 "%s: deleting unreferenced inode %lu",
2034 __func__, inode->i_ino);
2035 jbd_debug(2, "deleting unreferenced inode %lu\n",
2039 iput(inode); /* The delete magic happens here! */
2042 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2045 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2046 PLURAL(nr_orphans));
2048 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2049 PLURAL(nr_truncates));
2051 /* Turn quotas off */
2052 for (i = 0; i < MAXQUOTAS; i++) {
2053 if (sb_dqopt(sb)->files[i])
2054 vfs_quota_off(sb, i, 0);
2057 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2061 * Maximal extent format file size.
2062 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2063 * extent format containers, within a sector_t, and within i_blocks
2064 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2065 * so that won't be a limiting factor.
2067 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2069 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2072 loff_t upper_limit = MAX_LFS_FILESIZE;
2074 /* small i_blocks in vfs inode? */
2075 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2077 * CONFIG_LBDAF is not enabled implies the inode
2078 * i_block represent total blocks in 512 bytes
2079 * 32 == size of vfs inode i_blocks * 8
2081 upper_limit = (1LL << 32) - 1;
2083 /* total blocks in file system block size */
2084 upper_limit >>= (blkbits - 9);
2085 upper_limit <<= blkbits;
2088 /* 32-bit extent-start container, ee_block */
2093 /* Sanity check against vm- & vfs- imposed limits */
2094 if (res > upper_limit)
2101 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2102 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2103 * We need to be 1 filesystem block less than the 2^48 sector limit.
2105 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2107 loff_t res = EXT4_NDIR_BLOCKS;
2110 /* This is calculated to be the largest file size for a dense, block
2111 * mapped file such that the file's total number of 512-byte sectors,
2112 * including data and all indirect blocks, does not exceed (2^48 - 1).
2114 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2115 * number of 512-byte sectors of the file.
2118 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2120 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2121 * the inode i_block field represents total file blocks in
2122 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2124 upper_limit = (1LL << 32) - 1;
2126 /* total blocks in file system block size */
2127 upper_limit >>= (bits - 9);
2131 * We use 48 bit ext4_inode i_blocks
2132 * With EXT4_HUGE_FILE_FL set the i_blocks
2133 * represent total number of blocks in
2134 * file system block size
2136 upper_limit = (1LL << 48) - 1;
2140 /* indirect blocks */
2142 /* double indirect blocks */
2143 meta_blocks += 1 + (1LL << (bits-2));
2144 /* tripple indirect blocks */
2145 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2147 upper_limit -= meta_blocks;
2148 upper_limit <<= bits;
2150 res += 1LL << (bits-2);
2151 res += 1LL << (2*(bits-2));
2152 res += 1LL << (3*(bits-2));
2154 if (res > upper_limit)
2157 if (res > MAX_LFS_FILESIZE)
2158 res = MAX_LFS_FILESIZE;
2163 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2164 ext4_fsblk_t logical_sb_block, int nr)
2166 struct ext4_sb_info *sbi = EXT4_SB(sb);
2167 ext4_group_t bg, first_meta_bg;
2170 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2172 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2174 return logical_sb_block + nr + 1;
2175 bg = sbi->s_desc_per_block * nr;
2176 if (ext4_bg_has_super(sb, bg))
2179 return (has_super + ext4_group_first_block_no(sb, bg));
2183 * ext4_get_stripe_size: Get the stripe size.
2184 * @sbi: In memory super block info
2186 * If we have specified it via mount option, then
2187 * use the mount option value. If the value specified at mount time is
2188 * greater than the blocks per group use the super block value.
2189 * If the super block value is greater than blocks per group return 0.
2190 * Allocator needs it be less than blocks per group.
2193 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2195 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2196 unsigned long stripe_width =
2197 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2199 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2200 return sbi->s_stripe;
2202 if (stripe_width <= sbi->s_blocks_per_group)
2203 return stripe_width;
2205 if (stride <= sbi->s_blocks_per_group)
2214 struct attribute attr;
2215 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2216 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2217 const char *, size_t);
2221 static int parse_strtoul(const char *buf,
2222 unsigned long max, unsigned long *value)
2226 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2227 endp = skip_spaces(endp);
2228 if (*endp || *value > max)
2234 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2235 struct ext4_sb_info *sbi,
2238 return snprintf(buf, PAGE_SIZE, "%llu\n",
2239 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2242 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2243 struct ext4_sb_info *sbi, char *buf)
2245 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2247 return snprintf(buf, PAGE_SIZE, "%lu\n",
2248 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2249 sbi->s_sectors_written_start) >> 1);
2252 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2253 struct ext4_sb_info *sbi, char *buf)
2255 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2257 return snprintf(buf, PAGE_SIZE, "%llu\n",
2258 (unsigned long long)(sbi->s_kbytes_written +
2259 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2260 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2263 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2264 struct ext4_sb_info *sbi,
2265 const char *buf, size_t count)
2269 if (parse_strtoul(buf, 0x40000000, &t))
2272 if (!is_power_of_2(t))
2275 sbi->s_inode_readahead_blks = t;
2279 static ssize_t sbi_ui_show(struct ext4_attr *a,
2280 struct ext4_sb_info *sbi, char *buf)
2282 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2284 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2287 static ssize_t sbi_ui_store(struct ext4_attr *a,
2288 struct ext4_sb_info *sbi,
2289 const char *buf, size_t count)
2291 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2294 if (parse_strtoul(buf, 0xffffffff, &t))
2300 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2301 static struct ext4_attr ext4_attr_##_name = { \
2302 .attr = {.name = __stringify(_name), .mode = _mode }, \
2305 .offset = offsetof(struct ext4_sb_info, _elname), \
2307 #define EXT4_ATTR(name, mode, show, store) \
2308 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2310 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2311 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2312 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2313 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2314 #define ATTR_LIST(name) &ext4_attr_##name.attr
2316 EXT4_RO_ATTR(delayed_allocation_blocks);
2317 EXT4_RO_ATTR(session_write_kbytes);
2318 EXT4_RO_ATTR(lifetime_write_kbytes);
2319 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2320 inode_readahead_blks_store, s_inode_readahead_blks);
