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 del_timer(&sbi->s_err_report);
723 ext4_release_system_zone(sb);
725 ext4_ext_release(sb);
726 ext4_xattr_put_super(sb);
728 if (!(sb->s_flags & MS_RDONLY)) {
729 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
730 es->s_state = cpu_to_le16(sbi->s_mount_state);
731 ext4_commit_super(sb, 1);
734 remove_proc_entry(sb->s_id, ext4_proc_root);
736 kobject_del(&sbi->s_kobj);
738 for (i = 0; i < sbi->s_gdb_count; i++)
739 brelse(sbi->s_group_desc[i]);
740 kfree(sbi->s_group_desc);
741 if (is_vmalloc_addr(sbi->s_flex_groups))
742 vfree(sbi->s_flex_groups);
744 kfree(sbi->s_flex_groups);
745 percpu_counter_destroy(&sbi->s_freeblocks_counter);
746 percpu_counter_destroy(&sbi->s_freeinodes_counter);
747 percpu_counter_destroy(&sbi->s_dirs_counter);
748 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
751 for (i = 0; i < MAXQUOTAS; i++)
752 kfree(sbi->s_qf_names[i]);
755 /* Debugging code just in case the in-memory inode orphan list
756 * isn't empty. The on-disk one can be non-empty if we've
757 * detected an error and taken the fs readonly, but the
758 * in-memory list had better be clean by this point. */
759 if (!list_empty(&sbi->s_orphan))
760 dump_orphan_list(sb, sbi);
761 J_ASSERT(list_empty(&sbi->s_orphan));
763 invalidate_bdev(sb->s_bdev);
764 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
766 * Invalidate the journal device's buffers. We don't want them
767 * floating about in memory - the physical journal device may
768 * hotswapped, and it breaks the `ro-after' testing code.
770 sync_blockdev(sbi->journal_bdev);
771 invalidate_bdev(sbi->journal_bdev);
772 ext4_blkdev_remove(sbi);
774 sb->s_fs_info = NULL;
776 * Now that we are completely done shutting down the
777 * superblock, we need to actually destroy the kobject.
781 kobject_put(&sbi->s_kobj);
782 wait_for_completion(&sbi->s_kobj_unregister);
783 kfree(sbi->s_blockgroup_lock);
787 static struct kmem_cache *ext4_inode_cachep;
790 * Called inside transaction, so use GFP_NOFS
792 static struct inode *ext4_alloc_inode(struct super_block *sb)
794 struct ext4_inode_info *ei;
796 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
800 ei->vfs_inode.i_version = 1;
801 ei->vfs_inode.i_data.writeback_index = 0;
802 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
803 INIT_LIST_HEAD(&ei->i_prealloc_list);
804 spin_lock_init(&ei->i_prealloc_lock);
806 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
807 * therefore it can be null here. Don't check it, just initialize
810 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
811 ei->i_reserved_data_blocks = 0;
812 ei->i_reserved_meta_blocks = 0;
813 ei->i_allocated_meta_blocks = 0;
814 ei->i_da_metadata_calc_len = 0;
815 ei->i_delalloc_reserved_flag = 0;
816 spin_lock_init(&(ei->i_block_reservation_lock));
818 ei->i_reserved_quota = 0;
820 INIT_LIST_HEAD(&ei->i_completed_io_list);
821 spin_lock_init(&ei->i_completed_io_lock);
822 ei->cur_aio_dio = NULL;
824 ei->i_datasync_tid = 0;
826 return &ei->vfs_inode;
829 static void ext4_destroy_inode(struct inode *inode)
831 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
832 ext4_msg(inode->i_sb, KERN_ERR,
833 "Inode %lu (%p): orphan list check failed!",
834 inode->i_ino, EXT4_I(inode));
835 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
836 EXT4_I(inode), sizeof(struct ext4_inode_info),
840 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
843 static void init_once(void *foo)
845 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
847 INIT_LIST_HEAD(&ei->i_orphan);
848 #ifdef CONFIG_EXT4_FS_XATTR
849 init_rwsem(&ei->xattr_sem);
851 init_rwsem(&ei->i_data_sem);
852 inode_init_once(&ei->vfs_inode);
855 static int init_inodecache(void)
857 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
858 sizeof(struct ext4_inode_info),
859 0, (SLAB_RECLAIM_ACCOUNT|
862 if (ext4_inode_cachep == NULL)
867 static void destroy_inodecache(void)
869 kmem_cache_destroy(ext4_inode_cachep);
872 void ext4_clear_inode(struct inode *inode)
874 invalidate_inode_buffers(inode);
875 end_writeback(inode);
877 ext4_discard_preallocations(inode);
878 if (EXT4_JOURNAL(inode))
879 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
880 &EXT4_I(inode)->jinode);
883 static inline void ext4_show_quota_options(struct seq_file *seq,
884 struct super_block *sb)
886 #if defined(CONFIG_QUOTA)
887 struct ext4_sb_info *sbi = EXT4_SB(sb);
889 if (sbi->s_jquota_fmt) {
892 switch (sbi->s_jquota_fmt) {
903 seq_printf(seq, ",jqfmt=%s", fmtname);
906 if (sbi->s_qf_names[USRQUOTA])
907 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
909 if (sbi->s_qf_names[GRPQUOTA])
910 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
912 if (test_opt(sb, USRQUOTA))
913 seq_puts(seq, ",usrquota");
915 if (test_opt(sb, GRPQUOTA))
916 seq_puts(seq, ",grpquota");
922 * - it's set to a non-default value OR
923 * - if the per-sb default is different from the global default
925 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
928 unsigned long def_mount_opts;
929 struct super_block *sb = vfs->mnt_sb;
930 struct ext4_sb_info *sbi = EXT4_SB(sb);
931 struct ext4_super_block *es = sbi->s_es;
933 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
934 def_errors = le16_to_cpu(es->s_errors);
936 if (sbi->s_sb_block != 1)
937 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
938 if (test_opt(sb, MINIX_DF))
939 seq_puts(seq, ",minixdf");
940 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
941 seq_puts(seq, ",grpid");
942 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
943 seq_puts(seq, ",nogrpid");
944 if (sbi->s_resuid != EXT4_DEF_RESUID ||
945 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
946 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
948 if (sbi->s_resgid != EXT4_DEF_RESGID ||
949 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
950 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
952 if (test_opt(sb, ERRORS_RO)) {
953 if (def_errors == EXT4_ERRORS_PANIC ||
954 def_errors == EXT4_ERRORS_CONTINUE) {
955 seq_puts(seq, ",errors=remount-ro");
958 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
959 seq_puts(seq, ",errors=continue");
960 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
961 seq_puts(seq, ",errors=panic");
962 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
963 seq_puts(seq, ",nouid32");
964 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
965 seq_puts(seq, ",debug");
966 if (test_opt(sb, OLDALLOC))
967 seq_puts(seq, ",oldalloc");
968 #ifdef CONFIG_EXT4_FS_XATTR
969 if (test_opt(sb, XATTR_USER) &&
970 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
971 seq_puts(seq, ",user_xattr");
972 if (!test_opt(sb, XATTR_USER) &&
973 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
974 seq_puts(seq, ",nouser_xattr");
977 #ifdef CONFIG_EXT4_FS_POSIX_ACL
978 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
979 seq_puts(seq, ",acl");
980 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
981 seq_puts(seq, ",noacl");
983 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
984 seq_printf(seq, ",commit=%u",
985 (unsigned) (sbi->s_commit_interval / HZ));
987 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
988 seq_printf(seq, ",min_batch_time=%u",
989 (unsigned) sbi->s_min_batch_time);
991 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
992 seq_printf(seq, ",max_batch_time=%u",
993 (unsigned) sbi->s_min_batch_time);
997 * We're changing the default of barrier mount option, so
998 * let's always display its mount state so it's clear what its
1001 seq_puts(seq, ",barrier=");
1002 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1003 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1004 seq_puts(seq, ",journal_async_commit");
1005 else if (test_opt(sb, JOURNAL_CHECKSUM))
1006 seq_puts(seq, ",journal_checksum");
1007 if (test_opt(sb, I_VERSION))
1008 seq_puts(seq, ",i_version");
1009 if (!test_opt(sb, DELALLOC) &&
1010 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1011 seq_puts(seq, ",nodelalloc");
1014 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1016 * journal mode get enabled in different ways
1017 * So just print the value even if we didn't specify it
1019 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1020 seq_puts(seq, ",data=journal");
1021 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1022 seq_puts(seq, ",data=ordered");
1023 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1024 seq_puts(seq, ",data=writeback");
1026 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1027 seq_printf(seq, ",inode_readahead_blks=%u",
1028 sbi->s_inode_readahead_blks);
1030 if (test_opt(sb, DATA_ERR_ABORT))
1031 seq_puts(seq, ",data_err=abort");
1033 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1034 seq_puts(seq, ",noauto_da_alloc");
1036 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1037 seq_puts(seq, ",discard");
1039 if (test_opt(sb, NOLOAD))
1040 seq_puts(seq, ",norecovery");
1042 if (test_opt(sb, DIOREAD_NOLOCK))
1043 seq_puts(seq, ",dioread_nolock");
1045 if (test_opt(sb, BLOCK_VALIDITY) &&
1046 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1047 seq_puts(seq, ",block_validity");
1049 ext4_show_quota_options(seq, sb);
1054 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1055 u64 ino, u32 generation)
1057 struct inode *inode;
1059 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1060 return ERR_PTR(-ESTALE);
1061 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1062 return ERR_PTR(-ESTALE);
1064 /* iget isn't really right if the inode is currently unallocated!!
1066 * ext4_read_inode will return a bad_inode if the inode had been
1067 * deleted, so we should be safe.
1069 * Currently we don't know the generation for parent directory, so
1070 * a generation of 0 means "accept any"
1072 inode = ext4_iget(sb, ino);
1074 return ERR_CAST(inode);
1075 if (generation && inode->i_generation != generation) {
1077 return ERR_PTR(-ESTALE);
1083 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1084 int fh_len, int fh_type)
1086 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
1090 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1091 int fh_len, int fh_type)
1093 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1094 ext4_nfs_get_inode);
1098 * Try to release metadata pages (indirect blocks, directories) which are
1099 * mapped via the block device. Since these pages could have journal heads
1100 * which would prevent try_to_free_buffers() from freeing them, we must use
1101 * jbd2 layer's try_to_free_buffers() function to release them.
