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 vfs_check_frozen(sb, SB_FREEZE_TRANS);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal;
250 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, "Detected aborted journal");
252 return ERR_PTR(-EROFS);
254 return jbd2_journal_start(journal, nblocks);
256 return ext4_get_nojournal();
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
265 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
267 struct super_block *sb;
271 if (!ext4_handle_valid(handle)) {
272 ext4_put_nojournal(handle);
275 sb = handle->h_transaction->t_journal->j_private;
277 rc = jbd2_journal_stop(handle);
282 __ext4_std_error(sb, where, line, err);
286 void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
293 BUG_ON(!ext4_handle_valid(handle));
296 BUFFER_TRACE(bh, "abort");
301 if (is_handle_aborted(handle))
304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
307 jbd2_journal_abort_handle(handle);
310 static void __save_error_info(struct super_block *sb, const char *func,
313 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
315 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
316 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
317 es->s_last_error_time = cpu_to_le32(get_seconds());
318 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
319 es->s_last_error_line = cpu_to_le32(line);
320 if (!es->s_first_error_time) {
321 es->s_first_error_time = es->s_last_error_time;
322 strncpy(es->s_first_error_func, func,
323 sizeof(es->s_first_error_func));
324 es->s_first_error_line = cpu_to_le32(line);
325 es->s_first_error_ino = es->s_last_error_ino;
326 es->s_first_error_block = es->s_last_error_block;
329 * Start the daily error reporting function if it hasn't been
332 if (!es->s_error_count)
333 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
334 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
337 static void save_error_info(struct super_block *sb, const char *func,
340 __save_error_info(sb, func, line);
341 ext4_commit_super(sb, 1);
345 /* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
348 * On ext2, we can store the error state of the filesystem in the
349 * superblock. That is not possible on ext4, because we may have other
350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
355 * We'll just use the jbd2_journal_abort() error code to record an error in
356 * the journal instead. On recovery, the journal will complain about
357 * that error until we've noted it down and cleared it.
360 static void ext4_handle_error(struct super_block *sb)
362 if (sb->s_flags & MS_RDONLY)
365 if (!test_opt(sb, ERRORS_CONT)) {
366 journal_t *journal = EXT4_SB(sb)->s_journal;
368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
370 jbd2_journal_abort(journal, -EIO);
372 if (test_opt(sb, ERRORS_RO)) {
373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
374 sb->s_flags |= MS_RDONLY;
376 if (test_opt(sb, ERRORS_PANIC))
377 panic("EXT4-fs (device %s): panic forced after error\n",
381 void __ext4_error(struct super_block *sb, const char *function,
382 unsigned int line, const char *fmt, ...)
387 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
388 sb->s_id, function, line, current->comm);
393 ext4_handle_error(sb);
396 void ext4_error_inode(struct inode *inode, const char *function,
397 unsigned int line, ext4_fsblk_t block,
398 const char *fmt, ...)
401 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
403 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
404 es->s_last_error_block = cpu_to_le64(block);
405 save_error_info(inode->i_sb, function, line);
407 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
408 inode->i_sb->s_id, function, line, inode->i_ino);
410 printk("block %llu: ", block);
411 printk("comm %s: ", current->comm);
416 ext4_handle_error(inode->i_sb);
419 void ext4_error_file(struct file *file, const char *function,
420 unsigned int line, const char *fmt, ...)
423 struct ext4_super_block *es;
424 struct inode *inode = file->f_dentry->d_inode;
425 char pathname[80], *path;
427 es = EXT4_SB(inode->i_sb)->s_es;
428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 save_error_info(inode->i_sb, function, line);
431 path = d_path(&(file->f_path), pathname, sizeof(pathname));
435 "EXT4-fs error (device %s): %s:%d: inode #%lu "
436 "(comm %s path %s): ",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 current->comm, path);
443 ext4_handle_error(inode->i_sb);
446 static const char *ext4_decode_error(struct super_block *sb, int errno,
453 errstr = "IO failure";
456 errstr = "Out of memory";
459 if (!sb || (EXT4_SB(sb)->s_journal &&
460 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
461 errstr = "Journal has aborted";
463 errstr = "Readonly filesystem";
466 /* If the caller passed in an extra buffer for unknown
467 * errors, textualise them now. Else we just return
470 /* Check for truncated error codes... */
471 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
480 /* __ext4_std_error decodes expected errors from journaling functions
481 * automatically and invokes the appropriate error response. */
483 void __ext4_std_error(struct super_block *sb, const char *function,
484 unsigned int line, int errno)
489 /* Special case: if the error is EROFS, and we're not already
490 * inside a transaction, then there's really no point in logging
492 if (errno == -EROFS && journal_current_handle() == NULL &&
493 (sb->s_flags & MS_RDONLY))
496 errstr = ext4_decode_error(sb, errno, nbuf);
497 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
498 sb->s_id, function, line, errstr);
499 save_error_info(sb, function, line);
501 ext4_handle_error(sb);
505 * ext4_abort is a much stronger failure handler than ext4_error. The
506 * abort function may be used to deal with unrecoverable failures such
507 * as journal IO errors or ENOMEM at a critical moment in log management.
509 * We unconditionally force the filesystem into an ABORT|READONLY state,
510 * unless the error response on the fs has been set to panic in which
511 * case we take the easy way out and panic immediately.
514 void __ext4_abort(struct super_block *sb, const char *function,
515 unsigned int line, const char *fmt, ...)
519 save_error_info(sb, function, line);
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
527 if ((sb->s_flags & MS_RDONLY) == 0) {
528 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
529 sb->s_flags |= MS_RDONLY;
530 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
531 if (EXT4_SB(sb)->s_journal)
532 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
533 save_error_info(sb, function, line);
535 if (test_opt(sb, ERRORS_PANIC))
536 panic("EXT4-fs panic from previous error\n");
539 void ext4_msg (struct super_block * sb, const char *prefix,
540 const char *fmt, ...)
545 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
551 void __ext4_warning(struct super_block *sb, const char *function,
552 unsigned int line, const char *fmt, ...)
557 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
558 sb->s_id, function, line);
564 void __ext4_grp_locked_error(const char *function, unsigned int line,
565 struct super_block *sb, ext4_group_t grp,
566 unsigned long ino, ext4_fsblk_t block,
567 const char *fmt, ...)
572 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
574 es->s_last_error_ino = cpu_to_le32(ino);
575 es->s_last_error_block = cpu_to_le64(block);
576 __save_error_info(sb, function, line);
578 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
579 sb->s_id, function, line, grp);
581 printk("inode %lu: ", ino);
583 printk("block %llu:", (unsigned long long) block);
588 if (test_opt(sb, ERRORS_CONT)) {
589 ext4_commit_super(sb, 0);
593 ext4_unlock_group(sb, grp);
594 ext4_handle_error(sb);
596 * We only get here in the ERRORS_RO case; relocking the group
597 * may be dangerous, but nothing bad will happen since the
598 * filesystem will have already been marked read/only and the
599 * journal has been aborted. We return 1 as a hint to callers
600 * who might what to use the return value from
601 * ext4_grp_locked_error() to distinguish beween the
602 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
603 * aggressively from the ext4 function in question, with a
604 * more appropriate error code.
606 ext4_lock_group(sb, grp);
610 void ext4_update_dynamic_rev(struct super_block *sb)
612 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
614 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
618 "updating to rev %d because of new feature flag, "
619 "running e2fsck is recommended",
622 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
623 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
624 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
625 /* leave es->s_feature_*compat flags alone */
626 /* es->s_uuid will be set by e2fsck if empty */
629 * The rest of the superblock fields should be zero, and if not it
630 * means they are likely already in use, so leave them alone. We
631 * can leave it up to e2fsck to clean up any inconsistencies there.
636 * Open the external journal device
638 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
640 struct block_device *bdev;
641 char b[BDEVNAME_SIZE];
643 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
649 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
650 __bdevname(dev, b), PTR_ERR(bdev));
655 * Release the journal device
657 static int ext4_blkdev_put(struct block_device *bdev)
660 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
663 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
665 struct block_device *bdev;
668 bdev = sbi->journal_bdev;
670 ret = ext4_blkdev_put(bdev);
671 sbi->journal_bdev = NULL;
676 static inline struct inode *orphan_list_entry(struct list_head *l)
678 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
681 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
685 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
686 le32_to_cpu(sbi->s_es->s_last_orphan));
688 printk(KERN_ERR "sb_info orphan list:\n");
689 list_for_each(l, &sbi->s_orphan) {
690 struct inode *inode = orphan_list_entry(l);
692 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
693 inode->i_sb->s_id, inode->i_ino, inode,
694 inode->i_mode, inode->i_nlink,
699 static void ext4_put_super(struct super_block *sb)
701 struct ext4_sb_info *sbi = EXT4_SB(sb);
702 struct ext4_super_block *es = sbi->s_es;
705 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
707 flush_workqueue(sbi->dio_unwritten_wq);
708 destroy_workqueue(sbi->dio_unwritten_wq);
713 ext4_commit_super(sb, 1);
715 if (sbi->s_journal) {
716 err = jbd2_journal_destroy(sbi->s_journal);
717 sbi->s_journal = NULL;
719 ext4_abort(sb, "Couldn't clean up the journal");
722 ext4_release_system_zone(sb);
724 ext4_ext_release(sb);
725 ext4_xattr_put_super(sb);
727 if (!(sb->s_flags & MS_RDONLY)) {
728 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
729 es->s_state = cpu_to_le16(sbi->s_mount_state);
730 ext4_commit_super(sb, 1);
733 remove_proc_entry(sb->s_id, ext4_proc_root);
735 kobject_del(&sbi->s_kobj);
737 for (i = 0; i < sbi->s_gdb_count; i++)
738 brelse(sbi->s_group_desc[i]);
739 kfree(sbi->s_group_desc);
740 if (is_vmalloc_addr(sbi->s_flex_groups))
741 vfree(sbi->s_flex_groups);
743 kfree(sbi->s_flex_groups);
744 percpu_counter_destroy(&sbi->s_freeblocks_counter);
745 percpu_counter_destroy(&sbi->s_freeinodes_counter);
746 percpu_counter_destroy(&sbi->s_dirs_counter);
747 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
750 for (i = 0; i < MAXQUOTAS; i++)
751 kfree(sbi->s_qf_names[i]);
754 /* Debugging code just in case the in-memory inode orphan list
755 * isn't empty. The on-disk one can be non-empty if we've
756 * detected an error and taken the fs readonly, but the
757 * in-memory list had better be clean by this point. */
758 if (!list_empty(&sbi->s_orphan))
759 dump_orphan_list(sb, sbi);
760 J_ASSERT(list_empty(&sbi->s_orphan));
762 invalidate_bdev(sb->s_bdev);
763 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
765 * Invalidate the journal device's buffers. We don't want them
766 * floating about in memory - the physical journal device may
767 * hotswapped, and it breaks the `ro-after' testing code.
769 sync_blockdev(sbi->journal_bdev);
770 invalidate_bdev(sbi->journal_bdev);
771 ext4_blkdev_remove(sbi);
773 sb->s_fs_info = NULL;
775 * Now that we are completely done shutting down the
776 * superblock, we need to actually destroy the kobject.
