2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type = {
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #define IS_EXT3_SB(sb) (0)
87 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
95 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
103 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le32_to_cpu(bg->bg_inode_table_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
111 __u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
119 __u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
127 __u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
135 __u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
143 void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
151 void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
159 void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
167 void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
175 void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
183 void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
191 void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
200 /* Just increment the non-pointer handle value */
201 static handle_t *ext4_get_nojournal(void)
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
209 handle = (handle_t *)ref_cnt;
211 current->journal_info = handle;
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t *handle)
219 unsigned long ref_cnt = (unsigned long)handle;
221 BUG_ON(ref_cnt == 0);
224 handle = (handle_t *)ref_cnt;
226 current->journal_info = handle;
230 * Wrappers for jbd2_journal_start/end.
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
237 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
244 vfs_check_frozen(sb, SB_FREEZE_TRANS);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal;
250 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, "Detected aborted journal");
252 return ERR_PTR(-EROFS);
254 return jbd2_journal_start(journal, nblocks);
256 return ext4_get_nojournal();
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
265 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
267 struct super_block *sb;
271 if (!ext4_handle_valid(handle)) {
272 ext4_put_nojournal(handle);
275 sb = handle->h_transaction->t_journal->j_private;
277 rc = jbd2_journal_stop(handle);
282 __ext4_std_error(sb, where, line, err);
286 void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
293 BUG_ON(!ext4_handle_valid(handle));
296 BUFFER_TRACE(bh, "abort");
301 if (is_handle_aborted(handle))
304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
307 jbd2_journal_abort_handle(handle);
310 static void __save_error_info(struct super_block *sb, const char *func,
313 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
315 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
316 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
317 es->s_last_error_time = cpu_to_le32(get_seconds());
318 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
319 es->s_last_error_line = cpu_to_le32(line);
320 if (!es->s_first_error_time) {
321 es->s_first_error_time = es->s_last_error_time;
322 strncpy(es->s_first_error_func, func,
323 sizeof(es->s_first_error_func));
324 es->s_first_error_line = cpu_to_le32(line);
325 es->s_first_error_ino = es->s_last_error_ino;
326 es->s_first_error_block = es->s_last_error_block;
329 * Start the daily error reporting function if it hasn't been
332 if (!es->s_error_count)
333 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
334 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
337 static void save_error_info(struct super_block *sb, const char *func,
340 __save_error_info(sb, func, line);
341 ext4_commit_super(sb, 1);
345 /* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
348 * On ext2, we can store the error state of the filesystem in the
349 * superblock. That is not possible on ext4, because we may have other
350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
355 * We'll just use the jbd2_journal_abort() error code to record an error in
356 * the journal instead. On recovery, the journal will complain about
357 * that error until we've noted it down and cleared it.
360 static void ext4_handle_error(struct super_block *sb)
362 if (sb->s_flags & MS_RDONLY)
365 if (!test_opt(sb, ERRORS_CONT)) {
366 journal_t *journal = EXT4_SB(sb)->s_journal;
368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
370 jbd2_journal_abort(journal, -EIO);
372 if (test_opt(sb, ERRORS_RO)) {
373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
374 sb->s_flags |= MS_RDONLY;
376 if (test_opt(sb, ERRORS_PANIC))
377 panic("EXT4-fs (device %s): panic forced after error\n",
381 void __ext4_error(struct super_block *sb, const char *function,
382 unsigned int line, const char *fmt, ...)
387 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
388 sb->s_id, function, line, current->comm);
393 ext4_handle_error(sb);
396 void ext4_error_inode(struct inode *inode, const char *function,
397 unsigned int line, ext4_fsblk_t block,
398 const char *fmt, ...)
401 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
403 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
404 es->s_last_error_block = cpu_to_le64(block);
405 save_error_info(inode->i_sb, function, line);
407 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
408 inode->i_sb->s_id, function, line, inode->i_ino);
410 printk("block %llu: ", block);
411 printk("comm %s: ", current->comm);
416 ext4_handle_error(inode->i_sb);
419 void ext4_error_file(struct file *file, const char *function,
420 unsigned int line, const char *fmt, ...)
423 struct ext4_super_block *es;
424 struct inode *inode = file->f_dentry->d_inode;
425 char pathname[80], *path;
427 es = EXT4_SB(inode->i_sb)->s_es;
428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 save_error_info(inode->i_sb, function, line);
431 path = d_path(&(file->f_path), pathname, sizeof(pathname));
435 "EXT4-fs error (device %s): %s:%d: inode #%lu "
436 "(comm %s path %s): ",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 current->comm, path);
443 ext4_handle_error(inode->i_sb);
446 static const char *ext4_decode_error(struct super_block *sb, int errno,
453 errstr = "IO failure";
456 errstr = "Out of memory";
459 if (!sb || (EXT4_SB(sb)->s_journal &&
460 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
461 errstr = "Journal has aborted";
463 errstr = "Readonly filesystem";
466 /* If the caller passed in an extra buffer for unknown
467 * errors, textualise them now. Else we just return
470 /* Check for truncated error codes... */
471 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
480 /* __ext4_std_error decodes expected errors from journaling functions
481 * automatically and invokes the appropriate error response. */
483 void __ext4_std_error(struct super_block *sb, const char *function,
484 unsigned int line, int errno)
489 /* Special case: if the error is EROFS, and we're not already
490 * inside a transaction, then there's really no point in logging
492 if (errno == -EROFS && journal_current_handle() == NULL &&
493 (sb->s_flags & MS_RDONLY))
496 errstr = ext4_decode_error(sb, errno, nbuf);
497 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
498 sb->s_id, function, line, errstr);
499 save_error_info(sb, function, line);
501 ext4_handle_error(sb);
505 * ext4_abort is a much stronger failure handler than ext4_error. The
506 * abort function may be used to deal with unrecoverable failures such
507 * as journal IO errors or ENOMEM at a critical moment in log management.
509 * We unconditionally force the filesystem into an ABORT|READONLY state,
510 * unless the error response on the fs has been set to panic in which
511 * case we take the easy way out and panic immediately.
514 void __ext4_abort(struct super_block *sb, const char *function,
515 unsigned int line, const char *fmt, ...)
519 save_error_info(sb, function, line);
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
527 if ((sb->s_flags & MS_RDONLY) == 0) {
528 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
529 sb->s_flags |= MS_RDONLY;
530 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
531 if (EXT4_SB(sb)->s_journal)
532 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
533 save_error_info(sb, function, line);
535 if (test_opt(sb, ERRORS_PANIC))
536 panic("EXT4-fs panic from previous error\n");
539 void ext4_msg (struct super_block * sb, const char *prefix,
540 const char *fmt, ...)
545 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
551 void __ext4_warning(struct super_block *sb, const char *function,
552 unsigned int line, const char *fmt, ...)
557 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
558 sb->s_id, function, line);
564 void __ext4_grp_locked_error(const char *function, unsigned int line,
565 struct super_block *sb, ext4_group_t grp,
566 unsigned long ino, ext4_fsblk_t block,
567 const char *fmt, ...)
572 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
574 es->s_last_error_ino = cpu_to_le32(ino);
575 es->s_last_error_block = cpu_to_le64(block);
576 __save_error_info(sb, function, line);
578 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
579 sb->s_id, function, line, grp);
581 printk("inode %lu: ", ino);
583 printk("block %llu:", (unsigned long long) block);
588 if (test_opt(sb, ERRORS_CONT)) {
589 ext4_commit_super(sb, 0);
593 ext4_unlock_group(sb, grp);
594 ext4_handle_error(sb);
596 * We only get here in the ERRORS_RO case; relocking the group
597 * may be dangerous, but nothing bad will happen since the
598 * filesystem will have already been marked read/only and the
599 * journal has been aborted. We return 1 as a hint to callers
600 * who might what to use the return value from
601 * ext4_grp_locked_error() to distinguish beween the
602 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
603 * aggressively from the ext4 function in question, with a
604 * more appropriate error code.
606 ext4_lock_group(sb, grp);
610 void ext4_update_dynamic_rev(struct super_block *sb)
612 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
614 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
618 "updating to rev %d because of new feature flag, "
619 "running e2fsck is recommended",
622 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
623 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
624 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
625 /* leave es->s_feature_*compat flags alone */
626 /* es->s_uuid will be set by e2fsck if empty */
629 * The rest of the superblock fields should be zero, and if not it
630 * means they are likely already in use, so leave them alone. We
631 * can leave it up to e2fsck to clean up any inconsistencies there.
636 * Open the external journal device
638 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
640 struct block_device *bdev;
641 char b[BDEVNAME_SIZE];
643 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
649 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
650 __bdevname(dev, b), PTR_ERR(bdev));
655 * Release the journal device
657 static int ext4_blkdev_put(struct block_device *bdev)
660 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
663 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
665 struct block_device *bdev;
668 bdev = sbi->journal_bdev;
670 ret = ext4_blkdev_put(bdev);
671 sbi->journal_bdev = NULL;
676 static inline struct inode *orphan_list_entry(struct list_head *l)
678 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
681 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
685 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
686 le32_to_cpu(sbi->s_es->s_last_orphan));
688 printk(KERN_ERR "sb_info orphan list:\n");
689 list_for_each(l, &sbi->s_orphan) {
690 struct inode *inode = orphan_list_entry(l);
692 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
693 inode->i_sb->s_id, inode->i_ino, inode,
694 inode->i_mode, inode->i_nlink,
699 static void ext4_put_super(struct super_block *sb)
701 struct ext4_sb_info *sbi = EXT4_SB(sb);
702 struct ext4_super_block *es = sbi->s_es;
705 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
707 flush_workqueue(sbi->dio_unwritten_wq);
708 destroy_workqueue(sbi->dio_unwritten_wq);
713 ext4_commit_super(sb, 1);
715 if (sbi->s_journal) {
716 err = jbd2_journal_destroy(sbi->s_journal);
717 sbi->s_journal = NULL;
719 ext4_abort(sb, "Couldn't clean up the journal");
722 ext4_release_system_zone(sb);
724 ext4_ext_release(sb);
725 ext4_xattr_put_super(sb);
727 if (!(sb->s_flags & MS_RDONLY)) {
728 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
729 es->s_state = cpu_to_le16(sbi->s_mount_state);
730 ext4_commit_super(sb, 1);
733 remove_proc_entry(sb->s_id, ext4_proc_root);
735 kobject_del(&sbi->s_kobj);
737 for (i = 0; i < sbi->s_gdb_count; i++)
738 brelse(sbi->s_group_desc[i]);
739 kfree(sbi->s_group_desc);
740 if (is_vmalloc_addr(sbi->s_flex_groups))
741 vfree(sbi->s_flex_groups);
743 kfree(sbi->s_flex_groups);
744 percpu_counter_destroy(&sbi->s_freeblocks_counter);
745 percpu_counter_destroy(&sbi->s_freeinodes_counter);
746 percpu_counter_destroy(&sbi->s_dirs_counter);
747 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
750 for (i = 0; i < MAXQUOTAS; i++)
751 kfree(sbi->s_qf_names[i]);
754 /* Debugging code just in case the in-memory inode orphan list
755 * isn't empty. The on-disk one can be non-empty if we've
756 * detected an error and taken the fs readonly, but the
757 * in-memory list had better be clean by this point. */
758 if (!list_empty(&sbi->s_orphan))
759 dump_orphan_list(sb, sbi);
760 J_ASSERT(list_empty(&sbi->s_orphan));
762 invalidate_bdev(sb->s_bdev);
763 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
765 * Invalidate the journal device's buffers. We don't want them
766 * floating about in memory - the physical journal device may
767 * hotswapped, and it breaks the `ro-after' testing code.
769 sync_blockdev(sbi->journal_bdev);
770 invalidate_bdev(sbi->journal_bdev);
771 ext4_blkdev_remove(sbi);
773 sb->s_fs_info = NULL;
775 * Now that we are completely done shutting down the
776 * superblock, we need to actually destroy the kobject.