1103 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1106 journal_t *journal = EXT4_SB(sb)->s_journal;
1108 WARN_ON(PageChecked(page));
1109 if (!page_has_buffers(page))
1112 return jbd2_journal_try_to_free_buffers(journal, page,
1113 wait & ~__GFP_WAIT);
1114 return try_to_free_buffers(page);
1118 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1119 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1121 static int ext4_write_dquot(struct dquot *dquot);
1122 static int ext4_acquire_dquot(struct dquot *dquot);
1123 static int ext4_release_dquot(struct dquot *dquot);
1124 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1125 static int ext4_write_info(struct super_block *sb, int type);
1126 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1128 static int ext4_quota_off(struct super_block *sb, int type);
1129 static int ext4_quota_on_mount(struct super_block *sb, int type);
1130 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1131 size_t len, loff_t off);
1132 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1133 const char *data, size_t len, loff_t off);
1135 static const struct dquot_operations ext4_quota_operations = {
1137 .get_reserved_space = ext4_get_reserved_space,
1139 .write_dquot = ext4_write_dquot,
1140 .acquire_dquot = ext4_acquire_dquot,
1141 .release_dquot = ext4_release_dquot,
1142 .mark_dirty = ext4_mark_dquot_dirty,
1143 .write_info = ext4_write_info,
1144 .alloc_dquot = dquot_alloc,
1145 .destroy_dquot = dquot_destroy,
1148 static const struct quotactl_ops ext4_qctl_operations = {
1149 .quota_on = ext4_quota_on,
1150 .quota_off = ext4_quota_off,
1151 .quota_sync = dquot_quota_sync,
1152 .get_info = dquot_get_dqinfo,
1153 .set_info = dquot_set_dqinfo,
1154 .get_dqblk = dquot_get_dqblk,
1155 .set_dqblk = dquot_set_dqblk
1159 static const struct super_operations ext4_sops = {
1160 .alloc_inode = ext4_alloc_inode,
1161 .destroy_inode = ext4_destroy_inode,
1162 .write_inode = ext4_write_inode,
1163 .dirty_inode = ext4_dirty_inode,
1164 .evict_inode = ext4_evict_inode,
1165 .put_super = ext4_put_super,
1166 .sync_fs = ext4_sync_fs,
1167 .freeze_fs = ext4_freeze,
1168 .unfreeze_fs = ext4_unfreeze,
1169 .statfs = ext4_statfs,
1170 .remount_fs = ext4_remount,
1171 .show_options = ext4_show_options,
1173 .quota_read = ext4_quota_read,
1174 .quota_write = ext4_quota_write,
1176 .bdev_try_to_free_page = bdev_try_to_free_page,
1179 static const struct super_operations ext4_nojournal_sops = {
1180 .alloc_inode = ext4_alloc_inode,
1181 .destroy_inode = ext4_destroy_inode,
1182 .write_inode = ext4_write_inode,
1183 .dirty_inode = ext4_dirty_inode,
1184 .evict_inode = ext4_evict_inode,
1185 .write_super = ext4_write_super,
1186 .put_super = ext4_put_super,
1187 .statfs = ext4_statfs,
1188 .remount_fs = ext4_remount,
1189 .show_options = ext4_show_options,
1191 .quota_read = ext4_quota_read,
1192 .quota_write = ext4_quota_write,
1194 .bdev_try_to_free_page = bdev_try_to_free_page,
1197 static const struct export_operations ext4_export_ops = {
1198 .fh_to_dentry = ext4_fh_to_dentry,
1199 .fh_to_parent = ext4_fh_to_parent,
1200 .get_parent = ext4_get_parent,
1204 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1205 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1206 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1207 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1208 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1209 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1210 Opt_journal_update, Opt_journal_dev,
1211 Opt_journal_checksum, Opt_journal_async_commit,
1212 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1213 Opt_data_err_abort, Opt_data_err_ignore,
1214 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1215 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1216 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1217 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1218 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1219 Opt_block_validity, Opt_noblock_validity,
1220 Opt_inode_readahead_blks, Opt_journal_ioprio,
1221 Opt_dioread_nolock, Opt_dioread_lock,
1222 Opt_discard, Opt_nodiscard,
1225 static const match_table_t tokens = {
1226 {Opt_bsd_df, "bsddf"},
1227 {Opt_minix_df, "minixdf"},
1228 {Opt_grpid, "grpid"},
1229 {Opt_grpid, "bsdgroups"},
1230 {Opt_nogrpid, "nogrpid"},
1231 {Opt_nogrpid, "sysvgroups"},
1232 {Opt_resgid, "resgid=%u"},
1233 {Opt_resuid, "resuid=%u"},
1235 {Opt_err_cont, "errors=continue"},
1236 {Opt_err_panic, "errors=panic"},
1237 {Opt_err_ro, "errors=remount-ro"},
1238 {Opt_nouid32, "nouid32"},
1239 {Opt_debug, "debug"},
1240 {Opt_oldalloc, "oldalloc"},
1241 {Opt_orlov, "orlov"},
1242 {Opt_user_xattr, "user_xattr"},
1243 {Opt_nouser_xattr, "nouser_xattr"},
1245 {Opt_noacl, "noacl"},
1246 {Opt_noload, "noload"},
1247 {Opt_noload, "norecovery"},
1250 {Opt_commit, "commit=%u"},
1251 {Opt_min_batch_time, "min_batch_time=%u"},
1252 {Opt_max_batch_time, "max_batch_time=%u"},
1253 {Opt_journal_update, "journal=update"},
1254 {Opt_journal_dev, "journal_dev=%u"},
1255 {Opt_journal_checksum, "journal_checksum"},
1256 {Opt_journal_async_commit, "journal_async_commit"},
1257 {Opt_abort, "abort"},
1258 {Opt_data_journal, "data=journal"},
1259 {Opt_data_ordered, "data=ordered"},
1260 {Opt_data_writeback, "data=writeback"},
1261 {Opt_data_err_abort, "data_err=abort"},
1262 {Opt_data_err_ignore, "data_err=ignore"},
1263 {Opt_offusrjquota, "usrjquota="},
1264 {Opt_usrjquota, "usrjquota=%s"},
1265 {Opt_offgrpjquota, "grpjquota="},
1266 {Opt_grpjquota, "grpjquota=%s"},
1267 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1268 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1269 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1270 {Opt_grpquota, "grpquota"},
1271 {Opt_noquota, "noquota"},
1272 {Opt_quota, "quota"},
1273 {Opt_usrquota, "usrquota"},
1274 {Opt_barrier, "barrier=%u"},
1275 {Opt_barrier, "barrier"},
1276 {Opt_nobarrier, "nobarrier"},
1277 {Opt_i_version, "i_version"},
1278 {Opt_stripe, "stripe=%u"},
1279 {Opt_resize, "resize"},
1280 {Opt_delalloc, "delalloc"},
1281 {Opt_nodelalloc, "nodelalloc"},
1282 {Opt_block_validity, "block_validity"},
1283 {Opt_noblock_validity, "noblock_validity"},
1284 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1285 {Opt_journal_ioprio, "journal_ioprio=%u"},
1286 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1287 {Opt_auto_da_alloc, "auto_da_alloc"},
1288 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1289 {Opt_dioread_nolock, "dioread_nolock"},
1290 {Opt_dioread_lock, "dioread_lock"},
1291 {Opt_discard, "discard"},
1292 {Opt_nodiscard, "nodiscard"},
1296 static ext4_fsblk_t get_sb_block(void **data)
1298 ext4_fsblk_t sb_block;
1299 char *options = (char *) *data;
1301 if (!options || strncmp(options, "sb=", 3) != 0)
1302 return 1; /* Default location */
1305 /* TODO: use simple_strtoll with >32bit ext4 */
1306 sb_block = simple_strtoul(options, &options, 0);
1307 if (*options && *options != ',') {
1308 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1312 if (*options == ',')
1314 *data = (void *) options;
1319 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1320 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1321 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1324 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1326 struct ext4_sb_info *sbi = EXT4_SB(sb);
1329 if (sb_any_quota_loaded(sb) &&
1330 !sbi->s_qf_names[qtype]) {
1331 ext4_msg(sb, KERN_ERR,
1332 "Cannot change journaled "
1333 "quota options when quota turned on");
1336 qname = match_strdup(args);
1338 ext4_msg(sb, KERN_ERR,
1339 "Not enough memory for storing quotafile name");
1342 if (sbi->s_qf_names[qtype] &&
1343 strcmp(sbi->s_qf_names[qtype], qname)) {
1344 ext4_msg(sb, KERN_ERR,
1345 "%s quota file already specified", QTYPE2NAME(qtype));
1349 sbi->s_qf_names[qtype] = qname;
1350 if (strchr(sbi->s_qf_names[qtype], '/')) {
1351 ext4_msg(sb, KERN_ERR,
1352 "quotafile must be on filesystem root");
1353 kfree(sbi->s_qf_names[qtype]);
1354 sbi->s_qf_names[qtype] = NULL;
1357 set_opt(sbi->s_mount_opt, QUOTA);
1361 static int clear_qf_name(struct super_block *sb, int qtype)
1364 struct ext4_sb_info *sbi = EXT4_SB(sb);
1366 if (sb_any_quota_loaded(sb) &&
1367 sbi->s_qf_names[qtype]) {
1368 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1369 " when quota turned on");
1373 * The space will be released later when all options are confirmed
1376 sbi->s_qf_names[qtype] = NULL;
1381 static int parse_options(char *options, struct super_block *sb,
1382 unsigned long *journal_devnum,
1383 unsigned int *journal_ioprio,
1384 ext4_fsblk_t *n_blocks_count, int is_remount)
1386 struct ext4_sb_info *sbi = EXT4_SB(sb);
1388 substring_t args[MAX_OPT_ARGS];
1398 while ((p = strsep(&options, ",")) != NULL) {
1404 * Initialize args struct so we know whether arg was
1405 * found; some options take optional arguments.