780 kobject_put(&sbi->s_kobj);
781 wait_for_completion(&sbi->s_kobj_unregister);
782 kfree(sbi->s_blockgroup_lock);
786 static struct kmem_cache *ext4_inode_cachep;
789 * Called inside transaction, so use GFP_NOFS
791 static struct inode *ext4_alloc_inode(struct super_block *sb)
793 struct ext4_inode_info *ei;
795 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
799 ei->vfs_inode.i_version = 1;
800 ei->vfs_inode.i_data.writeback_index = 0;
801 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
802 INIT_LIST_HEAD(&ei->i_prealloc_list);
803 spin_lock_init(&ei->i_prealloc_lock);
805 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
806 * therefore it can be null here. Don't check it, just initialize
809 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
810 ei->i_reserved_data_blocks = 0;
811 ei->i_reserved_meta_blocks = 0;
812 ei->i_allocated_meta_blocks = 0;
813 ei->i_da_metadata_calc_len = 0;
814 ei->i_delalloc_reserved_flag = 0;
815 spin_lock_init(&(ei->i_block_reservation_lock));
817 ei->i_reserved_quota = 0;
819 INIT_LIST_HEAD(&ei->i_completed_io_list);
820 spin_lock_init(&ei->i_completed_io_lock);
821 ei->cur_aio_dio = NULL;
823 ei->i_datasync_tid = 0;
825 return &ei->vfs_inode;
828 static void ext4_destroy_inode(struct inode *inode)
830 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
831 ext4_msg(inode->i_sb, KERN_ERR,
832 "Inode %lu (%p): orphan list check failed!",
833 inode->i_ino, EXT4_I(inode));
834 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
835 EXT4_I(inode), sizeof(struct ext4_inode_info),
839 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
842 static void init_once(void *foo)
844 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
846 INIT_LIST_HEAD(&ei->i_orphan);
847 #ifdef CONFIG_EXT4_FS_XATTR
848 init_rwsem(&ei->xattr_sem);
850 init_rwsem(&ei->i_data_sem);
851 inode_init_once(&ei->vfs_inode);
854 static int init_inodecache(void)
856 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
857 sizeof(struct ext4_inode_info),
858 0, (SLAB_RECLAIM_ACCOUNT|
861 if (ext4_inode_cachep == NULL)
866 static void destroy_inodecache(void)
868 kmem_cache_destroy(ext4_inode_cachep);
871 static void ext4_clear_inode(struct inode *inode)
874 ext4_discard_preallocations(inode);
875 if (EXT4_JOURNAL(inode))
876 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
877 &EXT4_I(inode)->jinode);
880 static inline void ext4_show_quota_options(struct seq_file *seq,
881 struct super_block *sb)
883 #if defined(CONFIG_QUOTA)
884 struct ext4_sb_info *sbi = EXT4_SB(sb);
886 if (sbi->s_jquota_fmt) {
889 switch (sbi->s_jquota_fmt) {
900 seq_printf(seq, ",jqfmt=%s", fmtname);
903 if (sbi->s_qf_names[USRQUOTA])
904 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
906 if (sbi->s_qf_names[GRPQUOTA])
907 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
909 if (test_opt(sb, USRQUOTA))
910 seq_puts(seq, ",usrquota");
912 if (test_opt(sb, GRPQUOTA))
913 seq_puts(seq, ",grpquota");
919 * - it's set to a non-default value OR
920 * - if the per-sb default is different from the global default
922 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
925 unsigned long def_mount_opts;
926 struct super_block *sb = vfs->mnt_sb;
927 struct ext4_sb_info *sbi = EXT4_SB(sb);
928 struct ext4_super_block *es = sbi->s_es;
930 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
931 def_errors = le16_to_cpu(es->s_errors);
933 if (sbi->s_sb_block != 1)
934 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
935 if (test_opt(sb, MINIX_DF))
936 seq_puts(seq, ",minixdf");
937 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
938 seq_puts(seq, ",grpid");
939 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
940 seq_puts(seq, ",nogrpid");
941 if (sbi->s_resuid != EXT4_DEF_RESUID ||
942 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
943 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
945 if (sbi->s_resgid != EXT4_DEF_RESGID ||
946 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
947 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
949 if (test_opt(sb, ERRORS_RO)) {
950 if (def_errors == EXT4_ERRORS_PANIC ||
951 def_errors == EXT4_ERRORS_CONTINUE) {
952 seq_puts(seq, ",errors=remount-ro");
955 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
956 seq_puts(seq, ",errors=continue");
957 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
958 seq_puts(seq, ",errors=panic");
959 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
960 seq_puts(seq, ",nouid32");
961 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
962 seq_puts(seq, ",debug");
963 if (test_opt(sb, OLDALLOC))
964 seq_puts(seq, ",oldalloc");
965 #ifdef CONFIG_EXT4_FS_XATTR
966 if (test_opt(sb, XATTR_USER) &&
967 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
968 seq_puts(seq, ",user_xattr");
969 if (!test_opt(sb, XATTR_USER) &&
970 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
971 seq_puts(seq, ",nouser_xattr");
974 #ifdef CONFIG_EXT4_FS_POSIX_ACL
975 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
976 seq_puts(seq, ",acl");
977 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
978 seq_puts(seq, ",noacl");
980 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
981 seq_printf(seq, ",commit=%u",
982 (unsigned) (sbi->s_commit_interval / HZ));
984 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
985 seq_printf(seq, ",min_batch_time=%u",
986 (unsigned) sbi->s_min_batch_time);
988 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
989 seq_printf(seq, ",max_batch_time=%u",
990 (unsigned) sbi->s_min_batch_time);
994 * We're changing the default of barrier mount option, so
995 * let's always display its mount state so it's clear what its
998 seq_puts(seq, ",barrier=");
999 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1000 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1001 seq_puts(seq, ",journal_async_commit");
1002 else if (test_opt(sb, JOURNAL_CHECKSUM))
1003 seq_puts(seq, ",journal_checksum");
1004 if (test_opt(sb, I_VERSION))
1005 seq_puts(seq, ",i_version");
1006 if (!test_opt(sb, DELALLOC))
1007 seq_puts(seq, ",nodelalloc");
1011 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1013 * journal mode get enabled in different ways
1014 * So just print the value even if we didn't specify it
1016 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1017 seq_puts(seq, ",data=journal");
1018 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1019 seq_puts(seq, ",data=ordered");
1020 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1021 seq_puts(seq, ",data=writeback");
1023 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1024 seq_printf(seq, ",inode_readahead_blks=%u",
1025 sbi->s_inode_readahead_blks);
1027 if (test_opt(sb, DATA_ERR_ABORT))
1028 seq_puts(seq, ",data_err=abort");
1030 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1031 seq_puts(seq, ",noauto_da_alloc");
1033 if (test_opt(sb, DISCARD))
1034 seq_puts(seq, ",discard");
1036 if (test_opt(sb, NOLOAD))
1037 seq_puts(seq, ",norecovery");
1039 if (test_opt(sb, DIOREAD_NOLOCK))
1040 seq_puts(seq, ",dioread_nolock");
1042 ext4_show_quota_options(seq, sb);
1047 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1048 u64 ino, u32 generation)
1050 struct inode *inode;
1052 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1053 return ERR_PTR(-ESTALE);
1054 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1055 return ERR_PTR(-ESTALE);
1057 /* iget isn't really right if the inode is currently unallocated!!
1059 * ext4_read_inode will return a bad_inode if the inode had been
1060 * deleted, so we should be safe.
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1065 inode = ext4_iget(sb, ino);
1067 return ERR_CAST(inode);
1068 if (generation && inode->i_generation != generation) {
1070 return ERR_PTR(-ESTALE);
1076 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type)
1079 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1083 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1084 int fh_len, int fh_type)
1086 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1096 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1099 journal_t *journal = EXT4_SB(sb)->s_journal;
1101 WARN_ON(PageChecked(page));
1102 if (!page_has_buffers(page))
1105 return jbd2_journal_try_to_free_buffers(journal, page,
1106 wait & ~__GFP_WAIT);
1107 return try_to_free_buffers(page);
1111 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1112 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1114 static int ext4_write_dquot(struct dquot *dquot);
1115 static int ext4_acquire_dquot(struct dquot *dquot);
1116 static int ext4_release_dquot(struct dquot *dquot);
1117 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1118 static int ext4_write_info(struct super_block *sb, int type);
1119 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1121 static int ext4_quota_on_mount(struct super_block *sb, int type);
1122 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1123 size_t len, loff_t off);
1124 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1125 const char *data, size_t len, loff_t off);
1127 static const struct dquot_operations ext4_quota_operations = {
1129 .get_reserved_space = ext4_get_reserved_space,
1131 .write_dquot = ext4_write_dquot,
1132 .acquire_dquot = ext4_acquire_dquot,
1133 .release_dquot = ext4_release_dquot,
1134 .mark_dirty = ext4_mark_dquot_dirty,
1135 .write_info = ext4_write_info,
1136 .alloc_dquot = dquot_alloc,
1137 .destroy_dquot = dquot_destroy,
1140 static const struct quotactl_ops ext4_qctl_operations = {
1141 .quota_on = ext4_quota_on,
1142 .quota_off = dquot_quota_off,
1143 .quota_sync = dquot_quota_sync,
1144 .get_info = dquot_get_dqinfo,
1145 .set_info = dquot_set_dqinfo,
1146 .get_dqblk = dquot_get_dqblk,
1147 .set_dqblk = dquot_set_dqblk
1151 static const struct super_operations ext4_sops = {
1152 .alloc_inode = ext4_alloc_inode,
1153 .destroy_inode = ext4_destroy_inode,
1154 .write_inode = ext4_write_inode,
1155 .dirty_inode = ext4_dirty_inode,
1156 .delete_inode = ext4_delete_inode,
1157 .put_super = ext4_put_super,
1158 .sync_fs = ext4_sync_fs,
1159 .freeze_fs = ext4_freeze,
1160 .unfreeze_fs = ext4_unfreeze,
1161 .statfs = ext4_statfs,
1162 .remount_fs = ext4_remount,
1163 .clear_inode = ext4_clear_inode,
1164 .show_options = ext4_show_options,
1166 .quota_read = ext4_quota_read,
1167 .quota_write = ext4_quota_write,
1169 .bdev_try_to_free_page = bdev_try_to_free_page,
1172 static const struct super_operations ext4_nojournal_sops = {
1173 .alloc_inode = ext4_alloc_inode,
1174 .destroy_inode = ext4_destroy_inode,
1175 .write_inode = ext4_write_inode,
1176 .dirty_inode = ext4_dirty_inode,
1177 .delete_inode = ext4_delete_inode,
1178 .write_super = ext4_write_super,
1179 .put_super = ext4_put_super,
1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
1182 .clear_inode = ext4_clear_inode,
1183 .show_options = ext4_show_options,
1185 .quota_read = ext4_quota_read,
1186 .quota_write = ext4_quota_write,
1188 .bdev_try_to_free_page = bdev_try_to_free_page,
1191 static const struct export_operations ext4_export_ops = {
1192 .fh_to_dentry = ext4_fh_to_dentry,
1193 .fh_to_parent = ext4_fh_to_parent,
1194 .get_parent = ext4_get_parent,
1198 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1199 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1200 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1201 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1202 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1203 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1204 Opt_journal_update, Opt_journal_dev,
1205 Opt_journal_checksum, Opt_journal_async_commit,
1206 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1207 Opt_data_err_abort, Opt_data_err_ignore,
1208 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1209 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1210 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1211 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1212 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1213 Opt_block_validity, Opt_noblock_validity,
1214 Opt_inode_readahead_blks, Opt_journal_ioprio,
1215 Opt_dioread_nolock, Opt_dioread_lock,
1216 Opt_discard, Opt_nodiscard,
1219 static const match_table_t tokens = {
1220 {Opt_bsd_df, "bsddf"},
1221 {Opt_minix_df, "minixdf"},
1222 {Opt_grpid, "grpid"},
1223 {Opt_grpid, "bsdgroups"},
1224 {Opt_nogrpid, "nogrpid"},
1225 {Opt_nogrpid, "sysvgroups"},
1226 {Opt_resgid, "resgid=%u"},
1227 {Opt_resuid, "resuid=%u"},
1229 {Opt_err_cont, "errors=continue"},