780 kobject_put(&sbi->s_kobj);
781 wait_for_completion(&sbi->s_kobj_unregister);
782 kfree(sbi->s_blockgroup_lock);
786 static struct kmem_cache *ext4_inode_cachep;
789 * Called inside transaction, so use GFP_NOFS
791 static struct inode *ext4_alloc_inode(struct super_block *sb)
793 struct ext4_inode_info *ei;
795 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
799 ei->vfs_inode.i_version = 1;
800 ei->vfs_inode.i_data.writeback_index = 0;
801 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
802 INIT_LIST_HEAD(&ei->i_prealloc_list);
803 spin_lock_init(&ei->i_prealloc_lock);
805 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
806 * therefore it can be null here. Don't check it, just initialize
809 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
810 ei->i_reserved_data_blocks = 0;
811 ei->i_reserved_meta_blocks = 0;
812 ei->i_allocated_meta_blocks = 0;
813 ei->i_da_metadata_calc_len = 0;
814 ei->i_delalloc_reserved_flag = 0;
815 spin_lock_init(&(ei->i_block_reservation_lock));
817 ei->i_reserved_quota = 0;
819 INIT_LIST_HEAD(&ei->i_completed_io_list);
820 spin_lock_init(&ei->i_completed_io_lock);
821 ei->cur_aio_dio = NULL;
823 ei->i_datasync_tid = 0;
825 return &ei->vfs_inode;
828 static void ext4_destroy_inode(struct inode *inode)
830 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
831 ext4_msg(inode->i_sb, KERN_ERR,
832 "Inode %lu (%p): orphan list check failed!",
833 inode->i_ino, EXT4_I(inode));
834 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
835 EXT4_I(inode), sizeof(struct ext4_inode_info),
839 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
842 static void init_once(void *foo)
844 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
846 INIT_LIST_HEAD(&ei->i_orphan);
847 #ifdef CONFIG_EXT4_FS_XATTR
848 init_rwsem(&ei->xattr_sem);
850 init_rwsem(&ei->i_data_sem);
851 inode_init_once(&ei->vfs_inode);
854 static int init_inodecache(void)
856 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
857 sizeof(struct ext4_inode_info),
858 0, (SLAB_RECLAIM_ACCOUNT|
861 if (ext4_inode_cachep == NULL)
866 static void destroy_inodecache(void)
868 kmem_cache_destroy(ext4_inode_cachep);
871 static void ext4_clear_inode(struct inode *inode)
874 ext4_discard_preallocations(inode);
875 if (EXT4_JOURNAL(inode))
876 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
877 &EXT4_I(inode)->jinode);
880 static inline void ext4_show_quota_options(struct seq_file *seq,
881 struct super_block *sb)
883 #if defined(CONFIG_QUOTA)
884 struct ext4_sb_info *sbi = EXT4_SB(sb);
886 if (sbi->s_jquota_fmt) {
889 switch (sbi->s_jquota_fmt) {
900 seq_printf(seq, ",jqfmt=%s", fmtname);
903 if (sbi->s_qf_names[USRQUOTA])
904 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
906 if (sbi->s_qf_names[GRPQUOTA])
907 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
909 if (test_opt(sb, USRQUOTA))
910 seq_puts(seq, ",usrquota");
912 if (test_opt(sb, GRPQUOTA))
913 seq_puts(seq, ",grpquota");
919 * - it's set to a non-default value OR
920 * - if the per-sb default is different from the global default
922 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
925 unsigned long def_mount_opts;
926 struct super_block *sb = vfs->mnt_sb;
927 struct ext4_sb_info *sbi = EXT4_SB(sb);
928 struct ext4_super_block *es = sbi->s_es;
930 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
931 def_errors = le16_to_cpu(es->s_errors);
933 if (sbi->s_sb_block != 1)
934 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
935 if (test_opt(sb, MINIX_DF))
936 seq_puts(seq, ",minixdf");
937 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
938 seq_puts(seq, ",grpid");
939 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
940 seq_puts(seq, ",nogrpid");
941 if (sbi->s_resuid != EXT4_DEF_RESUID ||
942 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
943 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
945 if (sbi->s_resgid != EXT4_DEF_RESGID ||
946 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
947 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
949 if (test_opt(sb, ERRORS_RO)) {
950 if (def_errors == EXT4_ERRORS_PANIC ||
951 def_errors == EXT4_ERRORS_CONTINUE) {
952 seq_puts(seq, ",errors=remount-ro");
955 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
956 seq_puts(seq, ",errors=continue");
957 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
958 seq_puts(seq, ",errors=panic");
959 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
960 seq_puts(seq, ",nouid32");
961 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
962 seq_puts(seq, ",debug");
963 if (test_opt(sb, OLDALLOC))
964 seq_puts(seq, ",oldalloc");
965 #ifdef CONFIG_EXT4_FS_XATTR
966 if (test_opt(sb, XATTR_USER) &&
967 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
968 seq_puts(seq, ",user_xattr");
969 if (!test_opt(sb, XATTR_USER) &&
970 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
971 seq_puts(seq, ",nouser_xattr");
974 #ifdef CONFIG_EXT4_FS_POSIX_ACL
975 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
976 seq_puts(seq, ",acl");
977 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
978 seq_puts(seq, ",noacl");
980 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
981 seq_printf(seq, ",commit=%u",
982 (unsigned) (sbi->s_commit_interval / HZ));
984 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
985 seq_printf(seq, ",min_batch_time=%u",
986 (unsigned) sbi->s_min_batch_time);
988 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
989 seq_printf(seq, ",max_batch_time=%u",
990 (unsigned) sbi->s_min_batch_time);
994 * We're changing the default of barrier mount option, so
995 * let's always display its mount state so it's clear what its
998 seq_puts(seq, ",barrier=");
999 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1000 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1001 seq_puts(seq, ",journal_async_commit");
1002 else if (test_opt(sb, JOURNAL_CHECKSUM))
1003 seq_puts(seq, ",journal_checksum");
1004 if (test_opt(sb, I_VERSION))
1005 seq_puts(seq, ",i_version");
1006 if (!test_opt(sb, DELALLOC))
1007 seq_puts(seq, ",nodelalloc");
1011 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1013 * journal mode get enabled in different ways
1014 * So just print the value even if we didn't specify it
1016 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1017 seq_puts(seq, ",data=journal");
1018 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1019 seq_puts(seq, ",data=ordered");
1020 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1021 seq_puts(seq, ",data=writeback");
1023 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1024 seq_printf(seq, ",inode_readahead_blks=%u",
1025 sbi->s_inode_readahead_blks);
1027 if (test_opt(sb, DATA_ERR_ABORT))
1028 seq_puts(seq, ",data_err=abort");
1030 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1031 seq_puts(seq, ",noauto_da_alloc");
1033 if (test_opt(sb, DISCARD))
1034 seq_puts(seq, ",discard");
1036 if (test_opt(sb, NOLOAD))
1037 seq_puts(seq, ",norecovery");
1039 if (test_opt(sb, DIOREAD_NOLOCK))
1040 seq_puts(seq, ",dioread_nolock");
1042 ext4_show_quota_options(seq, sb);
1047 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1048 u64 ino, u32 generation)
1050 struct inode *inode;
1052 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1053 return ERR_PTR(-ESTALE);
1054 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1055 return ERR_PTR(-ESTALE);
1057 /* iget isn't really right if the inode is currently unallocated!!
1059 * ext4_read_inode will return a bad_inode if the inode had been
1060 * deleted, so we should be safe.
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1065 inode = ext4_iget(sb, ino);
1067 return ERR_CAST(inode);
1068 if (generation && inode->i_generation != generation) {
1070 return ERR_PTR(-ESTALE);
1076 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type)
1079 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1083 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1084 int fh_len, int fh_type)
1086 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1096 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1099 journal_t *journal = EXT4_SB(sb)->s_journal;
1101 WARN_ON(PageChecked(page));
1102 if (!page_has_buffers(page))
1105 return jbd2_journal_try_to_free_buffers(journal, page,
1106 wait & ~__GFP_WAIT);
1107 return try_to_free_buffers(page);
1111 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1112 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1114 static int ext4_write_dquot(struct dquot *dquot);
1115 static int ext4_acquire_dquot(struct dquot *dquot);
1116 static int ext4_release_dquot(struct dquot *dquot);
1117 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1118 static int ext4_write_info(struct super_block *sb, int type);
1119 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1121 static int ext4_quota_off(struct super_block *sb, int type);
1122 static int ext4_quota_on_mount(struct super_block *sb, int type);
1123 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1124 size_t len, loff_t off);
1125 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1126 const char *data, size_t len, loff_t off);
1128 static const struct dquot_operations ext4_quota_operations = {
1130 .get_reserved_space = ext4_get_reserved_space,
1132 .write_dquot = ext4_write_dquot,
1133 .acquire_dquot = ext4_acquire_dquot,
1134 .release_dquot = ext4_release_dquot,
1135 .mark_dirty = ext4_mark_dquot_dirty,
1136 .write_info = ext4_write_info,
1137 .alloc_dquot = dquot_alloc,
1138 .destroy_dquot = dquot_destroy,
1141 static const struct quotactl_ops ext4_qctl_operations = {
1142 .quota_on = ext4_quota_on,
1143 .quota_off = ext4_quota_off,
1144 .quota_sync = dquot_quota_sync,
1145 .get_info = dquot_get_dqinfo,
1146 .set_info = dquot_set_dqinfo,
1147 .get_dqblk = dquot_get_dqblk,
1148 .set_dqblk = dquot_set_dqblk
1152 static const struct super_operations ext4_sops = {
1153 .alloc_inode = ext4_alloc_inode,
1154 .destroy_inode = ext4_destroy_inode,
1155 .write_inode = ext4_write_inode,
1156 .dirty_inode = ext4_dirty_inode,
1157 .delete_inode = ext4_delete_inode,
1158 .put_super = ext4_put_super,
1159 .sync_fs = ext4_sync_fs,
1160 .freeze_fs = ext4_freeze,
1161 .unfreeze_fs = ext4_unfreeze,
1162 .statfs = ext4_statfs,
1163 .remount_fs = ext4_remount,
1164 .clear_inode = ext4_clear_inode,
1165 .show_options = ext4_show_options,
1167 .quota_read = ext4_quota_read,
1168 .quota_write = ext4_quota_write,
1170 .bdev_try_to_free_page = bdev_try_to_free_page,
1173 static const struct super_operations ext4_nojournal_sops = {
1174 .alloc_inode = ext4_alloc_inode,
1175 .destroy_inode = ext4_destroy_inode,
1176 .write_inode = ext4_write_inode,
1177 .dirty_inode = ext4_dirty_inode,
1178 .delete_inode = ext4_delete_inode,
1179 .write_super = ext4_write_super,
1180 .put_super = ext4_put_super,
1181 .statfs = ext4_statfs,
1182 .remount_fs = ext4_remount,
1183 .clear_inode = ext4_clear_inode,
1184 .show_options = ext4_show_options,
1186 .quota_read = ext4_quota_read,
1187 .quota_write = ext4_quota_write,
1189 .bdev_try_to_free_page = bdev_try_to_free_page,
1192 static const struct export_operations ext4_export_ops = {
1193 .fh_to_dentry = ext4_fh_to_dentry,
1194 .fh_to_parent = ext4_fh_to_parent,
1195 .get_parent = ext4_get_parent,
1199 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1200 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1201 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1202 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1203 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1204 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1205 Opt_journal_update, Opt_journal_dev,
1206 Opt_journal_checksum, Opt_journal_async_commit,
1207 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1208 Opt_data_err_abort, Opt_data_err_ignore,
1209 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1210 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1211 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1212 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1213 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1214 Opt_block_validity, Opt_noblock_validity,
1215 Opt_inode_readahead_blks, Opt_journal_ioprio,
1216 Opt_dioread_nolock, Opt_dioread_lock,
1217 Opt_discard, Opt_nodiscard,
1220 static const match_table_t tokens = {
1221 {Opt_bsd_df, "bsddf"},
1222 {Opt_minix_df, "minixdf"},
1223 {Opt_grpid, "grpid"},
1224 {Opt_grpid, "bsdgroups"},
1225 {Opt_nogrpid, "nogrpid"},
1226 {Opt_nogrpid, "sysvgroups"},
1227 {Opt_resgid, "resgid=%u"},
