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/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/ext4.h>
52 struct proc_dir_entry *ext4_proc_root;
53 static struct kset *ext4_kset;
55 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
56 unsigned long journal_devnum);
57 static int ext4_commit_super(struct super_block *sb, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static int ext4_unfreeze(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static int ext4_freeze(struct super_block *sb);
70 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
71 const char *dev_name, void *data, struct vfsmount *mnt);
73 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
74 static struct file_system_type ext3_fs_type = {
77 .get_sb = ext4_get_sb,
78 .kill_sb = kill_block_super,
79 .fs_flags = FS_REQUIRES_DEV,
81 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
83 #define IS_EXT3_SB(sb) (0)
86 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
87 struct ext4_group_desc *bg)
89 return le32_to_cpu(bg->bg_block_bitmap_lo) |
90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
91 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
94 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
95 struct ext4_group_desc *bg)
97 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
98 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
99 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
102 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
103 struct ext4_group_desc *bg)
105 return le32_to_cpu(bg->bg_inode_table_lo) |
106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
107 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
110 __u32 ext4_free_blks_count(struct super_block *sb,
111 struct ext4_group_desc *bg)
113 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
115 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
118 __u32 ext4_free_inodes_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
121 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
123 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
126 __u32 ext4_used_dirs_count(struct super_block *sb,
127 struct ext4_group_desc *bg)
129 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
130 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
131 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
134 __u32 ext4_itable_unused_count(struct super_block *sb,
135 struct ext4_group_desc *bg)
137 return le16_to_cpu(bg->bg_itable_unused_lo) |
138 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
139 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
142 void ext4_block_bitmap_set(struct super_block *sb,
143 struct ext4_group_desc *bg, ext4_fsblk_t blk)
145 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
146 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
147 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
150 void ext4_inode_bitmap_set(struct super_block *sb,
151 struct ext4_group_desc *bg, ext4_fsblk_t blk)
153 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
158 void ext4_inode_table_set(struct super_block *sb,
159 struct ext4_group_desc *bg, ext4_fsblk_t blk)
161 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
166 void ext4_free_blks_set(struct super_block *sb,
167 struct ext4_group_desc *bg, __u32 count)
169 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
174 void ext4_free_inodes_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
177 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
182 void ext4_used_dirs_set(struct super_block *sb,
183 struct ext4_group_desc *bg, __u32 count)
185 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
186 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
187 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
190 void ext4_itable_unused_set(struct super_block *sb,
191 struct ext4_group_desc *bg, __u32 count)
193 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
194 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
195 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
199 /* Just increment the non-pointer handle value */
200 static handle_t *ext4_get_nojournal(void)
202 handle_t *handle = current->journal_info;
203 unsigned long ref_cnt = (unsigned long)handle;
205 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
208 handle = (handle_t *)ref_cnt;
210 current->journal_info = handle;
215 /* Decrement the non-pointer handle value */
216 static void ext4_put_nojournal(handle_t *handle)
218 unsigned long ref_cnt = (unsigned long)handle;
220 BUG_ON(ref_cnt == 0);
223 handle = (handle_t *)ref_cnt;
225 current->journal_info = handle;
229 * Wrappers for jbd2_journal_start/end.
231 * The only special thing we need to do here is to make sure that all
232 * journal_end calls result in the superblock being marked dirty, so
233 * that sync() will call the filesystem's write_super callback if
236 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
240 if (sb->s_flags & MS_RDONLY)
241 return ERR_PTR(-EROFS);
243 vfs_check_frozen(sb, SB_FREEZE_TRANS);
244 /* Special case here: if the journal has aborted behind our
245 * backs (eg. EIO in the commit thread), then we still need to
246 * take the FS itself readonly cleanly. */
247 journal = EXT4_SB(sb)->s_journal;
249 if (is_journal_aborted(journal)) {
250 ext4_abort(sb, "Detected aborted journal");
251 return ERR_PTR(-EROFS);
253 return jbd2_journal_start(journal, nblocks);
255 return ext4_get_nojournal();
259 * The only special thing we need to do here is to make sure that all
260 * jbd2_journal_stop calls result in the superblock being marked dirty, so
261 * that sync() will call the filesystem's write_super callback if
264 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
266 struct super_block *sb;
270 if (!ext4_handle_valid(handle)) {
271 ext4_put_nojournal(handle);
274 sb = handle->h_transaction->t_journal->j_private;
276 rc = jbd2_journal_stop(handle);
281 __ext4_std_error(sb, where, line, err);
285 void ext4_journal_abort_handle(const char *caller, unsigned int line,
286 const char *err_fn, struct buffer_head *bh,
287 handle_t *handle, int err)
290 const char *errstr = ext4_decode_error(NULL, err, nbuf);
292 BUG_ON(!ext4_handle_valid(handle));
295 BUFFER_TRACE(bh, "abort");
300 if (is_handle_aborted(handle))
303 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
304 caller, line, errstr, err_fn);
306 jbd2_journal_abort_handle(handle);
309 static void __save_error_info(struct super_block *sb, const char *func,
312 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
314 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
315 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
316 es->s_last_error_time = cpu_to_le32(get_seconds());
317 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
318 es->s_last_error_line = cpu_to_le32(line);
319 if (!es->s_first_error_time) {
320 es->s_first_error_time = es->s_last_error_time;
321 strncpy(es->s_first_error_func, func,
322 sizeof(es->s_first_error_func));
323 es->s_first_error_line = cpu_to_le32(line);
324 es->s_first_error_ino = es->s_last_error_ino;
325 es->s_first_error_block = es->s_last_error_block;
328 * Start the daily error reporting function if it hasn't been
331 if (!es->s_error_count)
332 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
333 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
336 static void save_error_info(struct super_block *sb, const char *func,
339 __save_error_info(sb, func, line);
340 ext4_commit_super(sb, 1);
344 /* Deal with the reporting of failure conditions on a filesystem such as
345 * inconsistencies detected or read IO failures.
347 * On ext2, we can store the error state of the filesystem in the
348 * superblock. That is not possible on ext4, because we may have other
349 * write ordering constraints on the superblock which prevent us from
350 * writing it out straight away; and given that the journal is about to
351 * be aborted, we can't rely on the current, or future, transactions to
352 * write out the superblock safely.
354 * We'll just use the jbd2_journal_abort() error code to record an error in
355 * the journal instead. On recovery, the journal will complain about
356 * that error until we've noted it down and cleared it.
359 static void ext4_handle_error(struct super_block *sb)
361 if (sb->s_flags & MS_RDONLY)
364 if (!test_opt(sb, ERRORS_CONT)) {
365 journal_t *journal = EXT4_SB(sb)->s_journal;
367 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
369 jbd2_journal_abort(journal, -EIO);
371 if (test_opt(sb, ERRORS_RO)) {
372 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
373 sb->s_flags |= MS_RDONLY;
375 if (test_opt(sb, ERRORS_PANIC))
376 panic("EXT4-fs (device %s): panic forced after error\n",
380 void __ext4_error(struct super_block *sb, const char *function,
381 unsigned int line, const char *fmt, ...)
386 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
387 sb->s_id, function, line, current->comm);
392 ext4_handle_error(sb);
395 void ext4_error_inode(struct inode *inode, const char *function,
396 unsigned int line, ext4_fsblk_t block,
397 const char *fmt, ...)
400 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
402 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
403 es->s_last_error_block = cpu_to_le64(block);
404 save_error_info(inode->i_sb, function, line);
406 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
407 inode->i_sb->s_id, function, line, inode->i_ino);
409 printk("block %llu: ", block);
410 printk("comm %s: ", current->comm);
415 ext4_handle_error(inode->i_sb);
418 void ext4_error_file(struct file *file, const char *function,
419 unsigned int line, const char *fmt, ...)
422 struct ext4_super_block *es;
423 struct inode *inode = file->f_dentry->d_inode;
424 char pathname[80], *path;
426 es = EXT4_SB(inode->i_sb)->s_es;
427 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
428 save_error_info(inode->i_sb, function, line);
430 path = d_path(&(file->f_path), pathname, sizeof(pathname));
434 "EXT4-fs error (device %s): %s:%d: inode #%lu "
435 "(comm %s path %s): ",
436 inode->i_sb->s_id, function, line, inode->i_ino,
437 current->comm, path);
442 ext4_handle_error(inode->i_sb);
445 static const char *ext4_decode_error(struct super_block *sb, int errno,
452 errstr = "IO failure";
455 errstr = "Out of memory";
458 if (!sb || (EXT4_SB(sb)->s_journal &&
459 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
460 errstr = "Journal has aborted";
462 errstr = "Readonly filesystem";
465 /* If the caller passed in an extra buffer for unknown
466 * errors, textualise them now. Else we just return
469 /* Check for truncated error codes... */
470 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
479 /* __ext4_std_error decodes expected errors from journaling functions
480 * automatically and invokes the appropriate error response. */
482 void __ext4_std_error(struct super_block *sb, const char *function,
483 unsigned int line, int errno)
488 /* Special case: if the error is EROFS, and we're not already
489 * inside a transaction, then there's really no point in logging
491 if (errno == -EROFS && journal_current_handle() == NULL &&
492 (sb->s_flags & MS_RDONLY))
495 errstr = ext4_decode_error(sb, errno, nbuf);
496 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
497 sb->s_id, function, line, errstr);
498 save_error_info(sb, function, line);
500 ext4_handle_error(sb);
504 * ext4_abort is a much stronger failure handler than ext4_error. The
505 * abort function may be used to deal with unrecoverable failures such
506 * as journal IO errors or ENOMEM at a critical moment in log management.