1230 {Opt_err_panic, "errors=panic"},
1231 {Opt_err_ro, "errors=remount-ro"},
1232 {Opt_nouid32, "nouid32"},
1233 {Opt_debug, "debug"},
1234 {Opt_oldalloc, "oldalloc"},
1235 {Opt_orlov, "orlov"},
1236 {Opt_user_xattr, "user_xattr"},
1237 {Opt_nouser_xattr, "nouser_xattr"},
1239 {Opt_noacl, "noacl"},
1240 {Opt_noload, "noload"},
1241 {Opt_noload, "norecovery"},
1244 {Opt_commit, "commit=%u"},
1245 {Opt_min_batch_time, "min_batch_time=%u"},
1246 {Opt_max_batch_time, "max_batch_time=%u"},
1247 {Opt_journal_update, "journal=update"},
1248 {Opt_journal_dev, "journal_dev=%u"},
1249 {Opt_journal_checksum, "journal_checksum"},
1250 {Opt_journal_async_commit, "journal_async_commit"},
1251 {Opt_abort, "abort"},
1252 {Opt_data_journal, "data=journal"},
1253 {Opt_data_ordered, "data=ordered"},
1254 {Opt_data_writeback, "data=writeback"},
1255 {Opt_data_err_abort, "data_err=abort"},
1256 {Opt_data_err_ignore, "data_err=ignore"},
1257 {Opt_offusrjquota, "usrjquota="},
1258 {Opt_usrjquota, "usrjquota=%s"},
1259 {Opt_offgrpjquota, "grpjquota="},
1260 {Opt_grpjquota, "grpjquota=%s"},
1261 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1262 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1263 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1264 {Opt_grpquota, "grpquota"},
1265 {Opt_noquota, "noquota"},
1266 {Opt_quota, "quota"},
1267 {Opt_usrquota, "usrquota"},
1268 {Opt_barrier, "barrier=%u"},
1269 {Opt_barrier, "barrier"},
1270 {Opt_nobarrier, "nobarrier"},
1271 {Opt_i_version, "i_version"},
1272 {Opt_stripe, "stripe=%u"},
1273 {Opt_resize, "resize"},
1274 {Opt_delalloc, "delalloc"},
1275 {Opt_nodelalloc, "nodelalloc"},
1276 {Opt_block_validity, "block_validity"},
1277 {Opt_noblock_validity, "noblock_validity"},
1278 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1279 {Opt_journal_ioprio, "journal_ioprio=%u"},
1280 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1281 {Opt_auto_da_alloc, "auto_da_alloc"},
1282 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1283 {Opt_dioread_nolock, "dioread_nolock"},
1284 {Opt_dioread_lock, "dioread_lock"},
1285 {Opt_discard, "discard"},
1286 {Opt_nodiscard, "nodiscard"},
1290 static ext4_fsblk_t get_sb_block(void **data)
1292 ext4_fsblk_t sb_block;
1293 char *options = (char *) *data;
1295 if (!options || strncmp(options, "sb=", 3) != 0)
1296 return 1; /* Default location */
1299 /* TODO: use simple_strtoll with >32bit ext4 */
1300 sb_block = simple_strtoul(options, &options, 0);
1301 if (*options && *options != ',') {
1302 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1306 if (*options == ',')
1308 *data = (void *) options;
1313 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1314 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1315 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1318 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1320 struct ext4_sb_info *sbi = EXT4_SB(sb);
1323 if (sb_any_quota_loaded(sb) &&
1324 !sbi->s_qf_names[qtype]) {
1325 ext4_msg(sb, KERN_ERR,
1326 "Cannot change journaled "
1327 "quota options when quota turned on");
1330 qname = match_strdup(args);
1332 ext4_msg(sb, KERN_ERR,
1333 "Not enough memory for storing quotafile name");
1336 if (sbi->s_qf_names[qtype] &&
1337 strcmp(sbi->s_qf_names[qtype], qname)) {
1338 ext4_msg(sb, KERN_ERR,
1339 "%s quota file already specified", QTYPE2NAME(qtype));
1343 sbi->s_qf_names[qtype] = qname;
1344 if (strchr(sbi->s_qf_names[qtype], '/')) {
1345 ext4_msg(sb, KERN_ERR,
1346 "quotafile must be on filesystem root");
1347 kfree(sbi->s_qf_names[qtype]);
1348 sbi->s_qf_names[qtype] = NULL;
1351 set_opt(sbi->s_mount_opt, QUOTA);
1355 static int clear_qf_name(struct super_block *sb, int qtype)
1358 struct ext4_sb_info *sbi = EXT4_SB(sb);
1360 if (sb_any_quota_loaded(sb) &&
1361 sbi->s_qf_names[qtype]) {
1362 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1363 " when quota turned on");
1367 * The space will be released later when all options are confirmed
1370 sbi->s_qf_names[qtype] = NULL;
1375 static int parse_options(char *options, struct super_block *sb,
1376 unsigned long *journal_devnum,
1377 unsigned int *journal_ioprio,
1378 ext4_fsblk_t *n_blocks_count, int is_remount)
1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
1382 substring_t args[MAX_OPT_ARGS];
1392 while ((p = strsep(&options, ",")) != NULL) {
1398 * Initialize args struct so we know whether arg was
1399 * found; some options take optional arguments.
1401 args[0].to = args[0].from = 0;
1402 token = match_token(p, tokens, args);
1405 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1406 clear_opt(sbi->s_mount_opt, MINIX_DF);
1409 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1410 set_opt(sbi->s_mount_opt, MINIX_DF);
1414 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1415 set_opt(sbi->s_mount_opt, GRPID);
1419 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1420 clear_opt(sbi->s_mount_opt, GRPID);
1424 if (match_int(&args[0], &option))
1426 sbi->s_resuid = option;
1429 if (match_int(&args[0], &option))
1431 sbi->s_resgid = option;
1434 /* handled by get_sb_block() instead of here */
1435 /* *sb_block = match_int(&args[0]); */
1438 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1439 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1440 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1443 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1444 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1445 set_opt(sbi->s_mount_opt, ERRORS_RO);
1448 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1449 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1450 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1453 set_opt(sbi->s_mount_opt, NO_UID32);
1456 set_opt(sbi->s_mount_opt, DEBUG);
1459 set_opt(sbi->s_mount_opt, OLDALLOC);
1462 clear_opt(sbi->s_mount_opt, OLDALLOC);
1464 #ifdef CONFIG_EXT4_FS_XATTR
1465 case Opt_user_xattr:
1466 set_opt(sbi->s_mount_opt, XATTR_USER);
1468 case Opt_nouser_xattr:
1469 clear_opt(sbi->s_mount_opt, XATTR_USER);
1472 case Opt_user_xattr:
1473 case Opt_nouser_xattr:
1474 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1477 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1479 set_opt(sbi->s_mount_opt, POSIX_ACL);
1482 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1487 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1490 case Opt_journal_update:
1492 /* Eventually we will want to be able to create
1493 a journal file here. For now, only allow the
1494 user to specify an existing inode to be the
1497 ext4_msg(sb, KERN_ERR,
1498 "Cannot specify journal on remount");
1501 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1503 case Opt_journal_dev:
1505 ext4_msg(sb, KERN_ERR,
1506 "Cannot specify journal on remount");
1509 if (match_int(&args[0], &option))
1511 *journal_devnum = option;
1513 case Opt_journal_checksum:
1514 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1516 case Opt_journal_async_commit:
1517 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1518 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1521 set_opt(sbi->s_mount_opt, NOLOAD);
1524 if (match_int(&args[0], &option))
1529 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1530 sbi->s_commit_interval = HZ * option;
1532 case Opt_max_batch_time:
1533 if (match_int(&args[0], &option))
1538 option = EXT4_DEF_MAX_BATCH_TIME;
1539 sbi->s_max_batch_time = option;
1541 case Opt_min_batch_time:
1542 if (match_int(&args[0], &option))
1546 sbi->s_min_batch_time = option;
1548 case Opt_data_journal:
1549 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1551 case Opt_data_ordered:
1552 data_opt = EXT4_MOUNT_ORDERED_DATA;
1554 case Opt_data_writeback:
1555 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1558 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1559 ext4_msg(sb, KERN_ERR,
1560 "Cannot change data mode on remount");
1564 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1565 sbi->s_mount_opt |= data_opt;
1568 case Opt_data_err_abort:
1569 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1571 case Opt_data_err_ignore:
1572 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1576 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1580 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1583 case Opt_offusrjquota:
1584 if (!clear_qf_name(sb, USRQUOTA))
1587 case Opt_offgrpjquota:
1588 if (!clear_qf_name(sb, GRPQUOTA))
1592 case Opt_jqfmt_vfsold:
1593 qfmt = QFMT_VFS_OLD;
1595 case Opt_jqfmt_vfsv0:
1598 case Opt_jqfmt_vfsv1:
1601 if (sb_any_quota_loaded(sb) &&
1602 sbi->s_jquota_fmt != qfmt) {
1603 ext4_msg(sb, KERN_ERR, "Cannot change "
1604 "journaled quota options when "
1608 sbi->s_jquota_fmt = qfmt;
1612 set_opt(sbi->s_mount_opt, QUOTA);
1613 set_opt(sbi->s_mount_opt, USRQUOTA);
1616 set_opt(sbi->s_mount_opt, QUOTA);
1617 set_opt(sbi->s_mount_opt, GRPQUOTA);
1620 if (sb_any_quota_loaded(sb)) {
1621 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1622 "options when quota turned on");
1625 clear_opt(sbi->s_mount_opt, QUOTA);
1626 clear_opt(sbi->s_mount_opt, USRQUOTA);
1627 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1633 ext4_msg(sb, KERN_ERR,
1634 "quota options not supported");
1638 case Opt_offusrjquota:
1639 case Opt_offgrpjquota:
1640 case Opt_jqfmt_vfsold:
1641 case Opt_jqfmt_vfsv0:
1642 case Opt_jqfmt_vfsv1:
1643 ext4_msg(sb, KERN_ERR,
1644 "journaled quota options not supported");
1650 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1653 clear_opt(sbi->s_mount_opt, BARRIER);
1657 if (match_int(&args[0], &option))
1660 option = 1; /* No argument, default to 1 */
1662 set_opt(sbi->s_mount_opt, BARRIER);
1664 clear_opt(sbi->s_mount_opt, BARRIER);
1670 ext4_msg(sb, KERN_ERR,
1671 "resize option only available "
1675 if (match_int(&args[0], &option) != 0)
1677 *n_blocks_count = option;
1680 ext4_msg(sb, KERN_WARNING,
1681 "Ignoring deprecated nobh option");
1684 ext4_msg(sb, KERN_WARNING,
1685 "Ignoring deprecated bh option");
1688 set_opt(sbi->s_mount_opt, I_VERSION);
1689 sb->s_flags |= MS_I_VERSION;
1691 case Opt_nodelalloc:
1692 clear_opt(sbi->s_mount_opt, DELALLOC);
1695 if (match_int(&args[0], &option))
1699 sbi->s_stripe = option;
1702 set_opt(sbi->s_mount_opt, DELALLOC);
1704 case Opt_block_validity:
1705 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1707 case Opt_noblock_validity:
1708 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1710 case Opt_inode_readahead_blks:
1711 if (match_int(&args[0], &option))
1713 if (option < 0 || option > (1 << 30))
1715 if (!is_power_of_2(option)) {
1716 ext4_msg(sb, KERN_ERR,
1717 "EXT4-fs: inode_readahead_blks"
1718 " must be a power of 2");
1721 sbi->s_inode_readahead_blks = option;
1723 case Opt_journal_ioprio:
1724 if (match_int(&args[0], &option))
1726 if (option < 0 || option > 7)
1728 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1731 case Opt_noauto_da_alloc:
1732 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1734 case Opt_auto_da_alloc:
1736 if (match_int(&args[0], &option))
1739 option = 1; /* No argument, default to 1 */
1741 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1743 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1746 set_opt(sbi->s_mount_opt, DISCARD);
1749 clear_opt(sbi->s_mount_opt, DISCARD);
1751 case Opt_dioread_nolock:
1752 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1754 case Opt_dioread_lock:
1755 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1758 ext4_msg(sb, KERN_ERR,
1759 "Unrecognized mount option \"%s\" "
1760 "or missing value", p);
1765 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1766 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1767 clear_opt(sbi->s_mount_opt, USRQUOTA);
1769 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1770 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1772 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1773 ext4_msg(sb, KERN_ERR, "old and new quota "
1778 if (!sbi->s_jquota_fmt) {
1779 ext4_msg(sb, KERN_ERR, "journaled quota format "
1784 if (sbi->s_jquota_fmt) {
1785 ext4_msg(sb, KERN_ERR, "journaled quota format "
1786 "specified with no journaling "
1795 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1798 struct ext4_sb_info *sbi = EXT4_SB(sb);
1801 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1802 ext4_msg(sb, KERN_ERR, "revision level too high, "
1803 "forcing read-only mode");
1808 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1809 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1810 "running e2fsck is recommended");