1228 {Opt_resuid, "resuid=%u"},
1230 {Opt_err_cont, "errors=continue"},
1231 {Opt_err_panic, "errors=panic"},
1232 {Opt_err_ro, "errors=remount-ro"},
1233 {Opt_nouid32, "nouid32"},
1234 {Opt_debug, "debug"},
1235 {Opt_oldalloc, "oldalloc"},
1236 {Opt_orlov, "orlov"},
1237 {Opt_user_xattr, "user_xattr"},
1238 {Opt_nouser_xattr, "nouser_xattr"},
1240 {Opt_noacl, "noacl"},
1241 {Opt_noload, "noload"},
1242 {Opt_noload, "norecovery"},
1245 {Opt_commit, "commit=%u"},
1246 {Opt_min_batch_time, "min_batch_time=%u"},
1247 {Opt_max_batch_time, "max_batch_time=%u"},
1248 {Opt_journal_update, "journal=update"},
1249 {Opt_journal_dev, "journal_dev=%u"},
1250 {Opt_journal_checksum, "journal_checksum"},
1251 {Opt_journal_async_commit, "journal_async_commit"},
1252 {Opt_abort, "abort"},
1253 {Opt_data_journal, "data=journal"},
1254 {Opt_data_ordered, "data=ordered"},
1255 {Opt_data_writeback, "data=writeback"},
1256 {Opt_data_err_abort, "data_err=abort"},
1257 {Opt_data_err_ignore, "data_err=ignore"},
1258 {Opt_offusrjquota, "usrjquota="},
1259 {Opt_usrjquota, "usrjquota=%s"},
1260 {Opt_offgrpjquota, "grpjquota="},
1261 {Opt_grpjquota, "grpjquota=%s"},
1262 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1263 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1264 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1265 {Opt_grpquota, "grpquota"},
1266 {Opt_noquota, "noquota"},
1267 {Opt_quota, "quota"},
1268 {Opt_usrquota, "usrquota"},
1269 {Opt_barrier, "barrier=%u"},
1270 {Opt_barrier, "barrier"},
1271 {Opt_nobarrier, "nobarrier"},
1272 {Opt_i_version, "i_version"},
1273 {Opt_stripe, "stripe=%u"},
1274 {Opt_resize, "resize"},
1275 {Opt_delalloc, "delalloc"},
1276 {Opt_nodelalloc, "nodelalloc"},
1277 {Opt_block_validity, "block_validity"},
1278 {Opt_noblock_validity, "noblock_validity"},
1279 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1280 {Opt_journal_ioprio, "journal_ioprio=%u"},
1281 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1282 {Opt_auto_da_alloc, "auto_da_alloc"},
1283 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1284 {Opt_dioread_nolock, "dioread_nolock"},
1285 {Opt_dioread_lock, "dioread_lock"},
1286 {Opt_discard, "discard"},
1287 {Opt_nodiscard, "nodiscard"},
1291 static ext4_fsblk_t get_sb_block(void **data)
1293 ext4_fsblk_t sb_block;
1294 char *options = (char *) *data;
1296 if (!options || strncmp(options, "sb=", 3) != 0)
1297 return 1; /* Default location */
1300 /* TODO: use simple_strtoll with >32bit ext4 */
1301 sb_block = simple_strtoul(options, &options, 0);
1302 if (*options && *options != ',') {
1303 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1307 if (*options == ',')
1309 *data = (void *) options;
1314 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1315 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1316 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1319 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1321 struct ext4_sb_info *sbi = EXT4_SB(sb);
1324 if (sb_any_quota_loaded(sb) &&
1325 !sbi->s_qf_names[qtype]) {
1326 ext4_msg(sb, KERN_ERR,
1327 "Cannot change journaled "
1328 "quota options when quota turned on");
1331 qname = match_strdup(args);
1333 ext4_msg(sb, KERN_ERR,
1334 "Not enough memory for storing quotafile name");
1337 if (sbi->s_qf_names[qtype] &&
1338 strcmp(sbi->s_qf_names[qtype], qname)) {
1339 ext4_msg(sb, KERN_ERR,
1340 "%s quota file already specified", QTYPE2NAME(qtype));
1344 sbi->s_qf_names[qtype] = qname;
1345 if (strchr(sbi->s_qf_names[qtype], '/')) {
1346 ext4_msg(sb, KERN_ERR,
1347 "quotafile must be on filesystem root");
1348 kfree(sbi->s_qf_names[qtype]);
1349 sbi->s_qf_names[qtype] = NULL;
1352 set_opt(sbi->s_mount_opt, QUOTA);
1356 static int clear_qf_name(struct super_block *sb, int qtype)
1359 struct ext4_sb_info *sbi = EXT4_SB(sb);
1361 if (sb_any_quota_loaded(sb) &&
1362 sbi->s_qf_names[qtype]) {
1363 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1364 " when quota turned on");
1368 * The space will be released later when all options are confirmed
1371 sbi->s_qf_names[qtype] = NULL;
1376 static int parse_options(char *options, struct super_block *sb,
1377 unsigned long *journal_devnum,
1378 unsigned int *journal_ioprio,
1379 ext4_fsblk_t *n_blocks_count, int is_remount)
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1383 substring_t args[MAX_OPT_ARGS];
1393 while ((p = strsep(&options, ",")) != NULL) {
1399 * Initialize args struct so we know whether arg was
1400 * found; some options take optional arguments.
1402 args[0].to = args[0].from = 0;
1403 token = match_token(p, tokens, args);
1406 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1407 clear_opt(sbi->s_mount_opt, MINIX_DF);
1410 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1411 set_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, GRPID);
1420 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1421 clear_opt(sbi->s_mount_opt, GRPID);
1425 if (match_int(&args[0], &option))
1427 sbi->s_resuid = option;
1430 if (match_int(&args[0], &option))
1432 sbi->s_resgid = option;
1435 /* handled by get_sb_block() instead of here */
1436 /* *sb_block = match_int(&args[0]); */
1439 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1440 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1441 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1444 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1445 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1446 set_opt(sbi->s_mount_opt, ERRORS_RO);
1449 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1450 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1451 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1454 set_opt(sbi->s_mount_opt, NO_UID32);
1457 set_opt(sbi->s_mount_opt, DEBUG);
1460 set_opt(sbi->s_mount_opt, OLDALLOC);
1463 clear_opt(sbi->s_mount_opt, OLDALLOC);
1465 #ifdef CONFIG_EXT4_FS_XATTR
1466 case Opt_user_xattr:
1467 set_opt(sbi->s_mount_opt, XATTR_USER);
1469 case Opt_nouser_xattr:
1470 clear_opt(sbi->s_mount_opt, XATTR_USER);
1473 case Opt_user_xattr:
1474 case Opt_nouser_xattr:
1475 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1478 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1480 set_opt(sbi->s_mount_opt, POSIX_ACL);
1483 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1488 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1491 case Opt_journal_update:
1493 /* Eventually we will want to be able to create
1494 a journal file here. For now, only allow the
1495 user to specify an existing inode to be the
1498 ext4_msg(sb, KERN_ERR,
1499 "Cannot specify journal on remount");
1502 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1504 case Opt_journal_dev:
1506 ext4_msg(sb, KERN_ERR,
1507 "Cannot specify journal on remount");
1510 if (match_int(&args[0], &option))
1512 *journal_devnum = option;
1514 case Opt_journal_checksum:
1515 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1517 case Opt_journal_async_commit:
1518 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1519 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1522 set_opt(sbi->s_mount_opt, NOLOAD);
1525 if (match_int(&args[0], &option))
1530 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1531 sbi->s_commit_interval = HZ * option;
1533 case Opt_max_batch_time:
1534 if (match_int(&args[0], &option))
1539 option = EXT4_DEF_MAX_BATCH_TIME;
1540 sbi->s_max_batch_time = option;
1542 case Opt_min_batch_time:
1543 if (match_int(&args[0], &option))
1547 sbi->s_min_batch_time = option;
1549 case Opt_data_journal:
1550 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1552 case Opt_data_ordered:
1553 data_opt = EXT4_MOUNT_ORDERED_DATA;
1555 case Opt_data_writeback:
1556 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1559 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1560 ext4_msg(sb, KERN_ERR,
1561 "Cannot change data mode on remount");
1565 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1566 sbi->s_mount_opt |= data_opt;
1569 case Opt_data_err_abort:
1570 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1572 case Opt_data_err_ignore:
1573 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1577 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1581 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1584 case Opt_offusrjquota:
1585 if (!clear_qf_name(sb, USRQUOTA))
1588 case Opt_offgrpjquota:
1589 if (!clear_qf_name(sb, GRPQUOTA))
1593 case Opt_jqfmt_vfsold:
1594 qfmt = QFMT_VFS_OLD;
1596 case Opt_jqfmt_vfsv0:
1599 case Opt_jqfmt_vfsv1:
1602 if (sb_any_quota_loaded(sb) &&
1603 sbi->s_jquota_fmt != qfmt) {
1604 ext4_msg(sb, KERN_ERR, "Cannot change "
1605 "journaled quota options when "
1609 sbi->s_jquota_fmt = qfmt;
1613 set_opt(sbi->s_mount_opt, QUOTA);
1614 set_opt(sbi->s_mount_opt, USRQUOTA);
1617 set_opt(sbi->s_mount_opt, QUOTA);
1618 set_opt(sbi->s_mount_opt, GRPQUOTA);
1621 if (sb_any_quota_loaded(sb)) {
1622 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1623 "options when quota turned on");
1626 clear_opt(sbi->s_mount_opt, QUOTA);
1627 clear_opt(sbi->s_mount_opt, USRQUOTA);
1628 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1634 ext4_msg(sb, KERN_ERR,
1635 "quota options not supported");
1639 case Opt_offusrjquota:
1640 case Opt_offgrpjquota:
1641 case Opt_jqfmt_vfsold:
1642 case Opt_jqfmt_vfsv0:
1643 case Opt_jqfmt_vfsv1:
1644 ext4_msg(sb, KERN_ERR,
1645 "journaled quota options not supported");
1651 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1654 clear_opt(sbi->s_mount_opt, BARRIER);
1658 if (match_int(&args[0], &option))
1661 option = 1; /* No argument, default to 1 */
1663 set_opt(sbi->s_mount_opt, BARRIER);
1665 clear_opt(sbi->s_mount_opt, BARRIER);
1671 ext4_msg(sb, KERN_ERR,
1672 "resize option only available "
1676 if (match_int(&args[0], &option) != 0)
1678 *n_blocks_count = option;
1681 ext4_msg(sb, KERN_WARNING,
1682 "Ignoring deprecated nobh option");
1685 ext4_msg(sb, KERN_WARNING,
1686 "Ignoring deprecated bh option");
1689 set_opt(sbi->s_mount_opt, I_VERSION);
1690 sb->s_flags |= MS_I_VERSION;
1692 case Opt_nodelalloc:
1693 clear_opt(sbi->s_mount_opt, DELALLOC);
1696 if (match_int(&args[0], &option))
1700 sbi->s_stripe = option;
1703 set_opt(sbi->s_mount_opt, DELALLOC);
1705 case Opt_block_validity:
1706 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1708 case Opt_noblock_validity:
1709 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1711 case Opt_inode_readahead_blks:
1712 if (match_int(&args[0], &option))
1714 if (option < 0 || option > (1 << 30))
1716 if (!is_power_of_2(option)) {
1717 ext4_msg(sb, KERN_ERR,
1718 "EXT4-fs: inode_readahead_blks"
1719 " must be a power of 2");
1722 sbi->s_inode_readahead_blks = option;
1724 case Opt_journal_ioprio:
1725 if (match_int(&args[0], &option))
1727 if (option < 0 || option > 7)
1729 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1732 case Opt_noauto_da_alloc:
1733 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1735 case Opt_auto_da_alloc:
1737 if (match_int(&args[0], &option))
1740 option = 1; /* No argument, default to 1 */
1742 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1744 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1747 set_opt(sbi->s_mount_opt, DISCARD);
1750 clear_opt(sbi->s_mount_opt, DISCARD);
1752 case Opt_dioread_nolock:
1753 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1755 case Opt_dioread_lock:
1756 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1759 ext4_msg(sb, KERN_ERR,
1760 "Unrecognized mount option \"%s\" "
1761 "or missing value", p);
1766 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1767 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1768 clear_opt(sbi->s_mount_opt, USRQUOTA);
1770 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1771 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1773 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1774 ext4_msg(sb, KERN_ERR, "old and new quota "
1779 if (!sbi->s_jquota_fmt) {
1780 ext4_msg(sb, KERN_ERR, "journaled quota format "
1785 if (sbi->s_jquota_fmt) {
1786 ext4_msg(sb, KERN_ERR, "journaled quota format "
1787 "specified with no journaling "
1796 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1799 struct ext4_sb_info *sbi = EXT4_SB(sb);
1802 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1803 ext4_msg(sb, KERN_ERR, "revision level too high, "
1804 "forcing read-only mode");
1809 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1810 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1811 "running e2fsck is recommended");