508 * We unconditionally force the filesystem into an ABORT|READONLY state,
509 * unless the error response on the fs has been set to panic in which
510 * case we take the easy way out and panic immediately.
513 void __ext4_abort(struct super_block *sb, const char *function,
514 unsigned int line, const char *fmt, ...)
518 save_error_info(sb, function, line);
520 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
526 if ((sb->s_flags & MS_RDONLY) == 0) {
527 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
528 sb->s_flags |= MS_RDONLY;
529 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
530 if (EXT4_SB(sb)->s_journal)
531 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
532 save_error_info(sb, function, line);
534 if (test_opt(sb, ERRORS_PANIC))
535 panic("EXT4-fs panic from previous error\n");
538 void ext4_msg (struct super_block * sb, const char *prefix,
539 const char *fmt, ...)
544 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
550 void __ext4_warning(struct super_block *sb, const char *function,
551 unsigned int line, const char *fmt, ...)
556 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
557 sb->s_id, function, line);
563 void __ext4_grp_locked_error(const char *function, unsigned int line,
564 struct super_block *sb, ext4_group_t grp,
565 unsigned long ino, ext4_fsblk_t block,
566 const char *fmt, ...)
571 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
573 es->s_last_error_ino = cpu_to_le32(ino);
574 es->s_last_error_block = cpu_to_le64(block);
575 __save_error_info(sb, function, line);
577 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
578 sb->s_id, function, line, grp);
580 printk("inode %lu: ", ino);
582 printk("block %llu:", (unsigned long long) block);
587 if (test_opt(sb, ERRORS_CONT)) {
588 ext4_commit_super(sb, 0);
592 ext4_unlock_group(sb, grp);
593 ext4_handle_error(sb);
595 * We only get here in the ERRORS_RO case; relocking the group
596 * may be dangerous, but nothing bad will happen since the
597 * filesystem will have already been marked read/only and the
598 * journal has been aborted. We return 1 as a hint to callers
599 * who might what to use the return value from
600 * ext4_grp_locked_error() to distinguish beween the
601 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
602 * aggressively from the ext4 function in question, with a
603 * more appropriate error code.
605 ext4_lock_group(sb, grp);
609 void ext4_update_dynamic_rev(struct super_block *sb)
611 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
613 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
617 "updating to rev %d because of new feature flag, "
618 "running e2fsck is recommended",
621 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
622 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
623 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
624 /* leave es->s_feature_*compat flags alone */
625 /* es->s_uuid will be set by e2fsck if empty */
628 * The rest of the superblock fields should be zero, and if not it
629 * means they are likely already in use, so leave them alone. We
630 * can leave it up to e2fsck to clean up any inconsistencies there.
635 * Open the external journal device
637 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
639 struct block_device *bdev;
640 char b[BDEVNAME_SIZE];
642 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
648 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
649 __bdevname(dev, b), PTR_ERR(bdev));
654 * Release the journal device
656 static int ext4_blkdev_put(struct block_device *bdev)
659 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
662 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
664 struct block_device *bdev;
667 bdev = sbi->journal_bdev;
669 ret = ext4_blkdev_put(bdev);
670 sbi->journal_bdev = NULL;
675 static inline struct inode *orphan_list_entry(struct list_head *l)
677 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
680 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
684 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
685 le32_to_cpu(sbi->s_es->s_last_orphan));
687 printk(KERN_ERR "sb_info orphan list:\n");
688 list_for_each(l, &sbi->s_orphan) {
689 struct inode *inode = orphan_list_entry(l);
691 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
692 inode->i_sb->s_id, inode->i_ino, inode,
693 inode->i_mode, inode->i_nlink,
698 static void ext4_put_super(struct super_block *sb)
700 struct ext4_sb_info *sbi = EXT4_SB(sb);
701 struct ext4_super_block *es = sbi->s_es;
704 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
706 flush_workqueue(sbi->dio_unwritten_wq);
707 destroy_workqueue(sbi->dio_unwritten_wq);
711 ext4_commit_super(sb, 1);
713 if (sbi->s_journal) {
714 err = jbd2_journal_destroy(sbi->s_journal);
715 sbi->s_journal = NULL;
717 ext4_abort(sb, "Couldn't clean up the journal");
720 ext4_release_system_zone(sb);
722 ext4_ext_release(sb);
723 ext4_xattr_put_super(sb);
725 if (!(sb->s_flags & MS_RDONLY)) {
726 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
727 es->s_state = cpu_to_le16(sbi->s_mount_state);
728 ext4_commit_super(sb, 1);
731 remove_proc_entry(sb->s_id, ext4_proc_root);
733 kobject_del(&sbi->s_kobj);
735 for (i = 0; i < sbi->s_gdb_count; i++)
736 brelse(sbi->s_group_desc[i]);
737 kfree(sbi->s_group_desc);
738 if (is_vmalloc_addr(sbi->s_flex_groups))
739 vfree(sbi->s_flex_groups);
741 kfree(sbi->s_flex_groups);
742 percpu_counter_destroy(&sbi->s_freeblocks_counter);
743 percpu_counter_destroy(&sbi->s_freeinodes_counter);
744 percpu_counter_destroy(&sbi->s_dirs_counter);
745 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
748 for (i = 0; i < MAXQUOTAS; i++)
749 kfree(sbi->s_qf_names[i]);
752 /* Debugging code just in case the in-memory inode orphan list
753 * isn't empty. The on-disk one can be non-empty if we've
754 * detected an error and taken the fs readonly, but the
755 * in-memory list had better be clean by this point. */
756 if (!list_empty(&sbi->s_orphan))
757 dump_orphan_list(sb, sbi);
758 J_ASSERT(list_empty(&sbi->s_orphan));
760 invalidate_bdev(sb->s_bdev);
761 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
763 * Invalidate the journal device's buffers. We don't want them
764 * floating about in memory - the physical journal device may
765 * hotswapped, and it breaks the `ro-after' testing code.
767 sync_blockdev(sbi->journal_bdev);
768 invalidate_bdev(sbi->journal_bdev);
769 ext4_blkdev_remove(sbi);
771 sb->s_fs_info = NULL;
773 * Now that we are completely done shutting down the
774 * superblock, we need to actually destroy the kobject.
777 kobject_put(&sbi->s_kobj);
778 wait_for_completion(&sbi->s_kobj_unregister);
779 kfree(sbi->s_blockgroup_lock);
783 static struct kmem_cache *ext4_inode_cachep;
786 * Called inside transaction, so use GFP_NOFS
788 static struct inode *ext4_alloc_inode(struct super_block *sb)
790 struct ext4_inode_info *ei;
792 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
796 ei->vfs_inode.i_version = 1;
797 ei->vfs_inode.i_data.writeback_index = 0;
798 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
799 INIT_LIST_HEAD(&ei->i_prealloc_list);
800 spin_lock_init(&ei->i_prealloc_lock);
802 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
803 * therefore it can be null here. Don't check it, just initialize
806 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
807 ei->i_reserved_data_blocks = 0;
808 ei->i_reserved_meta_blocks = 0;
809 ei->i_allocated_meta_blocks = 0;
810 ei->i_da_metadata_calc_len = 0;
811 ei->i_delalloc_reserved_flag = 0;
812 spin_lock_init(&(ei->i_block_reservation_lock));
814 ei->i_reserved_quota = 0;
816 INIT_LIST_HEAD(&ei->i_completed_io_list);
817 spin_lock_init(&ei->i_completed_io_lock);
818 ei->cur_aio_dio = NULL;
820 ei->i_datasync_tid = 0;
822 return &ei->vfs_inode;
825 static void ext4_destroy_inode(struct inode *inode)
827 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
828 ext4_msg(inode->i_sb, KERN_ERR,
829 "Inode %lu (%p): orphan list check failed!",
830 inode->i_ino, EXT4_I(inode));
831 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
832 EXT4_I(inode), sizeof(struct ext4_inode_info),
836 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
839 static void init_once(void *foo)
841 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
843 INIT_LIST_HEAD(&ei->i_orphan);
844 #ifdef CONFIG_EXT4_FS_XATTR
845 init_rwsem(&ei->xattr_sem);
847 init_rwsem(&ei->i_data_sem);
848 inode_init_once(&ei->vfs_inode);
851 static int init_inodecache(void)
853 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
854 sizeof(struct ext4_inode_info),
855 0, (SLAB_RECLAIM_ACCOUNT|
858 if (ext4_inode_cachep == NULL)
863 static void destroy_inodecache(void)
865 kmem_cache_destroy(ext4_inode_cachep);
868 void ext4_clear_inode(struct inode *inode)
870 invalidate_inode_buffers(inode);
871 end_writeback(inode);
873 ext4_discard_preallocations(inode);
874 if (EXT4_JOURNAL(inode))
875 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
876 &EXT4_I(inode)->jinode);
879 static inline void ext4_show_quota_options(struct seq_file *seq,
880 struct super_block *sb)
882 #if defined(CONFIG_QUOTA)
883 struct ext4_sb_info *sbi = EXT4_SB(sb);
885 if (sbi->s_jquota_fmt) {
888 switch (sbi->s_jquota_fmt) {
899 seq_printf(seq, ",jqfmt=%s", fmtname);
902 if (sbi->s_qf_names[USRQUOTA])
903 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
905 if (sbi->s_qf_names[GRPQUOTA])
906 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
908 if (test_opt(sb, USRQUOTA))
909 seq_puts(seq, ",usrquota");
911 if (test_opt(sb, GRPQUOTA))
912 seq_puts(seq, ",grpquota");
918 * - it's set to a non-default value OR
919 * - if the per-sb default is different from the global default
921 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