1811 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1812 ext4_msg(sb, KERN_WARNING,
1813 "warning: mounting fs with errors, "
1814 "running e2fsck is recommended");
1815 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1816 le16_to_cpu(es->s_mnt_count) >=
1817 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1818 ext4_msg(sb, KERN_WARNING,
1819 "warning: maximal mount count reached, "
1820 "running e2fsck is recommended");
1821 else if (le32_to_cpu(es->s_checkinterval) &&
1822 (le32_to_cpu(es->s_lastcheck) +
1823 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1824 ext4_msg(sb, KERN_WARNING,
1825 "warning: checktime reached, "
1826 "running e2fsck is recommended");
1827 if (!sbi->s_journal)
1828 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1829 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1830 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1831 le16_add_cpu(&es->s_mnt_count, 1);
1832 es->s_mtime = cpu_to_le32(get_seconds());
1833 ext4_update_dynamic_rev(sb);
1835 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1837 ext4_commit_super(sb, 1);
1838 if (test_opt(sb, DEBUG))
1839 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1840 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1842 sbi->s_groups_count,
1843 EXT4_BLOCKS_PER_GROUP(sb),
1844 EXT4_INODES_PER_GROUP(sb),
1850 static int ext4_fill_flex_info(struct super_block *sb)
1852 struct ext4_sb_info *sbi = EXT4_SB(sb);
1853 struct ext4_group_desc *gdp = NULL;
1854 ext4_group_t flex_group_count;
1855 ext4_group_t flex_group;
1856 int groups_per_flex = 0;
1860 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1861 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1863 if (groups_per_flex < 2) {
1864 sbi->s_log_groups_per_flex = 0;
1868 /* We allocate both existing and potentially added groups */
1869 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1870 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1871 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1872 size = flex_group_count * sizeof(struct flex_groups);
1873 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1874 if (sbi->s_flex_groups == NULL) {
1875 sbi->s_flex_groups = vmalloc(size);
1876 if (sbi->s_flex_groups)
1877 memset(sbi->s_flex_groups, 0, size);
1879 if (sbi->s_flex_groups == NULL) {
1880 ext4_msg(sb, KERN_ERR, "not enough memory for "
1881 "%u flex groups", flex_group_count);
1885 for (i = 0; i < sbi->s_groups_count; i++) {
1886 gdp = ext4_get_group_desc(sb, i, NULL);
1888 flex_group = ext4_flex_group(sbi, i);
1889 atomic_add(ext4_free_inodes_count(sb, gdp),
1890 &sbi->s_flex_groups[flex_group].free_inodes);
1891 atomic_add(ext4_free_blks_count(sb, gdp),
1892 &sbi->s_flex_groups[flex_group].free_blocks);
1893 atomic_add(ext4_used_dirs_count(sb, gdp),
1894 &sbi->s_flex_groups[flex_group].used_dirs);
1902 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1903 struct ext4_group_desc *gdp)
1907 if (sbi->s_es->s_feature_ro_compat &
1908 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1909 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1910 __le32 le_group = cpu_to_le32(block_group);
1912 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1913 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1914 crc = crc16(crc, (__u8 *)gdp, offset);
1915 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1916 /* for checksum of struct ext4_group_desc do the rest...*/
1917 if ((sbi->s_es->s_feature_incompat &
1918 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1919 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1920 crc = crc16(crc, (__u8 *)gdp + offset,
1921 le16_to_cpu(sbi->s_es->s_desc_size) -
1925 return cpu_to_le16(crc);
1928 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1929 struct ext4_group_desc *gdp)
1931 if ((sbi->s_es->s_feature_ro_compat &
1932 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1933 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1939 /* Called at mount-time, super-block is locked */
1940 static int ext4_check_descriptors(struct super_block *sb)
1942 struct ext4_sb_info *sbi = EXT4_SB(sb);
1943 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1944 ext4_fsblk_t last_block;
1945 ext4_fsblk_t block_bitmap;
1946 ext4_fsblk_t inode_bitmap;
1947 ext4_fsblk_t inode_table;
1948 int flexbg_flag = 0;
1951 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1954 ext4_debug("Checking group descriptors");
1956 for (i = 0; i < sbi->s_groups_count; i++) {
1957 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1959 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1960 last_block = ext4_blocks_count(sbi->s_es) - 1;
1962 last_block = first_block +
1963 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1965 block_bitmap = ext4_block_bitmap(sb, gdp);
1966 if (block_bitmap < first_block || block_bitmap > last_block) {
1967 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1968 "Block bitmap for group %u not in group "
1969 "(block %llu)!", i, block_bitmap);
1972 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1973 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1974 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1975 "Inode bitmap for group %u not in group "
1976 "(block %llu)!", i, inode_bitmap);
1979 inode_table = ext4_inode_table(sb, gdp);
1980 if (inode_table < first_block ||
1981 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1982 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1983 "Inode table for group %u not in group "
1984 "(block %llu)!", i, inode_table);
1987 ext4_lock_group(sb, i);
1988 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1989 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1990 "Checksum for group %u failed (%u!=%u)",
1991 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1992 gdp)), le16_to_cpu(gdp->bg_checksum));
1993 if (!(sb->s_flags & MS_RDONLY)) {
1994 ext4_unlock_group(sb, i);
1998 ext4_unlock_group(sb, i);
2000 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2003 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2004 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2008 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2009 * the superblock) which were deleted from all directories, but held open by
2010 * a process at the time of a crash. We walk the list and try to delete these
2011 * inodes at recovery time (only with a read-write filesystem).
2013 * In order to keep the orphan inode chain consistent during traversal (in
2014 * case of crash during recovery), we link each inode into the superblock
2015 * orphan list_head and handle it the same way as an inode deletion during
2016 * normal operation (which journals the operations for us).
2018 * We only do an iget() and an iput() on each inode, which is very safe if we
2019 * accidentally point at an in-use or already deleted inode. The worst that
2020 * can happen in this case is that we get a "bit already cleared" message from
2021 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2022 * e2fsck was run on this filesystem, and it must have already done the orphan
2023 * inode cleanup for us, so we can safely abort without any further action.
2025 static void ext4_orphan_cleanup(struct super_block *sb,
2026 struct ext4_super_block *es)
2028 unsigned int s_flags = sb->s_flags;
2029 int nr_orphans = 0, nr_truncates = 0;
2033 if (!es->s_last_orphan) {
2034 jbd_debug(4, "no orphan inodes to clean up\n");
2038 if (bdev_read_only(sb->s_bdev)) {
2039 ext4_msg(sb, KERN_ERR, "write access "
2040 "unavailable, skipping orphan cleanup");
2044 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2045 if (es->s_last_orphan)
2046 jbd_debug(1, "Errors on filesystem, "
2047 "clearing orphan list.\n");
2048 es->s_last_orphan = 0;
2049 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2053 if (s_flags & MS_RDONLY) {
2054 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2055 sb->s_flags &= ~MS_RDONLY;
2058 /* Needed for iput() to work correctly and not trash data */
2059 sb->s_flags |= MS_ACTIVE;
2060 /* Turn on quotas so that they are updated correctly */
2061 for (i = 0; i < MAXQUOTAS; i++) {
2062 if (EXT4_SB(sb)->s_qf_names[i]) {
2063 int ret = ext4_quota_on_mount(sb, i);
2065 ext4_msg(sb, KERN_ERR,
2066 "Cannot turn on journaled "
2067 "quota: error %d", ret);
2072 while (es->s_last_orphan) {
2073 struct inode *inode;
2075 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2076 if (IS_ERR(inode)) {
2077 es->s_last_orphan = 0;
2081 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2082 dquot_initialize(inode);
2083 if (inode->i_nlink) {
2084 ext4_msg(sb, KERN_DEBUG,
2085 "%s: truncating inode %lu to %lld bytes",
2086 __func__, inode->i_ino, inode->i_size);
2087 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2088 inode->i_ino, inode->i_size);
2089 ext4_truncate(inode);
2092 ext4_msg(sb, KERN_DEBUG,
2093 "%s: deleting unreferenced inode %lu",
2094 __func__, inode->i_ino);
2095 jbd_debug(2, "deleting unreferenced inode %lu\n",
2099 iput(inode); /* The delete magic happens here! */
2102 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2105 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2106 PLURAL(nr_orphans));
2108 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2109 PLURAL(nr_truncates));
2111 /* Turn quotas off */
2112 for (i = 0; i < MAXQUOTAS; i++) {
2113 if (sb_dqopt(sb)->files[i])
2114 dquot_quota_off(sb, i);
2117 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2121 * Maximal extent format file size.
2122 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2123 * extent format containers, within a sector_t, and within i_blocks
2124 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2125 * so that won't be a limiting factor.
2127 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2129 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2132 loff_t upper_limit = MAX_LFS_FILESIZE;
2134 /* small i_blocks in vfs inode? */
2135 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2137 * CONFIG_LBDAF is not enabled implies the inode
2138 * i_block represent total blocks in 512 bytes
2139 * 32 == size of vfs inode i_blocks * 8
2141 upper_limit = (1LL << 32) - 1;
2143 /* total blocks in file system block size */
2144 upper_limit >>= (blkbits - 9);
2145 upper_limit <<= blkbits;
2148 /* 32-bit extent-start container, ee_block */
2153 /* Sanity check against vm- & vfs- imposed limits */
2154 if (res > upper_limit)
2161 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2162 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2163 * We need to be 1 filesystem block less than the 2^48 sector limit.
2165 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2167 loff_t res = EXT4_NDIR_BLOCKS;
2170 /* This is calculated to be the largest file size for a dense, block
2171 * mapped file such that the file's total number of 512-byte sectors,
2172 * including data and all indirect blocks, does not exceed (2^48 - 1).
2174 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2175 * number of 512-byte sectors of the file.
2178 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2180 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2181 * the inode i_block field represents total file blocks in
2182 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2184 upper_limit = (1LL << 32) - 1;
2186 /* total blocks in file system block size */
2187 upper_limit >>= (bits - 9);
2191 * We use 48 bit ext4_inode i_blocks
2192 * With EXT4_HUGE_FILE_FL set the i_blocks
2193 * represent total number of blocks in
2194 * file system block size
2196 upper_limit = (1LL << 48) - 1;
2200 /* indirect blocks */
2202 /* double indirect blocks */
2203 meta_blocks += 1 + (1LL << (bits-2));
2204 /* tripple indirect blocks */
2205 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2207 upper_limit -= meta_blocks;
2208 upper_limit <<= bits;
2210 res += 1LL << (bits-2);
2211 res += 1LL << (2*(bits-2));
2212 res += 1LL << (3*(bits-2));
2214 if (res > upper_limit)
2217 if (res > MAX_LFS_FILESIZE)
2218 res = MAX_LFS_FILESIZE;
2223 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2224 ext4_fsblk_t logical_sb_block, int nr)
2226 struct ext4_sb_info *sbi = EXT4_SB(sb);
2227 ext4_group_t bg, first_meta_bg;
2230 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2232 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2234 return logical_sb_block + nr + 1;
2235 bg = sbi->s_desc_per_block * nr;
2236 if (ext4_bg_has_super(sb, bg))
2239 return (has_super + ext4_group_first_block_no(sb, bg));
2243 * ext4_get_stripe_size: Get the stripe size.