1812 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1813 ext4_msg(sb, KERN_WARNING,
1814 "warning: mounting fs with errors, "
1815 "running e2fsck is recommended");
1816 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1817 le16_to_cpu(es->s_mnt_count) >=
1818 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1819 ext4_msg(sb, KERN_WARNING,
1820 "warning: maximal mount count reached, "
1821 "running e2fsck is recommended");
1822 else if (le32_to_cpu(es->s_checkinterval) &&
1823 (le32_to_cpu(es->s_lastcheck) +
1824 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1825 ext4_msg(sb, KERN_WARNING,
1826 "warning: checktime reached, "
1827 "running e2fsck is recommended");
1828 if (!sbi->s_journal)
1829 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1830 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1831 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1832 le16_add_cpu(&es->s_mnt_count, 1);
1833 es->s_mtime = cpu_to_le32(get_seconds());
1834 ext4_update_dynamic_rev(sb);
1836 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1838 ext4_commit_super(sb, 1);
1839 if (test_opt(sb, DEBUG))
1840 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1841 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1843 sbi->s_groups_count,
1844 EXT4_BLOCKS_PER_GROUP(sb),
1845 EXT4_INODES_PER_GROUP(sb),
1851 static int ext4_fill_flex_info(struct super_block *sb)
1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 struct ext4_group_desc *gdp = NULL;
1855 ext4_group_t flex_group_count;
1856 ext4_group_t flex_group;
1857 int groups_per_flex = 0;
1861 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1862 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1864 if (groups_per_flex < 2) {
1865 sbi->s_log_groups_per_flex = 0;
1869 /* We allocate both existing and potentially added groups */
1870 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1871 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1872 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1873 size = flex_group_count * sizeof(struct flex_groups);
1874 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1875 if (sbi->s_flex_groups == NULL) {
1876 sbi->s_flex_groups = vmalloc(size);
1877 if (sbi->s_flex_groups)
1878 memset(sbi->s_flex_groups, 0, size);
1880 if (sbi->s_flex_groups == NULL) {
1881 ext4_msg(sb, KERN_ERR, "not enough memory for "
1882 "%u flex groups", flex_group_count);
1886 for (i = 0; i < sbi->s_groups_count; i++) {
1887 gdp = ext4_get_group_desc(sb, i, NULL);
1889 flex_group = ext4_flex_group(sbi, i);
1890 atomic_add(ext4_free_inodes_count(sb, gdp),
1891 &sbi->s_flex_groups[flex_group].free_inodes);
1892 atomic_add(ext4_free_blks_count(sb, gdp),
1893 &sbi->s_flex_groups[flex_group].free_blocks);
1894 atomic_add(ext4_used_dirs_count(sb, gdp),
1895 &sbi->s_flex_groups[flex_group].used_dirs);
1903 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1904 struct ext4_group_desc *gdp)
1908 if (sbi->s_es->s_feature_ro_compat &
1909 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1910 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1911 __le32 le_group = cpu_to_le32(block_group);
1913 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1914 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1915 crc = crc16(crc, (__u8 *)gdp, offset);
1916 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1917 /* for checksum of struct ext4_group_desc do the rest...*/
1918 if ((sbi->s_es->s_feature_incompat &
1919 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1920 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1921 crc = crc16(crc, (__u8 *)gdp + offset,
1922 le16_to_cpu(sbi->s_es->s_desc_size) -
1926 return cpu_to_le16(crc);
1929 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1930 struct ext4_group_desc *gdp)
1932 if ((sbi->s_es->s_feature_ro_compat &
1933 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1934 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1940 /* Called at mount-time, super-block is locked */
1941 static int ext4_check_descriptors(struct super_block *sb)
1943 struct ext4_sb_info *sbi = EXT4_SB(sb);
1944 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1945 ext4_fsblk_t last_block;
1946 ext4_fsblk_t block_bitmap;
1947 ext4_fsblk_t inode_bitmap;
1948 ext4_fsblk_t inode_table;
1949 int flexbg_flag = 0;
1952 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1955 ext4_debug("Checking group descriptors");
1957 for (i = 0; i < sbi->s_groups_count; i++) {
1958 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1960 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1961 last_block = ext4_blocks_count(sbi->s_es) - 1;
1963 last_block = first_block +
1964 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1966 block_bitmap = ext4_block_bitmap(sb, gdp);
1967 if (block_bitmap < first_block || block_bitmap > last_block) {
1968 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1969 "Block bitmap for group %u not in group "
1970 "(block %llu)!", i, block_bitmap);
1973 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1974 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1975 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1976 "Inode bitmap for group %u not in group "
1977 "(block %llu)!", i, inode_bitmap);
1980 inode_table = ext4_inode_table(sb, gdp);
1981 if (inode_table < first_block ||
1982 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1983 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1984 "Inode table for group %u not in group "
1985 "(block %llu)!", i, inode_table);
1988 ext4_lock_group(sb, i);
1989 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1990 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1991 "Checksum for group %u failed (%u!=%u)",
1992 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1993 gdp)), le16_to_cpu(gdp->bg_checksum));
1994 if (!(sb->s_flags & MS_RDONLY)) {
1995 ext4_unlock_group(sb, i);
1999 ext4_unlock_group(sb, i);
2001 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2004 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2005 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2009 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2010 * the superblock) which were deleted from all directories, but held open by
2011 * a process at the time of a crash. We walk the list and try to delete these
2012 * inodes at recovery time (only with a read-write filesystem).
2014 * In order to keep the orphan inode chain consistent during traversal (in
2015 * case of crash during recovery), we link each inode into the superblock
2016 * orphan list_head and handle it the same way as an inode deletion during
2017 * normal operation (which journals the operations for us).
2019 * We only do an iget() and an iput() on each inode, which is very safe if we
2020 * accidentally point at an in-use or already deleted inode. The worst that
2021 * can happen in this case is that we get a "bit already cleared" message from
2022 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2023 * e2fsck was run on this filesystem, and it must have already done the orphan
2024 * inode cleanup for us, so we can safely abort without any further action.
2026 static void ext4_orphan_cleanup(struct super_block *sb,
2027 struct ext4_super_block *es)
2029 unsigned int s_flags = sb->s_flags;
2030 int nr_orphans = 0, nr_truncates = 0;
2034 if (!es->s_last_orphan) {
2035 jbd_debug(4, "no orphan inodes to clean up\n");
2039 if (bdev_read_only(sb->s_bdev)) {
2040 ext4_msg(sb, KERN_ERR, "write access "
2041 "unavailable, skipping orphan cleanup");
2045 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2046 if (es->s_last_orphan)
2047 jbd_debug(1, "Errors on filesystem, "
2048 "clearing orphan list.\n");
2049 es->s_last_orphan = 0;
2050 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2054 if (s_flags & MS_RDONLY) {
2055 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2056 sb->s_flags &= ~MS_RDONLY;
2059 /* Needed for iput() to work correctly and not trash data */
2060 sb->s_flags |= MS_ACTIVE;
2061 /* Turn on quotas so that they are updated correctly */
2062 for (i = 0; i < MAXQUOTAS; i++) {
2063 if (EXT4_SB(sb)->s_qf_names[i]) {
2064 int ret = ext4_quota_on_mount(sb, i);
2066 ext4_msg(sb, KERN_ERR,
2067 "Cannot turn on journaled "
2068 "quota: error %d", ret);
2073 while (es->s_last_orphan) {
2074 struct inode *inode;
2076 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2077 if (IS_ERR(inode)) {
2078 es->s_last_orphan = 0;
2082 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2083 dquot_initialize(inode);
2084 if (inode->i_nlink) {
2085 ext4_msg(sb, KERN_DEBUG,
2086 "%s: truncating inode %lu to %lld bytes",
2087 __func__, inode->i_ino, inode->i_size);
2088 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2089 inode->i_ino, inode->i_size);
2090 ext4_truncate(inode);
2093 ext4_msg(sb, KERN_DEBUG,
2094 "%s: deleting unreferenced inode %lu",
2095 __func__, inode->i_ino);
2096 jbd_debug(2, "deleting unreferenced inode %lu\n",
2100 iput(inode); /* The delete magic happens here! */
2103 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2106 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2107 PLURAL(nr_orphans));
2109 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2110 PLURAL(nr_truncates));
2112 /* Turn quotas off */
2113 for (i = 0; i < MAXQUOTAS; i++) {
2114 if (sb_dqopt(sb)->files[i])
2115 dquot_quota_off(sb, i);
2118 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2122 * Maximal extent format file size.
2123 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2124 * extent format containers, within a sector_t, and within i_blocks
2125 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2126 * so that won't be a limiting factor.
2128 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2130 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2133 loff_t upper_limit = MAX_LFS_FILESIZE;
2135 /* small i_blocks in vfs inode? */
2136 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2138 * CONFIG_LBDAF is not enabled implies the inode
2139 * i_block represent total blocks in 512 bytes
2140 * 32 == size of vfs inode i_blocks * 8
2142 upper_limit = (1LL << 32) - 1;
2144 /* total blocks in file system block size */
2145 upper_limit >>= (blkbits - 9);
2146 upper_limit <<= blkbits;
2149 /* 32-bit extent-start container, ee_block */
2154 /* Sanity check against vm- & vfs- imposed limits */
2155 if (res > upper_limit)
2162 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2163 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2164 * We need to be 1 filesystem block less than the 2^48 sector limit.
2166 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2168 loff_t res = EXT4_NDIR_BLOCKS;
2171 /* This is calculated to be the largest file size for a dense, block
2172 * mapped file such that the file's total number of 512-byte sectors,
2173 * including data and all indirect blocks, does not exceed (2^48 - 1).
2175 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2176 * number of 512-byte sectors of the file.
2179 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2181 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2182 * the inode i_block field represents total file blocks in
2183 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2185 upper_limit = (1LL << 32) - 1;
2187 /* total blocks in file system block size */
2188 upper_limit >>= (bits - 9);
2192 * We use 48 bit ext4_inode i_blocks
2193 * With EXT4_HUGE_FILE_FL set the i_blocks
2194 * represent total number of blocks in
2195 * file system block size
2197 upper_limit = (1LL << 48) - 1;
2201 /* indirect blocks */
2203 /* double indirect blocks */
2204 meta_blocks += 1 + (1LL << (bits-2));
2205 /* tripple indirect blocks */
2206 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2208 upper_limit -= meta_blocks;
2209 upper_limit <<= bits;
2211 res += 1LL << (bits-2);
2212 res += 1LL << (2*(bits-2));
2213 res += 1LL << (3*(bits-2));
2215 if (res > upper_limit)
2218 if (res > MAX_LFS_FILESIZE)
2219 res = MAX_LFS_FILESIZE;
2224 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2225 ext4_fsblk_t logical_sb_block, int nr)
2227 struct ext4_sb_info *sbi = EXT4_SB(sb);
2228 ext4_group_t bg, first_meta_bg;
2231 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2233 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2235 return logical_sb_block + nr + 1;
2236 bg = sbi->s_desc_per_block * nr;
2237 if (ext4_bg_has_super(sb, bg))
2240 return (has_super + ext4_group_first_block_no(sb, bg));
2244 * ext4_get_stripe_size: Get the stripe size.