1407 args[0].to = args[0].from = 0;
1408 token = match_token(p, tokens, args);
1411 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1412 clear_opt(sbi->s_mount_opt, MINIX_DF);
1415 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1416 set_opt(sbi->s_mount_opt, MINIX_DF);
1420 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1421 set_opt(sbi->s_mount_opt, GRPID);
1425 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1426 clear_opt(sbi->s_mount_opt, GRPID);
1430 if (match_int(&args[0], &option))
1432 sbi->s_resuid = option;
1435 if (match_int(&args[0], &option))
1437 sbi->s_resgid = option;
1440 /* handled by get_sb_block() instead of here */
1441 /* *sb_block = match_int(&args[0]); */
1444 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1445 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1446 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1449 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1450 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1451 set_opt(sbi->s_mount_opt, ERRORS_RO);
1454 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1455 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1456 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1459 set_opt(sbi->s_mount_opt, NO_UID32);
1462 set_opt(sbi->s_mount_opt, DEBUG);
1465 set_opt(sbi->s_mount_opt, OLDALLOC);
1468 clear_opt(sbi->s_mount_opt, OLDALLOC);
1470 #ifdef CONFIG_EXT4_FS_XATTR
1471 case Opt_user_xattr:
1472 set_opt(sbi->s_mount_opt, XATTR_USER);
1474 case Opt_nouser_xattr:
1475 clear_opt(sbi->s_mount_opt, XATTR_USER);
1478 case Opt_user_xattr:
1479 case Opt_nouser_xattr:
1480 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1483 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1485 set_opt(sbi->s_mount_opt, POSIX_ACL);
1488 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1493 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1496 case Opt_journal_update:
1498 /* Eventually we will want to be able to create
1499 a journal file here. For now, only allow the
1500 user to specify an existing inode to be the
1503 ext4_msg(sb, KERN_ERR,
1504 "Cannot specify journal on remount");
1507 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1509 case Opt_journal_dev:
1511 ext4_msg(sb, KERN_ERR,
1512 "Cannot specify journal on remount");
1515 if (match_int(&args[0], &option))
1517 *journal_devnum = option;
1519 case Opt_journal_checksum:
1520 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1522 case Opt_journal_async_commit:
1523 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1524 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1527 set_opt(sbi->s_mount_opt, NOLOAD);
1530 if (match_int(&args[0], &option))
1535 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1536 sbi->s_commit_interval = HZ * option;
1538 case Opt_max_batch_time:
1539 if (match_int(&args[0], &option))
1544 option = EXT4_DEF_MAX_BATCH_TIME;
1545 sbi->s_max_batch_time = option;
1547 case Opt_min_batch_time:
1548 if (match_int(&args[0], &option))
1552 sbi->s_min_batch_time = option;
1554 case Opt_data_journal:
1555 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1557 case Opt_data_ordered:
1558 data_opt = EXT4_MOUNT_ORDERED_DATA;
1560 case Opt_data_writeback:
1561 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1564 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1565 ext4_msg(sb, KERN_ERR,
1566 "Cannot change data mode on remount");
1570 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1571 sbi->s_mount_opt |= data_opt;
1574 case Opt_data_err_abort:
1575 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1577 case Opt_data_err_ignore:
1578 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1582 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1586 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1589 case Opt_offusrjquota:
1590 if (!clear_qf_name(sb, USRQUOTA))
1593 case Opt_offgrpjquota:
1594 if (!clear_qf_name(sb, GRPQUOTA))
1598 case Opt_jqfmt_vfsold:
1599 qfmt = QFMT_VFS_OLD;
1601 case Opt_jqfmt_vfsv0:
1604 case Opt_jqfmt_vfsv1:
1607 if (sb_any_quota_loaded(sb) &&
1608 sbi->s_jquota_fmt != qfmt) {
1609 ext4_msg(sb, KERN_ERR, "Cannot change "
1610 "journaled quota options when "
1614 sbi->s_jquota_fmt = qfmt;
1618 set_opt(sbi->s_mount_opt, QUOTA);
1619 set_opt(sbi->s_mount_opt, USRQUOTA);
1622 set_opt(sbi->s_mount_opt, QUOTA);
1623 set_opt(sbi->s_mount_opt, GRPQUOTA);
1626 if (sb_any_quota_loaded(sb)) {
1627 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1628 "options when quota turned on");
1631 clear_opt(sbi->s_mount_opt, QUOTA);
1632 clear_opt(sbi->s_mount_opt, USRQUOTA);
1633 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1639 ext4_msg(sb, KERN_ERR,
1640 "quota options not supported");
1644 case Opt_offusrjquota:
1645 case Opt_offgrpjquota:
1646 case Opt_jqfmt_vfsold:
1647 case Opt_jqfmt_vfsv0:
1648 case Opt_jqfmt_vfsv1:
1649 ext4_msg(sb, KERN_ERR,
1650 "journaled quota options not supported");
1656 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1659 clear_opt(sbi->s_mount_opt, BARRIER);
1663 if (match_int(&args[0], &option))
1666 option = 1; /* No argument, default to 1 */
1668 set_opt(sbi->s_mount_opt, BARRIER);
1670 clear_opt(sbi->s_mount_opt, BARRIER);
1676 ext4_msg(sb, KERN_ERR,
1677 "resize option only available "
1681 if (match_int(&args[0], &option) != 0)
1683 *n_blocks_count = option;
1686 ext4_msg(sb, KERN_WARNING,
1687 "Ignoring deprecated nobh option");
1690 ext4_msg(sb, KERN_WARNING,
1691 "Ignoring deprecated bh option");
1694 set_opt(sbi->s_mount_opt, I_VERSION);
1695 sb->s_flags |= MS_I_VERSION;
1697 case Opt_nodelalloc:
1698 clear_opt(sbi->s_mount_opt, DELALLOC);
1701 if (match_int(&args[0], &option))
1705 sbi->s_stripe = option;
1708 set_opt(sbi->s_mount_opt, DELALLOC);
1710 case Opt_block_validity:
1711 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1713 case Opt_noblock_validity:
1714 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1716 case Opt_inode_readahead_blks:
1717 if (match_int(&args[0], &option))
1719 if (option < 0 || option > (1 << 30))
1721 if (!is_power_of_2(option)) {
1722 ext4_msg(sb, KERN_ERR,
1723 "EXT4-fs: inode_readahead_blks"
1724 " must be a power of 2");
1727 sbi->s_inode_readahead_blks = option;
1729 case Opt_journal_ioprio:
1730 if (match_int(&args[0], &option))
1732 if (option < 0 || option > 7)
1734 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1737 case Opt_noauto_da_alloc:
1738 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1740 case Opt_auto_da_alloc:
1742 if (match_int(&args[0], &option))
1745 option = 1; /* No argument, default to 1 */
1747 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1749 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1752 set_opt(sbi->s_mount_opt, DISCARD);
1755 clear_opt(sbi->s_mount_opt, DISCARD);
1757 case Opt_dioread_nolock:
1758 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1760 case Opt_dioread_lock:
1761 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1764 ext4_msg(sb, KERN_ERR,
1765 "Unrecognized mount option \"%s\" "
1766 "or missing value", p);
1771 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1772 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1773 clear_opt(sbi->s_mount_opt, USRQUOTA);
1775 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1776 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1778 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1779 ext4_msg(sb, KERN_ERR, "old and new quota "
1784 if (!sbi->s_jquota_fmt) {
1785 ext4_msg(sb, KERN_ERR, "journaled quota format "
1790 if (sbi->s_jquota_fmt) {
1791 ext4_msg(sb, KERN_ERR, "journaled quota format "
1792 "specified with no journaling "
1801 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1804 struct ext4_sb_info *sbi = EXT4_SB(sb);
1807 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1808 ext4_msg(sb, KERN_ERR, "revision level too high, "
1809 "forcing read-only mode");
1814 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1815 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1816 "running e2fsck is recommended");
1817 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1818 ext4_msg(sb, KERN_WARNING,
1819 "warning: mounting fs with errors, "
1820 "running e2fsck is recommended");
1821 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1822 le16_to_cpu(es->s_mnt_count) >=
1823 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1824 ext4_msg(sb, KERN_WARNING,
1825 "warning: maximal mount count reached, "
1826 "running e2fsck is recommended");
1827 else if (le32_to_cpu(es->s_checkinterval) &&
1828 (le32_to_cpu(es->s_lastcheck) +
1829 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1830 ext4_msg(sb, KERN_WARNING,
1831 "warning: checktime reached, "
1832 "running e2fsck is recommended");
1833 if (!sbi->s_journal)
1834 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1835 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1836 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1837 le16_add_cpu(&es->s_mnt_count, 1);
1838 es->s_mtime = cpu_to_le32(get_seconds());
1839 ext4_update_dynamic_rev(sb);
1841 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1843 ext4_commit_super(sb, 1);
1844 if (test_opt(sb, DEBUG))
1845 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1846 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1848 sbi->s_groups_count,
1849 EXT4_BLOCKS_PER_GROUP(sb),
1850 EXT4_INODES_PER_GROUP(sb),
1856 static int ext4_fill_flex_info(struct super_block *sb)
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 struct ext4_group_desc *gdp = NULL;
1860 ext4_group_t flex_group_count;
1861 ext4_group_t flex_group;
1862 int groups_per_flex = 0;
1866 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1867 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1869 if (groups_per_flex < 2) {
1870 sbi->s_log_groups_per_flex = 0;
1874 /* We allocate both existing and potentially added groups */
1875 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1876 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1877 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1878 size = flex_group_count * sizeof(struct flex_groups);
1879 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1880 if (sbi->s_flex_groups == NULL) {
1881 sbi->s_flex_groups = vmalloc(size);
1882 if (sbi->s_flex_groups)
1883 memset(sbi->s_flex_groups, 0, size);
1885 if (sbi->s_flex_groups == NULL) {
1886 ext4_msg(sb, KERN_ERR, "not enough memory for "
1887 "%u flex groups", flex_group_count);
1891 for (i = 0; i < sbi->s_groups_count; i++) {
1892 gdp = ext4_get_group_desc(sb, i, NULL);
1894 flex_group = ext4_flex_group(sbi, i);
1895 atomic_add(ext4_free_inodes_count(sb, gdp),
1896 &sbi->s_flex_groups[flex_group].free_inodes);
1897 atomic_add(ext4_free_blks_count(sb, gdp),
1898 &sbi->s_flex_groups[flex_group].free_blocks);
1899 atomic_add(ext4_used_dirs_count(sb, gdp),
1900 &sbi->s_flex_groups[flex_group].used_dirs);
1908 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1909 struct ext4_group_desc *gdp)
1913 if (sbi->s_es->s_feature_ro_compat &
1914 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1915 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1916 __le32 le_group = cpu_to_le32(block_group);
1918 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1919 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1920 crc = crc16(crc, (__u8 *)gdp, offset);
1921 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1922 /* for checksum of struct ext4_group_desc do the rest...*/
1923 if ((sbi->s_es->s_feature_incompat &
1924 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1925 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1926 crc = crc16(crc, (__u8 *)gdp + offset,
1927 le16_to_cpu(sbi->s_es->s_desc_size) -
1931 return cpu_to_le16(crc);
1934 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1935 struct ext4_group_desc *gdp)
1937 if ((sbi->s_es->s_feature_ro_compat &
1938 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1939 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1945 /* Called at mount-time, super-block is locked */
1946 static int ext4_check_descriptors(struct super_block *sb)
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1950 ext4_fsblk_t last_block;
1951 ext4_fsblk_t block_bitmap;
1952 ext4_fsblk_t inode_bitmap;
1953 ext4_fsblk_t inode_table;
1954 int flexbg_flag = 0;
1957 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1960 ext4_debug("Checking group descriptors");
1962 for (i = 0; i < sbi->s_groups_count; i++) {
1963 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1965 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1966 last_block = ext4_blocks_count(sbi->s_es) - 1;
1968 last_block = first_block +
1969 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1971 block_bitmap = ext4_block_bitmap(sb, gdp);
1972 if (block_bitmap < first_block || block_bitmap > last_block) {
1973 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1974 "Block bitmap for group %u not in group "
1975 "(block %llu)!", i, block_bitmap);
1978 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1979 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1980 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1981 "Inode bitmap for group %u not in group "
1982 "(block %llu)!", i, inode_bitmap);
1985 inode_table = ext4_inode_table(sb, gdp);
1986 if (inode_table < first_block ||
1987 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1988 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1989 "Inode table for group %u not in group "
1990 "(block %llu)!", i, inode_table);
1993 ext4_lock_group(sb, i);
1994 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1995 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1996 "Checksum for group %u failed (%u!=%u)",
1997 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1998 gdp)), le16_to_cpu(gdp->bg_checksum));
1999 if (!(sb->s_flags & MS_RDONLY)) {
2000 ext4_unlock_group(sb, i);
2004 ext4_unlock_group(sb, i);
2006 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2009 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2010 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2014 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2015 * the superblock) which were deleted from all directories, but held open by
2016 * a process at the time of a crash. We walk the list and try to delete these
2017 * inodes at recovery time (only with a read-write filesystem).