2321 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2322 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2323 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2324 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2325 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2326 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2327 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2328 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2330 static struct attribute *ext4_attrs[] = {
2331 ATTR_LIST(delayed_allocation_blocks),
2332 ATTR_LIST(session_write_kbytes),
2333 ATTR_LIST(lifetime_write_kbytes),
2334 ATTR_LIST(inode_readahead_blks),
2335 ATTR_LIST(inode_goal),
2336 ATTR_LIST(mb_stats),
2337 ATTR_LIST(mb_max_to_scan),
2338 ATTR_LIST(mb_min_to_scan),
2339 ATTR_LIST(mb_order2_req),
2340 ATTR_LIST(mb_stream_req),
2341 ATTR_LIST(mb_group_prealloc),
2342 ATTR_LIST(max_writeback_mb_bump),
2346 static ssize_t ext4_attr_show(struct kobject *kobj,
2347 struct attribute *attr, char *buf)
2349 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2351 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2353 return a->show ? a->show(a, sbi, buf) : 0;
2356 static ssize_t ext4_attr_store(struct kobject *kobj,
2357 struct attribute *attr,
2358 const char *buf, size_t len)
2360 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2362 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2364 return a->store ? a->store(a, sbi, buf, len) : 0;
2367 static void ext4_sb_release(struct kobject *kobj)
2369 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2371 complete(&sbi->s_kobj_unregister);
2375 static const struct sysfs_ops ext4_attr_ops = {
2376 .show = ext4_attr_show,
2377 .store = ext4_attr_store,
2380 static struct kobj_type ext4_ktype = {
2381 .default_attrs = ext4_attrs,
2382 .sysfs_ops = &ext4_attr_ops,
2383 .release = ext4_sb_release,
2387 * Check whether this filesystem can be mounted based on
2388 * the features present and the RDONLY/RDWR mount requested.
2389 * Returns 1 if this filesystem can be mounted as requested,
2390 * 0 if it cannot be.
2392 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2394 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2395 ext4_msg(sb, KERN_ERR,
2396 "Couldn't mount because of "
2397 "unsupported optional features (%x)",
2398 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2399 ~EXT4_FEATURE_INCOMPAT_SUPP));
2406 /* Check that feature set is OK for a read-write mount */
2407 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2408 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2409 "unsupported optional features (%x)",
2410 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2411 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2415 * Large file size enabled file system can only be mounted
2416 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2418 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2419 if (sizeof(blkcnt_t) < sizeof(u64)) {
2420 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2421 "cannot be mounted RDWR without "
2429 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2430 __releases(kernel_lock)
2431 __acquires(kernel_lock)
2433 struct buffer_head *bh;
2434 struct ext4_super_block *es = NULL;
2435 struct ext4_sb_info *sbi;
2437 ext4_fsblk_t sb_block = get_sb_block(&data);
2438 ext4_fsblk_t logical_sb_block;
2439 unsigned long offset = 0;
2440 unsigned long journal_devnum = 0;
2441 unsigned long def_mount_opts;
2447 unsigned int db_count;
2449 int needs_recovery, has_huge_files;
2452 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2454 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2458 sbi->s_blockgroup_lock =
2459 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2460 if (!sbi->s_blockgroup_lock) {
2464 sb->s_fs_info = sbi;
2465 sbi->s_mount_opt = 0;
2466 sbi->s_resuid = EXT4_DEF_RESUID;
2467 sbi->s_resgid = EXT4_DEF_RESGID;
2468 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2469 sbi->s_sb_block = sb_block;
2470 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2475 /* Cleanup superblock name */
2476 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2479 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2481 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2486 * The ext4 superblock will not be buffer aligned for other than 1kB
2487 * block sizes. We need to calculate the offset from buffer start.
2489 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2490 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2491 offset = do_div(logical_sb_block, blocksize);
2493 logical_sb_block = sb_block;
2496 if (!(bh = sb_bread(sb, logical_sb_block))) {
2497 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2501 * Note: s_es must be initialized as soon as possible because
2502 * some ext4 macro-instructions depend on its value
2504 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2506 sb->s_magic = le16_to_cpu(es->s_magic);
2507 if (sb->s_magic != EXT4_SUPER_MAGIC)
2509 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2511 /* Set defaults before we parse the mount options */
2512 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2513 if (def_mount_opts & EXT4_DEFM_DEBUG)
2514 set_opt(sbi->s_mount_opt, DEBUG);
2515 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2516 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2518 set_opt(sbi->s_mount_opt, GRPID);
2520 if (def_mount_opts & EXT4_DEFM_UID16)
2521 set_opt(sbi->s_mount_opt, NO_UID32);
2522 #ifdef CONFIG_EXT4_FS_XATTR
2523 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2524 set_opt(sbi->s_mount_opt, XATTR_USER);
2526 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2527 if (def_mount_opts & EXT4_DEFM_ACL)
2528 set_opt(sbi->s_mount_opt, POSIX_ACL);
2530 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2531 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2532 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2533 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2534 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2535 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2537 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2538 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2539 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2540 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2542 set_opt(sbi->s_mount_opt, ERRORS_RO);
2544 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2545 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2546 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2547 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2548 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2550 set_opt(sbi->s_mount_opt, BARRIER);
2553 * enable delayed allocation by default
2554 * Use -o nodelalloc to turn it off
2556 if (!IS_EXT3_SB(sb))
2557 set_opt(sbi->s_mount_opt, DELALLOC);
2559 if (!parse_options((char *) data, sb, &journal_devnum,
2560 &journal_ioprio, NULL, 0))
2563 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2564 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2566 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2567 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2568 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2569 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2570 ext4_msg(sb, KERN_WARNING,
2571 "feature flags set on rev 0 fs, "
2572 "running e2fsck is recommended");
2575 * Check feature flags regardless of the revision level, since we
2576 * previously didn't change the revision level when setting the flags,
2577 * so there is a chance incompat flags are set on a rev 0 filesystem.