2245 * @sbi: In memory super block info
2247 * If we have specified it via mount option, then
2248 * use the mount option value. If the value specified at mount time is
2249 * greater than the blocks per group use the super block value.
2250 * If the super block value is greater than blocks per group return 0.
2251 * Allocator needs it be less than blocks per group.
2254 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2256 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2257 unsigned long stripe_width =
2258 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2260 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2261 return sbi->s_stripe;
2263 if (stripe_width <= sbi->s_blocks_per_group)
2264 return stripe_width;
2266 if (stride <= sbi->s_blocks_per_group)
2275 struct attribute attr;
2276 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2277 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2278 const char *, size_t);
2282 static int parse_strtoul(const char *buf,
2283 unsigned long max, unsigned long *value)
2287 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2288 endp = skip_spaces(endp);
2289 if (*endp || *value > max)
2295 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2296 struct ext4_sb_info *sbi,
2299 return snprintf(buf, PAGE_SIZE, "%llu\n",
2300 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2303 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2304 struct ext4_sb_info *sbi, char *buf)
2306 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2308 if (!sb->s_bdev->bd_part)
2309 return snprintf(buf, PAGE_SIZE, "0\n");
2310 return snprintf(buf, PAGE_SIZE, "%lu\n",
2311 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2312 sbi->s_sectors_written_start) >> 1);
2315 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2316 struct ext4_sb_info *sbi, char *buf)
2318 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2320 if (!sb->s_bdev->bd_part)
2321 return snprintf(buf, PAGE_SIZE, "0\n");
2322 return snprintf(buf, PAGE_SIZE, "%llu\n",
2323 (unsigned long long)(sbi->s_kbytes_written +
2324 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2325 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2328 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2329 struct ext4_sb_info *sbi,
2330 const char *buf, size_t count)
2334 if (parse_strtoul(buf, 0x40000000, &t))
2337 if (!is_power_of_2(t))
2340 sbi->s_inode_readahead_blks = t;
2344 static ssize_t sbi_ui_show(struct ext4_attr *a,
2345 struct ext4_sb_info *sbi, char *buf)
2347 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2349 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2352 static ssize_t sbi_ui_store(struct ext4_attr *a,
2353 struct ext4_sb_info *sbi,
2354 const char *buf, size_t count)
2356 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2359 if (parse_strtoul(buf, 0xffffffff, &t))
2365 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2366 static struct ext4_attr ext4_attr_##_name = { \
2367 .attr = {.name = __stringify(_name), .mode = _mode }, \
2370 .offset = offsetof(struct ext4_sb_info, _elname), \
2372 #define EXT4_ATTR(name, mode, show, store) \
2373 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2375 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2376 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2377 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2378 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2379 #define ATTR_LIST(name) &ext4_attr_##name.attr
2381 EXT4_RO_ATTR(delayed_allocation_blocks);
2382 EXT4_RO_ATTR(session_write_kbytes);
2383 EXT4_RO_ATTR(lifetime_write_kbytes);
2384 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2385 inode_readahead_blks_store, s_inode_readahead_blks);
2386 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2387 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2388 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2389 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2390 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2391 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2392 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2393 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2395 static struct attribute *ext4_attrs[] = {
2396 ATTR_LIST(delayed_allocation_blocks),
2397 ATTR_LIST(session_write_kbytes),
2398 ATTR_LIST(lifetime_write_kbytes),
2399 ATTR_LIST(inode_readahead_blks),
2400 ATTR_LIST(inode_goal),
2401 ATTR_LIST(mb_stats),
2402 ATTR_LIST(mb_max_to_scan),
2403 ATTR_LIST(mb_min_to_scan),
2404 ATTR_LIST(mb_order2_req),
2405 ATTR_LIST(mb_stream_req),
2406 ATTR_LIST(mb_group_prealloc),
2407 ATTR_LIST(max_writeback_mb_bump),
2411 static ssize_t ext4_attr_show(struct kobject *kobj,
2412 struct attribute *attr, char *buf)
2414 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2416 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2418 return a->show ? a->show(a, sbi, buf) : 0;
2421 static ssize_t ext4_attr_store(struct kobject *kobj,
2422 struct attribute *attr,
2423 const char *buf, size_t len)
2425 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2427 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2429 return a->store ? a->store(a, sbi, buf, len) : 0;
2432 static void ext4_sb_release(struct kobject *kobj)
2434 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2436 complete(&sbi->s_kobj_unregister);
2440 static const struct sysfs_ops ext4_attr_ops = {
2441 .show = ext4_attr_show,
2442 .store = ext4_attr_store,
2445 static struct kobj_type ext4_ktype = {
2446 .default_attrs = ext4_attrs,
2447 .sysfs_ops = &ext4_attr_ops,
2448 .release = ext4_sb_release,
2452 * Check whether this filesystem can be mounted based on
2453 * the features present and the RDONLY/RDWR mount requested.
2454 * Returns 1 if this filesystem can be mounted as requested,
2455 * 0 if it cannot be.
2457 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2459 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2460 ext4_msg(sb, KERN_ERR,
2461 "Couldn't mount because of "
2462 "unsupported optional features (%x)",
2463 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2464 ~EXT4_FEATURE_INCOMPAT_SUPP));
2471 /* Check that feature set is OK for a read-write mount */
2472 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2473 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2474 "unsupported optional features (%x)",
2475 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2476 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2480 * Large file size enabled file system can only be mounted
2481 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2484 if (sizeof(blkcnt_t) < sizeof(u64)) {
2485 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2486 "cannot be mounted RDWR without "
2495 * This function is called once a day if we have errors logged
2496 * on the file system
2498 static void print_daily_error_info(unsigned long arg)
2500 struct super_block *sb = (struct super_block *) arg;
2501 struct ext4_sb_info *sbi;
2502 struct ext4_super_block *es;
2507 if (es->s_error_count)
2508 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2509 le32_to_cpu(es->s_error_count));
2510 if (es->s_first_error_time) {
2511 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2512 sb->s_id, le32_to_cpu(es->s_first_error_time),
2513 (int) sizeof(es->s_first_error_func),
2514 es->s_first_error_func,
2515 le32_to_cpu(es->s_first_error_line));
2516 if (es->s_first_error_ino)
2517 printk(": inode %u",
2518 le32_to_cpu(es->s_first_error_ino));
2519 if (es->s_first_error_block)
2520 printk(": block %llu", (unsigned long long)
2521 le64_to_cpu(es->s_first_error_block));
2524 if (es->s_last_error_time) {
2525 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2526 sb->s_id, le32_to_cpu(es->s_last_error_time),
2527 (int) sizeof(es->s_last_error_func),
2528 es->s_last_error_func,
2529 le32_to_cpu(es->s_last_error_line));
2530 if (es->s_last_error_ino)
2531 printk(": inode %u",
2532 le32_to_cpu(es->s_last_error_ino));
2533 if (es->s_last_error_block)
2534 printk(": block %llu", (unsigned long long)
2535 le64_to_cpu(es->s_last_error_block));
2538 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2541 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2542 __releases(kernel_lock)
2543 __acquires(kernel_lock)
2545 char *orig_data = kstrdup(data, GFP_KERNEL);
2546 struct buffer_head *bh;
2547 struct ext4_super_block *es = NULL;
2548 struct ext4_sb_info *sbi;
2550 ext4_fsblk_t sb_block = get_sb_block(&data);
2551 ext4_fsblk_t logical_sb_block;
2552 unsigned long offset = 0;
2553 unsigned long journal_devnum = 0;
2554 unsigned long def_mount_opts;
2560 unsigned int db_count;
2562 int needs_recovery, has_huge_files;
2565 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2567 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2571 sbi->s_blockgroup_lock =
2572 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2573 if (!sbi->s_blockgroup_lock) {
2577 sb->s_fs_info = sbi;
2578 sbi->s_mount_opt = 0;
2579 sbi->s_resuid = EXT4_DEF_RESUID;
2580 sbi->s_resgid = EXT4_DEF_RESGID;
2581 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2582 sbi->s_sb_block = sb_block;
2583 if (sb->s_bdev->bd_part)
2584 sbi->s_sectors_written_start =
2585 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2589 /* Cleanup superblock name */
2590 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2594 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2596 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2601 * The ext4 superblock will not be buffer aligned for other than 1kB
2602 * block sizes. We need to calculate the offset from buffer start.
2604 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2605 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2606 offset = do_div(logical_sb_block, blocksize);
2608 logical_sb_block = sb_block;
2611 if (!(bh = sb_bread(sb, logical_sb_block))) {
2612 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2616 * Note: s_es must be initialized as soon as possible because
2617 * some ext4 macro-instructions depend on its value
2619 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2621 sb->s_magic = le16_to_cpu(es->s_magic);
2622 if (sb->s_magic != EXT4_SUPER_MAGIC)
2624 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2626 /* Set defaults before we parse the mount options */
2627 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2628 if (def_mount_opts & EXT4_DEFM_DEBUG)
2629 set_opt(sbi->s_mount_opt, DEBUG);
2630 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2631 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2633 set_opt(sbi->s_mount_opt, GRPID);
2635 if (def_mount_opts & EXT4_DEFM_UID16)
2636 set_opt(sbi->s_mount_opt, NO_UID32);
2637 #ifdef CONFIG_EXT4_FS_XATTR
2638 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2639 set_opt(sbi->s_mount_opt, XATTR_USER);
2641 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2642 if (def_mount_opts & EXT4_DEFM_ACL)
2643 set_opt(sbi->s_mount_opt, POSIX_ACL);
2645 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2646 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2647 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2648 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2649 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2650 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2652 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2653 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2654 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2655 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2657 set_opt(sbi->s_mount_opt, ERRORS_RO);
2659 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2660 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2661 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2662 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2663 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2665 set_opt(sbi->s_mount_opt, BARRIER);
2668 * enable delayed allocation by default
2669 * Use -o nodelalloc to turn it off
2671 if (!IS_EXT3_SB(sb))
2672 set_opt(sbi->s_mount_opt, DELALLOC);
2674 if (!parse_options((char *) data, sb, &journal_devnum,
2675 &journal_ioprio, NULL, 0))
2678 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2679 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2681 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2682 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2683 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2684 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2685 ext4_msg(sb, KERN_WARNING,
2686 "feature flags set on rev 0 fs, "
2687 "running e2fsck is recommended");
2690 * Check feature flags regardless of the revision level, since we
2691 * previously didn't change the revision level when setting the flags,
2692 * so there is a chance incompat flags are set on a rev 0 filesystem.