2244 * @sbi: In memory super block info
2246 * If we have specified it via mount option, then
2247 * use the mount option value. If the value specified at mount time is
2248 * greater than the blocks per group use the super block value.
2249 * If the super block value is greater than blocks per group return 0.
2250 * Allocator needs it be less than blocks per group.
2253 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2255 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2256 unsigned long stripe_width =
2257 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2259 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2260 return sbi->s_stripe;
2262 if (stripe_width <= sbi->s_blocks_per_group)
2263 return stripe_width;
2265 if (stride <= sbi->s_blocks_per_group)
2274 struct attribute attr;
2275 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2276 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2277 const char *, size_t);
2281 static int parse_strtoul(const char *buf,
2282 unsigned long max, unsigned long *value)
2286 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2287 endp = skip_spaces(endp);
2288 if (*endp || *value > max)
2294 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2295 struct ext4_sb_info *sbi,
2298 return snprintf(buf, PAGE_SIZE, "%llu\n",
2299 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2302 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2303 struct ext4_sb_info *sbi, char *buf)
2305 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2307 if (!sb->s_bdev->bd_part)
2308 return snprintf(buf, PAGE_SIZE, "0\n");
2309 return snprintf(buf, PAGE_SIZE, "%lu\n",
2310 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2311 sbi->s_sectors_written_start) >> 1);
2314 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2315 struct ext4_sb_info *sbi, char *buf)
2317 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2319 if (!sb->s_bdev->bd_part)
2320 return snprintf(buf, PAGE_SIZE, "0\n");
2321 return snprintf(buf, PAGE_SIZE, "%llu\n",
2322 (unsigned long long)(sbi->s_kbytes_written +
2323 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2324 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2327 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2328 struct ext4_sb_info *sbi,
2329 const char *buf, size_t count)
2333 if (parse_strtoul(buf, 0x40000000, &t))
2336 if (!is_power_of_2(t))
2339 sbi->s_inode_readahead_blks = t;
2343 static ssize_t sbi_ui_show(struct ext4_attr *a,
2344 struct ext4_sb_info *sbi, char *buf)
2346 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2348 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2351 static ssize_t sbi_ui_store(struct ext4_attr *a,
2352 struct ext4_sb_info *sbi,
2353 const char *buf, size_t count)
2355 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2358 if (parse_strtoul(buf, 0xffffffff, &t))
2364 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2365 static struct ext4_attr ext4_attr_##_name = { \
2366 .attr = {.name = __stringify(_name), .mode = _mode }, \
2369 .offset = offsetof(struct ext4_sb_info, _elname), \
2371 #define EXT4_ATTR(name, mode, show, store) \
2372 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2374 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2375 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2376 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2377 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2378 #define ATTR_LIST(name) &ext4_attr_##name.attr
2380 EXT4_RO_ATTR(delayed_allocation_blocks);
2381 EXT4_RO_ATTR(session_write_kbytes);
2382 EXT4_RO_ATTR(lifetime_write_kbytes);
2383 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2384 inode_readahead_blks_store, s_inode_readahead_blks);
2385 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2386 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2387 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2388 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2389 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2390 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2391 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2392 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2394 static struct attribute *ext4_attrs[] = {
2395 ATTR_LIST(delayed_allocation_blocks),
2396 ATTR_LIST(session_write_kbytes),
2397 ATTR_LIST(lifetime_write_kbytes),
2398 ATTR_LIST(inode_readahead_blks),
2399 ATTR_LIST(inode_goal),
2400 ATTR_LIST(mb_stats),
2401 ATTR_LIST(mb_max_to_scan),
2402 ATTR_LIST(mb_min_to_scan),
2403 ATTR_LIST(mb_order2_req),
2404 ATTR_LIST(mb_stream_req),
2405 ATTR_LIST(mb_group_prealloc),
2406 ATTR_LIST(max_writeback_mb_bump),
2410 static ssize_t ext4_attr_show(struct kobject *kobj,
2411 struct attribute *attr, char *buf)
2413 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2415 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2417 return a->show ? a->show(a, sbi, buf) : 0;
2420 static ssize_t ext4_attr_store(struct kobject *kobj,
2421 struct attribute *attr,
2422 const char *buf, size_t len)
2424 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2426 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2428 return a->store ? a->store(a, sbi, buf, len) : 0;
2431 static void ext4_sb_release(struct kobject *kobj)
2433 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2435 complete(&sbi->s_kobj_unregister);
2439 static const struct sysfs_ops ext4_attr_ops = {
2440 .show = ext4_attr_show,
2441 .store = ext4_attr_store,
2444 static struct kobj_type ext4_ktype = {
2445 .default_attrs = ext4_attrs,
2446 .sysfs_ops = &ext4_attr_ops,
2447 .release = ext4_sb_release,
2451 * Check whether this filesystem can be mounted based on
2452 * the features present and the RDONLY/RDWR mount requested.
2453 * Returns 1 if this filesystem can be mounted as requested,
2454 * 0 if it cannot be.
2456 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2458 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2459 ext4_msg(sb, KERN_ERR,
2460 "Couldn't mount because of "
2461 "unsupported optional features (%x)",
2462 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2463 ~EXT4_FEATURE_INCOMPAT_SUPP));
2470 /* Check that feature set is OK for a read-write mount */
2471 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2472 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2473 "unsupported optional features (%x)",
2474 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2475 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2479 * Large file size enabled file system can only be mounted
2480 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2482 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2483 if (sizeof(blkcnt_t) < sizeof(u64)) {
2484 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2485 "cannot be mounted RDWR without "
2494 * This function is called once a day if we have errors logged
2495 * on the file system
2497 static void print_daily_error_info(unsigned long arg)
2499 struct super_block *sb = (struct super_block *) arg;
2500 struct ext4_sb_info *sbi;
2501 struct ext4_super_block *es;
2506 if (es->s_error_count)
2507 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2508 le32_to_cpu(es->s_error_count));
2509 if (es->s_first_error_time) {
2510 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2511 sb->s_id, le32_to_cpu(es->s_first_error_time),
2512 (int) sizeof(es->s_first_error_func),
2513 es->s_first_error_func,
2514 le32_to_cpu(es->s_first_error_line));
2515 if (es->s_first_error_ino)
2516 printk(": inode %u",
2517 le32_to_cpu(es->s_first_error_ino));
2518 if (es->s_first_error_block)
2519 printk(": block %llu", (unsigned long long)
2520 le64_to_cpu(es->s_first_error_block));
2523 if (es->s_last_error_time) {
2524 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2525 sb->s_id, le32_to_cpu(es->s_last_error_time),
2526 (int) sizeof(es->s_last_error_func),
2527 es->s_last_error_func,
2528 le32_to_cpu(es->s_last_error_line));
2529 if (es->s_last_error_ino)
2530 printk(": inode %u",
2531 le32_to_cpu(es->s_last_error_ino));
2532 if (es->s_last_error_block)
2533 printk(": block %llu", (unsigned long long)
2534 le64_to_cpu(es->s_last_error_block));
2537 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2540 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2541 __releases(kernel_lock)
2542 __acquires(kernel_lock)
2544 char *orig_data = kstrdup(data, GFP_KERNEL);
2545 struct buffer_head *bh;
2546 struct ext4_super_block *es = NULL;
2547 struct ext4_sb_info *sbi;
2549 ext4_fsblk_t sb_block = get_sb_block(&data);
2550 ext4_fsblk_t logical_sb_block;
2551 unsigned long offset = 0;
2552 unsigned long journal_devnum = 0;
2553 unsigned long def_mount_opts;
2559 unsigned int db_count;
2561 int needs_recovery, has_huge_files;
2564 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2566 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2570 sbi->s_blockgroup_lock =
2571 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2572 if (!sbi->s_blockgroup_lock) {
2576 sb->s_fs_info = sbi;
2577 sbi->s_mount_opt = 0;
2578 sbi->s_resuid = EXT4_DEF_RESUID;
2579 sbi->s_resgid = EXT4_DEF_RESGID;
2580 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2581 sbi->s_sb_block = sb_block;
2582 if (sb->s_bdev->bd_part)
2583 sbi->s_sectors_written_start =
2584 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2588 /* Cleanup superblock name */
2589 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2593 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2595 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2600 * The ext4 superblock will not be buffer aligned for other than 1kB
2601 * block sizes. We need to calculate the offset from buffer start.
2603 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2604 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2605 offset = do_div(logical_sb_block, blocksize);
2607 logical_sb_block = sb_block;
2610 if (!(bh = sb_bread(sb, logical_sb_block))) {
2611 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2615 * Note: s_es must be initialized as soon as possible because
2616 * some ext4 macro-instructions depend on its value
2618 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2620 sb->s_magic = le16_to_cpu(es->s_magic);
2621 if (sb->s_magic != EXT4_SUPER_MAGIC)
2623 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2625 /* Set defaults before we parse the mount options */
2626 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2627 if (def_mount_opts & EXT4_DEFM_DEBUG)
2628 set_opt(sbi->s_mount_opt, DEBUG);
2629 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2630 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2632 set_opt(sbi->s_mount_opt, GRPID);
2634 if (def_mount_opts & EXT4_DEFM_UID16)
2635 set_opt(sbi->s_mount_opt, NO_UID32);
2636 #ifdef CONFIG_EXT4_FS_XATTR
2637 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2638 set_opt(sbi->s_mount_opt, XATTR_USER);
2640 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2641 if (def_mount_opts & EXT4_DEFM_ACL)
2642 set_opt(sbi->s_mount_opt, POSIX_ACL);
2644 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2645 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2646 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2647 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2648 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2649 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2651 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2652 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2653 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2654 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2656 set_opt(sbi->s_mount_opt, ERRORS_RO);
2658 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2659 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2660 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2661 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2662 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2664 set_opt(sbi->s_mount_opt, BARRIER);
2667 * enable delayed allocation by default
2668 * Use -o nodelalloc to turn it off
2670 if (!IS_EXT3_SB(sb))
2671 set_opt(sbi->s_mount_opt, DELALLOC);
2673 if (!parse_options((char *) data, sb, &journal_devnum,
2674 &journal_ioprio, NULL, 0))
2677 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2678 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2680 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2681 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2682 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2683 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2684 ext4_msg(sb, KERN_WARNING,
2685 "feature flags set on rev 0 fs, "
2686 "running e2fsck is recommended");
2689 * Check feature flags regardless of the revision level, since we
2690 * previously didn't change the revision level when setting the flags,
2691 * so there is a chance incompat flags are set on a rev 0 filesystem.