924 unsigned long def_mount_opts;
925 struct super_block *sb = vfs->mnt_sb;
926 struct ext4_sb_info *sbi = EXT4_SB(sb);
927 struct ext4_super_block *es = sbi->s_es;
929 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
930 def_errors = le16_to_cpu(es->s_errors);
932 if (sbi->s_sb_block != 1)
933 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
934 if (test_opt(sb, MINIX_DF))
935 seq_puts(seq, ",minixdf");
936 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
937 seq_puts(seq, ",grpid");
938 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
939 seq_puts(seq, ",nogrpid");
940 if (sbi->s_resuid != EXT4_DEF_RESUID ||
941 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
942 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
944 if (sbi->s_resgid != EXT4_DEF_RESGID ||
945 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
946 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
948 if (test_opt(sb, ERRORS_RO)) {
949 if (def_errors == EXT4_ERRORS_PANIC ||
950 def_errors == EXT4_ERRORS_CONTINUE) {
951 seq_puts(seq, ",errors=remount-ro");
954 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
955 seq_puts(seq, ",errors=continue");
956 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
957 seq_puts(seq, ",errors=panic");
958 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
959 seq_puts(seq, ",nouid32");
960 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
961 seq_puts(seq, ",debug");
962 if (test_opt(sb, OLDALLOC))
963 seq_puts(seq, ",oldalloc");
964 #ifdef CONFIG_EXT4_FS_XATTR
965 if (test_opt(sb, XATTR_USER) &&
966 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
967 seq_puts(seq, ",user_xattr");
968 if (!test_opt(sb, XATTR_USER) &&
969 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
970 seq_puts(seq, ",nouser_xattr");
973 #ifdef CONFIG_EXT4_FS_POSIX_ACL
974 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
975 seq_puts(seq, ",acl");
976 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
977 seq_puts(seq, ",noacl");
979 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
980 seq_printf(seq, ",commit=%u",
981 (unsigned) (sbi->s_commit_interval / HZ));
983 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
984 seq_printf(seq, ",min_batch_time=%u",
985 (unsigned) sbi->s_min_batch_time);
987 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
988 seq_printf(seq, ",max_batch_time=%u",
989 (unsigned) sbi->s_min_batch_time);
993 * We're changing the default of barrier mount option, so
994 * let's always display its mount state so it's clear what its
997 seq_puts(seq, ",barrier=");
998 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
999 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1000 seq_puts(seq, ",journal_async_commit");
1001 else if (test_opt(sb, JOURNAL_CHECKSUM))
1002 seq_puts(seq, ",journal_checksum");
1003 if (test_opt(sb, I_VERSION))
1004 seq_puts(seq, ",i_version");
1005 if (!test_opt(sb, DELALLOC) &&
1006 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1007 seq_puts(seq, ",nodelalloc");
1010 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1012 * journal mode get enabled in different ways
1013 * So just print the value even if we didn't specify it
1015 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1016 seq_puts(seq, ",data=journal");
1017 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1018 seq_puts(seq, ",data=ordered");
1019 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1020 seq_puts(seq, ",data=writeback");
1022 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1023 seq_printf(seq, ",inode_readahead_blks=%u",
1024 sbi->s_inode_readahead_blks);
1026 if (test_opt(sb, DATA_ERR_ABORT))
1027 seq_puts(seq, ",data_err=abort");
1029 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1030 seq_puts(seq, ",noauto_da_alloc");
1032 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1033 seq_puts(seq, ",discard");
1035 if (test_opt(sb, NOLOAD))
1036 seq_puts(seq, ",norecovery");
1038 if (test_opt(sb, DIOREAD_NOLOCK))
1039 seq_puts(seq, ",dioread_nolock");
1041 if (test_opt(sb, BLOCK_VALIDITY) &&
1042 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1043 seq_puts(seq, ",block_validity");
1045 ext4_show_quota_options(seq, sb);
1050 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1051 u64 ino, u32 generation)
1053 struct inode *inode;
1055 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1056 return ERR_PTR(-ESTALE);
1057 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1058 return ERR_PTR(-ESTALE);
1060 /* iget isn't really right if the inode is currently unallocated!!
1062 * ext4_read_inode will return a bad_inode if the inode had been
1063 * deleted, so we should be safe.
1065 * Currently we don't know the generation for parent directory, so
1066 * a generation of 0 means "accept any"
1068 inode = ext4_iget(sb, ino);
1070 return ERR_CAST(inode);
1071 if (generation && inode->i_generation != generation) {
1073 return ERR_PTR(-ESTALE);
1079 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1080 int fh_len, int fh_type)
1082 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1083 ext4_nfs_get_inode);
1086 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1087 int fh_len, int fh_type)
1089 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1090 ext4_nfs_get_inode);
1094 * Try to release metadata pages (indirect blocks, directories) which are
1095 * mapped via the block device. Since these pages could have journal heads
1096 * which would prevent try_to_free_buffers() from freeing them, we must use
1097 * jbd2 layer's try_to_free_buffers() function to release them.
1099 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1102 journal_t *journal = EXT4_SB(sb)->s_journal;
1104 WARN_ON(PageChecked(page));
1105 if (!page_has_buffers(page))
1108 return jbd2_journal_try_to_free_buffers(journal, page,
1109 wait & ~__GFP_WAIT);
1110 return try_to_free_buffers(page);
1114 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1115 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1117 static int ext4_write_dquot(struct dquot *dquot);
1118 static int ext4_acquire_dquot(struct dquot *dquot);
1119 static int ext4_release_dquot(struct dquot *dquot);
1120 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1121 static int ext4_write_info(struct super_block *sb, int type);
1122 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1124 static int ext4_quota_off(struct super_block *sb, int type);
1125 static int ext4_quota_on_mount(struct super_block *sb, int type);
1126 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1127 size_t len, loff_t off);
1128 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1129 const char *data, size_t len, loff_t off);
1131 static const struct dquot_operations ext4_quota_operations = {
1133 .get_reserved_space = ext4_get_reserved_space,
1135 .write_dquot = ext4_write_dquot,
1136 .acquire_dquot = ext4_acquire_dquot,
1137 .release_dquot = ext4_release_dquot,
1138 .mark_dirty = ext4_mark_dquot_dirty,
1139 .write_info = ext4_write_info,
1140 .alloc_dquot = dquot_alloc,
1141 .destroy_dquot = dquot_destroy,
1144 static const struct quotactl_ops ext4_qctl_operations = {
1145 .quota_on = ext4_quota_on,
1146 .quota_off = ext4_quota_off,
1147 .quota_sync = dquot_quota_sync,
1148 .get_info = dquot_get_dqinfo,
1149 .set_info = dquot_set_dqinfo,
1150 .get_dqblk = dquot_get_dqblk,
1151 .set_dqblk = dquot_set_dqblk
1155 static const struct super_operations ext4_sops = {
1156 .alloc_inode = ext4_alloc_inode,
1157 .destroy_inode = ext4_destroy_inode,
1158 .write_inode = ext4_write_inode,
1159 .dirty_inode = ext4_dirty_inode,
1160 .evict_inode = ext4_evict_inode,
1161 .put_super = ext4_put_super,
1162 .sync_fs = ext4_sync_fs,
1163 .freeze_fs = ext4_freeze,
1164 .unfreeze_fs = ext4_unfreeze,
1165 .statfs = ext4_statfs,
1166 .remount_fs = ext4_remount,
1167 .show_options = ext4_show_options,
1169 .quota_read = ext4_quota_read,
1170 .quota_write = ext4_quota_write,
1172 .bdev_try_to_free_page = bdev_try_to_free_page,
1175 static const struct super_operations ext4_nojournal_sops = {
1176 .alloc_inode = ext4_alloc_inode,
1177 .destroy_inode = ext4_destroy_inode,
1178 .write_inode = ext4_write_inode,
1179 .dirty_inode = ext4_dirty_inode,
1180 .evict_inode = ext4_evict_inode,
1181 .write_super = ext4_write_super,
1182 .put_super = ext4_put_super,
1183 .statfs = ext4_statfs,
1184 .remount_fs = ext4_remount,
1185 .show_options = ext4_show_options,
1187 .quota_read = ext4_quota_read,
1188 .quota_write = ext4_quota_write,
1190 .bdev_try_to_free_page = bdev_try_to_free_page,
1193 static const struct export_operations ext4_export_ops = {
1194 .fh_to_dentry = ext4_fh_to_dentry,
1195 .fh_to_parent = ext4_fh_to_parent,
1196 .get_parent = ext4_get_parent,
1200 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1201 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1202 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1203 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1204 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1205 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1206 Opt_journal_update, Opt_journal_dev,
1207 Opt_journal_checksum, Opt_journal_async_commit,
1208 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1209 Opt_data_err_abort, Opt_data_err_ignore,
1210 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1211 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1212 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1213 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1214 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1215 Opt_block_validity, Opt_noblock_validity,
1216 Opt_inode_readahead_blks, Opt_journal_ioprio,
1217 Opt_dioread_nolock, Opt_dioread_lock,
1218 Opt_discard, Opt_nodiscard,