2694 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2697 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2699 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2700 blocksize > EXT4_MAX_BLOCK_SIZE) {
2701 ext4_msg(sb, KERN_ERR,
2702 "Unsupported filesystem blocksize %d", blocksize);
2706 if (sb->s_blocksize != blocksize) {
2707 /* Validate the filesystem blocksize */
2708 if (!sb_set_blocksize(sb, blocksize)) {
2709 ext4_msg(sb, KERN_ERR, "bad block size %d",
2715 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2716 offset = do_div(logical_sb_block, blocksize);
2717 bh = sb_bread(sb, logical_sb_block);
2719 ext4_msg(sb, KERN_ERR,
2720 "Can't read superblock on 2nd try");
2723 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2725 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2726 ext4_msg(sb, KERN_ERR,
2727 "Magic mismatch, very weird!");
2732 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2733 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2734 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2736 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2738 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2739 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2740 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2742 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2743 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2744 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2745 (!is_power_of_2(sbi->s_inode_size)) ||
2746 (sbi->s_inode_size > blocksize)) {
2747 ext4_msg(sb, KERN_ERR,
2748 "unsupported inode size: %d",
2752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2753 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2756 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2758 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2759 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2760 !is_power_of_2(sbi->s_desc_size)) {
2761 ext4_msg(sb, KERN_ERR,
2762 "unsupported descriptor size %lu",
2767 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2769 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2771 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2774 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2775 if (sbi->s_inodes_per_block == 0)
2777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2778 sbi->s_inodes_per_block;
2779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2781 sbi->s_mount_state = le16_to_cpu(es->s_state);
2782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2785 for (i = 0; i < 4; i++)
2786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2787 sbi->s_def_hash_version = es->s_def_hash_version;
2788 i = le32_to_cpu(es->s_flags);
2789 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2790 sbi->s_hash_unsigned = 3;
2791 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2792 #ifdef __CHAR_UNSIGNED__
2793 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2794 sbi->s_hash_unsigned = 3;
2796 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2801 if (sbi->s_blocks_per_group > blocksize * 8) {
2802 ext4_msg(sb, KERN_ERR,
2803 "#blocks per group too big: %lu",
2804 sbi->s_blocks_per_group);
2807 if (sbi->s_inodes_per_group > blocksize * 8) {
2808 ext4_msg(sb, KERN_ERR,
2809 "#inodes per group too big: %lu",
2810 sbi->s_inodes_per_group);
2815 * Test whether we have more sectors than will fit in sector_t,
2816 * and whether the max offset is addressable by the page cache.
2818 if ((ext4_blocks_count(es) >
2819 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2820 (ext4_blocks_count(es) >
2821 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2822 ext4_msg(sb, KERN_ERR, "filesystem"
2823 " too large to mount safely on this system");
2824 if (sizeof(sector_t) < 8)
2825 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2830 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2833 /* check blocks count against device size */
2834 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2835 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2836 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2837 "exceeds size of device (%llu blocks)",
2838 ext4_blocks_count(es), blocks_count);
2843 * It makes no sense for the first data block to be beyond the end
2844 * of the filesystem.
2846 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2847 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2848 "block %u is beyond end of filesystem (%llu)",
2849 le32_to_cpu(es->s_first_data_block),
2850 ext4_blocks_count(es));
2853 blocks_count = (ext4_blocks_count(es) -
2854 le32_to_cpu(es->s_first_data_block) +
2855 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2856 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2857 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2858 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2859 "(block count %llu, first data block %u, "
2860 "blocks per group %lu)", sbi->s_groups_count,
2861 ext4_blocks_count(es),
2862 le32_to_cpu(es->s_first_data_block),
2863 EXT4_BLOCKS_PER_GROUP(sb));
2866 sbi->s_groups_count = blocks_count;
2867 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2868 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2869 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2870 EXT4_DESC_PER_BLOCK(sb);
2871 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2873 if (sbi->s_group_desc == NULL) {
2874 ext4_msg(sb, KERN_ERR, "not enough memory");
2878 #ifdef CONFIG_PROC_FS
2880 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2883 bgl_lock_init(sbi->s_blockgroup_lock);
2885 for (i = 0; i < db_count; i++) {
2886 block = descriptor_loc(sb, logical_sb_block, i);
2887 sbi->s_group_desc[i] = sb_bread(sb, block);
2888 if (!sbi->s_group_desc[i]) {
2889 ext4_msg(sb, KERN_ERR,
2890 "can't read group descriptor %d", i);
2895 if (!ext4_check_descriptors(sb)) {
2896 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2899 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2900 if (!ext4_fill_flex_info(sb)) {
2901 ext4_msg(sb, KERN_ERR,
2902 "unable to initialize "
2903 "flex_bg meta info!");
2907 sbi->s_gdb_count = db_count;
2908 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2909 spin_lock_init(&sbi->s_next_gen_lock);
2911 sbi->s_stripe = ext4_get_stripe_size(sbi);
2912 sbi->s_max_writeback_mb_bump = 128;
2915 * set up enough so that it can read an inode
2917 if (!test_opt(sb, NOLOAD) &&
2918 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2919 sb->s_op = &ext4_sops;
2921 sb->s_op = &ext4_nojournal_sops;
2922 sb->s_export_op = &ext4_export_ops;
2923 sb->s_xattr = ext4_xattr_handlers;
2925 sb->s_qcop = &ext4_qctl_operations;
2926 sb->dq_op = &ext4_quota_operations;
2928 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2929 mutex_init(&sbi->s_orphan_lock);
2930 mutex_init(&sbi->s_resize_lock);
2934 needs_recovery = (es->s_last_orphan != 0 ||
2935 EXT4_HAS_INCOMPAT_FEATURE(sb,
2936 EXT4_FEATURE_INCOMPAT_RECOVER));
2939 * The first inode we look at is the journal inode. Don't try
2940 * root first: it may be modified in the journal!
2942 if (!test_opt(sb, NOLOAD) &&
2943 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2944 if (ext4_load_journal(sb, es, journal_devnum))
2946 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2947 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2948 ext4_msg(sb, KERN_ERR, "required journal recovery "
2949 "suppressed and not mounted read-only");
2950 goto failed_mount_wq;
2952 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2953 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2954 sbi->s_journal = NULL;
2959 if (ext4_blocks_count(es) > 0xffffffffULL &&
2960 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2961 JBD2_FEATURE_INCOMPAT_64BIT)) {
2962 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2963 goto failed_mount_wq;
2966 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2967 jbd2_journal_set_features(sbi->s_journal,
2968 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2969 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2970 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2971 jbd2_journal_set_features(sbi->s_journal,
2972 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2973 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2974 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2976 jbd2_journal_clear_features(sbi->s_journal,
2977 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2978 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2981 /* We have now updated the journal if required, so we can
2982 * validate the data journaling mode. */
2983 switch (test_opt(sb, DATA_FLAGS)) {
2985 /* No mode set, assume a default based on the journal
2986 * capabilities: ORDERED_DATA if the journal can
2987 * cope, else JOURNAL_DATA
2989 if (jbd2_journal_check_available_features
2990 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2991 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2993 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2996 case EXT4_MOUNT_ORDERED_DATA:
2997 case EXT4_MOUNT_WRITEBACK_DATA:
2998 if (!jbd2_journal_check_available_features
2999 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3000 ext4_msg(sb, KERN_ERR, "Journal does not support "
3001 "requested data journaling mode");
3002 goto failed_mount_wq;
3007 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3010 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3011 ext4_count_free_blocks(sb));
3013 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3014 ext4_count_free_inodes(sb));
3016 err = percpu_counter_init(&sbi->s_dirs_counter,
3017 ext4_count_dirs(sb));
3019 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3021 ext4_msg(sb, KERN_ERR, "insufficient memory");
3022 goto failed_mount_wq;
3025 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3026 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3027 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3028 goto failed_mount_wq;
3032 * The jbd2_journal_load will have done any necessary log recovery,
3033 * so we can safely mount the rest of the filesystem now.
3036 root = ext4_iget(sb, EXT4_ROOT_INO);
3038 ext4_msg(sb, KERN_ERR, "get root inode failed");
3039 ret = PTR_ERR(root);
3042 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3044 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3047 sb->s_root = d_alloc_root(root);
3049 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3055 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3057 /* determine the minimum size of new large inodes, if present */
3058 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3059 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3060 EXT4_GOOD_OLD_INODE_SIZE;
3061 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3062 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3063 if (sbi->s_want_extra_isize <
3064 le16_to_cpu(es->s_want_extra_isize))
3065 sbi->s_want_extra_isize =
3066 le16_to_cpu(es->s_want_extra_isize);
3067 if (sbi->s_want_extra_isize <
3068 le16_to_cpu(es->s_min_extra_isize))
3069 sbi->s_want_extra_isize =
3070 le16_to_cpu(es->s_min_extra_isize);
3073 /* Check if enough inode space is available */
3074 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3075 sbi->s_inode_size) {
3076 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3077 EXT4_GOOD_OLD_INODE_SIZE;
3078 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3082 if (test_opt(sb, DELALLOC) &&
3083 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3084 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3085 "requested data journaling mode");
3086 clear_opt(sbi->s_mount_opt, DELALLOC);
3088 if (test_opt(sb, DIOREAD_NOLOCK)) {
3089 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3090 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3091 "option - requested data journaling mode");
3092 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3094 if (sb->s_blocksize < PAGE_SIZE) {
3095 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3096 "option - block size is too small");
3097 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3101 err = ext4_setup_system_zone(sb);
3103 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3109 err = ext4_mb_init(sb, needs_recovery);
3111 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3116 sbi->s_kobj.kset = ext4_kset;
3117 init_completion(&sbi->s_kobj_unregister);
3118 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3121 ext4_mb_release(sb);
3122 ext4_ext_release(sb);
3126 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3127 ext4_orphan_cleanup(sb, es);
3128 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3129 if (needs_recovery) {
3130 ext4_msg(sb, KERN_INFO, "recovery complete");
3131 ext4_mark_recovery_complete(sb, es);
3133 if (EXT4_SB(sb)->s_journal) {
3134 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3135 descr = " journalled data mode";
3136 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3137 descr = " ordered data mode";
3139 descr = " writeback data mode";
3141 descr = "out journal";
3143 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3144 "Opts: %s", descr, orig_data);
3146 init_timer(&sbi->s_err_report);
3147 sbi->s_err_report.function = print_daily_error_info;
3148 sbi->s_err_report.data = (unsigned long) sb;
3149 if (es->s_error_count)
3150 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3158 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3162 ext4_msg(sb, KERN_ERR, "mount failed");
3163 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3165 ext4_release_system_zone(sb);
3166 if (sbi->s_journal) {
3167 jbd2_journal_destroy(sbi->s_journal);
3168 sbi->s_journal = NULL;
3170 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3171 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3172 percpu_counter_destroy(&sbi->s_dirs_counter);
3173 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3175 if (sbi->s_flex_groups) {
3176 if (is_vmalloc_addr(sbi->s_flex_groups))
3177 vfree(sbi->s_flex_groups);
3179 kfree(sbi->s_flex_groups);
3182 for (i = 0; i < db_count; i++)
3183 brelse(sbi->s_group_desc[i]);
3184 kfree(sbi->s_group_desc);
3187 remove_proc_entry(sb->s_id, ext4_proc_root);
3190 for (i = 0; i < MAXQUOTAS; i++)
3191 kfree(sbi->s_qf_names[i]);
3193 ext4_blkdev_remove(sbi);
3196 sb->s_fs_info = NULL;
3197 kfree(sbi->s_blockgroup_lock);
3206 * Setup any per-fs journal parameters now. We'll do this both on
3207 * initial mount, once the journal has been initialised but before we've
3208 * done any recovery; and again on any subsequent remount.