2693 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2696 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2698 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2699 blocksize > EXT4_MAX_BLOCK_SIZE) {
2700 ext4_msg(sb, KERN_ERR,
2701 "Unsupported filesystem blocksize %d", blocksize);
2705 if (sb->s_blocksize != blocksize) {
2706 /* Validate the filesystem blocksize */
2707 if (!sb_set_blocksize(sb, blocksize)) {
2708 ext4_msg(sb, KERN_ERR, "bad block size %d",
2714 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2715 offset = do_div(logical_sb_block, blocksize);
2716 bh = sb_bread(sb, logical_sb_block);
2718 ext4_msg(sb, KERN_ERR,
2719 "Can't read superblock on 2nd try");
2722 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2724 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2725 ext4_msg(sb, KERN_ERR,
2726 "Magic mismatch, very weird!");
2731 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2732 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2733 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2735 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2737 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2738 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2739 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2741 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2742 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2743 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2744 (!is_power_of_2(sbi->s_inode_size)) ||
2745 (sbi->s_inode_size > blocksize)) {
2746 ext4_msg(sb, KERN_ERR,
2747 "unsupported inode size: %d",
2751 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2752 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2755 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2756 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2757 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2758 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2759 !is_power_of_2(sbi->s_desc_size)) {
2760 ext4_msg(sb, KERN_ERR,
2761 "unsupported descriptor size %lu",
2766 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2768 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2769 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2770 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2773 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2774 if (sbi->s_inodes_per_block == 0)
2776 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2777 sbi->s_inodes_per_block;
2778 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2780 sbi->s_mount_state = le16_to_cpu(es->s_state);
2781 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2782 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2784 for (i = 0; i < 4; i++)
2785 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2786 sbi->s_def_hash_version = es->s_def_hash_version;
2787 i = le32_to_cpu(es->s_flags);
2788 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2789 sbi->s_hash_unsigned = 3;
2790 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2791 #ifdef __CHAR_UNSIGNED__
2792 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2793 sbi->s_hash_unsigned = 3;
2795 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2800 if (sbi->s_blocks_per_group > blocksize * 8) {
2801 ext4_msg(sb, KERN_ERR,
2802 "#blocks per group too big: %lu",
2803 sbi->s_blocks_per_group);
2806 if (sbi->s_inodes_per_group > blocksize * 8) {
2807 ext4_msg(sb, KERN_ERR,
2808 "#inodes per group too big: %lu",
2809 sbi->s_inodes_per_group);
2814 * Test whether we have more sectors than will fit in sector_t,
2815 * and whether the max offset is addressable by the page cache.
2817 if ((ext4_blocks_count(es) >
2818 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2819 (ext4_blocks_count(es) >
2820 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2821 ext4_msg(sb, KERN_ERR, "filesystem"
2822 " too large to mount safely on this system");
2823 if (sizeof(sector_t) < 8)
2824 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2829 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2832 /* check blocks count against device size */
2833 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2834 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2835 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2836 "exceeds size of device (%llu blocks)",
2837 ext4_blocks_count(es), blocks_count);
2842 * It makes no sense for the first data block to be beyond the end
2843 * of the filesystem.
2845 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2846 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2847 "block %u is beyond end of filesystem (%llu)",
2848 le32_to_cpu(es->s_first_data_block),
2849 ext4_blocks_count(es));
2852 blocks_count = (ext4_blocks_count(es) -
2853 le32_to_cpu(es->s_first_data_block) +
2854 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2855 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2856 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2857 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2858 "(block count %llu, first data block %u, "
2859 "blocks per group %lu)", sbi->s_groups_count,
2860 ext4_blocks_count(es),
2861 le32_to_cpu(es->s_first_data_block),
2862 EXT4_BLOCKS_PER_GROUP(sb));
2865 sbi->s_groups_count = blocks_count;
2866 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2867 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2868 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2869 EXT4_DESC_PER_BLOCK(sb);
2870 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2872 if (sbi->s_group_desc == NULL) {
2873 ext4_msg(sb, KERN_ERR, "not enough memory");
2877 #ifdef CONFIG_PROC_FS
2879 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2882 bgl_lock_init(sbi->s_blockgroup_lock);
2884 for (i = 0; i < db_count; i++) {
2885 block = descriptor_loc(sb, logical_sb_block, i);
2886 sbi->s_group_desc[i] = sb_bread(sb, block);
2887 if (!sbi->s_group_desc[i]) {
2888 ext4_msg(sb, KERN_ERR,
2889 "can't read group descriptor %d", i);
2894 if (!ext4_check_descriptors(sb)) {
2895 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2898 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2899 if (!ext4_fill_flex_info(sb)) {
2900 ext4_msg(sb, KERN_ERR,
2901 "unable to initialize "
2902 "flex_bg meta info!");
2906 sbi->s_gdb_count = db_count;
2907 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2908 spin_lock_init(&sbi->s_next_gen_lock);
2910 sbi->s_stripe = ext4_get_stripe_size(sbi);
2911 sbi->s_max_writeback_mb_bump = 128;
2914 * set up enough so that it can read an inode
2916 if (!test_opt(sb, NOLOAD) &&
2917 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2918 sb->s_op = &ext4_sops;
2920 sb->s_op = &ext4_nojournal_sops;
2921 sb->s_export_op = &ext4_export_ops;
2922 sb->s_xattr = ext4_xattr_handlers;
2924 sb->s_qcop = &ext4_qctl_operations;
2925 sb->dq_op = &ext4_quota_operations;
2927 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2928 mutex_init(&sbi->s_orphan_lock);
2929 mutex_init(&sbi->s_resize_lock);
2933 needs_recovery = (es->s_last_orphan != 0 ||
2934 EXT4_HAS_INCOMPAT_FEATURE(sb,
2935 EXT4_FEATURE_INCOMPAT_RECOVER));
2938 * The first inode we look at is the journal inode. Don't try
2939 * root first: it may be modified in the journal!
2941 if (!test_opt(sb, NOLOAD) &&
2942 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2943 if (ext4_load_journal(sb, es, journal_devnum))
2945 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2946 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2947 ext4_msg(sb, KERN_ERR, "required journal recovery "
2948 "suppressed and not mounted read-only");
2949 goto failed_mount_wq;
2951 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2952 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2953 sbi->s_journal = NULL;
2958 if (ext4_blocks_count(es) > 0xffffffffULL &&
2959 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2960 JBD2_FEATURE_INCOMPAT_64BIT)) {
2961 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2962 goto failed_mount_wq;
2965 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2966 jbd2_journal_set_features(sbi->s_journal,
2967 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2968 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2969 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2970 jbd2_journal_set_features(sbi->s_journal,
2971 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2972 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2973 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2975 jbd2_journal_clear_features(sbi->s_journal,
2976 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2977 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2980 /* We have now updated the journal if required, so we can
2981 * validate the data journaling mode. */
2982 switch (test_opt(sb, DATA_FLAGS)) {
2984 /* No mode set, assume a default based on the journal
2985 * capabilities: ORDERED_DATA if the journal can
2986 * cope, else JOURNAL_DATA
2988 if (jbd2_journal_check_available_features
2989 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2990 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2992 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2995 case EXT4_MOUNT_ORDERED_DATA:
2996 case EXT4_MOUNT_WRITEBACK_DATA:
2997 if (!jbd2_journal_check_available_features
2998 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2999 ext4_msg(sb, KERN_ERR, "Journal does not support "
3000 "requested data journaling mode");
3001 goto failed_mount_wq;
3006 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3009 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3010 ext4_count_free_blocks(sb));
3012 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3013 ext4_count_free_inodes(sb));
3015 err = percpu_counter_init(&sbi->s_dirs_counter,
3016 ext4_count_dirs(sb));
3018 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3020 ext4_msg(sb, KERN_ERR, "insufficient memory");
3021 goto failed_mount_wq;
3024 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3025 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3026 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3027 goto failed_mount_wq;
3031 * The jbd2_journal_load will have done any necessary log recovery,
3032 * so we can safely mount the rest of the filesystem now.
3035 root = ext4_iget(sb, EXT4_ROOT_INO);
3037 ext4_msg(sb, KERN_ERR, "get root inode failed");
3038 ret = PTR_ERR(root);
3041 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3043 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3046 sb->s_root = d_alloc_root(root);
3048 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3054 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3056 /* determine the minimum size of new large inodes, if present */
3057 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3058 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3059 EXT4_GOOD_OLD_INODE_SIZE;
3060 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3061 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3062 if (sbi->s_want_extra_isize <
3063 le16_to_cpu(es->s_want_extra_isize))
3064 sbi->s_want_extra_isize =
3065 le16_to_cpu(es->s_want_extra_isize);
3066 if (sbi->s_want_extra_isize <
3067 le16_to_cpu(es->s_min_extra_isize))
3068 sbi->s_want_extra_isize =
3069 le16_to_cpu(es->s_min_extra_isize);
3072 /* Check if enough inode space is available */
3073 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3074 sbi->s_inode_size) {
3075 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3076 EXT4_GOOD_OLD_INODE_SIZE;
3077 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3081 if (test_opt(sb, DELALLOC) &&
3082 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3083 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3084 "requested data journaling mode");
3085 clear_opt(sbi->s_mount_opt, DELALLOC);
3087 if (test_opt(sb, DIOREAD_NOLOCK)) {
3088 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3089 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3090 "option - requested data journaling mode");
3091 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3093 if (sb->s_blocksize < PAGE_SIZE) {
3094 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3095 "option - block size is too small");
3096 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3100 err = ext4_setup_system_zone(sb);
3102 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3108 err = ext4_mb_init(sb, needs_recovery);
3110 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3115 sbi->s_kobj.kset = ext4_kset;
3116 init_completion(&sbi->s_kobj_unregister);
3117 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3120 ext4_mb_release(sb);
3121 ext4_ext_release(sb);
3125 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3126 ext4_orphan_cleanup(sb, es);
3127 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3128 if (needs_recovery) {
3129 ext4_msg(sb, KERN_INFO, "recovery complete");
3130 ext4_mark_recovery_complete(sb, es);
3132 if (EXT4_SB(sb)->s_journal) {
3133 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3134 descr = " journalled data mode";
3135 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3136 descr = " ordered data mode";
3138 descr = " writeback data mode";
3140 descr = "out journal";
3142 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3143 "Opts: %s", descr, orig_data);
3145 init_timer(&sbi->s_err_report);
3146 sbi->s_err_report.function = print_daily_error_info;
3147 sbi->s_err_report.data = (unsigned long) sb;
3148 if (es->s_error_count)
3149 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3157 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3161 ext4_msg(sb, KERN_ERR, "mount failed");
3162 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3164 ext4_release_system_zone(sb);
3165 if (sbi->s_journal) {
3166 jbd2_journal_destroy(sbi->s_journal);
3167 sbi->s_journal = NULL;
3169 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3170 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3171 percpu_counter_destroy(&sbi->s_dirs_counter);
3172 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3174 if (sbi->s_flex_groups) {
3175 if (is_vmalloc_addr(sbi->s_flex_groups))
3176 vfree(sbi->s_flex_groups);
3178 kfree(sbi->s_flex_groups);
3181 for (i = 0; i < db_count; i++)
3182 brelse(sbi->s_group_desc[i]);
3183 kfree(sbi->s_group_desc);
3186 remove_proc_entry(sb->s_id, ext4_proc_root);
3189 for (i = 0; i < MAXQUOTAS; i++)
3190 kfree(sbi->s_qf_names[i]);
3192 ext4_blkdev_remove(sbi);
3195 sb->s_fs_info = NULL;
3196 kfree(sbi->s_blockgroup_lock);
3205 * Setup any per-fs journal parameters now. We'll do this both on
3206 * initial mount, once the journal has been initialised but before we've
3207 * done any recovery; and again on any subsequent remount.