2579 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2582 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2584 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2585 blocksize > EXT4_MAX_BLOCK_SIZE) {
2586 ext4_msg(sb, KERN_ERR,
2587 "Unsupported filesystem blocksize %d", blocksize);
2591 if (sb->s_blocksize != blocksize) {
2592 /* Validate the filesystem blocksize */
2593 if (!sb_set_blocksize(sb, blocksize)) {
2594 ext4_msg(sb, KERN_ERR, "bad block size %d",
2600 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2601 offset = do_div(logical_sb_block, blocksize);
2602 bh = sb_bread(sb, logical_sb_block);
2604 ext4_msg(sb, KERN_ERR,
2605 "Can't read superblock on 2nd try");
2608 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2610 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2611 ext4_msg(sb, KERN_ERR,
2612 "Magic mismatch, very weird!");
2617 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2618 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2619 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2621 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2623 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2624 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2625 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2627 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2628 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2629 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2630 (!is_power_of_2(sbi->s_inode_size)) ||
2631 (sbi->s_inode_size > blocksize)) {
2632 ext4_msg(sb, KERN_ERR,
2633 "unsupported inode size: %d",
2637 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2638 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2641 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2642 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2643 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2644 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2645 !is_power_of_2(sbi->s_desc_size)) {
2646 ext4_msg(sb, KERN_ERR,
2647 "unsupported descriptor size %lu",
2652 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2654 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2655 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2656 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2659 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2660 if (sbi->s_inodes_per_block == 0)
2662 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2663 sbi->s_inodes_per_block;
2664 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2666 sbi->s_mount_state = le16_to_cpu(es->s_state);
2667 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2668 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2670 for (i = 0; i < 4; i++)
2671 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2672 sbi->s_def_hash_version = es->s_def_hash_version;
2673 i = le32_to_cpu(es->s_flags);
2674 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2675 sbi->s_hash_unsigned = 3;
2676 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2677 #ifdef __CHAR_UNSIGNED__
2678 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2679 sbi->s_hash_unsigned = 3;
2681 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2686 if (sbi->s_blocks_per_group > blocksize * 8) {
2687 ext4_msg(sb, KERN_ERR,
2688 "#blocks per group too big: %lu",
2689 sbi->s_blocks_per_group);
2692 if (sbi->s_inodes_per_group > blocksize * 8) {
2693 ext4_msg(sb, KERN_ERR,
2694 "#inodes per group too big: %lu",
2695 sbi->s_inodes_per_group);
2700 * Test whether we have more sectors than will fit in sector_t,
2701 * and whether the max offset is addressable by the page cache.
2703 if ((ext4_blocks_count(es) >
2704 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2705 (ext4_blocks_count(es) >
2706 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2707 ext4_msg(sb, KERN_ERR, "filesystem"
2708 " too large to mount safely on this system");
2709 if (sizeof(sector_t) < 8)
2710 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2715 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2718 /* check blocks count against device size */
2719 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2720 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2721 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2722 "exceeds size of device (%llu blocks)",
2723 ext4_blocks_count(es), blocks_count);
2728 * It makes no sense for the first data block to be beyond the end
2729 * of the filesystem.
2731 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2732 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2733 "block %u is beyond end of filesystem (%llu)",
2734 le32_to_cpu(es->s_first_data_block),
2735 ext4_blocks_count(es));
2738 blocks_count = (ext4_blocks_count(es) -
2739 le32_to_cpu(es->s_first_data_block) +
2740 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2741 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2742 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2743 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2744 "(block count %llu, first data block %u, "
2745 "blocks per group %lu)", sbi->s_groups_count,
2746 ext4_blocks_count(es),
2747 le32_to_cpu(es->s_first_data_block),
2748 EXT4_BLOCKS_PER_GROUP(sb));
2751 sbi->s_groups_count = blocks_count;
2752 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2753 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2754 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2755 EXT4_DESC_PER_BLOCK(sb);
2756 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2758 if (sbi->s_group_desc == NULL) {
2759 ext4_msg(sb, KERN_ERR, "not enough memory");
2763 #ifdef CONFIG_PROC_FS
2765 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2768 bgl_lock_init(sbi->s_blockgroup_lock);
2770 for (i = 0; i < db_count; i++) {
2771 block = descriptor_loc(sb, logical_sb_block, i);
2772 sbi->s_group_desc[i] = sb_bread(sb, block);
2773 if (!sbi->s_group_desc[i]) {
2774 ext4_msg(sb, KERN_ERR,
2775 "can't read group descriptor %d", i);
2780 if (!ext4_check_descriptors(sb)) {
2781 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2784 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2785 if (!ext4_fill_flex_info(sb)) {
2786 ext4_msg(sb, KERN_ERR,
2787 "unable to initialize "
2788 "flex_bg meta info!");
2792 sbi->s_gdb_count = db_count;
2793 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2794 spin_lock_init(&sbi->s_next_gen_lock);
2796 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2797 ext4_count_free_blocks(sb));
2799 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2800 ext4_count_free_inodes(sb));
2803 err = percpu_counter_init(&sbi->s_dirs_counter,
2804 ext4_count_dirs(sb));
2807 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2810 ext4_msg(sb, KERN_ERR, "insufficient memory");
2814 sbi->s_stripe = ext4_get_stripe_size(sbi);
2815 sbi->s_max_writeback_mb_bump = 128;
2818 * set up enough so that it can read an inode
2820 if (!test_opt(sb, NOLOAD) &&
2821 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2822 sb->s_op = &ext4_sops;
2824 sb->s_op = &ext4_nojournal_sops;
2825 sb->s_export_op = &ext4_export_ops;
2826 sb->s_xattr = ext4_xattr_handlers;
2828 sb->s_qcop = &ext4_qctl_operations;
2829 sb->dq_op = &ext4_quota_operations;
2831 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2832 mutex_init(&sbi->s_orphan_lock);
2833 mutex_init(&sbi->s_resize_lock);
2837 needs_recovery = (es->s_last_orphan != 0 ||
2838 EXT4_HAS_INCOMPAT_FEATURE(sb,
2839 EXT4_FEATURE_INCOMPAT_RECOVER));
2842 * The first inode we look at is the journal inode. Don't try
2843 * root first: it may be modified in the journal!