1221 static const match_table_t tokens = {
1222 {Opt_bsd_df, "bsddf"},
1223 {Opt_minix_df, "minixdf"},
1224 {Opt_grpid, "grpid"},
1225 {Opt_grpid, "bsdgroups"},
1226 {Opt_nogrpid, "nogrpid"},
1227 {Opt_nogrpid, "sysvgroups"},
1228 {Opt_resgid, "resgid=%u"},
1229 {Opt_resuid, "resuid=%u"},
1231 {Opt_err_cont, "errors=continue"},
1232 {Opt_err_panic, "errors=panic"},
1233 {Opt_err_ro, "errors=remount-ro"},
1234 {Opt_nouid32, "nouid32"},
1235 {Opt_debug, "debug"},
1236 {Opt_oldalloc, "oldalloc"},
1237 {Opt_orlov, "orlov"},
1238 {Opt_user_xattr, "user_xattr"},
1239 {Opt_nouser_xattr, "nouser_xattr"},
1241 {Opt_noacl, "noacl"},
1242 {Opt_noload, "noload"},
1243 {Opt_noload, "norecovery"},
1246 {Opt_commit, "commit=%u"},
1247 {Opt_min_batch_time, "min_batch_time=%u"},
1248 {Opt_max_batch_time, "max_batch_time=%u"},
1249 {Opt_journal_update, "journal=update"},
1250 {Opt_journal_dev, "journal_dev=%u"},
1251 {Opt_journal_checksum, "journal_checksum"},
1252 {Opt_journal_async_commit, "journal_async_commit"},
1253 {Opt_abort, "abort"},
1254 {Opt_data_journal, "data=journal"},
1255 {Opt_data_ordered, "data=ordered"},
1256 {Opt_data_writeback, "data=writeback"},
1257 {Opt_data_err_abort, "data_err=abort"},
1258 {Opt_data_err_ignore, "data_err=ignore"},
1259 {Opt_offusrjquota, "usrjquota="},
1260 {Opt_usrjquota, "usrjquota=%s"},
1261 {Opt_offgrpjquota, "grpjquota="},
1262 {Opt_grpjquota, "grpjquota=%s"},
1263 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1264 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1265 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1266 {Opt_grpquota, "grpquota"},
1267 {Opt_noquota, "noquota"},
1268 {Opt_quota, "quota"},
1269 {Opt_usrquota, "usrquota"},
1270 {Opt_barrier, "barrier=%u"},
1271 {Opt_barrier, "barrier"},
1272 {Opt_nobarrier, "nobarrier"},
1273 {Opt_i_version, "i_version"},
1274 {Opt_stripe, "stripe=%u"},
1275 {Opt_resize, "resize"},
1276 {Opt_delalloc, "delalloc"},
1277 {Opt_nodelalloc, "nodelalloc"},
1278 {Opt_block_validity, "block_validity"},
1279 {Opt_noblock_validity, "noblock_validity"},
1280 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1281 {Opt_journal_ioprio, "journal_ioprio=%u"},
1282 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1283 {Opt_auto_da_alloc, "auto_da_alloc"},
1284 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1285 {Opt_dioread_nolock, "dioread_nolock"},
1286 {Opt_dioread_lock, "dioread_lock"},
1287 {Opt_discard, "discard"},
1288 {Opt_nodiscard, "nodiscard"},
1292 static ext4_fsblk_t get_sb_block(void **data)
1294 ext4_fsblk_t sb_block;
1295 char *options = (char *) *data;
1297 if (!options || strncmp(options, "sb=", 3) != 0)
1298 return 1; /* Default location */
1301 /* TODO: use simple_strtoll with >32bit ext4 */
1302 sb_block = simple_strtoul(options, &options, 0);
1303 if (*options && *options != ',') {
1304 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1308 if (*options == ',')
1310 *data = (void *) options;
1315 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1316 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1317 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1320 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1322 struct ext4_sb_info *sbi = EXT4_SB(sb);
1325 if (sb_any_quota_loaded(sb) &&
1326 !sbi->s_qf_names[qtype]) {
1327 ext4_msg(sb, KERN_ERR,
1328 "Cannot change journaled "
1329 "quota options when quota turned on");
1332 qname = match_strdup(args);
1334 ext4_msg(sb, KERN_ERR,
1335 "Not enough memory for storing quotafile name");
1338 if (sbi->s_qf_names[qtype] &&
1339 strcmp(sbi->s_qf_names[qtype], qname)) {
1340 ext4_msg(sb, KERN_ERR,
1341 "%s quota file already specified", QTYPE2NAME(qtype));
1345 sbi->s_qf_names[qtype] = qname;
1346 if (strchr(sbi->s_qf_names[qtype], '/')) {
1347 ext4_msg(sb, KERN_ERR,
1348 "quotafile must be on filesystem root");
1349 kfree(sbi->s_qf_names[qtype]);
1350 sbi->s_qf_names[qtype] = NULL;
1353 set_opt(sbi->s_mount_opt, QUOTA);
1357 static int clear_qf_name(struct super_block *sb, int qtype)
1360 struct ext4_sb_info *sbi = EXT4_SB(sb);
1362 if (sb_any_quota_loaded(sb) &&
1363 sbi->s_qf_names[qtype]) {
1364 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1365 " when quota turned on");
1369 * The space will be released later when all options are confirmed
1372 sbi->s_qf_names[qtype] = NULL;
1377 static int parse_options(char *options, struct super_block *sb,
1378 unsigned long *journal_devnum,
1379 unsigned int *journal_ioprio,
1380 ext4_fsblk_t *n_blocks_count, int is_remount)
1382 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384 substring_t args[MAX_OPT_ARGS];
1394 while ((p = strsep(&options, ",")) != NULL) {
1400 * Initialize args struct so we know whether arg was
1401 * found; some options take optional arguments.
1403 args[0].to = args[0].from = 0;
1404 token = match_token(p, tokens, args);
1407 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1408 clear_opt(sbi->s_mount_opt, MINIX_DF);
1411 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1412 set_opt(sbi->s_mount_opt, MINIX_DF);
1416 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1417 set_opt(sbi->s_mount_opt, GRPID);
1421 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1422 clear_opt(sbi->s_mount_opt, GRPID);
1426 if (match_int(&args[0], &option))
1428 sbi->s_resuid = option;
1431 if (match_int(&args[0], &option))
1433 sbi->s_resgid = option;
1436 /* handled by get_sb_block() instead of here */
1437 /* *sb_block = match_int(&args[0]); */
1440 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1441 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1442 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1445 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1446 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1447 set_opt(sbi->s_mount_opt, ERRORS_RO);
1450 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1451 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1452 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1455 set_opt(sbi->s_mount_opt, NO_UID32);
1458 set_opt(sbi->s_mount_opt, DEBUG);
1461 set_opt(sbi->s_mount_opt, OLDALLOC);
1464 clear_opt(sbi->s_mount_opt, OLDALLOC);
1466 #ifdef CONFIG_EXT4_FS_XATTR
1467 case Opt_user_xattr:
1468 set_opt(sbi->s_mount_opt, XATTR_USER);
1470 case Opt_nouser_xattr:
1471 clear_opt(sbi->s_mount_opt, XATTR_USER);
1474 case Opt_user_xattr:
1475 case Opt_nouser_xattr:
1476 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1479 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1481 set_opt(sbi->s_mount_opt, POSIX_ACL);
1484 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1489 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1492 case Opt_journal_update:
1494 /* Eventually we will want to be able to create
1495 a journal file here. For now, only allow the
1496 user to specify an existing inode to be the
1499 ext4_msg(sb, KERN_ERR,
1500 "Cannot specify journal on remount");
1503 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1505 case Opt_journal_dev:
1507 ext4_msg(sb, KERN_ERR,
1508 "Cannot specify journal on remount");
1511 if (match_int(&args[0], &option))
1513 *journal_devnum = option;
1515 case Opt_journal_checksum:
1516 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1518 case Opt_journal_async_commit:
1519 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1520 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1523 set_opt(sbi->s_mount_opt, NOLOAD);
1526 if (match_int(&args[0], &option))
1531 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1532 sbi->s_commit_interval = HZ * option;
1534 case Opt_max_batch_time:
1535 if (match_int(&args[0], &option))
1540 option = EXT4_DEF_MAX_BATCH_TIME;
1541 sbi->s_max_batch_time = option;
1543 case Opt_min_batch_time:
1544 if (match_int(&args[0], &option))
1548 sbi->s_min_batch_time = option;
1550 case Opt_data_journal:
1551 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1553 case Opt_data_ordered:
1554 data_opt = EXT4_MOUNT_ORDERED_DATA;
1556 case Opt_data_writeback:
1557 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1560 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1561 ext4_msg(sb, KERN_ERR,
1562 "Cannot change data mode on remount");
1566 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1567 sbi->s_mount_opt |= data_opt;
1570 case Opt_data_err_abort:
1571 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1573 case Opt_data_err_ignore:
1574 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1578 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1582 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1585 case Opt_offusrjquota:
1586 if (!clear_qf_name(sb, USRQUOTA))
1589 case Opt_offgrpjquota:
1590 if (!clear_qf_name(sb, GRPQUOTA))
1594 case Opt_jqfmt_vfsold:
1595 qfmt = QFMT_VFS_OLD;
1597 case Opt_jqfmt_vfsv0:
1600 case Opt_jqfmt_vfsv1:
1603 if (sb_any_quota_loaded(sb) &&
1604 sbi->s_jquota_fmt != qfmt) {
1605 ext4_msg(sb, KERN_ERR, "Cannot change "
1606 "journaled quota options when "
1610 sbi->s_jquota_fmt = qfmt;
1614 set_opt(sbi->s_mount_opt, QUOTA);
1615 set_opt(sbi->s_mount_opt, USRQUOTA);
1618 set_opt(sbi->s_mount_opt, QUOTA);
1619 set_opt(sbi->s_mount_opt, GRPQUOTA);
1622 if (sb_any_quota_loaded(sb)) {
1623 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1624 "options when quota turned on");
1627 clear_opt(sbi->s_mount_opt, QUOTA);
1628 clear_opt(sbi->s_mount_opt, USRQUOTA);
1629 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1635 ext4_msg(sb, KERN_ERR,
1636 "quota options not supported");
1640 case Opt_offusrjquota:
1641 case Opt_offgrpjquota:
1642 case Opt_jqfmt_vfsold:
1643 case Opt_jqfmt_vfsv0:
1644 case Opt_jqfmt_vfsv1:
1645 ext4_msg(sb, KERN_ERR,
1646 "journaled quota options not supported");