3209 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3213 journal->j_commit_interval = sbi->s_commit_interval;
3214 journal->j_min_batch_time = sbi->s_min_batch_time;
3215 journal->j_max_batch_time = sbi->s_max_batch_time;
3217 spin_lock(&journal->j_state_lock);
3218 if (test_opt(sb, BARRIER))
3219 journal->j_flags |= JBD2_BARRIER;
3221 journal->j_flags &= ~JBD2_BARRIER;
3222 if (test_opt(sb, DATA_ERR_ABORT))
3223 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3225 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3226 spin_unlock(&journal->j_state_lock);
3229 static journal_t *ext4_get_journal(struct super_block *sb,
3230 unsigned int journal_inum)
3232 struct inode *journal_inode;
3235 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3237 /* First, test for the existence of a valid inode on disk. Bad
3238 * things happen if we iget() an unused inode, as the subsequent
3239 * iput() will try to delete it. */
3241 journal_inode = ext4_iget(sb, journal_inum);
3242 if (IS_ERR(journal_inode)) {
3243 ext4_msg(sb, KERN_ERR, "no journal found");
3246 if (!journal_inode->i_nlink) {
3247 make_bad_inode(journal_inode);
3248 iput(journal_inode);
3249 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3253 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3254 journal_inode, journal_inode->i_size);
3255 if (!S_ISREG(journal_inode->i_mode)) {
3256 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3257 iput(journal_inode);
3261 journal = jbd2_journal_init_inode(journal_inode);
3263 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3264 iput(journal_inode);
3267 journal->j_private = sb;
3268 ext4_init_journal_params(sb, journal);
3272 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3275 struct buffer_head *bh;
3279 int hblock, blocksize;
3280 ext4_fsblk_t sb_block;
3281 unsigned long offset;
3282 struct ext4_super_block *es;
3283 struct block_device *bdev;
3285 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3287 bdev = ext4_blkdev_get(j_dev, sb);
3291 if (bd_claim(bdev, sb)) {
3292 ext4_msg(sb, KERN_ERR,
3293 "failed to claim external journal device");
3294 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3298 blocksize = sb->s_blocksize;
3299 hblock = bdev_logical_block_size(bdev);
3300 if (blocksize < hblock) {
3301 ext4_msg(sb, KERN_ERR,
3302 "blocksize too small for journal device");
3306 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3307 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3308 set_blocksize(bdev, blocksize);
3309 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3310 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3311 "external journal");
3315 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3316 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3317 !(le32_to_cpu(es->s_feature_incompat) &
3318 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3319 ext4_msg(sb, KERN_ERR, "external journal has "
3325 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3326 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3331 len = ext4_blocks_count(es);
3332 start = sb_block + 1;
3333 brelse(bh); /* we're done with the superblock */
3335 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3336 start, len, blocksize);
3338 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3341 journal->j_private = sb;
3342 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3343 wait_on_buffer(journal->j_sb_buffer);
3344 if (!buffer_uptodate(journal->j_sb_buffer)) {
3345 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3348 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3349 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3350 "user (unsupported) - %d",
3351 be32_to_cpu(journal->j_superblock->s_nr_users));
3354 EXT4_SB(sb)->journal_bdev = bdev;
3355 ext4_init_journal_params(sb, journal);
3359 jbd2_journal_destroy(journal);
3361 ext4_blkdev_put(bdev);
3365 static int ext4_load_journal(struct super_block *sb,
3366 struct ext4_super_block *es,
3367 unsigned long journal_devnum)
3370 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3373 int really_read_only;
3375 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3377 if (journal_devnum &&
3378 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3379 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3380 "numbers have changed");
3381 journal_dev = new_decode_dev(journal_devnum);
3383 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3385 really_read_only = bdev_read_only(sb->s_bdev);
3388 * Are we loading a blank journal or performing recovery after a
3389 * crash? For recovery, we need to check in advance whether we
3390 * can get read-write access to the device.
3392 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3393 if (sb->s_flags & MS_RDONLY) {
3394 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3395 "required on readonly filesystem");
3396 if (really_read_only) {
3397 ext4_msg(sb, KERN_ERR, "write access "
3398 "unavailable, cannot proceed");
3401 ext4_msg(sb, KERN_INFO, "write access will "
3402 "be enabled during recovery");
3406 if (journal_inum && journal_dev) {
3407 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3408 "and inode journals!");
3413 if (!(journal = ext4_get_journal(sb, journal_inum)))
3416 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3420 if (!(journal->j_flags & JBD2_BARRIER))
3421 ext4_msg(sb, KERN_INFO, "barriers disabled");
3423 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3424 err = jbd2_journal_update_format(journal);
3426 ext4_msg(sb, KERN_ERR, "error updating journal");
3427 jbd2_journal_destroy(journal);
3432 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3433 err = jbd2_journal_wipe(journal, !really_read_only);
3435 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3437 memcpy(save, ((char *) es) +
3438 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3439 err = jbd2_journal_load(journal);
3441 memcpy(((char *) es) + EXT4_S_ERR_START,
3442 save, EXT4_S_ERR_LEN);
3447 ext4_msg(sb, KERN_ERR, "error loading journal");
3448 jbd2_journal_destroy(journal);
3452 EXT4_SB(sb)->s_journal = journal;
3453 ext4_clear_journal_err(sb, es);
3455 if (journal_devnum &&
3456 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3457 es->s_journal_dev = cpu_to_le32(journal_devnum);
3459 /* Make sure we flush the recovery flag to disk. */
3460 ext4_commit_super(sb, 1);
3466 static int ext4_commit_super(struct super_block *sb, int sync)
3468 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3469 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3474 if (buffer_write_io_error(sbh)) {
3476 * Oh, dear. A previous attempt to write the
3477 * superblock failed. This could happen because the
3478 * USB device was yanked out. Or it could happen to
3479 * be a transient write error and maybe the block will
3480 * be remapped. Nothing we can do but to retry the
3481 * write and hope for the best.
3483 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3484 "superblock detected");
3485 clear_buffer_write_io_error(sbh);
3486 set_buffer_uptodate(sbh);
3489 * If the file system is mounted read-only, don't update the
3490 * superblock write time. This avoids updating the superblock
3491 * write time when we are mounting the root file system
3492 * read/only but we need to replay the journal; at that point,
3493 * for people who are east of GMT and who make their clock
3494 * tick in localtime for Windows bug-for-bug compatibility,
3495 * the clock is set in the future, and this will cause e2fsck
3496 * to complain and force a full file system check.
3498 if (!(sb->s_flags & MS_RDONLY))
3499 es->s_wtime = cpu_to_le32(get_seconds());
3500 if (sb->s_bdev->bd_part)
3501 es->s_kbytes_written =
3502 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3503 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3504 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3506 es->s_kbytes_written =
3507 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3508 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3509 &EXT4_SB(sb)->s_freeblocks_counter));
3510 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3511 &EXT4_SB(sb)->s_freeinodes_counter));
3513 BUFFER_TRACE(sbh, "marking dirty");
3514 mark_buffer_dirty(sbh);
3516 error = sync_dirty_buffer(sbh);
3520 error = buffer_write_io_error(sbh);
3522 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3524 clear_buffer_write_io_error(sbh);
3525 set_buffer_uptodate(sbh);
3532 * Have we just finished recovery? If so, and if we are mounting (or
3533 * remounting) the filesystem readonly, then we will end up with a
3534 * consistent fs on disk. Record that fact.
3536 static void ext4_mark_recovery_complete(struct super_block *sb,
3537 struct ext4_super_block *es)
3539 journal_t *journal = EXT4_SB(sb)->s_journal;
3541 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3542 BUG_ON(journal != NULL);
3545 jbd2_journal_lock_updates(journal);
3546 if (jbd2_journal_flush(journal) < 0)
3549 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3550 sb->s_flags & MS_RDONLY) {
3551 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3552 ext4_commit_super(sb, 1);
3556 jbd2_journal_unlock_updates(journal);
3560 * If we are mounting (or read-write remounting) a filesystem whose journal
3561 * has recorded an error from a previous lifetime, move that error to the
3562 * main filesystem now.
3564 static void ext4_clear_journal_err(struct super_block *sb,
3565 struct ext4_super_block *es)
3571 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3573 journal = EXT4_SB(sb)->s_journal;
3576 * Now check for any error status which may have been recorded in the
3577 * journal by a prior ext4_error() or ext4_abort()
3580 j_errno = jbd2_journal_errno(journal);
3584 errstr = ext4_decode_error(sb, j_errno, nbuf);
3585 ext4_warning(sb, "Filesystem error recorded "
3586 "from previous mount: %s", errstr);
3587 ext4_warning(sb, "Marking fs in need of filesystem check.");
3589 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3590 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3591 ext4_commit_super(sb, 1);
3593 jbd2_journal_clear_err(journal);
3598 * Force the running and committing transactions to commit,
3599 * and wait on the commit.
3601 int ext4_force_commit(struct super_block *sb)
3606 if (sb->s_flags & MS_RDONLY)
3609 journal = EXT4_SB(sb)->s_journal;
3611 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3612 ret = ext4_journal_force_commit(journal);
3618 static void ext4_write_super(struct super_block *sb)
3621 ext4_commit_super(sb, 1);
3625 static int ext4_sync_fs(struct super_block *sb, int wait)
3629 struct ext4_sb_info *sbi = EXT4_SB(sb);
3631 trace_ext4_sync_fs(sb, wait);
3632 flush_workqueue(sbi->dio_unwritten_wq);
3633 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3635 jbd2_log_wait_commit(sbi->s_journal, target);
3641 * LVM calls this function before a (read-only) snapshot is created. This
3642 * gives us a chance to flush the journal completely and mark the fs clean.
3644 static int ext4_freeze(struct super_block *sb)
3649 if (sb->s_flags & MS_RDONLY)
3652 journal = EXT4_SB(sb)->s_journal;
3654 /* Now we set up the journal barrier. */
3655 jbd2_journal_lock_updates(journal);
3658 * Don't clear the needs_recovery flag if we failed to flush
3661 error = jbd2_journal_flush(journal);
3665 /* Journal blocked and flushed, clear needs_recovery flag. */
3666 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3667 error = ext4_commit_super(sb, 1);
3669 /* we rely on s_frozen to stop further updates */
3670 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3675 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3676 * flag here, even though the filesystem is not technically dirty yet.
3678 static int ext4_unfreeze(struct super_block *sb)
3680 if (sb->s_flags & MS_RDONLY)
3684 /* Reset the needs_recovery flag before the fs is unlocked. */
3685 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3686 ext4_commit_super(sb, 1);
3691 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3693 struct ext4_super_block *es;
3694 struct ext4_sb_info *sbi = EXT4_SB(sb);
3695 ext4_fsblk_t n_blocks_count = 0;
3696 unsigned long old_sb_flags;
3697 struct ext4_mount_options old_opts;
3698 int enable_quota = 0;
3700 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3705 char *orig_data = kstrdup(data, GFP_KERNEL);
3709 /* Store the original options */
3711 old_sb_flags = sb->s_flags;
3712 old_opts.s_mount_opt = sbi->s_mount_opt;
3713 old_opts.s_resuid = sbi->s_resuid;
3714 old_opts.s_resgid = sbi->s_resgid;
3715 old_opts.s_commit_interval = sbi->s_commit_interval;
3716 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3717 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3719 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3720 for (i = 0; i < MAXQUOTAS; i++)
3721 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3723 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3724 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3727 * Allow the "check" option to be passed as a remount option.
3729 if (!parse_options(data, sb, NULL, &journal_ioprio,
3730 &n_blocks_count, 1)) {
3735 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3736 ext4_abort(sb, "Abort forced by user");
3738 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3739 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3743 if (sbi->s_journal) {
3744 ext4_init_journal_params(sb, sbi->s_journal);
3745 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3748 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3749 n_blocks_count > ext4_blocks_count(es)) {
3750 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3755 if (*flags & MS_RDONLY) {
3756 err = dquot_suspend(sb, -1);
3761 * First of all, the unconditional stuff we have to do
3762 * to disable replay of the journal when we next remount
3764 sb->s_flags |= MS_RDONLY;
3767 * OK, test if we are remounting a valid rw partition
3768 * readonly, and if so set the rdonly flag and then
3769 * mark the partition as valid again.
3771 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3772 (sbi->s_mount_state & EXT4_VALID_FS))
3773 es->s_state = cpu_to_le16(sbi->s_mount_state);
3776 ext4_mark_recovery_complete(sb, es);
3778 /* Make sure we can mount this feature set readwrite */
3779 if (!ext4_feature_set_ok(sb, 0)) {
3784 * Make sure the group descriptor checksums
3785 * are sane. If they aren't, refuse to remount r/w.