2845 if (!test_opt(sb, NOLOAD) &&
2846 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2847 if (ext4_load_journal(sb, es, journal_devnum))
2849 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2850 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2851 ext4_msg(sb, KERN_ERR, "required journal recovery "
2852 "suppressed and not mounted read-only");
2853 goto failed_mount_wq;
2855 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2856 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2857 sbi->s_journal = NULL;
2862 if (ext4_blocks_count(es) > 0xffffffffULL &&
2863 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2864 JBD2_FEATURE_INCOMPAT_64BIT)) {
2865 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2866 goto failed_mount_wq;
2869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2870 jbd2_journal_set_features(sbi->s_journal,
2871 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2872 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2873 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2874 jbd2_journal_set_features(sbi->s_journal,
2875 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2876 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2877 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2879 jbd2_journal_clear_features(sbi->s_journal,
2880 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2881 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2884 /* We have now updated the journal if required, so we can
2885 * validate the data journaling mode. */
2886 switch (test_opt(sb, DATA_FLAGS)) {
2888 /* No mode set, assume a default based on the journal
2889 * capabilities: ORDERED_DATA if the journal can
2890 * cope, else JOURNAL_DATA
2892 if (jbd2_journal_check_available_features
2893 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2894 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2896 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2899 case EXT4_MOUNT_ORDERED_DATA:
2900 case EXT4_MOUNT_WRITEBACK_DATA:
2901 if (!jbd2_journal_check_available_features
2902 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2903 ext4_msg(sb, KERN_ERR, "Journal does not support "
2904 "requested data journaling mode");
2905 goto failed_mount_wq;
2910 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
2913 if (test_opt(sb, NOBH)) {
2914 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2915 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2916 "its supported only with writeback mode");
2917 clear_opt(sbi->s_mount_opt, NOBH);
2919 if (test_opt(sb, DIOREAD_NOLOCK)) {
2920 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2921 "not supported with nobh mode");
2922 goto failed_mount_wq;
2925 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2926 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2927 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2928 goto failed_mount_wq;
2932 * The jbd2_journal_load will have done any necessary log recovery,
2933 * so we can safely mount the rest of the filesystem now.
2936 root = ext4_iget(sb, EXT4_ROOT_INO);
2938 ext4_msg(sb, KERN_ERR, "get root inode failed");
2939 ret = PTR_ERR(root);
2942 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2944 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
2947 sb->s_root = d_alloc_root(root);
2949 ext4_msg(sb, KERN_ERR, "get root dentry failed");
2955 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2957 /* determine the minimum size of new large inodes, if present */
2958 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2959 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2960 EXT4_GOOD_OLD_INODE_SIZE;
2961 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2962 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2963 if (sbi->s_want_extra_isize <
2964 le16_to_cpu(es->s_want_extra_isize))
2965 sbi->s_want_extra_isize =
2966 le16_to_cpu(es->s_want_extra_isize);
2967 if (sbi->s_want_extra_isize <
2968 le16_to_cpu(es->s_min_extra_isize))
2969 sbi->s_want_extra_isize =
2970 le16_to_cpu(es->s_min_extra_isize);
2973 /* Check if enough inode space is available */
2974 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2975 sbi->s_inode_size) {
2976 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2977 EXT4_GOOD_OLD_INODE_SIZE;
2978 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2982 if (test_opt(sb, DELALLOC) &&
2983 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2984 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2985 "requested data journaling mode");
2986 clear_opt(sbi->s_mount_opt, DELALLOC);
2988 if (test_opt(sb, DIOREAD_NOLOCK)) {
2989 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2990 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2991 "option - requested data journaling mode");
2992 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2994 if (sb->s_blocksize < PAGE_SIZE) {
2995 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2996 "option - block size is too small");
2997 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3001 err = ext4_setup_system_zone(sb);
3003 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3004 "zone (%d)\n", err);
3009 err = ext4_mb_init(sb, needs_recovery);
3011 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3016 sbi->s_kobj.kset = ext4_kset;
3017 init_completion(&sbi->s_kobj_unregister);
3018 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3021 ext4_mb_release(sb);
3022 ext4_ext_release(sb);
3026 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3027 ext4_orphan_cleanup(sb, es);
3028 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3029 if (needs_recovery) {
3030 ext4_msg(sb, KERN_INFO, "recovery complete");
3031 ext4_mark_recovery_complete(sb, es);
3033 if (EXT4_SB(sb)->s_journal) {
3034 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3035 descr = " journalled data mode";
3036 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3037 descr = " ordered data mode";
3039 descr = " writeback data mode";
3041 descr = "out journal";
3043 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
3050 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3054 ext4_msg(sb, KERN_ERR, "mount failed");
3055 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3057 ext4_release_system_zone(sb);
3058 if (sbi->s_journal) {
3059 jbd2_journal_destroy(sbi->s_journal);
3060 sbi->s_journal = NULL;
3063 if (sbi->s_flex_groups) {
3064 if (is_vmalloc_addr(sbi->s_flex_groups))
3065 vfree(sbi->s_flex_groups);
3067 kfree(sbi->s_flex_groups);
3069 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3070 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3071 percpu_counter_destroy(&sbi->s_dirs_counter);
3072 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3074 for (i = 0; i < db_count; i++)
3075 brelse(sbi->s_group_desc[i]);
3076 kfree(sbi->s_group_desc);
3079 remove_proc_entry(sb->s_id, ext4_proc_root);
3082 for (i = 0; i < MAXQUOTAS; i++)
3083 kfree(sbi->s_qf_names[i]);
3085 ext4_blkdev_remove(sbi);
3088 sb->s_fs_info = NULL;
3089 kfree(sbi->s_blockgroup_lock);
3096 * Setup any per-fs journal parameters now. We'll do this both on
3097 * initial mount, once the journal has been initialised but before we've
3098 * done any recovery; and again on any subsequent remount.