1652 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1655 clear_opt(sbi->s_mount_opt, BARRIER);
1659 if (match_int(&args[0], &option))
1662 option = 1; /* No argument, default to 1 */
1664 set_opt(sbi->s_mount_opt, BARRIER);
1666 clear_opt(sbi->s_mount_opt, BARRIER);
1672 ext4_msg(sb, KERN_ERR,
1673 "resize option only available "
1677 if (match_int(&args[0], &option) != 0)
1679 *n_blocks_count = option;
1682 ext4_msg(sb, KERN_WARNING,
1683 "Ignoring deprecated nobh option");
1686 ext4_msg(sb, KERN_WARNING,
1687 "Ignoring deprecated bh option");
1690 set_opt(sbi->s_mount_opt, I_VERSION);
1691 sb->s_flags |= MS_I_VERSION;
1693 case Opt_nodelalloc:
1694 clear_opt(sbi->s_mount_opt, DELALLOC);
1697 if (match_int(&args[0], &option))
1701 sbi->s_stripe = option;
1704 set_opt(sbi->s_mount_opt, DELALLOC);
1706 case Opt_block_validity:
1707 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1709 case Opt_noblock_validity:
1710 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1712 case Opt_inode_readahead_blks:
1713 if (match_int(&args[0], &option))
1715 if (option < 0 || option > (1 << 30))
1717 if (!is_power_of_2(option)) {
1718 ext4_msg(sb, KERN_ERR,
1719 "EXT4-fs: inode_readahead_blks"
1720 " must be a power of 2");
1723 sbi->s_inode_readahead_blks = option;
1725 case Opt_journal_ioprio:
1726 if (match_int(&args[0], &option))
1728 if (option < 0 || option > 7)
1730 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1733 case Opt_noauto_da_alloc:
1734 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1736 case Opt_auto_da_alloc:
1738 if (match_int(&args[0], &option))
1741 option = 1; /* No argument, default to 1 */
1743 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1745 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1748 set_opt(sbi->s_mount_opt, DISCARD);
1751 clear_opt(sbi->s_mount_opt, DISCARD);
1753 case Opt_dioread_nolock:
1754 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1756 case Opt_dioread_lock:
1757 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1760 ext4_msg(sb, KERN_ERR,
1761 "Unrecognized mount option \"%s\" "
1762 "or missing value", p);
1767 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1768 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1769 clear_opt(sbi->s_mount_opt, USRQUOTA);
1771 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1772 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1774 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1775 ext4_msg(sb, KERN_ERR, "old and new quota "
1780 if (!sbi->s_jquota_fmt) {
1781 ext4_msg(sb, KERN_ERR, "journaled quota format "
1786 if (sbi->s_jquota_fmt) {
1787 ext4_msg(sb, KERN_ERR, "journaled quota format "
1788 "specified with no journaling "
1797 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1800 struct ext4_sb_info *sbi = EXT4_SB(sb);
1803 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1804 ext4_msg(sb, KERN_ERR, "revision level too high, "
1805 "forcing read-only mode");
1810 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1811 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1812 "running e2fsck is recommended");
1813 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1814 ext4_msg(sb, KERN_WARNING,
1815 "warning: mounting fs with errors, "
1816 "running e2fsck is recommended");
1817 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1818 le16_to_cpu(es->s_mnt_count) >=
1819 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1820 ext4_msg(sb, KERN_WARNING,
1821 "warning: maximal mount count reached, "
1822 "running e2fsck is recommended");
1823 else if (le32_to_cpu(es->s_checkinterval) &&
1824 (le32_to_cpu(es->s_lastcheck) +
1825 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1826 ext4_msg(sb, KERN_WARNING,
1827 "warning: checktime reached, "
1828 "running e2fsck is recommended");
1829 if (!sbi->s_journal)
1830 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1831 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1832 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1833 le16_add_cpu(&es->s_mnt_count, 1);
1834 es->s_mtime = cpu_to_le32(get_seconds());
1835 ext4_update_dynamic_rev(sb);
1837 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1839 ext4_commit_super(sb, 1);
1840 if (test_opt(sb, DEBUG))
1841 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1842 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1844 sbi->s_groups_count,
1845 EXT4_BLOCKS_PER_GROUP(sb),
1846 EXT4_INODES_PER_GROUP(sb),
1852 static int ext4_fill_flex_info(struct super_block *sb)
1854 struct ext4_sb_info *sbi = EXT4_SB(sb);
1855 struct ext4_group_desc *gdp = NULL;
1856 ext4_group_t flex_group_count;
1857 ext4_group_t flex_group;
1858 int groups_per_flex = 0;
1862 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1863 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1865 if (groups_per_flex < 2) {
1866 sbi->s_log_groups_per_flex = 0;
1870 /* We allocate both existing and potentially added groups */
1871 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1872 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1873 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1874 size = flex_group_count * sizeof(struct flex_groups);
1875 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1876 if (sbi->s_flex_groups == NULL) {
1877 sbi->s_flex_groups = vmalloc(size);
1878 if (sbi->s_flex_groups)
1879 memset(sbi->s_flex_groups, 0, size);
1881 if (sbi->s_flex_groups == NULL) {
1882 ext4_msg(sb, KERN_ERR, "not enough memory for "
1883 "%u flex groups", flex_group_count);
1887 for (i = 0; i < sbi->s_groups_count; i++) {
1888 gdp = ext4_get_group_desc(sb, i, NULL);
1890 flex_group = ext4_flex_group(sbi, i);
1891 atomic_add(ext4_free_inodes_count(sb, gdp),
1892 &sbi->s_flex_groups[flex_group].free_inodes);
1893 atomic_add(ext4_free_blks_count(sb, gdp),
1894 &sbi->s_flex_groups[flex_group].free_blocks);
1895 atomic_add(ext4_used_dirs_count(sb, gdp),
1896 &sbi->s_flex_groups[flex_group].used_dirs);
1904 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1905 struct ext4_group_desc *gdp)
1909 if (sbi->s_es->s_feature_ro_compat &
1910 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1911 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1912 __le32 le_group = cpu_to_le32(block_group);
1914 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1915 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1916 crc = crc16(crc, (__u8 *)gdp, offset);
1917 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1918 /* for checksum of struct ext4_group_desc do the rest...*/
1919 if ((sbi->s_es->s_feature_incompat &
1920 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1921 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1922 crc = crc16(crc, (__u8 *)gdp + offset,
1923 le16_to_cpu(sbi->s_es->s_desc_size) -
1927 return cpu_to_le16(crc);
1930 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1931 struct ext4_group_desc *gdp)
1933 if ((sbi->s_es->s_feature_ro_compat &
1934 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1935 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1941 /* Called at mount-time, super-block is locked */
1942 static int ext4_check_descriptors(struct super_block *sb)
1944 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1946 ext4_fsblk_t last_block;
1947 ext4_fsblk_t block_bitmap;
1948 ext4_fsblk_t inode_bitmap;
1949 ext4_fsblk_t inode_table;
1950 int flexbg_flag = 0;
1953 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1956 ext4_debug("Checking group descriptors");
1958 for (i = 0; i < sbi->s_groups_count; i++) {
1959 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1961 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1962 last_block = ext4_blocks_count(sbi->s_es) - 1;
1964 last_block = first_block +
1965 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1967 block_bitmap = ext4_block_bitmap(sb, gdp);
1968 if (block_bitmap < first_block || block_bitmap > last_block) {
1969 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1970 "Block bitmap for group %u not in group "
1971 "(block %llu)!", i, block_bitmap);
1974 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1975 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1976 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1977 "Inode bitmap for group %u not in group "
1978 "(block %llu)!", i, inode_bitmap);
1981 inode_table = ext4_inode_table(sb, gdp);
1982 if (inode_table < first_block ||
1983 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1984 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1985 "Inode table for group %u not in group "
1986 "(block %llu)!", i, inode_table);
1989 ext4_lock_group(sb, i);
1990 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1991 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1992 "Checksum for group %u failed (%u!=%u)",
1993 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1994 gdp)), le16_to_cpu(gdp->bg_checksum));
1995 if (!(sb->s_flags & MS_RDONLY)) {
1996 ext4_unlock_group(sb, i);
2000 ext4_unlock_group(sb, i);
2002 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2005 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2006 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2010 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2011 * the superblock) which were deleted from all directories, but held open by
2012 * a process at the time of a crash. We walk the list and try to delete these
2013 * inodes at recovery time (only with a read-write filesystem).
2015 * In order to keep the orphan inode chain consistent during traversal (in
2016 * case of crash during recovery), we link each inode into the superblock
2017 * orphan list_head and handle it the same way as an inode deletion during
2018 * normal operation (which journals the operations for us).