2019 * In order to keep the orphan inode chain consistent during traversal (in
2020 * case of crash during recovery), we link each inode into the superblock
2021 * orphan list_head and handle it the same way as an inode deletion during
2022 * normal operation (which journals the operations for us).
2024 * We only do an iget() and an iput() on each inode, which is very safe if we
2025 * accidentally point at an in-use or already deleted inode. The worst that
2026 * can happen in this case is that we get a "bit already cleared" message from
2027 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2028 * e2fsck was run on this filesystem, and it must have already done the orphan
2029 * inode cleanup for us, so we can safely abort without any further action.
2031 static void ext4_orphan_cleanup(struct super_block *sb,
2032 struct ext4_super_block *es)
2034 unsigned int s_flags = sb->s_flags;
2035 int nr_orphans = 0, nr_truncates = 0;
2039 if (!es->s_last_orphan) {
2040 jbd_debug(4, "no orphan inodes to clean up\n");
2044 if (bdev_read_only(sb->s_bdev)) {
2045 ext4_msg(sb, KERN_ERR, "write access "
2046 "unavailable, skipping orphan cleanup");
2050 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2051 if (es->s_last_orphan)
2052 jbd_debug(1, "Errors on filesystem, "
2053 "clearing orphan list.\n");
2054 es->s_last_orphan = 0;
2055 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2059 if (s_flags & MS_RDONLY) {
2060 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2061 sb->s_flags &= ~MS_RDONLY;
2064 /* Needed for iput() to work correctly and not trash data */
2065 sb->s_flags |= MS_ACTIVE;
2066 /* Turn on quotas so that they are updated correctly */
2067 for (i = 0; i < MAXQUOTAS; i++) {
2068 if (EXT4_SB(sb)->s_qf_names[i]) {
2069 int ret = ext4_quota_on_mount(sb, i);
2071 ext4_msg(sb, KERN_ERR,
2072 "Cannot turn on journaled "
2073 "quota: error %d", ret);
2078 while (es->s_last_orphan) {
2079 struct inode *inode;
2081 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2082 if (IS_ERR(inode)) {
2083 es->s_last_orphan = 0;
2087 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2088 dquot_initialize(inode);
2089 if (inode->i_nlink) {
2090 ext4_msg(sb, KERN_DEBUG,
2091 "%s: truncating inode %lu to %lld bytes",
2092 __func__, inode->i_ino, inode->i_size);
2093 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2094 inode->i_ino, inode->i_size);
2095 ext4_truncate(inode);
2098 ext4_msg(sb, KERN_DEBUG,
2099 "%s: deleting unreferenced inode %lu",
2100 __func__, inode->i_ino);
2101 jbd_debug(2, "deleting unreferenced inode %lu\n",
2105 iput(inode); /* The delete magic happens here! */
2108 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2111 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2112 PLURAL(nr_orphans));
2114 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2115 PLURAL(nr_truncates));
2117 /* Turn quotas off */
2118 for (i = 0; i < MAXQUOTAS; i++) {
2119 if (sb_dqopt(sb)->files[i])
2120 dquot_quota_off(sb, i);
2123 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2127 * Maximal extent format file size.
2128 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2129 * extent format containers, within a sector_t, and within i_blocks
2130 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2131 * so that won't be a limiting factor.
2133 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2135 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2138 loff_t upper_limit = MAX_LFS_FILESIZE;
2140 /* small i_blocks in vfs inode? */
2141 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2143 * CONFIG_LBDAF is not enabled implies the inode
2144 * i_block represent total blocks in 512 bytes
2145 * 32 == size of vfs inode i_blocks * 8
2147 upper_limit = (1LL << 32) - 1;
2149 /* total blocks in file system block size */
2150 upper_limit >>= (blkbits - 9);
2151 upper_limit <<= blkbits;
2154 /* 32-bit extent-start container, ee_block */
2159 /* Sanity check against vm- & vfs- imposed limits */
2160 if (res > upper_limit)
2167 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2168 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2169 * We need to be 1 filesystem block less than the 2^48 sector limit.
2171 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2173 loff_t res = EXT4_NDIR_BLOCKS;
2176 /* This is calculated to be the largest file size for a dense, block
2177 * mapped file such that the file's total number of 512-byte sectors,
2178 * including data and all indirect blocks, does not exceed (2^48 - 1).
2180 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2181 * number of 512-byte sectors of the file.
2184 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2186 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2187 * the inode i_block field represents total file blocks in
2188 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2190 upper_limit = (1LL << 32) - 1;
2192 /* total blocks in file system block size */
2193 upper_limit >>= (bits - 9);
2197 * We use 48 bit ext4_inode i_blocks
2198 * With EXT4_HUGE_FILE_FL set the i_blocks
2199 * represent total number of blocks in
2200 * file system block size
2202 upper_limit = (1LL << 48) - 1;
2206 /* indirect blocks */
2208 /* double indirect blocks */
2209 meta_blocks += 1 + (1LL << (bits-2));
2210 /* tripple indirect blocks */
2211 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2213 upper_limit -= meta_blocks;
2214 upper_limit <<= bits;
2216 res += 1LL << (bits-2);
2217 res += 1LL << (2*(bits-2));
2218 res += 1LL << (3*(bits-2));
2220 if (res > upper_limit)
2223 if (res > MAX_LFS_FILESIZE)
2224 res = MAX_LFS_FILESIZE;
2229 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2230 ext4_fsblk_t logical_sb_block, int nr)
2232 struct ext4_sb_info *sbi = EXT4_SB(sb);
2233 ext4_group_t bg, first_meta_bg;
2236 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2238 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2240 return logical_sb_block + nr + 1;
2241 bg = sbi->s_desc_per_block * nr;
2242 if (ext4_bg_has_super(sb, bg))
2245 return (has_super + ext4_group_first_block_no(sb, bg));
2249 * ext4_get_stripe_size: Get the stripe size.
2250 * @sbi: In memory super block info
2252 * If we have specified it via mount option, then
2253 * use the mount option value. If the value specified at mount time is
2254 * greater than the blocks per group use the super block value.
2255 * If the super block value is greater than blocks per group return 0.
2256 * Allocator needs it be less than blocks per group.
2259 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2261 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2262 unsigned long stripe_width =
2263 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2265 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2266 return sbi->s_stripe;
2268 if (stripe_width <= sbi->s_blocks_per_group)
2269 return stripe_width;
2271 if (stride <= sbi->s_blocks_per_group)
2280 struct attribute attr;
2281 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2282 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2283 const char *, size_t);
2287 static int parse_strtoul(const char *buf,
2288 unsigned long max, unsigned long *value)
2292 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2293 endp = skip_spaces(endp);
2294 if (*endp || *value > max)
2300 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2301 struct ext4_sb_info *sbi,
2304 return snprintf(buf, PAGE_SIZE, "%llu\n",
2305 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2308 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2309 struct ext4_sb_info *sbi, char *buf)
2311 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2313 if (!sb->s_bdev->bd_part)
2314 return snprintf(buf, PAGE_SIZE, "0\n");
2315 return snprintf(buf, PAGE_SIZE, "%lu\n",
2316 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2317 sbi->s_sectors_written_start) >> 1);
2320 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2321 struct ext4_sb_info *sbi, char *buf)
2323 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2325 if (!sb->s_bdev->bd_part)
2326 return snprintf(buf, PAGE_SIZE, "0\n");
2327 return snprintf(buf, PAGE_SIZE, "%llu\n",
2328 (unsigned long long)(sbi->s_kbytes_written +
2329 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2330 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2333 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2334 struct ext4_sb_info *sbi,
2335 const char *buf, size_t count)
2339 if (parse_strtoul(buf, 0x40000000, &t))
2342 if (!is_power_of_2(t))
2345 sbi->s_inode_readahead_blks = t;
2349 static ssize_t sbi_ui_show(struct ext4_attr *a,
2350 struct ext4_sb_info *sbi, char *buf)
2352 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2354 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2357 static ssize_t sbi_ui_store(struct ext4_attr *a,
2358 struct ext4_sb_info *sbi,
2359 const char *buf, size_t count)
2361 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2364 if (parse_strtoul(buf, 0xffffffff, &t))
2370 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2371 static struct ext4_attr ext4_attr_##_name = { \
2372 .attr = {.name = __stringify(_name), .mode = _mode }, \
2375 .offset = offsetof(struct ext4_sb_info, _elname), \
2377 #define EXT4_ATTR(name, mode, show, store) \
2378 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2380 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2381 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2382 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2383 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2384 #define ATTR_LIST(name) &ext4_attr_##name.attr
2386 EXT4_RO_ATTR(delayed_allocation_blocks);
2387 EXT4_RO_ATTR(session_write_kbytes);
2388 EXT4_RO_ATTR(lifetime_write_kbytes);
2389 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2390 inode_readahead_blks_store, s_inode_readahead_blks);
2391 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2392 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2393 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2394 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2395 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2396 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2397 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2398 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2400 static struct attribute *ext4_attrs[] = {
2401 ATTR_LIST(delayed_allocation_blocks),
2402 ATTR_LIST(session_write_kbytes),
2403 ATTR_LIST(lifetime_write_kbytes),
2404 ATTR_LIST(inode_readahead_blks),
2405 ATTR_LIST(inode_goal),
2406 ATTR_LIST(mb_stats),
2407 ATTR_LIST(mb_max_to_scan),
2408 ATTR_LIST(mb_min_to_scan),
2409 ATTR_LIST(mb_order2_req),
2410 ATTR_LIST(mb_stream_req),
2411 ATTR_LIST(mb_group_prealloc),
2412 ATTR_LIST(max_writeback_mb_bump),
2416 static ssize_t ext4_attr_show(struct kobject *kobj,
2417 struct attribute *attr, char *buf)
2419 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2421 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2423 return a->show ? a->show(a, sbi, buf) : 0;
2426 static ssize_t ext4_attr_store(struct kobject *kobj,
2427 struct attribute *attr,
2428 const char *buf, size_t len)
2430 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2432 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2434 return a->store ? a->store(a, sbi, buf, len) : 0;
2437 static void ext4_sb_release(struct kobject *kobj)
2439 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2441 complete(&sbi->s_kobj_unregister);
2445 static const struct sysfs_ops ext4_attr_ops = {
2446 .show = ext4_attr_show,
2447 .store = ext4_attr_store,
2450 static struct kobj_type ext4_ktype = {
2451 .default_attrs = ext4_attrs,
2452 .sysfs_ops = &ext4_attr_ops,
2453 .release = ext4_sb_release,
2457 * Check whether this filesystem can be mounted based on
2458 * the features present and the RDONLY/RDWR mount requested.