3100 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3102 struct ext4_sb_info *sbi = EXT4_SB(sb);
3104 journal->j_commit_interval = sbi->s_commit_interval;
3105 journal->j_min_batch_time = sbi->s_min_batch_time;
3106 journal->j_max_batch_time = sbi->s_max_batch_time;
3108 spin_lock(&journal->j_state_lock);
3109 if (test_opt(sb, BARRIER))
3110 journal->j_flags |= JBD2_BARRIER;
3112 journal->j_flags &= ~JBD2_BARRIER;
3113 if (test_opt(sb, DATA_ERR_ABORT))
3114 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3116 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3117 spin_unlock(&journal->j_state_lock);
3120 static journal_t *ext4_get_journal(struct super_block *sb,
3121 unsigned int journal_inum)
3123 struct inode *journal_inode;
3126 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3128 /* First, test for the existence of a valid inode on disk. Bad
3129 * things happen if we iget() an unused inode, as the subsequent
3130 * iput() will try to delete it. */
3132 journal_inode = ext4_iget(sb, journal_inum);
3133 if (IS_ERR(journal_inode)) {
3134 ext4_msg(sb, KERN_ERR, "no journal found");
3137 if (!journal_inode->i_nlink) {
3138 make_bad_inode(journal_inode);
3139 iput(journal_inode);
3140 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3144 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3145 journal_inode, journal_inode->i_size);
3146 if (!S_ISREG(journal_inode->i_mode)) {
3147 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3148 iput(journal_inode);
3152 journal = jbd2_journal_init_inode(journal_inode);
3154 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3155 iput(journal_inode);
3158 journal->j_private = sb;
3159 ext4_init_journal_params(sb, journal);
3163 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3166 struct buffer_head *bh;
3170 int hblock, blocksize;
3171 ext4_fsblk_t sb_block;
3172 unsigned long offset;
3173 struct ext4_super_block *es;
3174 struct block_device *bdev;
3176 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3178 bdev = ext4_blkdev_get(j_dev, sb);
3182 if (bd_claim(bdev, sb)) {
3183 ext4_msg(sb, KERN_ERR,
3184 "failed to claim external journal device");
3185 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3189 blocksize = sb->s_blocksize;
3190 hblock = bdev_logical_block_size(bdev);
3191 if (blocksize < hblock) {
3192 ext4_msg(sb, KERN_ERR,
3193 "blocksize too small for journal device");
3197 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3198 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3199 set_blocksize(bdev, blocksize);
3200 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3201 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3202 "external journal");
3206 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3207 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3208 !(le32_to_cpu(es->s_feature_incompat) &
3209 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3210 ext4_msg(sb, KERN_ERR, "external journal has "
3216 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3217 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3222 len = ext4_blocks_count(es);
3223 start = sb_block + 1;
3224 brelse(bh); /* we're done with the superblock */
3226 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3227 start, len, blocksize);
3229 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3232 journal->j_private = sb;
3233 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3234 wait_on_buffer(journal->j_sb_buffer);
3235 if (!buffer_uptodate(journal->j_sb_buffer)) {
3236 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3239 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3240 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3241 "user (unsupported) - %d",
3242 be32_to_cpu(journal->j_superblock->s_nr_users));
3245 EXT4_SB(sb)->journal_bdev = bdev;
3246 ext4_init_journal_params(sb, journal);
3250 jbd2_journal_destroy(journal);
3252 ext4_blkdev_put(bdev);
3256 static int ext4_load_journal(struct super_block *sb,
3257 struct ext4_super_block *es,
3258 unsigned long journal_devnum)
3261 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3264 int really_read_only;
3266 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3268 if (journal_devnum &&
3269 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3270 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3271 "numbers have changed");
3272 journal_dev = new_decode_dev(journal_devnum);
3274 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3276 really_read_only = bdev_read_only(sb->s_bdev);
3279 * Are we loading a blank journal or performing recovery after a
3280 * crash? For recovery, we need to check in advance whether we
3281 * can get read-write access to the device.
3283 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3284 if (sb->s_flags & MS_RDONLY) {
3285 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3286 "required on readonly filesystem");
3287 if (really_read_only) {
3288 ext4_msg(sb, KERN_ERR, "write access "
3289 "unavailable, cannot proceed");
3292 ext4_msg(sb, KERN_INFO, "write access will "
3293 "be enabled during recovery");
3297 if (journal_inum && journal_dev) {
3298 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3299 "and inode journals!");
3304 if (!(journal = ext4_get_journal(sb, journal_inum)))
3307 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3311 if (!(journal->j_flags & JBD2_BARRIER))
3312 ext4_msg(sb, KERN_INFO, "barriers disabled");
3314 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3315 err = jbd2_journal_update_format(journal);
3317 ext4_msg(sb, KERN_ERR, "error updating journal");
3318 jbd2_journal_destroy(journal);
3323 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3324 err = jbd2_journal_wipe(journal, !really_read_only);
3326 err = jbd2_journal_load(journal);
3329 ext4_msg(sb, KERN_ERR, "error loading journal");
3330 jbd2_journal_destroy(journal);
3334 EXT4_SB(sb)->s_journal = journal;
3335 ext4_clear_journal_err(sb, es);
3337 if (journal_devnum &&
3338 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3339 es->s_journal_dev = cpu_to_le32(journal_devnum);
3341 /* Make sure we flush the recovery flag to disk. */
3342 ext4_commit_super(sb, 1);
3348 static int ext4_commit_super(struct super_block *sb, int sync)
3350 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3351 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3356 if (buffer_write_io_error(sbh)) {
3358 * Oh, dear. A previous attempt to write the
3359 * superblock failed. This could happen because the
3360 * USB device was yanked out. Or it could happen to
3361 * be a transient write error and maybe the block will
3362 * be remapped. Nothing we can do but to retry the
3363 * write and hope for the best.
3365 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3366 "superblock detected");
3367 clear_buffer_write_io_error(sbh);
3368 set_buffer_uptodate(sbh);
3371 * If the file system is mounted read-only, don't update the
3372 * superblock write time. This avoids updating the superblock
3373 * write time when we are mounting the root file system
3374 * read/only but we need to replay the journal; at that point,
3375 * for people who are east of GMT and who make their clock
3376 * tick in localtime for Windows bug-for-bug compatibility,
3377 * the clock is set in the future, and this will cause e2fsck
3378 * to complain and force a full file system check.