2020 * We only do an iget() and an iput() on each inode, which is very safe if we
2021 * accidentally point at an in-use or already deleted inode. The worst that
2022 * can happen in this case is that we get a "bit already cleared" message from
2023 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2024 * e2fsck was run on this filesystem, and it must have already done the orphan
2025 * inode cleanup for us, so we can safely abort without any further action.
2027 static void ext4_orphan_cleanup(struct super_block *sb,
2028 struct ext4_super_block *es)
2030 unsigned int s_flags = sb->s_flags;
2031 int nr_orphans = 0, nr_truncates = 0;
2035 if (!es->s_last_orphan) {
2036 jbd_debug(4, "no orphan inodes to clean up\n");
2040 if (bdev_read_only(sb->s_bdev)) {
2041 ext4_msg(sb, KERN_ERR, "write access "
2042 "unavailable, skipping orphan cleanup");
2046 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2047 if (es->s_last_orphan)
2048 jbd_debug(1, "Errors on filesystem, "
2049 "clearing orphan list.\n");
2050 es->s_last_orphan = 0;
2051 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2055 if (s_flags & MS_RDONLY) {
2056 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2057 sb->s_flags &= ~MS_RDONLY;
2060 /* Needed for iput() to work correctly and not trash data */
2061 sb->s_flags |= MS_ACTIVE;
2062 /* Turn on quotas so that they are updated correctly */
2063 for (i = 0; i < MAXQUOTAS; i++) {
2064 if (EXT4_SB(sb)->s_qf_names[i]) {
2065 int ret = ext4_quota_on_mount(sb, i);
2067 ext4_msg(sb, KERN_ERR,
2068 "Cannot turn on journaled "
2069 "quota: error %d", ret);
2074 while (es->s_last_orphan) {
2075 struct inode *inode;
2077 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2078 if (IS_ERR(inode)) {
2079 es->s_last_orphan = 0;
2083 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2084 dquot_initialize(inode);
2085 if (inode->i_nlink) {
2086 ext4_msg(sb, KERN_DEBUG,
2087 "%s: truncating inode %lu to %lld bytes",
2088 __func__, inode->i_ino, inode->i_size);
2089 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2090 inode->i_ino, inode->i_size);
2091 ext4_truncate(inode);
2094 ext4_msg(sb, KERN_DEBUG,
2095 "%s: deleting unreferenced inode %lu",
2096 __func__, inode->i_ino);
2097 jbd_debug(2, "deleting unreferenced inode %lu\n",
2101 iput(inode); /* The delete magic happens here! */
2104 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2107 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2108 PLURAL(nr_orphans));
2110 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2111 PLURAL(nr_truncates));
2113 /* Turn quotas off */
2114 for (i = 0; i < MAXQUOTAS; i++) {
2115 if (sb_dqopt(sb)->files[i])
2116 dquot_quota_off(sb, i);
2119 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2123 * Maximal extent format file size.
2124 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2125 * extent format containers, within a sector_t, and within i_blocks
2126 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2127 * so that won't be a limiting factor.
2129 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2131 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2134 loff_t upper_limit = MAX_LFS_FILESIZE;
2136 /* small i_blocks in vfs inode? */
2137 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2139 * CONFIG_LBDAF is not enabled implies the inode
2140 * i_block represent total blocks in 512 bytes
2141 * 32 == size of vfs inode i_blocks * 8
2143 upper_limit = (1LL << 32) - 1;
2145 /* total blocks in file system block size */
2146 upper_limit >>= (blkbits - 9);
2147 upper_limit <<= blkbits;
2150 /* 32-bit extent-start container, ee_block */
2155 /* Sanity check against vm- & vfs- imposed limits */
2156 if (res > upper_limit)
2163 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2164 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2165 * We need to be 1 filesystem block less than the 2^48 sector limit.
2167 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2169 loff_t res = EXT4_NDIR_BLOCKS;
2172 /* This is calculated to be the largest file size for a dense, block
2173 * mapped file such that the file's total number of 512-byte sectors,
2174 * including data and all indirect blocks, does not exceed (2^48 - 1).
2176 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2177 * number of 512-byte sectors of the file.
2180 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2182 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2183 * the inode i_block field represents total file blocks in
2184 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2186 upper_limit = (1LL << 32) - 1;
2188 /* total blocks in file system block size */
2189 upper_limit >>= (bits - 9);
2193 * We use 48 bit ext4_inode i_blocks
2194 * With EXT4_HUGE_FILE_FL set the i_blocks
2195 * represent total number of blocks in
2196 * file system block size
2198 upper_limit = (1LL << 48) - 1;
2202 /* indirect blocks */
2204 /* double indirect blocks */
2205 meta_blocks += 1 + (1LL << (bits-2));
2206 /* tripple indirect blocks */
2207 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2209 upper_limit -= meta_blocks;
2210 upper_limit <<= bits;
2212 res += 1LL << (bits-2);
2213 res += 1LL << (2*(bits-2));
2214 res += 1LL << (3*(bits-2));
2216 if (res > upper_limit)
2219 if (res > MAX_LFS_FILESIZE)
2220 res = MAX_LFS_FILESIZE;
2225 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2226 ext4_fsblk_t logical_sb_block, int nr)
2228 struct ext4_sb_info *sbi = EXT4_SB(sb);
2229 ext4_group_t bg, first_meta_bg;
2232 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2234 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2236 return logical_sb_block + nr + 1;
2237 bg = sbi->s_desc_per_block * nr;
2238 if (ext4_bg_has_super(sb, bg))
2241 return (has_super + ext4_group_first_block_no(sb, bg));
2245 * ext4_get_stripe_size: Get the stripe size.
2246 * @sbi: In memory super block info
2248 * If we have specified it via mount option, then
2249 * use the mount option value. If the value specified at mount time is
2250 * greater than the blocks per group use the super block value.
2251 * If the super block value is greater than blocks per group return 0.
2252 * Allocator needs it be less than blocks per group.
2255 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2257 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2258 unsigned long stripe_width =
2259 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2261 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2262 return sbi->s_stripe;
2264 if (stripe_width <= sbi->s_blocks_per_group)
2265 return stripe_width;
2267 if (stride <= sbi->s_blocks_per_group)
2276 struct attribute attr;
2277 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2278 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2279 const char *, size_t);
2283 static int parse_strtoul(const char *buf,
2284 unsigned long max, unsigned long *value)
2288 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2289 endp = skip_spaces(endp);
2290 if (*endp || *value > max)
2296 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2297 struct ext4_sb_info *sbi,
2300 return snprintf(buf, PAGE_SIZE, "%llu\n",
2301 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2304 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2305 struct ext4_sb_info *sbi, char *buf)
2307 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2309 if (!sb->s_bdev->bd_part)
2310 return snprintf(buf, PAGE_SIZE, "0\n");
2311 return snprintf(buf, PAGE_SIZE, "%lu\n",
2312 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2313 sbi->s_sectors_written_start) >> 1);
2316 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2317 struct ext4_sb_info *sbi, char *buf)
2319 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2321 if (!sb->s_bdev->bd_part)
2322 return snprintf(buf, PAGE_SIZE, "0\n");
2323 return snprintf(buf, PAGE_SIZE, "%llu\n",
2324 (unsigned long long)(sbi->s_kbytes_written +
2325 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2326 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2329 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2330 struct ext4_sb_info *sbi,
2331 const char *buf, size_t count)
2335 if (parse_strtoul(buf, 0x40000000, &t))
2338 if (!is_power_of_2(t))
2341 sbi->s_inode_readahead_blks = t;
2345 static ssize_t sbi_ui_show(struct ext4_attr *a,
2346 struct ext4_sb_info *sbi, char *buf)
2348 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2350 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2353 static ssize_t sbi_ui_store(struct ext4_attr *a,
2354 struct ext4_sb_info *sbi,
2355 const char *buf, size_t count)
2357 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2360 if (parse_strtoul(buf, 0xffffffff, &t))
2366 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2367 static struct ext4_attr ext4_attr_##_name = { \
2368 .attr = {.name = __stringify(_name), .mode = _mode }, \
2371 .offset = offsetof(struct ext4_sb_info, _elname), \
2373 #define EXT4_ATTR(name, mode, show, store) \
2374 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2376 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2377 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2378 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2379 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2380 #define ATTR_LIST(name) &ext4_attr_##name.attr
2382 EXT4_RO_ATTR(delayed_allocation_blocks);
2383 EXT4_RO_ATTR(session_write_kbytes);
2384 EXT4_RO_ATTR(lifetime_write_kbytes);
2385 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2386 inode_readahead_blks_store, s_inode_readahead_blks);
2387 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2388 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2389 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2390 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2391 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2392 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2393 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2394 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2396 static struct attribute *ext4_attrs[] = {
2397 ATTR_LIST(delayed_allocation_blocks),
2398 ATTR_LIST(session_write_kbytes),
2399 ATTR_LIST(lifetime_write_kbytes),
2400 ATTR_LIST(inode_readahead_blks),
2401 ATTR_LIST(inode_goal),
2402 ATTR_LIST(mb_stats),
2403 ATTR_LIST(mb_max_to_scan),
2404 ATTR_LIST(mb_min_to_scan),
2405 ATTR_LIST(mb_order2_req),
2406 ATTR_LIST(mb_stream_req),
2407 ATTR_LIST(mb_group_prealloc),
2408 ATTR_LIST(max_writeback_mb_bump),
2412 static ssize_t ext4_attr_show(struct kobject *kobj,
2413 struct attribute *attr, char *buf)
2415 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2417 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2419 return a->show ? a->show(a, sbi, buf) : 0;
2422 static ssize_t ext4_attr_store(struct kobject *kobj,
2423 struct attribute *attr,
2424 const char *buf, size_t len)
2426 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2428 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2430 return a->store ? a->store(a, sbi, buf, len) : 0;
2433 static void ext4_sb_release(struct kobject *kobj)
2435 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2437 complete(&sbi->s_kobj_unregister);
2441 static const struct sysfs_ops ext4_attr_ops = {
2442 .show = ext4_attr_show,
2443 .store = ext4_attr_store,
2446 static struct kobj_type ext4_ktype = {
2447 .default_attrs = ext4_attrs,
2448 .sysfs_ops = &ext4_attr_ops,
2449 .release = ext4_sb_release,
2453 * Check whether this filesystem can be mounted based on
2454 * the features present and the RDONLY/RDWR mount requested.