2459 * Returns 1 if this filesystem can be mounted as requested,
2460 * 0 if it cannot be.
2462 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2464 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2465 ext4_msg(sb, KERN_ERR,
2466 "Couldn't mount because of "
2467 "unsupported optional features (%x)",
2468 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2469 ~EXT4_FEATURE_INCOMPAT_SUPP));
2476 /* Check that feature set is OK for a read-write mount */
2477 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2478 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2479 "unsupported optional features (%x)",
2480 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2481 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2485 * Large file size enabled file system can only be mounted
2486 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2488 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2489 if (sizeof(blkcnt_t) < sizeof(u64)) {
2490 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2491 "cannot be mounted RDWR without "
2500 * This function is called once a day if we have errors logged
2501 * on the file system
2503 static void print_daily_error_info(unsigned long arg)
2505 struct super_block *sb = (struct super_block *) arg;
2506 struct ext4_sb_info *sbi;
2507 struct ext4_super_block *es;
2512 if (es->s_error_count)
2513 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2514 le32_to_cpu(es->s_error_count));
2515 if (es->s_first_error_time) {
2516 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2517 sb->s_id, le32_to_cpu(es->s_first_error_time),
2518 (int) sizeof(es->s_first_error_func),
2519 es->s_first_error_func,
2520 le32_to_cpu(es->s_first_error_line));
2521 if (es->s_first_error_ino)
2522 printk(": inode %u",
2523 le32_to_cpu(es->s_first_error_ino));
2524 if (es->s_first_error_block)
2525 printk(": block %llu", (unsigned long long)
2526 le64_to_cpu(es->s_first_error_block));
2529 if (es->s_last_error_time) {
2530 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2531 sb->s_id, le32_to_cpu(es->s_last_error_time),
2532 (int) sizeof(es->s_last_error_func),
2533 es->s_last_error_func,
2534 le32_to_cpu(es->s_last_error_line));
2535 if (es->s_last_error_ino)
2536 printk(": inode %u",
2537 le32_to_cpu(es->s_last_error_ino));
2538 if (es->s_last_error_block)
2539 printk(": block %llu", (unsigned long long)
2540 le64_to_cpu(es->s_last_error_block));
2543 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2546 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2547 __releases(kernel_lock)
2548 __acquires(kernel_lock)
2550 char *orig_data = kstrdup(data, GFP_KERNEL);
2551 struct buffer_head *bh;
2552 struct ext4_super_block *es = NULL;
2553 struct ext4_sb_info *sbi;
2555 ext4_fsblk_t sb_block = get_sb_block(&data);
2556 ext4_fsblk_t logical_sb_block;
2557 unsigned long offset = 0;
2558 unsigned long journal_devnum = 0;
2559 unsigned long def_mount_opts;
2565 unsigned int db_count;
2567 int needs_recovery, has_huge_files;
2570 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2572 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2576 sbi->s_blockgroup_lock =
2577 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2578 if (!sbi->s_blockgroup_lock) {
2582 sb->s_fs_info = sbi;
2583 sbi->s_mount_opt = 0;
2584 sbi->s_resuid = EXT4_DEF_RESUID;
2585 sbi->s_resgid = EXT4_DEF_RESGID;
2586 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2587 sbi->s_sb_block = sb_block;
2588 if (sb->s_bdev->bd_part)
2589 sbi->s_sectors_written_start =
2590 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2594 /* Cleanup superblock name */
2595 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2599 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2601 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2606 * The ext4 superblock will not be buffer aligned for other than 1kB
2607 * block sizes. We need to calculate the offset from buffer start.
2609 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2610 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2611 offset = do_div(logical_sb_block, blocksize);
2613 logical_sb_block = sb_block;
2616 if (!(bh = sb_bread(sb, logical_sb_block))) {
2617 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2621 * Note: s_es must be initialized as soon as possible because
2622 * some ext4 macro-instructions depend on its value
2624 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2626 sb->s_magic = le16_to_cpu(es->s_magic);
2627 if (sb->s_magic != EXT4_SUPER_MAGIC)
2629 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2631 /* Set defaults before we parse the mount options */
2632 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2633 if (def_mount_opts & EXT4_DEFM_DEBUG)
2634 set_opt(sbi->s_mount_opt, DEBUG);
2635 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2636 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2638 set_opt(sbi->s_mount_opt, GRPID);
2640 if (def_mount_opts & EXT4_DEFM_UID16)
2641 set_opt(sbi->s_mount_opt, NO_UID32);
2642 #ifdef CONFIG_EXT4_FS_XATTR
2643 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2644 set_opt(sbi->s_mount_opt, XATTR_USER);
2646 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2647 if (def_mount_opts & EXT4_DEFM_ACL)
2648 set_opt(sbi->s_mount_opt, POSIX_ACL);
2650 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2651 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2652 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2653 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2654 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2655 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2657 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2658 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2659 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2660 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2662 set_opt(sbi->s_mount_opt, ERRORS_RO);
2663 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
2664 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
2665 if (def_mount_opts & EXT4_DEFM_DISCARD)
2666 set_opt(sbi->s_mount_opt, DISCARD);
2668 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2669 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2670 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2671 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2672 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2674 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
2675 set_opt(sbi->s_mount_opt, BARRIER);
2678 * enable delayed allocation by default
2679 * Use -o nodelalloc to turn it off
2681 if (!IS_EXT3_SB(sb) &&
2682 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
2683 set_opt(sbi->s_mount_opt, DELALLOC);
2685 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
2686 &journal_devnum, &journal_ioprio, NULL, 0)) {
2687 ext4_msg(sb, KERN_WARNING,
2688 "failed to parse options in superblock: %s",
2689 sbi->s_es->s_mount_opts);
2691 if (!parse_options((char *) data, sb, &journal_devnum,
2692 &journal_ioprio, NULL, 0))
2695 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2696 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2698 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2699 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2700 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2701 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2702 ext4_msg(sb, KERN_WARNING,
2703 "feature flags set on rev 0 fs, "
2704 "running e2fsck is recommended");
2707 * Check feature flags regardless of the revision level, since we
2708 * previously didn't change the revision level when setting the flags,
2709 * so there is a chance incompat flags are set on a rev 0 filesystem.
2711 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2714 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2716 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2717 blocksize > EXT4_MAX_BLOCK_SIZE) {
2718 ext4_msg(sb, KERN_ERR,
2719 "Unsupported filesystem blocksize %d", blocksize);
2723 if (sb->s_blocksize != blocksize) {
2724 /* Validate the filesystem blocksize */
2725 if (!sb_set_blocksize(sb, blocksize)) {
2726 ext4_msg(sb, KERN_ERR, "bad block size %d",
2732 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2733 offset = do_div(logical_sb_block, blocksize);
2734 bh = sb_bread(sb, logical_sb_block);
2736 ext4_msg(sb, KERN_ERR,
2737 "Can't read superblock on 2nd try");
2740 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2742 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2743 ext4_msg(sb, KERN_ERR,
2744 "Magic mismatch, very weird!");
2749 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2750 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2751 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2753 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2755 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2756 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2757 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2759 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2760 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2761 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2762 (!is_power_of_2(sbi->s_inode_size)) ||
2763 (sbi->s_inode_size > blocksize)) {
2764 ext4_msg(sb, KERN_ERR,
2765 "unsupported inode size: %d",
2769 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2770 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2773 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2774 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2775 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2776 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2777 !is_power_of_2(sbi->s_desc_size)) {
2778 ext4_msg(sb, KERN_ERR,
2779 "unsupported descriptor size %lu",
2784 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2786 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2787 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2788 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2791 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2792 if (sbi->s_inodes_per_block == 0)
2794 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2795 sbi->s_inodes_per_block;
2796 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2798 sbi->s_mount_state = le16_to_cpu(es->s_state);
2799 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2800 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2802 for (i = 0; i < 4; i++)
2803 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2804 sbi->s_def_hash_version = es->s_def_hash_version;
2805 i = le32_to_cpu(es->s_flags);
2806 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2807 sbi->s_hash_unsigned = 3;
2808 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2809 #ifdef __CHAR_UNSIGNED__
2810 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2811 sbi->s_hash_unsigned = 3;
2813 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2818 if (sbi->s_blocks_per_group > blocksize * 8) {
2819 ext4_msg(sb, KERN_ERR,
2820 "#blocks per group too big: %lu",
2821 sbi->s_blocks_per_group);
2824 if (sbi->s_inodes_per_group > blocksize * 8) {
2825 ext4_msg(sb, KERN_ERR,
2826 "#inodes per group too big: %lu",
2827 sbi->s_inodes_per_group);
2832 * Test whether we have more sectors than will fit in sector_t,
2833 * and whether the max offset is addressable by the page cache.
2835 if ((ext4_blocks_count(es) >
2836 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2837 (ext4_blocks_count(es) >
2838 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2839 ext4_msg(sb, KERN_ERR, "filesystem"
2840 " too large to mount safely on this system");
2841 if (sizeof(sector_t) < 8)
2842 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2847 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2850 /* check blocks count against device size */
2851 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2852 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2853 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2854 "exceeds size of device (%llu blocks)",
2855 ext4_blocks_count(es), blocks_count);
2860 * It makes no sense for the first data block to be beyond the end
2861 * of the filesystem.