3380 if (!(sb->s_flags & MS_RDONLY))
3381 es->s_wtime = cpu_to_le32(get_seconds());
3382 es->s_kbytes_written =
3383 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3384 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3385 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3386 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3387 &EXT4_SB(sb)->s_freeblocks_counter));
3388 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3389 &EXT4_SB(sb)->s_freeinodes_counter));
3391 BUFFER_TRACE(sbh, "marking dirty");
3392 mark_buffer_dirty(sbh);
3394 error = sync_dirty_buffer(sbh);
3398 error = buffer_write_io_error(sbh);
3400 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3402 clear_buffer_write_io_error(sbh);
3403 set_buffer_uptodate(sbh);
3410 * Have we just finished recovery? If so, and if we are mounting (or
3411 * remounting) the filesystem readonly, then we will end up with a
3412 * consistent fs on disk. Record that fact.
3414 static void ext4_mark_recovery_complete(struct super_block *sb,
3415 struct ext4_super_block *es)
3417 journal_t *journal = EXT4_SB(sb)->s_journal;
3419 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3420 BUG_ON(journal != NULL);
3423 jbd2_journal_lock_updates(journal);
3424 if (jbd2_journal_flush(journal) < 0)
3427 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3428 sb->s_flags & MS_RDONLY) {
3429 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3430 ext4_commit_super(sb, 1);
3434 jbd2_journal_unlock_updates(journal);
3438 * If we are mounting (or read-write remounting) a filesystem whose journal
3439 * has recorded an error from a previous lifetime, move that error to the
3440 * main filesystem now.
3442 static void ext4_clear_journal_err(struct super_block *sb,
3443 struct ext4_super_block *es)
3449 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3451 journal = EXT4_SB(sb)->s_journal;
3454 * Now check for any error status which may have been recorded in the
3455 * journal by a prior ext4_error() or ext4_abort()
3458 j_errno = jbd2_journal_errno(journal);
3462 errstr = ext4_decode_error(sb, j_errno, nbuf);
3463 ext4_warning(sb, "Filesystem error recorded "
3464 "from previous mount: %s", errstr);
3465 ext4_warning(sb, "Marking fs in need of filesystem check.");
3467 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3468 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3469 ext4_commit_super(sb, 1);
3471 jbd2_journal_clear_err(journal);
3476 * Force the running and committing transactions to commit,
3477 * and wait on the commit.
3479 int ext4_force_commit(struct super_block *sb)
3484 if (sb->s_flags & MS_RDONLY)
3487 journal = EXT4_SB(sb)->s_journal;
3489 ret = ext4_journal_force_commit(journal);
3494 static void ext4_write_super(struct super_block *sb)
3497 ext4_commit_super(sb, 1);
3501 static int ext4_sync_fs(struct super_block *sb, int wait)
3505 struct ext4_sb_info *sbi = EXT4_SB(sb);
3507 trace_ext4_sync_fs(sb, wait);
3508 flush_workqueue(sbi->dio_unwritten_wq);
3509 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3511 jbd2_log_wait_commit(sbi->s_journal, target);
3517 * LVM calls this function before a (read-only) snapshot is created. This
3518 * gives us a chance to flush the journal completely and mark the fs clean.
3520 static int ext4_freeze(struct super_block *sb)
3525 if (sb->s_flags & MS_RDONLY)
3528 journal = EXT4_SB(sb)->s_journal;
3530 /* Now we set up the journal barrier. */
3531 jbd2_journal_lock_updates(journal);
3534 * Don't clear the needs_recovery flag if we failed to flush
3537 error = jbd2_journal_flush(journal);
3540 jbd2_journal_unlock_updates(journal);
3544 /* Journal blocked and flushed, clear needs_recovery flag. */
3545 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3546 error = ext4_commit_super(sb, 1);
3553 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3554 * flag here, even though the filesystem is not technically dirty yet.
3556 static int ext4_unfreeze(struct super_block *sb)
3558 if (sb->s_flags & MS_RDONLY)
3562 /* Reset the needs_recovery flag before the fs is unlocked. */
3563 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3564 ext4_commit_super(sb, 1);
3566 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3570 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3572 struct ext4_super_block *es;
3573 struct ext4_sb_info *sbi = EXT4_SB(sb);
3574 ext4_fsblk_t n_blocks_count = 0;
3575 unsigned long old_sb_flags;
3576 struct ext4_mount_options old_opts;
3578 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3586 /* Store the original options */
3588 old_sb_flags = sb->s_flags;
3589 old_opts.s_mount_opt = sbi->s_mount_opt;
3590 old_opts.s_resuid = sbi->s_resuid;
3591 old_opts.s_resgid = sbi->s_resgid;
3592 old_opts.s_commit_interval = sbi->s_commit_interval;
3593 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3594 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3596 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3597 for (i = 0; i < MAXQUOTAS; i++)
3598 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3600 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3601 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3604 * Allow the "check" option to be passed as a remount option.
3606 if (!parse_options(data, sb, NULL, &journal_ioprio,
3607 &n_blocks_count, 1)) {
3612 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3613 ext4_abort(sb, __func__, "Abort forced by user");
3615 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3616 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3620 if (sbi->s_journal) {
3621 ext4_init_journal_params(sb, sbi->s_journal);
3622 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3625 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3626 n_blocks_count > ext4_blocks_count(es)) {
3627 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3632 if (*flags & MS_RDONLY) {
3634 * First of all, the unconditional stuff we have to do
3635 * to disable replay of the journal when we next remount
3637 sb->s_flags |= MS_RDONLY;
3640 * OK, test if we are remounting a valid rw partition
3641 * readonly, and if so set the rdonly flag and then
3642 * mark the partition as valid again.
3644 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3645 (sbi->s_mount_state & EXT4_VALID_FS))
3646 es->s_state = cpu_to_le16(sbi->s_mount_state);
3649 ext4_mark_recovery_complete(sb, es);
3651 /* Make sure we can mount this feature set readwrite */
3652 if (!ext4_feature_set_ok(sb, 0)) {
3657 * Make sure the group descriptor checksums
3658 * are sane. If they aren't, refuse to remount r/w.