3787 for (g = 0; g < sbi->s_groups_count; g++) {
3788 struct ext4_group_desc *gdp =
3789 ext4_get_group_desc(sb, g, NULL);
3791 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3792 ext4_msg(sb, KERN_ERR,
3793 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3794 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3795 le16_to_cpu(gdp->bg_checksum));
3802 * If we have an unprocessed orphan list hanging
3803 * around from a previously readonly bdev mount,
3804 * require a full umount/remount for now.
3806 if (es->s_last_orphan) {
3807 ext4_msg(sb, KERN_WARNING, "Couldn't "
3808 "remount RDWR because of unprocessed "
3809 "orphan inode list. Please "
3810 "umount/remount instead");
3816 * Mounting a RDONLY partition read-write, so reread
3817 * and store the current valid flag. (It may have
3818 * been changed by e2fsck since we originally mounted
3822 ext4_clear_journal_err(sb, es);
3823 sbi->s_mount_state = le16_to_cpu(es->s_state);
3824 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3826 if (!ext4_setup_super(sb, es, 0))
3827 sb->s_flags &= ~MS_RDONLY;
3831 ext4_setup_system_zone(sb);
3832 if (sbi->s_journal == NULL)
3833 ext4_commit_super(sb, 1);
3836 /* Release old quota file names */
3837 for (i = 0; i < MAXQUOTAS; i++)
3838 if (old_opts.s_qf_names[i] &&
3839 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3840 kfree(old_opts.s_qf_names[i]);
3845 dquot_resume(sb, -1);
3847 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3852 sb->s_flags = old_sb_flags;
3853 sbi->s_mount_opt = old_opts.s_mount_opt;
3854 sbi->s_resuid = old_opts.s_resuid;
3855 sbi->s_resgid = old_opts.s_resgid;
3856 sbi->s_commit_interval = old_opts.s_commit_interval;
3857 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3858 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3860 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3861 for (i = 0; i < MAXQUOTAS; i++) {
3862 if (sbi->s_qf_names[i] &&
3863 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3864 kfree(sbi->s_qf_names[i]);
3865 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3874 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3876 struct super_block *sb = dentry->d_sb;
3877 struct ext4_sb_info *sbi = EXT4_SB(sb);
3878 struct ext4_super_block *es = sbi->s_es;
3881 if (test_opt(sb, MINIX_DF)) {
3882 sbi->s_overhead_last = 0;
3883 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3884 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3885 ext4_fsblk_t overhead = 0;
3888 * Compute the overhead (FS structures). This is constant
3889 * for a given filesystem unless the number of block groups
3890 * changes so we cache the previous value until it does.
3894 * All of the blocks before first_data_block are
3897 overhead = le32_to_cpu(es->s_first_data_block);
3900 * Add the overhead attributed to the superblock and
3901 * block group descriptors. If the sparse superblocks
3902 * feature is turned on, then not all groups have this.
3904 for (i = 0; i < ngroups; i++) {
3905 overhead += ext4_bg_has_super(sb, i) +
3906 ext4_bg_num_gdb(sb, i);
3911 * Every block group has an inode bitmap, a block
3912 * bitmap, and an inode table.
3914 overhead += ngroups * (2 + sbi->s_itb_per_group);
3915 sbi->s_overhead_last = overhead;
3917 sbi->s_blocks_last = ext4_blocks_count(es);
3920 buf->f_type = EXT4_SUPER_MAGIC;
3921 buf->f_bsize = sb->s_blocksize;
3922 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3923 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3924 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3925 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3926 if (buf->f_bfree < ext4_r_blocks_count(es))
3928 buf->f_files = le32_to_cpu(es->s_inodes_count);
3929 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3930 buf->f_namelen = EXT4_NAME_LEN;
3931 fsid = le64_to_cpup((void *)es->s_uuid) ^
3932 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3933 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3934 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3939 /* Helper function for writing quotas on sync - we need to start transaction
3940 * before quota file is locked for write. Otherwise the are possible deadlocks:
3941 * Process 1 Process 2
3942 * ext4_create() quota_sync()
3943 * jbd2_journal_start() write_dquot()
3944 * dquot_initialize() down(dqio_mutex)
3945 * down(dqio_mutex) jbd2_journal_start()
3951 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3953 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3956 static int ext4_write_dquot(struct dquot *dquot)
3960 struct inode *inode;
3962 inode = dquot_to_inode(dquot);
3963 handle = ext4_journal_start(inode,
3964 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3966 return PTR_ERR(handle);
3967 ret = dquot_commit(dquot);
3968 err = ext4_journal_stop(handle);
3974 static int ext4_acquire_dquot(struct dquot *dquot)
3979 handle = ext4_journal_start(dquot_to_inode(dquot),
3980 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3982 return PTR_ERR(handle);
3983 ret = dquot_acquire(dquot);
3984 err = ext4_journal_stop(handle);
3990 static int ext4_release_dquot(struct dquot *dquot)
3995 handle = ext4_journal_start(dquot_to_inode(dquot),
3996 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3997 if (IS_ERR(handle)) {
3998 /* Release dquot anyway to avoid endless cycle in dqput() */
3999 dquot_release(dquot);
4000 return PTR_ERR(handle);
4002 ret = dquot_release(dquot);
4003 err = ext4_journal_stop(handle);
4009 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4011 /* Are we journaling quotas? */
4012 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4013 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4014 dquot_mark_dquot_dirty(dquot);
4015 return ext4_write_dquot(dquot);
4017 return dquot_mark_dquot_dirty(dquot);
4021 static int ext4_write_info(struct super_block *sb, int type)
4026 /* Data block + inode block */
4027 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4029 return PTR_ERR(handle);
4030 ret = dquot_commit_info(sb, type);
4031 err = ext4_journal_stop(handle);
4038 * Turn on quotas during mount time - we need to find
4039 * the quota file and such...
4041 static int ext4_quota_on_mount(struct super_block *sb, int type)
4043 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4044 EXT4_SB(sb)->s_jquota_fmt, type);
4048 * Standard function to be called on quota_on
4050 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4056 if (!test_opt(sb, QUOTA))
4059 err = kern_path(name, LOOKUP_FOLLOW, &path);
4063 /* Quotafile not on the same filesystem? */
4064 if (path.mnt->mnt_sb != sb) {
4068 /* Journaling quota? */
4069 if (EXT4_SB(sb)->s_qf_names[type]) {
4070 /* Quotafile not in fs root? */
4071 if (path.dentry->d_parent != sb->s_root)
4072 ext4_msg(sb, KERN_WARNING,
4073 "Quota file not on filesystem root. "
4074 "Journaled quota will not work");
4078 * When we journal data on quota file, we have to flush journal to see
4079 * all updates to the file when we bypass pagecache...
4081 if (EXT4_SB(sb)->s_journal &&
4082 ext4_should_journal_data(path.dentry->d_inode)) {
4084 * We don't need to lock updates but journal_flush() could
4085 * otherwise be livelocked...
4087 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4088 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4089 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4096 err = dquot_quota_on_path(sb, type, format_id, &path);
4101 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4102 * acquiring the locks... As quota files are never truncated and quota code
4103 * itself serializes the operations (and noone else should touch the files)
4104 * we don't have to be afraid of races */
4105 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4106 size_t len, loff_t off)
4108 struct inode *inode = sb_dqopt(sb)->files[type];
4109 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4111 int offset = off & (sb->s_blocksize - 1);
4114 struct buffer_head *bh;
4115 loff_t i_size = i_size_read(inode);
4119 if (off+len > i_size)
4122 while (toread > 0) {
4123 tocopy = sb->s_blocksize - offset < toread ?
4124 sb->s_blocksize - offset : toread;
4125 bh = ext4_bread(NULL, inode, blk, 0, &err);
4128 if (!bh) /* A hole? */
4129 memset(data, 0, tocopy);
4131 memcpy(data, bh->b_data+offset, tocopy);
4141 /* Write to quotafile (we know the transaction is already started and has
4142 * enough credits) */
4143 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4144 const char *data, size_t len, loff_t off)
4146 struct inode *inode = sb_dqopt(sb)->files[type];
4147 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4149 int offset = off & (sb->s_blocksize - 1);
4150 struct buffer_head *bh;
4151 handle_t *handle = journal_current_handle();
4153 if (EXT4_SB(sb)->s_journal && !handle) {
4154 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4155 " cancelled because transaction is not started",
4156 (unsigned long long)off, (unsigned long long)len);
4160 * Since we account only one data block in transaction credits,
4161 * then it is impossible to cross a block boundary.
4163 if (sb->s_blocksize - offset < len) {
4164 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4165 " cancelled because not block aligned",
4166 (unsigned long long)off, (unsigned long long)len);
4170 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4171 bh = ext4_bread(handle, inode, blk, 1, &err);
4174 err = ext4_journal_get_write_access(handle, bh);
4180 memcpy(bh->b_data+offset, data, len);
4181 flush_dcache_page(bh->b_page);
4183 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4187 mutex_unlock(&inode->i_mutex);
4190 if (inode->i_size < off + len) {
4191 i_size_write(inode, off + len);
4192 EXT4_I(inode)->i_disksize = inode->i_size;
4194 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4195 ext4_mark_inode_dirty(handle, inode);
4196 mutex_unlock(&inode->i_mutex);
4202 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4203 const char *dev_name, void *data, struct vfsmount *mnt)
4205 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4208 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4209 static struct file_system_type ext2_fs_type = {
4210 .owner = THIS_MODULE,
4212 .get_sb = ext4_get_sb,
4213 .kill_sb = kill_block_super,
4214 .fs_flags = FS_REQUIRES_DEV,
4217 static inline void register_as_ext2(void)
4219 int err = register_filesystem(&ext2_fs_type);
4222 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4225 static inline void unregister_as_ext2(void)
4227 unregister_filesystem(&ext2_fs_type);
4229 MODULE_ALIAS("ext2");
4231 static inline void register_as_ext2(void) { }
4232 static inline void unregister_as_ext2(void) { }
4235 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4236 static inline void register_as_ext3(void)
4238 int err = register_filesystem(&ext3_fs_type);
4241 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4244 static inline void unregister_as_ext3(void)
4246 unregister_filesystem(&ext3_fs_type);
4248 MODULE_ALIAS("ext3");
4250 static inline void register_as_ext3(void) { }
4251 static inline void unregister_as_ext3(void) { }
4254 static struct file_system_type ext4_fs_type = {
4255 .owner = THIS_MODULE,
4257 .get_sb = ext4_get_sb,
4258 .kill_sb = kill_block_super,
4259 .fs_flags = FS_REQUIRES_DEV,
4262 static int __init init_ext4_fs(void)
4266 ext4_check_flag_values();
4267 err = init_ext4_system_zone();
4270 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4273 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4274 err = init_ext4_mballoc();
4278 err = init_ext4_xattr();
4281 err = init_inodecache();
4286 err = register_filesystem(&ext4_fs_type);
4291 unregister_as_ext2();
4292 unregister_as_ext3();
4293 destroy_inodecache();
4297 exit_ext4_mballoc();
4299 remove_proc_entry("fs/ext4", NULL);
4300 kset_unregister(ext4_kset);
4302 exit_ext4_system_zone();
4306 static void __exit exit_ext4_fs(void)
4308 unregister_as_ext2();
4309 unregister_as_ext3();
4310 unregister_filesystem(&ext4_fs_type);
4311 destroy_inodecache();
4313 exit_ext4_mballoc();
4314 remove_proc_entry("fs/ext4", NULL);
4315 kset_unregister(ext4_kset);
4316 exit_ext4_system_zone();
4319 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4320 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4321 MODULE_LICENSE("GPL");
4322 module_init(init_ext4_fs)
4323 module_exit(exit_ext4_fs)