3210 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3212 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214 journal->j_commit_interval = sbi->s_commit_interval;
3215 journal->j_min_batch_time = sbi->s_min_batch_time;
3216 journal->j_max_batch_time = sbi->s_max_batch_time;
3218 spin_lock(&journal->j_state_lock);
3219 if (test_opt(sb, BARRIER))
3220 journal->j_flags |= JBD2_BARRIER;
3222 journal->j_flags &= ~JBD2_BARRIER;
3223 if (test_opt(sb, DATA_ERR_ABORT))
3224 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3226 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3227 spin_unlock(&journal->j_state_lock);
3230 static journal_t *ext4_get_journal(struct super_block *sb,
3231 unsigned int journal_inum)
3233 struct inode *journal_inode;
3236 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3238 /* First, test for the existence of a valid inode on disk. Bad
3239 * things happen if we iget() an unused inode, as the subsequent
3240 * iput() will try to delete it. */
3242 journal_inode = ext4_iget(sb, journal_inum);
3243 if (IS_ERR(journal_inode)) {
3244 ext4_msg(sb, KERN_ERR, "no journal found");
3247 if (!journal_inode->i_nlink) {
3248 make_bad_inode(journal_inode);
3249 iput(journal_inode);
3250 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3254 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3255 journal_inode, journal_inode->i_size);
3256 if (!S_ISREG(journal_inode->i_mode)) {
3257 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3258 iput(journal_inode);
3262 journal = jbd2_journal_init_inode(journal_inode);
3264 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3265 iput(journal_inode);
3268 journal->j_private = sb;
3269 ext4_init_journal_params(sb, journal);
3273 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3276 struct buffer_head *bh;
3280 int hblock, blocksize;
3281 ext4_fsblk_t sb_block;
3282 unsigned long offset;
3283 struct ext4_super_block *es;
3284 struct block_device *bdev;
3286 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3288 bdev = ext4_blkdev_get(j_dev, sb);
3292 if (bd_claim(bdev, sb)) {
3293 ext4_msg(sb, KERN_ERR,
3294 "failed to claim external journal device");
3295 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3299 blocksize = sb->s_blocksize;
3300 hblock = bdev_logical_block_size(bdev);
3301 if (blocksize < hblock) {
3302 ext4_msg(sb, KERN_ERR,
3303 "blocksize too small for journal device");
3307 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3308 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3309 set_blocksize(bdev, blocksize);
3310 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3311 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3312 "external journal");
3316 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3317 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3318 !(le32_to_cpu(es->s_feature_incompat) &
3319 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3320 ext4_msg(sb, KERN_ERR, "external journal has "
3326 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3327 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3332 len = ext4_blocks_count(es);
3333 start = sb_block + 1;
3334 brelse(bh); /* we're done with the superblock */
3336 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3337 start, len, blocksize);
3339 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3342 journal->j_private = sb;
3343 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3344 wait_on_buffer(journal->j_sb_buffer);
3345 if (!buffer_uptodate(journal->j_sb_buffer)) {
3346 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3349 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3350 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3351 "user (unsupported) - %d",
3352 be32_to_cpu(journal->j_superblock->s_nr_users));
3355 EXT4_SB(sb)->journal_bdev = bdev;
3356 ext4_init_journal_params(sb, journal);
3360 jbd2_journal_destroy(journal);
3362 ext4_blkdev_put(bdev);
3366 static int ext4_load_journal(struct super_block *sb,
3367 struct ext4_super_block *es,
3368 unsigned long journal_devnum)
3371 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3374 int really_read_only;
3376 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3378 if (journal_devnum &&
3379 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3380 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3381 "numbers have changed");
3382 journal_dev = new_decode_dev(journal_devnum);
3384 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3386 really_read_only = bdev_read_only(sb->s_bdev);
3389 * Are we loading a blank journal or performing recovery after a
3390 * crash? For recovery, we need to check in advance whether we
3391 * can get read-write access to the device.
3393 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3394 if (sb->s_flags & MS_RDONLY) {
3395 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3396 "required on readonly filesystem");
3397 if (really_read_only) {
3398 ext4_msg(sb, KERN_ERR, "write access "
3399 "unavailable, cannot proceed");
3402 ext4_msg(sb, KERN_INFO, "write access will "
3403 "be enabled during recovery");
3407 if (journal_inum && journal_dev) {
3408 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3409 "and inode journals!");
3414 if (!(journal = ext4_get_journal(sb, journal_inum)))
3417 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3421 if (!(journal->j_flags & JBD2_BARRIER))
3422 ext4_msg(sb, KERN_INFO, "barriers disabled");
3424 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3425 err = jbd2_journal_update_format(journal);
3427 ext4_msg(sb, KERN_ERR, "error updating journal");
3428 jbd2_journal_destroy(journal);
3433 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3434 err = jbd2_journal_wipe(journal, !really_read_only);
3436 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3438 memcpy(save, ((char *) es) +
3439 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3440 err = jbd2_journal_load(journal);
3442 memcpy(((char *) es) + EXT4_S_ERR_START,
3443 save, EXT4_S_ERR_LEN);
3448 ext4_msg(sb, KERN_ERR, "error loading journal");
3449 jbd2_journal_destroy(journal);
3453 EXT4_SB(sb)->s_journal = journal;
3454 ext4_clear_journal_err(sb, es);
3456 if (journal_devnum &&
3457 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3458 es->s_journal_dev = cpu_to_le32(journal_devnum);
3460 /* Make sure we flush the recovery flag to disk. */
3461 ext4_commit_super(sb, 1);
3467 static int ext4_commit_super(struct super_block *sb, int sync)
3469 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3470 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3475 if (buffer_write_io_error(sbh)) {
3477 * Oh, dear. A previous attempt to write the
3478 * superblock failed. This could happen because the
3479 * USB device was yanked out. Or it could happen to
3480 * be a transient write error and maybe the block will
3481 * be remapped. Nothing we can do but to retry the
3482 * write and hope for the best.
3484 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3485 "superblock detected");
3486 clear_buffer_write_io_error(sbh);
3487 set_buffer_uptodate(sbh);
3490 * If the file system is mounted read-only, don't update the
3491 * superblock write time. This avoids updating the superblock
3492 * write time when we are mounting the root file system
3493 * read/only but we need to replay the journal; at that point,
3494 * for people who are east of GMT and who make their clock
3495 * tick in localtime for Windows bug-for-bug compatibility,
3496 * the clock is set in the future, and this will cause e2fsck
3497 * to complain and force a full file system check.
3499 if (!(sb->s_flags & MS_RDONLY))
3500 es->s_wtime = cpu_to_le32(get_seconds());
3501 if (sb->s_bdev->bd_part)
3502 es->s_kbytes_written =
3503 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3504 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3505 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3507 es->s_kbytes_written =
3508 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3509 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3510 &EXT4_SB(sb)->s_freeblocks_counter));
3511 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3512 &EXT4_SB(sb)->s_freeinodes_counter));
3514 BUFFER_TRACE(sbh, "marking dirty");
3515 mark_buffer_dirty(sbh);
3517 error = sync_dirty_buffer(sbh);
3521 error = buffer_write_io_error(sbh);
3523 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3525 clear_buffer_write_io_error(sbh);
3526 set_buffer_uptodate(sbh);
3533 * Have we just finished recovery? If so, and if we are mounting (or
3534 * remounting) the filesystem readonly, then we will end up with a
3535 * consistent fs on disk. Record that fact.
3537 static void ext4_mark_recovery_complete(struct super_block *sb,
3538 struct ext4_super_block *es)
3540 journal_t *journal = EXT4_SB(sb)->s_journal;
3542 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3543 BUG_ON(journal != NULL);
3546 jbd2_journal_lock_updates(journal);
3547 if (jbd2_journal_flush(journal) < 0)
3550 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3551 sb->s_flags & MS_RDONLY) {
3552 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3553 ext4_commit_super(sb, 1);
3557 jbd2_journal_unlock_updates(journal);
3561 * If we are mounting (or read-write remounting) a filesystem whose journal
3562 * has recorded an error from a previous lifetime, move that error to the
3563 * main filesystem now.
3565 static void ext4_clear_journal_err(struct super_block *sb,
3566 struct ext4_super_block *es)
3572 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3574 journal = EXT4_SB(sb)->s_journal;
3577 * Now check for any error status which may have been recorded in the
3578 * journal by a prior ext4_error() or ext4_abort()
3581 j_errno = jbd2_journal_errno(journal);
3585 errstr = ext4_decode_error(sb, j_errno, nbuf);
3586 ext4_warning(sb, "Filesystem error recorded "
3587 "from previous mount: %s", errstr);
3588 ext4_warning(sb, "Marking fs in need of filesystem check.");
3590 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3591 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3592 ext4_commit_super(sb, 1);
3594 jbd2_journal_clear_err(journal);
3599 * Force the running and committing transactions to commit,
3600 * and wait on the commit.
3602 int ext4_force_commit(struct super_block *sb)
3607 if (sb->s_flags & MS_RDONLY)
3610 journal = EXT4_SB(sb)->s_journal;
3612 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3613 ret = ext4_journal_force_commit(journal);
3619 static void ext4_write_super(struct super_block *sb)
3622 ext4_commit_super(sb, 1);
3626 static int ext4_sync_fs(struct super_block *sb, int wait)
3630 struct ext4_sb_info *sbi = EXT4_SB(sb);
3632 trace_ext4_sync_fs(sb, wait);
3633 flush_workqueue(sbi->dio_unwritten_wq);
3634 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3636 jbd2_log_wait_commit(sbi->s_journal, target);
3642 * LVM calls this function before a (read-only) snapshot is created. This
3643 * gives us a chance to flush the journal completely and mark the fs clean.
3645 static int ext4_freeze(struct super_block *sb)
3650 if (sb->s_flags & MS_RDONLY)
3653 journal = EXT4_SB(sb)->s_journal;
3655 /* Now we set up the journal barrier. */
3656 jbd2_journal_lock_updates(journal);
3659 * Don't clear the needs_recovery flag if we failed to flush
3662 error = jbd2_journal_flush(journal);
3666 /* Journal blocked and flushed, clear needs_recovery flag. */
3667 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3668 error = ext4_commit_super(sb, 1);
3670 /* we rely on s_frozen to stop further updates */
3671 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3676 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3677 * flag here, even though the filesystem is not technically dirty yet.
3679 static int ext4_unfreeze(struct super_block *sb)
3681 if (sb->s_flags & MS_RDONLY)
3685 /* Reset the needs_recovery flag before the fs is unlocked. */
3686 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3687 ext4_commit_super(sb, 1);
3692 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3694 struct ext4_super_block *es;
3695 struct ext4_sb_info *sbi = EXT4_SB(sb);
3696 ext4_fsblk_t n_blocks_count = 0;
3697 unsigned long old_sb_flags;
3698 struct ext4_mount_options old_opts;
3699 int enable_quota = 0;
3701 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3706 char *orig_data = kstrdup(data, GFP_KERNEL);
3710 /* Store the original options */
3712 old_sb_flags = sb->s_flags;
3713 old_opts.s_mount_opt = sbi->s_mount_opt;
3714 old_opts.s_resuid = sbi->s_resuid;
3715 old_opts.s_resgid = sbi->s_resgid;
3716 old_opts.s_commit_interval = sbi->s_commit_interval;
3717 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3718 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3720 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3721 for (i = 0; i < MAXQUOTAS; i++)
3722 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3724 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3725 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3728 * Allow the "check" option to be passed as a remount option.
3730 if (!parse_options(data, sb, NULL, &journal_ioprio,
3731 &n_blocks_count, 1)) {
3736 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3737 ext4_abort(sb, "Abort forced by user");
3739 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3740 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3744 if (sbi->s_journal) {
3745 ext4_init_journal_params(sb, sbi->s_journal);
3746 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3749 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3750 n_blocks_count > ext4_blocks_count(es)) {
3751 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3756 if (*flags & MS_RDONLY) {
3757 err = dquot_suspend(sb, -1);
3762 * First of all, the unconditional stuff we have to do
3763 * to disable replay of the journal when we next remount
3765 sb->s_flags |= MS_RDONLY;
3768 * OK, test if we are remounting a valid rw partition
3769 * readonly, and if so set the rdonly flag and then
3770 * mark the partition as valid again.
3772 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3773 (sbi->s_mount_state & EXT4_VALID_FS))
3774 es->s_state = cpu_to_le16(sbi->s_mount_state);
3777 ext4_mark_recovery_complete(sb, es);
3779 /* Make sure we can mount this feature set readwrite */
3780 if (!ext4_feature_set_ok(sb, 0)) {
3785 * Make sure the group descriptor checksums
3786 * are sane. If they aren't, refuse to remount r/w.