3660 for (g = 0; g < sbi->s_groups_count; g++) {
3661 struct ext4_group_desc *gdp =
3662 ext4_get_group_desc(sb, g, NULL);
3664 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3665 ext4_msg(sb, KERN_ERR,
3666 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3667 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3668 le16_to_cpu(gdp->bg_checksum));
3675 * If we have an unprocessed orphan list hanging
3676 * around from a previously readonly bdev mount,
3677 * require a full umount/remount for now.
3679 if (es->s_last_orphan) {
3680 ext4_msg(sb, KERN_WARNING, "Couldn't "
3681 "remount RDWR because of unprocessed "
3682 "orphan inode list. Please "
3683 "umount/remount instead");
3689 * Mounting a RDONLY partition read-write, so reread
3690 * and store the current valid flag. (It may have
3691 * been changed by e2fsck since we originally mounted
3695 ext4_clear_journal_err(sb, es);
3696 sbi->s_mount_state = le16_to_cpu(es->s_state);
3697 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3699 if (!ext4_setup_super(sb, es, 0))
3700 sb->s_flags &= ~MS_RDONLY;
3703 ext4_setup_system_zone(sb);
3704 if (sbi->s_journal == NULL)
3705 ext4_commit_super(sb, 1);
3708 /* Release old quota file names */
3709 for (i = 0; i < MAXQUOTAS; i++)
3710 if (old_opts.s_qf_names[i] &&
3711 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3712 kfree(old_opts.s_qf_names[i]);
3719 sb->s_flags = old_sb_flags;
3720 sbi->s_mount_opt = old_opts.s_mount_opt;
3721 sbi->s_resuid = old_opts.s_resuid;
3722 sbi->s_resgid = old_opts.s_resgid;
3723 sbi->s_commit_interval = old_opts.s_commit_interval;
3724 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3725 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3727 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3728 for (i = 0; i < MAXQUOTAS; i++) {
3729 if (sbi->s_qf_names[i] &&
3730 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3731 kfree(sbi->s_qf_names[i]);
3732 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3740 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3742 struct super_block *sb = dentry->d_sb;
3743 struct ext4_sb_info *sbi = EXT4_SB(sb);
3744 struct ext4_super_block *es = sbi->s_es;
3747 if (test_opt(sb, MINIX_DF)) {
3748 sbi->s_overhead_last = 0;
3749 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3750 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3751 ext4_fsblk_t overhead = 0;
3754 * Compute the overhead (FS structures). This is constant
3755 * for a given filesystem unless the number of block groups
3756 * changes so we cache the previous value until it does.
3760 * All of the blocks before first_data_block are
3763 overhead = le32_to_cpu(es->s_first_data_block);
3766 * Add the overhead attributed to the superblock and
3767 * block group descriptors. If the sparse superblocks
3768 * feature is turned on, then not all groups have this.
3770 for (i = 0; i < ngroups; i++) {
3771 overhead += ext4_bg_has_super(sb, i) +
3772 ext4_bg_num_gdb(sb, i);
3777 * Every block group has an inode bitmap, a block
3778 * bitmap, and an inode table.
3780 overhead += ngroups * (2 + sbi->s_itb_per_group);
3781 sbi->s_overhead_last = overhead;
3783 sbi->s_blocks_last = ext4_blocks_count(es);
3786 buf->f_type = EXT4_SUPER_MAGIC;
3787 buf->f_bsize = sb->s_blocksize;
3788 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3789 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3790 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3791 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3792 if (buf->f_bfree < ext4_r_blocks_count(es))
3794 buf->f_files = le32_to_cpu(es->s_inodes_count);
3795 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3796 buf->f_namelen = EXT4_NAME_LEN;
3797 fsid = le64_to_cpup((void *)es->s_uuid) ^
3798 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3799 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3800 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3805 /* Helper function for writing quotas on sync - we need to start transaction
3806 * before quota file is locked for write. Otherwise the are possible deadlocks:
3807 * Process 1 Process 2
3808 * ext4_create() quota_sync()
3809 * jbd2_journal_start() write_dquot()
3810 * dquot_initialize() down(dqio_mutex)
3811 * down(dqio_mutex) jbd2_journal_start()
3817 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3819 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3822 static int ext4_write_dquot(struct dquot *dquot)
3826 struct inode *inode;
3828 inode = dquot_to_inode(dquot);
3829 handle = ext4_journal_start(inode,
3830 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3832 return PTR_ERR(handle);
3833 ret = dquot_commit(dquot);
3834 err = ext4_journal_stop(handle);
3840 static int ext4_acquire_dquot(struct dquot *dquot)
3845 handle = ext4_journal_start(dquot_to_inode(dquot),
3846 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3848 return PTR_ERR(handle);
3849 ret = dquot_acquire(dquot);
3850 err = ext4_journal_stop(handle);
3856 static int ext4_release_dquot(struct dquot *dquot)
3861 handle = ext4_journal_start(dquot_to_inode(dquot),
3862 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3863 if (IS_ERR(handle)) {
3864 /* Release dquot anyway to avoid endless cycle in dqput() */
3865 dquot_release(dquot);
3866 return PTR_ERR(handle);
3868 ret = dquot_release(dquot);
3869 err = ext4_journal_stop(handle);
3875 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3877 /* Are we journaling quotas? */
3878 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3879 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3880 dquot_mark_dquot_dirty(dquot);
3881 return ext4_write_dquot(dquot);
3883 return dquot_mark_dquot_dirty(dquot);
3887 static int ext4_write_info(struct super_block *sb, int type)
3892 /* Data block + inode block */
3893 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3895 return PTR_ERR(handle);
3896 ret = dquot_commit_info(sb, type);
3897 err = ext4_journal_stop(handle);
3904 * Turn on quotas during mount time - we need to find
3905 * the quota file and such...