2863 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2864 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2865 "block %u is beyond end of filesystem (%llu)",
2866 le32_to_cpu(es->s_first_data_block),
2867 ext4_blocks_count(es));
2870 blocks_count = (ext4_blocks_count(es) -
2871 le32_to_cpu(es->s_first_data_block) +
2872 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2873 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2874 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2875 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2876 "(block count %llu, first data block %u, "
2877 "blocks per group %lu)", sbi->s_groups_count,
2878 ext4_blocks_count(es),
2879 le32_to_cpu(es->s_first_data_block),
2880 EXT4_BLOCKS_PER_GROUP(sb));
2883 sbi->s_groups_count = blocks_count;
2884 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2885 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2886 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2887 EXT4_DESC_PER_BLOCK(sb);
2888 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2890 if (sbi->s_group_desc == NULL) {
2891 ext4_msg(sb, KERN_ERR, "not enough memory");
2895 #ifdef CONFIG_PROC_FS
2897 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2900 bgl_lock_init(sbi->s_blockgroup_lock);
2902 for (i = 0; i < db_count; i++) {
2903 block = descriptor_loc(sb, logical_sb_block, i);
2904 sbi->s_group_desc[i] = sb_bread(sb, block);
2905 if (!sbi->s_group_desc[i]) {
2906 ext4_msg(sb, KERN_ERR,
2907 "can't read group descriptor %d", i);
2912 if (!ext4_check_descriptors(sb)) {
2913 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2916 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2917 if (!ext4_fill_flex_info(sb)) {
2918 ext4_msg(sb, KERN_ERR,
2919 "unable to initialize "
2920 "flex_bg meta info!");
2924 sbi->s_gdb_count = db_count;
2925 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2926 spin_lock_init(&sbi->s_next_gen_lock);
2928 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2929 ext4_count_free_blocks(sb));
2931 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2932 ext4_count_free_inodes(sb));
2935 err = percpu_counter_init(&sbi->s_dirs_counter,
2936 ext4_count_dirs(sb));
2939 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2942 ext4_msg(sb, KERN_ERR, "insufficient memory");
2946 sbi->s_stripe = ext4_get_stripe_size(sbi);
2947 sbi->s_max_writeback_mb_bump = 128;
2950 * set up enough so that it can read an inode
2952 if (!test_opt(sb, NOLOAD) &&
2953 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2954 sb->s_op = &ext4_sops;
2956 sb->s_op = &ext4_nojournal_sops;
2957 sb->s_export_op = &ext4_export_ops;
2958 sb->s_xattr = ext4_xattr_handlers;
2960 sb->s_qcop = &ext4_qctl_operations;
2961 sb->dq_op = &ext4_quota_operations;
2963 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2964 mutex_init(&sbi->s_orphan_lock);
2965 mutex_init(&sbi->s_resize_lock);
2969 needs_recovery = (es->s_last_orphan != 0 ||
2970 EXT4_HAS_INCOMPAT_FEATURE(sb,
2971 EXT4_FEATURE_INCOMPAT_RECOVER));
2974 * The first inode we look at is the journal inode. Don't try
2975 * root first: it may be modified in the journal!
2977 if (!test_opt(sb, NOLOAD) &&
2978 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2979 if (ext4_load_journal(sb, es, journal_devnum))
2981 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2982 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2983 ext4_msg(sb, KERN_ERR, "required journal recovery "
2984 "suppressed and not mounted read-only");
2985 goto failed_mount_wq;
2987 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2988 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2989 sbi->s_journal = NULL;
2994 if (ext4_blocks_count(es) > 0xffffffffULL &&
2995 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2996 JBD2_FEATURE_INCOMPAT_64BIT)) {
2997 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2998 goto failed_mount_wq;
3001 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3002 jbd2_journal_set_features(sbi->s_journal,
3003 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3004 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3005 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3006 jbd2_journal_set_features(sbi->s_journal,
3007 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3008 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3009 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3011 jbd2_journal_clear_features(sbi->s_journal,
3012 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3013 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3016 /* We have now updated the journal if required, so we can
3017 * validate the data journaling mode. */
3018 switch (test_opt(sb, DATA_FLAGS)) {
3020 /* No mode set, assume a default based on the journal
3021 * capabilities: ORDERED_DATA if the journal can
3022 * cope, else JOURNAL_DATA
3024 if (jbd2_journal_check_available_features
3025 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3026 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3028 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3031 case EXT4_MOUNT_ORDERED_DATA:
3032 case EXT4_MOUNT_WRITEBACK_DATA:
3033 if (!jbd2_journal_check_available_features
3034 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3035 ext4_msg(sb, KERN_ERR, "Journal does not support "
3036 "requested data journaling mode");
3037 goto failed_mount_wq;
3042 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3045 * The journal may have updated the bg summary counts, so we
3046 * need to update the global counters.
3048 percpu_counter_set(&sbi->s_freeblocks_counter,
3049 ext4_count_free_blocks(sb));
3050 percpu_counter_set(&sbi->s_freeinodes_counter,
3051 ext4_count_free_inodes(sb));
3052 percpu_counter_set(&sbi->s_dirs_counter,
3053 ext4_count_dirs(sb));
3054 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3057 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3058 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3059 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3060 goto failed_mount_wq;
3064 * The jbd2_journal_load will have done any necessary log recovery,
3065 * so we can safely mount the rest of the filesystem now.
3068 root = ext4_iget(sb, EXT4_ROOT_INO);
3070 ext4_msg(sb, KERN_ERR, "get root inode failed");
3071 ret = PTR_ERR(root);
3074 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3076 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3079 sb->s_root = d_alloc_root(root);
3081 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3087 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3089 /* determine the minimum size of new large inodes, if present */
3090 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3091 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3092 EXT4_GOOD_OLD_INODE_SIZE;
3093 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3094 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3095 if (sbi->s_want_extra_isize <
3096 le16_to_cpu(es->s_want_extra_isize))
3097 sbi->s_want_extra_isize =
3098 le16_to_cpu(es->s_want_extra_isize);
3099 if (sbi->s_want_extra_isize <
3100 le16_to_cpu(es->s_min_extra_isize))
3101 sbi->s_want_extra_isize =
3102 le16_to_cpu(es->s_min_extra_isize);
3105 /* Check if enough inode space is available */
3106 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3107 sbi->s_inode_size) {
3108 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3109 EXT4_GOOD_OLD_INODE_SIZE;
3110 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3114 if (test_opt(sb, DELALLOC) &&
3115 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3116 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3117 "requested data journaling mode");
3118 clear_opt(sbi->s_mount_opt, DELALLOC);
3120 if (test_opt(sb, DIOREAD_NOLOCK)) {
3121 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3122 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3123 "option - requested data journaling mode");
3124 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3126 if (sb->s_blocksize < PAGE_SIZE) {
3127 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3128 "option - block size is too small");
3129 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3133 err = ext4_setup_system_zone(sb);
3135 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3141 err = ext4_mb_init(sb, needs_recovery);
3143 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3148 sbi->s_kobj.kset = ext4_kset;
3149 init_completion(&sbi->s_kobj_unregister);
3150 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3153 ext4_mb_release(sb);
3154 ext4_ext_release(sb);
3158 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3159 ext4_orphan_cleanup(sb, es);
3160 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3161 if (needs_recovery) {
3162 ext4_msg(sb, KERN_INFO, "recovery complete");
3163 ext4_mark_recovery_complete(sb, es);
3165 if (EXT4_SB(sb)->s_journal) {
3166 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3167 descr = " journalled data mode";
3168 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3169 descr = " ordered data mode";
3171 descr = " writeback data mode";
3173 descr = "out journal";
3175 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3176 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3177 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3179 init_timer(&sbi->s_err_report);
3180 sbi->s_err_report.function = print_daily_error_info;
3181 sbi->s_err_report.data = (unsigned long) sb;
3182 if (es->s_error_count)
3183 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3191 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3195 ext4_msg(sb, KERN_ERR, "mount failed");
3196 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3198 ext4_release_system_zone(sb);
3199 if (sbi->s_journal) {
3200 jbd2_journal_destroy(sbi->s_journal);
3201 sbi->s_journal = NULL;
3204 if (sbi->s_flex_groups) {
3205 if (is_vmalloc_addr(sbi->s_flex_groups))
3206 vfree(sbi->s_flex_groups);
3208 kfree(sbi->s_flex_groups);
3210 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3211 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3212 percpu_counter_destroy(&sbi->s_dirs_counter);
3213 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3215 for (i = 0; i < db_count; i++)
3216 brelse(sbi->s_group_desc[i]);
3217 kfree(sbi->s_group_desc);
3220 remove_proc_entry(sb->s_id, ext4_proc_root);
3223 for (i = 0; i < MAXQUOTAS; i++)
3224 kfree(sbi->s_qf_names[i]);
3226 ext4_blkdev_remove(sbi);
3229 sb->s_fs_info = NULL;
3230 kfree(sbi->s_blockgroup_lock);
3239 * Setup any per-fs journal parameters now. We'll do this both on
3240 * initial mount, once the journal has been initialised but before we've
3241 * done any recovery; and again on any subsequent remount.
3243 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3245 struct ext4_sb_info *sbi = EXT4_SB(sb);
3247 journal->j_commit_interval = sbi->s_commit_interval;
3248 journal->j_min_batch_time = sbi->s_min_batch_time;
3249 journal->j_max_batch_time = sbi->s_max_batch_time;
3251 write_lock(&journal->j_state_lock);
3252 if (test_opt(sb, BARRIER))
3253 journal->j_flags |= JBD2_BARRIER;
3255 journal->j_flags &= ~JBD2_BARRIER;
3256 if (test_opt(sb, DATA_ERR_ABORT))
3257 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3259 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3260 write_unlock(&journal->j_state_lock);
3263 static journal_t *ext4_get_journal(struct super_block *sb,
3264 unsigned int journal_inum)
3266 struct inode *journal_inode;
3269 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3271 /* First, test for the existence of a valid inode on disk. Bad
3272 * things happen if we iget() an unused inode, as the subsequent
3273 * iput() will try to delete it. */
3275 journal_inode = ext4_iget(sb, journal_inum);
3276 if (IS_ERR(journal_inode)) {
3277 ext4_msg(sb, KERN_ERR, "no journal found");
3280 if (!journal_inode->i_nlink) {
3281 make_bad_inode(journal_inode);
3282 iput(journal_inode);
3283 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3287 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3288 journal_inode, journal_inode->i_size);
3289 if (!S_ISREG(journal_inode->i_mode)) {
3290 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3291 iput(journal_inode);
3295 journal = jbd2_journal_init_inode(journal_inode);
3297 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3298 iput(journal_inode);
3301 journal->j_private = sb;
3302 ext4_init_journal_params(sb, journal);
3306 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3309 struct buffer_head *bh;
3313 int hblock, blocksize;
3314 ext4_fsblk_t sb_block;
3315 unsigned long offset;
3316 struct ext4_super_block *es;
3317 struct block_device *bdev;
3319 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3321 bdev = ext4_blkdev_get(j_dev, sb);
3325 if (bd_claim(bdev, sb)) {
3326 ext4_msg(sb, KERN_ERR,
3327 "failed to claim external journal device");
3328 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3332 blocksize = sb->s_blocksize;
3333 hblock = bdev_logical_block_size(bdev);
3334 if (blocksize < hblock) {
3335 ext4_msg(sb, KERN_ERR,
3336 "blocksize too small for journal device");
3340 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3341 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3342 set_blocksize(bdev, blocksize);
3343 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3344 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3345 "external journal");
3349 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3350 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3351 !(le32_to_cpu(es->s_feature_incompat) &
3352 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3353 ext4_msg(sb, KERN_ERR, "external journal has "
3359 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3360 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3365 len = ext4_blocks_count(es);
3366 start = sb_block + 1;
3367 brelse(bh); /* we're done with the superblock */
3369 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3370 start, len, blocksize);
3372 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3375 journal->j_private = sb;
3376 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3377 wait_on_buffer(journal->j_sb_buffer);
3378 if (!buffer_uptodate(journal->j_sb_buffer)) {
3379 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3382 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3383 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3384 "user (unsupported) - %d",
3385 be32_to_cpu(journal->j_superblock->s_nr_users));
3388 EXT4_SB(sb)->journal_bdev = bdev;
3389 ext4_init_journal_params(sb, journal);
3393 jbd2_journal_destroy(journal);
3395 ext4_blkdev_put(bdev);
3399 static int ext4_load_journal(struct super_block *sb,
3400 struct ext4_super_block *es,
3401 unsigned long journal_devnum)
3404 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3407 int really_read_only;
3409 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3411 if (journal_devnum &&
3412 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3413 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3414 "numbers have changed");
3415 journal_dev = new_decode_dev(journal_devnum);
3417 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3419 really_read_only = bdev_read_only(sb->s_bdev);
3422 * Are we loading a blank journal or performing recovery after a
3423 * crash? For recovery, we need to check in advance whether we
3424 * can get read-write access to the device.