2455 * Returns 1 if this filesystem can be mounted as requested,
2456 * 0 if it cannot be.
2458 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2460 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2461 ext4_msg(sb, KERN_ERR,
2462 "Couldn't mount because of "
2463 "unsupported optional features (%x)",
2464 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2465 ~EXT4_FEATURE_INCOMPAT_SUPP));
2472 /* Check that feature set is OK for a read-write mount */
2473 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2474 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2475 "unsupported optional features (%x)",
2476 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2477 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2481 * Large file size enabled file system can only be mounted
2482 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2484 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2485 if (sizeof(blkcnt_t) < sizeof(u64)) {
2486 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2487 "cannot be mounted RDWR without "
2496 * This function is called once a day if we have errors logged
2497 * on the file system
2499 static void print_daily_error_info(unsigned long arg)
2501 struct super_block *sb = (struct super_block *) arg;
2502 struct ext4_sb_info *sbi;
2503 struct ext4_super_block *es;
2508 if (es->s_error_count)
2509 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2510 le32_to_cpu(es->s_error_count));
2511 if (es->s_first_error_time) {
2512 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2513 sb->s_id, le32_to_cpu(es->s_first_error_time),
2514 (int) sizeof(es->s_first_error_func),
2515 es->s_first_error_func,
2516 le32_to_cpu(es->s_first_error_line));
2517 if (es->s_first_error_ino)
2518 printk(": inode %u",
2519 le32_to_cpu(es->s_first_error_ino));
2520 if (es->s_first_error_block)
2521 printk(": block %llu", (unsigned long long)
2522 le64_to_cpu(es->s_first_error_block));
2525 if (es->s_last_error_time) {
2526 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2527 sb->s_id, le32_to_cpu(es->s_last_error_time),
2528 (int) sizeof(es->s_last_error_func),
2529 es->s_last_error_func,
2530 le32_to_cpu(es->s_last_error_line));
2531 if (es->s_last_error_ino)
2532 printk(": inode %u",
2533 le32_to_cpu(es->s_last_error_ino));
2534 if (es->s_last_error_block)
2535 printk(": block %llu", (unsigned long long)
2536 le64_to_cpu(es->s_last_error_block));
2539 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2542 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2543 __releases(kernel_lock)
2544 __acquires(kernel_lock)
2546 char *orig_data = kstrdup(data, GFP_KERNEL);
2547 struct buffer_head *bh;
2548 struct ext4_super_block *es = NULL;
2549 struct ext4_sb_info *sbi;
2551 ext4_fsblk_t sb_block = get_sb_block(&data);
2552 ext4_fsblk_t logical_sb_block;
2553 unsigned long offset = 0;
2554 unsigned long journal_devnum = 0;
2555 unsigned long def_mount_opts;
2561 unsigned int db_count;
2563 int needs_recovery, has_huge_files;
2566 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2568 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2572 sbi->s_blockgroup_lock =
2573 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2574 if (!sbi->s_blockgroup_lock) {
2578 sb->s_fs_info = sbi;
2579 sbi->s_mount_opt = 0;
2580 sbi->s_resuid = EXT4_DEF_RESUID;
2581 sbi->s_resgid = EXT4_DEF_RESGID;
2582 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2583 sbi->s_sb_block = sb_block;
2584 if (sb->s_bdev->bd_part)
2585 sbi->s_sectors_written_start =
2586 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2588 /* Cleanup superblock name */
2589 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2593 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2595 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2600 * The ext4 superblock will not be buffer aligned for other than 1kB
2601 * block sizes. We need to calculate the offset from buffer start.
2603 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2604 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2605 offset = do_div(logical_sb_block, blocksize);
2607 logical_sb_block = sb_block;
2610 if (!(bh = sb_bread(sb, logical_sb_block))) {
2611 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2615 * Note: s_es must be initialized as soon as possible because
2616 * some ext4 macro-instructions depend on its value
2618 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2620 sb->s_magic = le16_to_cpu(es->s_magic);
2621 if (sb->s_magic != EXT4_SUPER_MAGIC)
2623 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2625 /* Set defaults before we parse the mount options */
2626 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2627 if (def_mount_opts & EXT4_DEFM_DEBUG)
2628 set_opt(sbi->s_mount_opt, DEBUG);
2629 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2630 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2632 set_opt(sbi->s_mount_opt, GRPID);
2634 if (def_mount_opts & EXT4_DEFM_UID16)
2635 set_opt(sbi->s_mount_opt, NO_UID32);
2636 #ifdef CONFIG_EXT4_FS_XATTR
2637 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2638 set_opt(sbi->s_mount_opt, XATTR_USER);
2640 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2641 if (def_mount_opts & EXT4_DEFM_ACL)
2642 set_opt(sbi->s_mount_opt, POSIX_ACL);
2644 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2645 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2646 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2647 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2648 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2649 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2651 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2652 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2653 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2654 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2656 set_opt(sbi->s_mount_opt, ERRORS_RO);
2657 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
2658 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
2659 if (def_mount_opts & EXT4_DEFM_DISCARD)
2660 set_opt(sbi->s_mount_opt, DISCARD);
2662 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2663 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2664 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2665 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2666 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2668 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
2669 set_opt(sbi->s_mount_opt, BARRIER);
2672 * enable delayed allocation by default
2673 * Use -o nodelalloc to turn it off
2675 if (!IS_EXT3_SB(sb) &&
2676 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
2677 set_opt(sbi->s_mount_opt, DELALLOC);
2679 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
2680 &journal_devnum, &journal_ioprio, NULL, 0)) {
2681 ext4_msg(sb, KERN_WARNING,
2682 "failed to parse options in superblock: %s",
2683 sbi->s_es->s_mount_opts);
2685 if (!parse_options((char *) data, sb, &journal_devnum,
2686 &journal_ioprio, NULL, 0))
2689 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2690 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2692 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2693 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2694 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2695 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2696 ext4_msg(sb, KERN_WARNING,
2697 "feature flags set on rev 0 fs, "
2698 "running e2fsck is recommended");
2701 * Check feature flags regardless of the revision level, since we
2702 * previously didn't change the revision level when setting the flags,
2703 * so there is a chance incompat flags are set on a rev 0 filesystem.
2705 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2708 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2710 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2711 blocksize > EXT4_MAX_BLOCK_SIZE) {
2712 ext4_msg(sb, KERN_ERR,
2713 "Unsupported filesystem blocksize %d", blocksize);
2717 if (sb->s_blocksize != blocksize) {
2718 /* Validate the filesystem blocksize */
2719 if (!sb_set_blocksize(sb, blocksize)) {
2720 ext4_msg(sb, KERN_ERR, "bad block size %d",
2726 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2727 offset = do_div(logical_sb_block, blocksize);
2728 bh = sb_bread(sb, logical_sb_block);
2730 ext4_msg(sb, KERN_ERR,
2731 "Can't read superblock on 2nd try");
2734 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2736 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2737 ext4_msg(sb, KERN_ERR,
2738 "Magic mismatch, very weird!");
2743 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2744 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2745 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2747 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2749 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2750 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2751 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2753 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2754 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2755 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2756 (!is_power_of_2(sbi->s_inode_size)) ||
2757 (sbi->s_inode_size > blocksize)) {
2758 ext4_msg(sb, KERN_ERR,
2759 "unsupported inode size: %d",
2763 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2764 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2767 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2768 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2769 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2770 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2771 !is_power_of_2(sbi->s_desc_size)) {
2772 ext4_msg(sb, KERN_ERR,
2773 "unsupported descriptor size %lu",
2778 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2780 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2781 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2782 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2785 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2786 if (sbi->s_inodes_per_block == 0)
2788 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2789 sbi->s_inodes_per_block;
2790 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2792 sbi->s_mount_state = le16_to_cpu(es->s_state);
2793 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2794 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2796 for (i = 0; i < 4; i++)
2797 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2798 sbi->s_def_hash_version = es->s_def_hash_version;
2799 i = le32_to_cpu(es->s_flags);
2800 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2801 sbi->s_hash_unsigned = 3;
2802 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2803 #ifdef __CHAR_UNSIGNED__
2804 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2805 sbi->s_hash_unsigned = 3;
2807 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2812 if (sbi->s_blocks_per_group > blocksize * 8) {
2813 ext4_msg(sb, KERN_ERR,
2814 "#blocks per group too big: %lu",
2815 sbi->s_blocks_per_group);
2818 if (sbi->s_inodes_per_group > blocksize * 8) {
2819 ext4_msg(sb, KERN_ERR,
2820 "#inodes per group too big: %lu",
2821 sbi->s_inodes_per_group);
2826 * Test whether we have more sectors than will fit in sector_t,
2827 * and whether the max offset is addressable by the page cache.