3907 static int ext4_quota_on_mount(struct super_block *sb, int type)
3909 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3910 EXT4_SB(sb)->s_jquota_fmt, type);
3914 * Standard function to be called on quota_on
3916 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3917 char *name, int remount)
3922 if (!test_opt(sb, QUOTA))
3924 /* When remounting, no checks are needed and in fact, name is NULL */
3926 return vfs_quota_on(sb, type, format_id, name, remount);
3928 err = kern_path(name, LOOKUP_FOLLOW, &path);
3932 /* Quotafile not on the same filesystem? */
3933 if (path.mnt->mnt_sb != sb) {
3937 /* Journaling quota? */
3938 if (EXT4_SB(sb)->s_qf_names[type]) {
3939 /* Quotafile not in fs root? */
3940 if (path.dentry->d_parent != sb->s_root)
3941 ext4_msg(sb, KERN_WARNING,
3942 "Quota file not on filesystem root. "
3943 "Journaled quota will not work");
3947 * When we journal data on quota file, we have to flush journal to see
3948 * all updates to the file when we bypass pagecache...
3950 if (EXT4_SB(sb)->s_journal &&
3951 ext4_should_journal_data(path.dentry->d_inode)) {
3953 * We don't need to lock updates but journal_flush() could
3954 * otherwise be livelocked...
3956 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3957 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3958 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3965 err = vfs_quota_on_path(sb, type, format_id, &path);
3970 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3971 * acquiring the locks... As quota files are never truncated and quota code
3972 * itself serializes the operations (and noone else should touch the files)
3973 * we don't have to be afraid of races */
3974 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3975 size_t len, loff_t off)
3977 struct inode *inode = sb_dqopt(sb)->files[type];
3978 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3980 int offset = off & (sb->s_blocksize - 1);
3983 struct buffer_head *bh;
3984 loff_t i_size = i_size_read(inode);
3988 if (off+len > i_size)
3991 while (toread > 0) {
3992 tocopy = sb->s_blocksize - offset < toread ?
3993 sb->s_blocksize - offset : toread;
3994 bh = ext4_bread(NULL, inode, blk, 0, &err);
3997 if (!bh) /* A hole? */
3998 memset(data, 0, tocopy);
4000 memcpy(data, bh->b_data+offset, tocopy);
4010 /* Write to quotafile (we know the transaction is already started and has
4011 * enough credits) */
4012 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4013 const char *data, size_t len, loff_t off)
4015 struct inode *inode = sb_dqopt(sb)->files[type];
4016 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4018 int offset = off & (sb->s_blocksize - 1);
4019 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4020 struct buffer_head *bh;
4021 handle_t *handle = journal_current_handle();
4023 if (EXT4_SB(sb)->s_journal && !handle) {
4024 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4025 " cancelled because transaction is not started",
4026 (unsigned long long)off, (unsigned long long)len);
4030 * Since we account only one data block in transaction credits,
4031 * then it is impossible to cross a block boundary.
4033 if (sb->s_blocksize - offset < len) {
4034 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4035 " cancelled because not block aligned",
4036 (unsigned long long)off, (unsigned long long)len);
4040 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4041 bh = ext4_bread(handle, inode, blk, 1, &err);
4044 if (journal_quota) {
4045 err = ext4_journal_get_write_access(handle, bh);
4052 memcpy(bh->b_data+offset, data, len);
4053 flush_dcache_page(bh->b_page);
4056 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4058 /* Always do at least ordered writes for quotas */
4059 err = ext4_jbd2_file_inode(handle, inode);
4060 mark_buffer_dirty(bh);
4065 mutex_unlock(&inode->i_mutex);
4068 if (inode->i_size < off + len) {
4069 i_size_write(inode, off + len);
4070 EXT4_I(inode)->i_disksize = inode->i_size;
4072 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4073 ext4_mark_inode_dirty(handle, inode);
4074 mutex_unlock(&inode->i_mutex);
4080 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4081 const char *dev_name, void *data, struct vfsmount *mnt)
4083 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4086 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4087 static struct file_system_type ext2_fs_type = {
4088 .owner = THIS_MODULE,
4090 .get_sb = ext4_get_sb,
4091 .kill_sb = kill_block_super,
4092 .fs_flags = FS_REQUIRES_DEV,
4095 static inline void register_as_ext2(void)
4097 int err = register_filesystem(&ext2_fs_type);
4100 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4103 static inline void unregister_as_ext2(void)
4105 unregister_filesystem(&ext2_fs_type);
4107 MODULE_ALIAS("ext2");
4109 static inline void register_as_ext2(void) { }
4110 static inline void unregister_as_ext2(void) { }
4113 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4114 static inline void register_as_ext3(void)
4116 int err = register_filesystem(&ext3_fs_type);
4119 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4122 static inline void unregister_as_ext3(void)
4124 unregister_filesystem(&ext3_fs_type);
4126 MODULE_ALIAS("ext3");
4128 static inline void register_as_ext3(void) { }
4129 static inline void unregister_as_ext3(void) { }
4132 static struct file_system_type ext4_fs_type = {
4133 .owner = THIS_MODULE,
4135 .get_sb = ext4_get_sb,
4136 .kill_sb = kill_block_super,
4137 .fs_flags = FS_REQUIRES_DEV,
4140 static int __init init_ext4_fs(void)
4144 err = init_ext4_system_zone();
4147 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4150 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4151 err = init_ext4_mballoc();
4155 err = init_ext4_xattr();
4158 err = init_inodecache();
4163 err = register_filesystem(&ext4_fs_type);
4168 unregister_as_ext2();
4169 unregister_as_ext3();
4170 destroy_inodecache();
4174 exit_ext4_mballoc();
4176 remove_proc_entry("fs/ext4", NULL);
4177 kset_unregister(ext4_kset);
4179 exit_ext4_system_zone();
4183 static void __exit exit_ext4_fs(void)
4185 unregister_as_ext2();
4186 unregister_as_ext3();
4187 unregister_filesystem(&ext4_fs_type);
4188 destroy_inodecache();
4190 exit_ext4_mballoc();
4191 remove_proc_entry("fs/ext4", NULL);
4192 kset_unregister(ext4_kset);
4193 exit_ext4_system_zone();
4196 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4197 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4198 MODULE_LICENSE("GPL");
4199 module_init(init_ext4_fs)
4200 module_exit(exit_ext4_fs)