3426 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3427 if (sb->s_flags & MS_RDONLY) {
3428 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3429 "required on readonly filesystem");
3430 if (really_read_only) {
3431 ext4_msg(sb, KERN_ERR, "write access "
3432 "unavailable, cannot proceed");
3435 ext4_msg(sb, KERN_INFO, "write access will "
3436 "be enabled during recovery");
3440 if (journal_inum && journal_dev) {
3441 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3442 "and inode journals!");
3447 if (!(journal = ext4_get_journal(sb, journal_inum)))
3450 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3454 if (!(journal->j_flags & JBD2_BARRIER))
3455 ext4_msg(sb, KERN_INFO, "barriers disabled");
3457 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3458 err = jbd2_journal_update_format(journal);
3460 ext4_msg(sb, KERN_ERR, "error updating journal");
3461 jbd2_journal_destroy(journal);
3466 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3467 err = jbd2_journal_wipe(journal, !really_read_only);
3469 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3471 memcpy(save, ((char *) es) +
3472 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3473 err = jbd2_journal_load(journal);
3475 memcpy(((char *) es) + EXT4_S_ERR_START,
3476 save, EXT4_S_ERR_LEN);
3481 ext4_msg(sb, KERN_ERR, "error loading journal");
3482 jbd2_journal_destroy(journal);
3486 EXT4_SB(sb)->s_journal = journal;
3487 ext4_clear_journal_err(sb, es);
3489 if (journal_devnum &&
3490 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3491 es->s_journal_dev = cpu_to_le32(journal_devnum);
3493 /* Make sure we flush the recovery flag to disk. */
3494 ext4_commit_super(sb, 1);
3500 static int ext4_commit_super(struct super_block *sb, int sync)
3502 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3503 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3508 if (buffer_write_io_error(sbh)) {
3510 * Oh, dear. A previous attempt to write the
3511 * superblock failed. This could happen because the
3512 * USB device was yanked out. Or it could happen to
3513 * be a transient write error and maybe the block will
3514 * be remapped. Nothing we can do but to retry the
3515 * write and hope for the best.
3517 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3518 "superblock detected");
3519 clear_buffer_write_io_error(sbh);
3520 set_buffer_uptodate(sbh);
3523 * If the file system is mounted read-only, don't update the
3524 * superblock write time. This avoids updating the superblock
3525 * write time when we are mounting the root file system
3526 * read/only but we need to replay the journal; at that point,
3527 * for people who are east of GMT and who make their clock
3528 * tick in localtime for Windows bug-for-bug compatibility,
3529 * the clock is set in the future, and this will cause e2fsck
3530 * to complain and force a full file system check.
3532 if (!(sb->s_flags & MS_RDONLY))
3533 es->s_wtime = cpu_to_le32(get_seconds());
3534 if (sb->s_bdev->bd_part)
3535 es->s_kbytes_written =
3536 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3537 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3538 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3540 es->s_kbytes_written =
3541 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3542 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3543 &EXT4_SB(sb)->s_freeblocks_counter));
3544 es->s_free_inodes_count =
3545 cpu_to_le32(percpu_counter_sum_positive(
3546 &EXT4_SB(sb)->s_freeinodes_counter));
3548 BUFFER_TRACE(sbh, "marking dirty");
3549 mark_buffer_dirty(sbh);
3551 error = sync_dirty_buffer(sbh);
3555 error = buffer_write_io_error(sbh);
3557 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3559 clear_buffer_write_io_error(sbh);
3560 set_buffer_uptodate(sbh);
3567 * Have we just finished recovery? If so, and if we are mounting (or
3568 * remounting) the filesystem readonly, then we will end up with a
3569 * consistent fs on disk. Record that fact.
3571 static void ext4_mark_recovery_complete(struct super_block *sb,
3572 struct ext4_super_block *es)
3574 journal_t *journal = EXT4_SB(sb)->s_journal;
3576 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3577 BUG_ON(journal != NULL);
3580 jbd2_journal_lock_updates(journal);
3581 if (jbd2_journal_flush(journal) < 0)
3584 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3585 sb->s_flags & MS_RDONLY) {
3586 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3587 ext4_commit_super(sb, 1);
3591 jbd2_journal_unlock_updates(journal);
3595 * If we are mounting (or read-write remounting) a filesystem whose journal
3596 * has recorded an error from a previous lifetime, move that error to the
3597 * main filesystem now.
3599 static void ext4_clear_journal_err(struct super_block *sb,
3600 struct ext4_super_block *es)
3606 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3608 journal = EXT4_SB(sb)->s_journal;
3611 * Now check for any error status which may have been recorded in the
3612 * journal by a prior ext4_error() or ext4_abort()
3615 j_errno = jbd2_journal_errno(journal);
3619 errstr = ext4_decode_error(sb, j_errno, nbuf);
3620 ext4_warning(sb, "Filesystem error recorded "
3621 "from previous mount: %s", errstr);
3622 ext4_warning(sb, "Marking fs in need of filesystem check.");
3624 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3625 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3626 ext4_commit_super(sb, 1);
3628 jbd2_journal_clear_err(journal);
3633 * Force the running and committing transactions to commit,
3634 * and wait on the commit.
3636 int ext4_force_commit(struct super_block *sb)
3641 if (sb->s_flags & MS_RDONLY)
3644 journal = EXT4_SB(sb)->s_journal;
3646 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3647 ret = ext4_journal_force_commit(journal);
3653 static void ext4_write_super(struct super_block *sb)
3656 ext4_commit_super(sb, 1);
3660 static int ext4_sync_fs(struct super_block *sb, int wait)
3664 struct ext4_sb_info *sbi = EXT4_SB(sb);
3666 trace_ext4_sync_fs(sb, wait);
3667 flush_workqueue(sbi->dio_unwritten_wq);
3668 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3670 jbd2_log_wait_commit(sbi->s_journal, target);
3676 * LVM calls this function before a (read-only) snapshot is created. This
3677 * gives us a chance to flush the journal completely and mark the fs clean.
3679 static int ext4_freeze(struct super_block *sb)
3684 if (sb->s_flags & MS_RDONLY)
3687 journal = EXT4_SB(sb)->s_journal;
3689 /* Now we set up the journal barrier. */
3690 jbd2_journal_lock_updates(journal);
3693 * Don't clear the needs_recovery flag if we failed to flush
3696 error = jbd2_journal_flush(journal);
3700 /* Journal blocked and flushed, clear needs_recovery flag. */
3701 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3702 error = ext4_commit_super(sb, 1);
3704 /* we rely on s_frozen to stop further updates */
3705 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3710 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3711 * flag here, even though the filesystem is not technically dirty yet.
3713 static int ext4_unfreeze(struct super_block *sb)
3715 if (sb->s_flags & MS_RDONLY)
3719 /* Reset the needs_recovery flag before the fs is unlocked. */
3720 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3721 ext4_commit_super(sb, 1);
3726 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3728 struct ext4_super_block *es;
3729 struct ext4_sb_info *sbi = EXT4_SB(sb);
3730 ext4_fsblk_t n_blocks_count = 0;
3731 unsigned long old_sb_flags;
3732 struct ext4_mount_options old_opts;
3733 int enable_quota = 0;
3735 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3740 char *orig_data = kstrdup(data, GFP_KERNEL);
3744 /* Store the original options */
3746 old_sb_flags = sb->s_flags;
3747 old_opts.s_mount_opt = sbi->s_mount_opt;
3748 old_opts.s_resuid = sbi->s_resuid;
3749 old_opts.s_resgid = sbi->s_resgid;
3750 old_opts.s_commit_interval = sbi->s_commit_interval;
3751 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3752 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3754 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3755 for (i = 0; i < MAXQUOTAS; i++)
3756 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3758 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3759 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3762 * Allow the "check" option to be passed as a remount option.
3764 if (!parse_options(data, sb, NULL, &journal_ioprio,
3765 &n_blocks_count, 1)) {
3770 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3771 ext4_abort(sb, "Abort forced by user");
3773 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3774 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3778 if (sbi->s_journal) {
3779 ext4_init_journal_params(sb, sbi->s_journal);
3780 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3783 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3784 n_blocks_count > ext4_blocks_count(es)) {
3785 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3790 if (*flags & MS_RDONLY) {
3791 err = dquot_suspend(sb, -1);
3796 * First of all, the unconditional stuff we have to do
3797 * to disable replay of the journal when we next remount
3799 sb->s_flags |= MS_RDONLY;
3802 * OK, test if we are remounting a valid rw partition
3803 * readonly, and if so set the rdonly flag and then
3804 * mark the partition as valid again.
3806 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3807 (sbi->s_mount_state & EXT4_VALID_FS))
3808 es->s_state = cpu_to_le16(sbi->s_mount_state);
3811 ext4_mark_recovery_complete(sb, es);
3813 /* Make sure we can mount this feature set readwrite */
3814 if (!ext4_feature_set_ok(sb, 0)) {
3819 * Make sure the group descriptor checksums
3820 * are sane. If they aren't, refuse to remount r/w.
3822 for (g = 0; g < sbi->s_groups_count; g++) {
3823 struct ext4_group_desc *gdp =
3824 ext4_get_group_desc(sb, g, NULL);
3826 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3827 ext4_msg(sb, KERN_ERR,
3828 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3829 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3830 le16_to_cpu(gdp->bg_checksum));
3837 * If we have an unprocessed orphan list hanging
3838 * around from a previously readonly bdev mount,
3839 * require a full umount/remount for now.