3788 for (g = 0; g < sbi->s_groups_count; g++) {
3789 struct ext4_group_desc *gdp =
3790 ext4_get_group_desc(sb, g, NULL);
3792 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3793 ext4_msg(sb, KERN_ERR,
3794 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3795 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3796 le16_to_cpu(gdp->bg_checksum));
3803 * If we have an unprocessed orphan list hanging
3804 * around from a previously readonly bdev mount,
3805 * require a full umount/remount for now.
3807 if (es->s_last_orphan) {
3808 ext4_msg(sb, KERN_WARNING, "Couldn't "
3809 "remount RDWR because of unprocessed "
3810 "orphan inode list. Please "
3811 "umount/remount instead");
3817 * Mounting a RDONLY partition read-write, so reread
3818 * and store the current valid flag. (It may have
3819 * been changed by e2fsck since we originally mounted
3823 ext4_clear_journal_err(sb, es);
3824 sbi->s_mount_state = le16_to_cpu(es->s_state);
3825 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3827 if (!ext4_setup_super(sb, es, 0))
3828 sb->s_flags &= ~MS_RDONLY;
3832 ext4_setup_system_zone(sb);
3833 if (sbi->s_journal == NULL)
3834 ext4_commit_super(sb, 1);
3837 /* Release old quota file names */
3838 for (i = 0; i < MAXQUOTAS; i++)
3839 if (old_opts.s_qf_names[i] &&
3840 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3841 kfree(old_opts.s_qf_names[i]);
3846 dquot_resume(sb, -1);
3848 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3853 sb->s_flags = old_sb_flags;
3854 sbi->s_mount_opt = old_opts.s_mount_opt;
3855 sbi->s_resuid = old_opts.s_resuid;
3856 sbi->s_resgid = old_opts.s_resgid;
3857 sbi->s_commit_interval = old_opts.s_commit_interval;
3858 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3859 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3861 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3862 for (i = 0; i < MAXQUOTAS; i++) {
3863 if (sbi->s_qf_names[i] &&
3864 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3865 kfree(sbi->s_qf_names[i]);
3866 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3875 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3877 struct super_block *sb = dentry->d_sb;
3878 struct ext4_sb_info *sbi = EXT4_SB(sb);
3879 struct ext4_super_block *es = sbi->s_es;
3882 if (test_opt(sb, MINIX_DF)) {
3883 sbi->s_overhead_last = 0;
3884 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3885 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3886 ext4_fsblk_t overhead = 0;
3889 * Compute the overhead (FS structures). This is constant
3890 * for a given filesystem unless the number of block groups
3891 * changes so we cache the previous value until it does.
3895 * All of the blocks before first_data_block are
3898 overhead = le32_to_cpu(es->s_first_data_block);
3901 * Add the overhead attributed to the superblock and
3902 * block group descriptors. If the sparse superblocks
3903 * feature is turned on, then not all groups have this.
3905 for (i = 0; i < ngroups; i++) {
3906 overhead += ext4_bg_has_super(sb, i) +
3907 ext4_bg_num_gdb(sb, i);
3912 * Every block group has an inode bitmap, a block
3913 * bitmap, and an inode table.
3915 overhead += ngroups * (2 + sbi->s_itb_per_group);
3916 sbi->s_overhead_last = overhead;
3918 sbi->s_blocks_last = ext4_blocks_count(es);
3921 buf->f_type = EXT4_SUPER_MAGIC;
3922 buf->f_bsize = sb->s_blocksize;
3923 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3924 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3925 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3926 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3927 if (buf->f_bfree < ext4_r_blocks_count(es))
3929 buf->f_files = le32_to_cpu(es->s_inodes_count);
3930 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3931 buf->f_namelen = EXT4_NAME_LEN;
3932 fsid = le64_to_cpup((void *)es->s_uuid) ^
3933 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3934 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3935 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3940 /* Helper function for writing quotas on sync - we need to start transaction
3941 * before quota file is locked for write. Otherwise the are possible deadlocks:
3942 * Process 1 Process 2
3943 * ext4_create() quota_sync()
3944 * jbd2_journal_start() write_dquot()
3945 * dquot_initialize() down(dqio_mutex)
3946 * down(dqio_mutex) jbd2_journal_start()
3952 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3954 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3957 static int ext4_write_dquot(struct dquot *dquot)
3961 struct inode *inode;
3963 inode = dquot_to_inode(dquot);
3964 handle = ext4_journal_start(inode,
3965 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3967 return PTR_ERR(handle);
3968 ret = dquot_commit(dquot);
3969 err = ext4_journal_stop(handle);
3975 static int ext4_acquire_dquot(struct dquot *dquot)
3980 handle = ext4_journal_start(dquot_to_inode(dquot),
3981 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3983 return PTR_ERR(handle);
3984 ret = dquot_acquire(dquot);
3985 err = ext4_journal_stop(handle);
3991 static int ext4_release_dquot(struct dquot *dquot)
3996 handle = ext4_journal_start(dquot_to_inode(dquot),
3997 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3998 if (IS_ERR(handle)) {
3999 /* Release dquot anyway to avoid endless cycle in dqput() */
4000 dquot_release(dquot);
4001 return PTR_ERR(handle);
4003 ret = dquot_release(dquot);
4004 err = ext4_journal_stop(handle);
4010 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4012 /* Are we journaling quotas? */
4013 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4014 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4015 dquot_mark_dquot_dirty(dquot);
4016 return ext4_write_dquot(dquot);
4018 return dquot_mark_dquot_dirty(dquot);
4022 static int ext4_write_info(struct super_block *sb, int type)
4027 /* Data block + inode block */
4028 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4030 return PTR_ERR(handle);
4031 ret = dquot_commit_info(sb, type);
4032 err = ext4_journal_stop(handle);
4039 * Turn on quotas during mount time - we need to find
4040 * the quota file and such...
4042 static int ext4_quota_on_mount(struct super_block *sb, int type)
4044 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4045 EXT4_SB(sb)->s_jquota_fmt, type);
4049 * Standard function to be called on quota_on
4051 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4057 if (!test_opt(sb, QUOTA))
4060 err = kern_path(name, LOOKUP_FOLLOW, &path);
4064 /* Quotafile not on the same filesystem? */
4065 if (path.mnt->mnt_sb != sb) {
4069 /* Journaling quota? */
4070 if (EXT4_SB(sb)->s_qf_names[type]) {
4071 /* Quotafile not in fs root? */
4072 if (path.dentry->d_parent != sb->s_root)
4073 ext4_msg(sb, KERN_WARNING,
4074 "Quota file not on filesystem root. "
4075 "Journaled quota will not work");
4079 * When we journal data on quota file, we have to flush journal to see
4080 * all updates to the file when we bypass pagecache...
4082 if (EXT4_SB(sb)->s_journal &&
4083 ext4_should_journal_data(path.dentry->d_inode)) {
4085 * We don't need to lock updates but journal_flush() could
4086 * otherwise be livelocked...
4088 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4089 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4090 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4097 err = dquot_quota_on_path(sb, type, format_id, &path);
4102 static int ext4_quota_off(struct super_block *sb, int type)
4104 /* Force all delayed allocation blocks to be allocated */
4105 if (test_opt(sb, DELALLOC)) {
4106 down_read(&sb->s_umount);
4107 sync_filesystem(sb);
4108 up_read(&sb->s_umount);
4111 return dquot_quota_off(sb, type);
4114 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4115 * acquiring the locks... As quota files are never truncated and quota code
4116 * itself serializes the operations (and noone else should touch the files)
4117 * we don't have to be afraid of races */
4118 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4119 size_t len, loff_t off)
4121 struct inode *inode = sb_dqopt(sb)->files[type];
4122 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4124 int offset = off & (sb->s_blocksize - 1);
4127 struct buffer_head *bh;
4128 loff_t i_size = i_size_read(inode);
4132 if (off+len > i_size)
4135 while (toread > 0) {
4136 tocopy = sb->s_blocksize - offset < toread ?
4137 sb->s_blocksize - offset : toread;
4138 bh = ext4_bread(NULL, inode, blk, 0, &err);
4141 if (!bh) /* A hole? */
4142 memset(data, 0, tocopy);
4144 memcpy(data, bh->b_data+offset, tocopy);
4154 /* Write to quotafile (we know the transaction is already started and has
4155 * enough credits) */
4156 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4157 const char *data, size_t len, loff_t off)
4159 struct inode *inode = sb_dqopt(sb)->files[type];
4160 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4162 int offset = off & (sb->s_blocksize - 1);
4163 struct buffer_head *bh;
4164 handle_t *handle = journal_current_handle();
4166 if (EXT4_SB(sb)->s_journal && !handle) {
4167 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4168 " cancelled because transaction is not started",
4169 (unsigned long long)off, (unsigned long long)len);
4173 * Since we account only one data block in transaction credits,
4174 * then it is impossible to cross a block boundary.
4176 if (sb->s_blocksize - offset < len) {
4177 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4178 " cancelled because not block aligned",
4179 (unsigned long long)off, (unsigned long long)len);
4183 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4184 bh = ext4_bread(handle, inode, blk, 1, &err);
4187 err = ext4_journal_get_write_access(handle, bh);
4193 memcpy(bh->b_data+offset, data, len);
4194 flush_dcache_page(bh->b_page);
4196 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4200 mutex_unlock(&inode->i_mutex);
4203 if (inode->i_size < off + len) {
4204 i_size_write(inode, off + len);
4205 EXT4_I(inode)->i_disksize = inode->i_size;
4207 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4208 ext4_mark_inode_dirty(handle, inode);
4209 mutex_unlock(&inode->i_mutex);
4215 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4216 const char *dev_name, void *data, struct vfsmount *mnt)
4218 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4221 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4222 static struct file_system_type ext2_fs_type = {
4223 .owner = THIS_MODULE,
4225 .get_sb = ext4_get_sb,
4226 .kill_sb = kill_block_super,
4227 .fs_flags = FS_REQUIRES_DEV,
4230 static inline void register_as_ext2(void)
4232 int err = register_filesystem(&ext2_fs_type);
4235 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4238 static inline void unregister_as_ext2(void)
4240 unregister_filesystem(&ext2_fs_type);
4242 MODULE_ALIAS("ext2");
4244 static inline void register_as_ext2(void) { }
4245 static inline void unregister_as_ext2(void) { }
4248 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4249 static inline void register_as_ext3(void)
4251 int err = register_filesystem(&ext3_fs_type);
4254 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4257 static inline void unregister_as_ext3(void)
4259 unregister_filesystem(&ext3_fs_type);
4261 MODULE_ALIAS("ext3");
4263 static inline void register_as_ext3(void) { }
4264 static inline void unregister_as_ext3(void) { }
4267 static struct file_system_type ext4_fs_type = {
4268 .owner = THIS_MODULE,
4270 .get_sb = ext4_get_sb,
4271 .kill_sb = kill_block_super,
4272 .fs_flags = FS_REQUIRES_DEV,
4275 static int __init init_ext4_fs(void)
4279 ext4_check_flag_values();
4280 err = init_ext4_system_zone();
4283 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4286 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4287 err = init_ext4_mballoc();
4291 err = init_ext4_xattr();
4294 err = init_inodecache();
4299 err = register_filesystem(&ext4_fs_type);
4304 unregister_as_ext2();
4305 unregister_as_ext3();
4306 destroy_inodecache();
4310 exit_ext4_mballoc();
4312 remove_proc_entry("fs/ext4", NULL);
4313 kset_unregister(ext4_kset);
4315 exit_ext4_system_zone();
4319 static void __exit exit_ext4_fs(void)
4321 unregister_as_ext2();
4322 unregister_as_ext3();
4323 unregister_filesystem(&ext4_fs_type);
4324 destroy_inodecache();
4326 exit_ext4_mballoc();
4327 remove_proc_entry("fs/ext4", NULL);
4328 kset_unregister(ext4_kset);
4329 exit_ext4_system_zone();
4332 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4333 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4334 MODULE_LICENSE("GPL");
4335 module_init(init_ext4_fs)
4336 module_exit(exit_ext4_fs)