3841 if (es->s_last_orphan) {
3842 ext4_msg(sb, KERN_WARNING, "Couldn't "
3843 "remount RDWR because of unprocessed "
3844 "orphan inode list. Please "
3845 "umount/remount instead");
3851 * Mounting a RDONLY partition read-write, so reread
3852 * and store the current valid flag. (It may have
3853 * been changed by e2fsck since we originally mounted
3857 ext4_clear_journal_err(sb, es);
3858 sbi->s_mount_state = le16_to_cpu(es->s_state);
3859 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3861 if (!ext4_setup_super(sb, es, 0))
3862 sb->s_flags &= ~MS_RDONLY;
3866 ext4_setup_system_zone(sb);
3867 if (sbi->s_journal == NULL)
3868 ext4_commit_super(sb, 1);
3871 /* Release old quota file names */
3872 for (i = 0; i < MAXQUOTAS; i++)
3873 if (old_opts.s_qf_names[i] &&
3874 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3875 kfree(old_opts.s_qf_names[i]);
3880 dquot_resume(sb, -1);
3882 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3887 sb->s_flags = old_sb_flags;
3888 sbi->s_mount_opt = old_opts.s_mount_opt;
3889 sbi->s_resuid = old_opts.s_resuid;
3890 sbi->s_resgid = old_opts.s_resgid;
3891 sbi->s_commit_interval = old_opts.s_commit_interval;
3892 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3893 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3895 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3896 for (i = 0; i < MAXQUOTAS; i++) {
3897 if (sbi->s_qf_names[i] &&
3898 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3899 kfree(sbi->s_qf_names[i]);
3900 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3909 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3911 struct super_block *sb = dentry->d_sb;
3912 struct ext4_sb_info *sbi = EXT4_SB(sb);
3913 struct ext4_super_block *es = sbi->s_es;
3916 if (test_opt(sb, MINIX_DF)) {
3917 sbi->s_overhead_last = 0;
3918 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3919 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3920 ext4_fsblk_t overhead = 0;
3923 * Compute the overhead (FS structures). This is constant
3924 * for a given filesystem unless the number of block groups
3925 * changes so we cache the previous value until it does.
3929 * All of the blocks before first_data_block are
3932 overhead = le32_to_cpu(es->s_first_data_block);
3935 * Add the overhead attributed to the superblock and
3936 * block group descriptors. If the sparse superblocks
3937 * feature is turned on, then not all groups have this.
3939 for (i = 0; i < ngroups; i++) {
3940 overhead += ext4_bg_has_super(sb, i) +
3941 ext4_bg_num_gdb(sb, i);
3946 * Every block group has an inode bitmap, a block
3947 * bitmap, and an inode table.
3949 overhead += ngroups * (2 + sbi->s_itb_per_group);
3950 sbi->s_overhead_last = overhead;
3952 sbi->s_blocks_last = ext4_blocks_count(es);
3955 buf->f_type = EXT4_SUPER_MAGIC;
3956 buf->f_bsize = sb->s_blocksize;
3957 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3958 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3959 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3960 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3961 if (buf->f_bfree < ext4_r_blocks_count(es))
3963 buf->f_files = le32_to_cpu(es->s_inodes_count);
3964 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3965 buf->f_namelen = EXT4_NAME_LEN;
3966 fsid = le64_to_cpup((void *)es->s_uuid) ^
3967 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3968 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3969 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3974 /* Helper function for writing quotas on sync - we need to start transaction
3975 * before quota file is locked for write. Otherwise the are possible deadlocks:
3976 * Process 1 Process 2
3977 * ext4_create() quota_sync()
3978 * jbd2_journal_start() write_dquot()
3979 * dquot_initialize() down(dqio_mutex)
3980 * down(dqio_mutex) jbd2_journal_start()
3986 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3988 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3991 static int ext4_write_dquot(struct dquot *dquot)
3995 struct inode *inode;
3997 inode = dquot_to_inode(dquot);
3998 handle = ext4_journal_start(inode,
3999 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4001 return PTR_ERR(handle);
4002 ret = dquot_commit(dquot);
4003 err = ext4_journal_stop(handle);
4009 static int ext4_acquire_dquot(struct dquot *dquot)
4014 handle = ext4_journal_start(dquot_to_inode(dquot),
4015 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4017 return PTR_ERR(handle);
4018 ret = dquot_acquire(dquot);
4019 err = ext4_journal_stop(handle);
4025 static int ext4_release_dquot(struct dquot *dquot)
4030 handle = ext4_journal_start(dquot_to_inode(dquot),
4031 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4032 if (IS_ERR(handle)) {
4033 /* Release dquot anyway to avoid endless cycle in dqput() */
4034 dquot_release(dquot);
4035 return PTR_ERR(handle);
4037 ret = dquot_release(dquot);
4038 err = ext4_journal_stop(handle);
4044 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4046 /* Are we journaling quotas? */
4047 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4048 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4049 dquot_mark_dquot_dirty(dquot);
4050 return ext4_write_dquot(dquot);
4052 return dquot_mark_dquot_dirty(dquot);
4056 static int ext4_write_info(struct super_block *sb, int type)
4061 /* Data block + inode block */
4062 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4064 return PTR_ERR(handle);
4065 ret = dquot_commit_info(sb, type);
4066 err = ext4_journal_stop(handle);
4073 * Turn on quotas during mount time - we need to find
4074 * the quota file and such...
4076 static int ext4_quota_on_mount(struct super_block *sb, int type)
4078 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4079 EXT4_SB(sb)->s_jquota_fmt, type);
4083 * Standard function to be called on quota_on
4085 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4091 if (!test_opt(sb, QUOTA))
4094 err = kern_path(name, LOOKUP_FOLLOW, &path);
4098 /* Quotafile not on the same filesystem? */
4099 if (path.mnt->mnt_sb != sb) {
4103 /* Journaling quota? */
4104 if (EXT4_SB(sb)->s_qf_names[type]) {
4105 /* Quotafile not in fs root? */
4106 if (path.dentry->d_parent != sb->s_root)
4107 ext4_msg(sb, KERN_WARNING,
4108 "Quota file not on filesystem root. "
4109 "Journaled quota will not work");
4113 * When we journal data on quota file, we have to flush journal to see
4114 * all updates to the file when we bypass pagecache...
4116 if (EXT4_SB(sb)->s_journal &&
4117 ext4_should_journal_data(path.dentry->d_inode)) {
4119 * We don't need to lock updates but journal_flush() could
4120 * otherwise be livelocked...
4122 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4123 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4124 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4131 err = dquot_quota_on_path(sb, type, format_id, &path);
4136 static int ext4_quota_off(struct super_block *sb, int type)
4138 /* Force all delayed allocation blocks to be allocated */
4139 if (test_opt(sb, DELALLOC)) {
4140 down_read(&sb->s_umount);
4141 sync_filesystem(sb);
4142 up_read(&sb->s_umount);
4145 return dquot_quota_off(sb, type);
4148 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4149 * acquiring the locks... As quota files are never truncated and quota code
4150 * itself serializes the operations (and noone else should touch the files)
4151 * we don't have to be afraid of races */
4152 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4153 size_t len, loff_t off)
4155 struct inode *inode = sb_dqopt(sb)->files[type];
4156 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4158 int offset = off & (sb->s_blocksize - 1);
4161 struct buffer_head *bh;
4162 loff_t i_size = i_size_read(inode);
4166 if (off+len > i_size)
4169 while (toread > 0) {
4170 tocopy = sb->s_blocksize - offset < toread ?
4171 sb->s_blocksize - offset : toread;
4172 bh = ext4_bread(NULL, inode, blk, 0, &err);
4175 if (!bh) /* A hole? */
4176 memset(data, 0, tocopy);
4178 memcpy(data, bh->b_data+offset, tocopy);
4188 /* Write to quotafile (we know the transaction is already started and has
4189 * enough credits) */
4190 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4191 const char *data, size_t len, loff_t off)
4193 struct inode *inode = sb_dqopt(sb)->files[type];
4194 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4196 int offset = off & (sb->s_blocksize - 1);
4197 struct buffer_head *bh;
4198 handle_t *handle = journal_current_handle();
4200 if (EXT4_SB(sb)->s_journal && !handle) {
4201 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4202 " cancelled because transaction is not started",
4203 (unsigned long long)off, (unsigned long long)len);
4207 * Since we account only one data block in transaction credits,
4208 * then it is impossible to cross a block boundary.
4210 if (sb->s_blocksize - offset < len) {
4211 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4212 " cancelled because not block aligned",
4213 (unsigned long long)off, (unsigned long long)len);
4217 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4218 bh = ext4_bread(handle, inode, blk, 1, &err);
4221 err = ext4_journal_get_write_access(handle, bh);
4227 memcpy(bh->b_data+offset, data, len);
4228 flush_dcache_page(bh->b_page);
4230 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4234 mutex_unlock(&inode->i_mutex);
4237 if (inode->i_size < off + len) {
4238 i_size_write(inode, off + len);
4239 EXT4_I(inode)->i_disksize = inode->i_size;
4241 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4242 ext4_mark_inode_dirty(handle, inode);
4243 mutex_unlock(&inode->i_mutex);
4249 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4250 const char *dev_name, void *data, struct vfsmount *mnt)
4252 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4255 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4256 static struct file_system_type ext2_fs_type = {
4257 .owner = THIS_MODULE,
4259 .get_sb = ext4_get_sb,
4260 .kill_sb = kill_block_super,
4261 .fs_flags = FS_REQUIRES_DEV,
4264 static inline void register_as_ext2(void)
4266 int err = register_filesystem(&ext2_fs_type);
4269 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4272 static inline void unregister_as_ext2(void)
4274 unregister_filesystem(&ext2_fs_type);
4276 MODULE_ALIAS("ext2");
4278 static inline void register_as_ext2(void) { }
4279 static inline void unregister_as_ext2(void) { }
4282 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4283 static inline void register_as_ext3(void)
4285 int err = register_filesystem(&ext3_fs_type);
4288 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4291 static inline void unregister_as_ext3(void)
4293 unregister_filesystem(&ext3_fs_type);
4295 MODULE_ALIAS("ext3");
4297 static inline void register_as_ext3(void) { }
4298 static inline void unregister_as_ext3(void) { }
4301 static struct file_system_type ext4_fs_type = {
4302 .owner = THIS_MODULE,
4304 .get_sb = ext4_get_sb,
4305 .kill_sb = kill_block_super,
4306 .fs_flags = FS_REQUIRES_DEV,
4309 static int __init init_ext4_fs(void)
4313 ext4_check_flag_values();
4314 err = init_ext4_system_zone();
4317 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4320 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4321 err = init_ext4_mballoc();
4325 err = init_ext4_xattr();
4328 err = init_inodecache();
4333 err = register_filesystem(&ext4_fs_type);
4338 unregister_as_ext2();
4339 unregister_as_ext3();
4340 destroy_inodecache();
4344 exit_ext4_mballoc();
4346 remove_proc_entry("fs/ext4", NULL);
4347 kset_unregister(ext4_kset);
4349 exit_ext4_system_zone();
4353 static void __exit exit_ext4_fs(void)
4355 unregister_as_ext2();
4356 unregister_as_ext3();
4357 unregister_filesystem(&ext4_fs_type);
4358 destroy_inodecache();
4360 exit_ext4_mballoc();
4361 remove_proc_entry("fs/ext4", NULL);
4362 kset_unregister(ext4_kset);
4363 exit_ext4_system_zone();
4366 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4367 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4368 MODULE_LICENSE("GPL");
4369 module_init(init_ext4_fs)
4370 module_exit(exit_ext4_fs)