2829 if ((ext4_blocks_count(es) >
2830 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2831 (ext4_blocks_count(es) >
2832 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2833 ext4_msg(sb, KERN_ERR, "filesystem"
2834 " too large to mount safely on this system");
2835 if (sizeof(sector_t) < 8)
2836 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2841 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2844 /* check blocks count against device size */
2845 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2846 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2847 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2848 "exceeds size of device (%llu blocks)",
2849 ext4_blocks_count(es), blocks_count);
2854 * It makes no sense for the first data block to be beyond the end
2855 * of the filesystem.
2857 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2858 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2859 "block %u is beyond end of filesystem (%llu)",
2860 le32_to_cpu(es->s_first_data_block),
2861 ext4_blocks_count(es));
2864 blocks_count = (ext4_blocks_count(es) -
2865 le32_to_cpu(es->s_first_data_block) +
2866 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2867 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2868 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2869 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2870 "(block count %llu, first data block %u, "
2871 "blocks per group %lu)", sbi->s_groups_count,
2872 ext4_blocks_count(es),
2873 le32_to_cpu(es->s_first_data_block),
2874 EXT4_BLOCKS_PER_GROUP(sb));
2877 sbi->s_groups_count = blocks_count;
2878 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2879 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2880 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2881 EXT4_DESC_PER_BLOCK(sb);
2882 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2884 if (sbi->s_group_desc == NULL) {
2885 ext4_msg(sb, KERN_ERR, "not enough memory");
2889 #ifdef CONFIG_PROC_FS
2891 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2894 bgl_lock_init(sbi->s_blockgroup_lock);
2896 for (i = 0; i < db_count; i++) {
2897 block = descriptor_loc(sb, logical_sb_block, i);
2898 sbi->s_group_desc[i] = sb_bread(sb, block);
2899 if (!sbi->s_group_desc[i]) {
2900 ext4_msg(sb, KERN_ERR,
2901 "can't read group descriptor %d", i);
2906 if (!ext4_check_descriptors(sb)) {
2907 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2910 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2911 if (!ext4_fill_flex_info(sb)) {
2912 ext4_msg(sb, KERN_ERR,
2913 "unable to initialize "
2914 "flex_bg meta info!");
2918 sbi->s_gdb_count = db_count;
2919 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2920 spin_lock_init(&sbi->s_next_gen_lock);
2922 sbi->s_stripe = ext4_get_stripe_size(sbi);
2923 sbi->s_max_writeback_mb_bump = 128;
2926 * set up enough so that it can read an inode
2928 if (!test_opt(sb, NOLOAD) &&
2929 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2930 sb->s_op = &ext4_sops;
2932 sb->s_op = &ext4_nojournal_sops;
2933 sb->s_export_op = &ext4_export_ops;
2934 sb->s_xattr = ext4_xattr_handlers;
2936 sb->s_qcop = &ext4_qctl_operations;
2937 sb->dq_op = &ext4_quota_operations;
2939 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2940 mutex_init(&sbi->s_orphan_lock);
2941 mutex_init(&sbi->s_resize_lock);
2945 needs_recovery = (es->s_last_orphan != 0 ||
2946 EXT4_HAS_INCOMPAT_FEATURE(sb,
2947 EXT4_FEATURE_INCOMPAT_RECOVER));
2950 * The first inode we look at is the journal inode. Don't try
2951 * root first: it may be modified in the journal!
2953 if (!test_opt(sb, NOLOAD) &&
2954 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2955 if (ext4_load_journal(sb, es, journal_devnum))
2957 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2958 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2959 ext4_msg(sb, KERN_ERR, "required journal recovery "
2960 "suppressed and not mounted read-only");
2961 goto failed_mount_wq;
2963 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2964 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2965 sbi->s_journal = NULL;
2970 if (ext4_blocks_count(es) > 0xffffffffULL &&
2971 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2972 JBD2_FEATURE_INCOMPAT_64BIT)) {
2973 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2974 goto failed_mount_wq;
2977 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2978 jbd2_journal_set_features(sbi->s_journal,
2979 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2980 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2981 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2982 jbd2_journal_set_features(sbi->s_journal,
2983 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2984 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2985 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2987 jbd2_journal_clear_features(sbi->s_journal,
2988 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2989 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2992 /* We have now updated the journal if required, so we can
2993 * validate the data journaling mode. */
2994 switch (test_opt(sb, DATA_FLAGS)) {
2996 /* No mode set, assume a default based on the journal
2997 * capabilities: ORDERED_DATA if the journal can
2998 * cope, else JOURNAL_DATA
3000 if (jbd2_journal_check_available_features
3001 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3002 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3004 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3007 case EXT4_MOUNT_ORDERED_DATA:
3008 case EXT4_MOUNT_WRITEBACK_DATA:
3009 if (!jbd2_journal_check_available_features
3010 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3011 ext4_msg(sb, KERN_ERR, "Journal does not support "
3012 "requested data journaling mode");
3013 goto failed_mount_wq;
3018 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3021 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3022 ext4_count_free_blocks(sb));
3024 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3025 ext4_count_free_inodes(sb));
3027 err = percpu_counter_init(&sbi->s_dirs_counter,
3028 ext4_count_dirs(sb));
3030 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3032 ext4_msg(sb, KERN_ERR, "insufficient memory");
3033 goto failed_mount_wq;
3036 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3037 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3038 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3039 goto failed_mount_wq;
3043 * The jbd2_journal_load will have done any necessary log recovery,
3044 * so we can safely mount the rest of the filesystem now.
3047 root = ext4_iget(sb, EXT4_ROOT_INO);
3049 ext4_msg(sb, KERN_ERR, "get root inode failed");
3050 ret = PTR_ERR(root);
3053 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3055 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3058 sb->s_root = d_alloc_root(root);
3060 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3066 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3068 /* determine the minimum size of new large inodes, if present */
3069 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3070 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3071 EXT4_GOOD_OLD_INODE_SIZE;
3072 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3073 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3074 if (sbi->s_want_extra_isize <
3075 le16_to_cpu(es->s_want_extra_isize))
3076 sbi->s_want_extra_isize =
3077 le16_to_cpu(es->s_want_extra_isize);
3078 if (sbi->s_want_extra_isize <
3079 le16_to_cpu(es->s_min_extra_isize))
3080 sbi->s_want_extra_isize =
3081 le16_to_cpu(es->s_min_extra_isize);
3084 /* Check if enough inode space is available */
3085 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3086 sbi->s_inode_size) {
3087 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3088 EXT4_GOOD_OLD_INODE_SIZE;
3089 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3093 if (test_opt(sb, DELALLOC) &&
3094 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3095 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3096 "requested data journaling mode");
3097 clear_opt(sbi->s_mount_opt, DELALLOC);
3099 if (test_opt(sb, DIOREAD_NOLOCK)) {
3100 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3101 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3102 "option - requested data journaling mode");
3103 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3105 if (sb->s_blocksize < PAGE_SIZE) {
3106 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3107 "option - block size is too small");
3108 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3112 err = ext4_setup_system_zone(sb);
3114 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3120 err = ext4_mb_init(sb, needs_recovery);
3122 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3127 sbi->s_kobj.kset = ext4_kset;
3128 init_completion(&sbi->s_kobj_unregister);
3129 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3132 ext4_mb_release(sb);
3133 ext4_ext_release(sb);
3137 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3138 ext4_orphan_cleanup(sb, es);
3139 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3140 if (needs_recovery) {
3141 ext4_msg(sb, KERN_INFO, "recovery complete");
3142 ext4_mark_recovery_complete(sb, es);
3144 if (EXT4_SB(sb)->s_journal) {
3145 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3146 descr = " journalled data mode";
3147 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3148 descr = " ordered data mode";
3150 descr = " writeback data mode";
3152 descr = "out journal";
3154 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3155 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3156 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3158 init_timer(&sbi->s_err_report);
3159 sbi->s_err_report.function = print_daily_error_info;
3160 sbi->s_err_report.data = (unsigned long) sb;
3161 if (es->s_error_count)
3162 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3169 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3173 ext4_msg(sb, KERN_ERR, "mount failed");
3174 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3176 ext4_release_system_zone(sb);
3177 if (sbi->s_journal) {
3178 jbd2_journal_destroy(sbi->s_journal);
3179 sbi->s_journal = NULL;
3181 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3182 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3183 percpu_counter_destroy(&sbi->s_dirs_counter);
3184 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3186 if (sbi->s_flex_groups) {
3187 if (is_vmalloc_addr(sbi->s_flex_groups))
3188 vfree(sbi->s_flex_groups);
3190 kfree(sbi->s_flex_groups);
3193 for (i = 0; i < db_count; i++)
3194 brelse(sbi->s_group_desc[i]);
3195 kfree(sbi->s_group_desc);
3198 remove_proc_entry(sb->s_id, ext4_proc_root);
3201 for (i = 0; i < MAXQUOTAS; i++)
3202 kfree(sbi->s_qf_names[i]);
3204 ext4_blkdev_remove(sbi);
3207 sb->s_fs_info = NULL;
3208 kfree(sbi->s_blockgroup_lock);
3216 * Setup any per-fs journal parameters now. We'll do this both on
3217 * initial mount, once the journal has been initialised but before we've
3218 * done any recovery; and again on any subsequent remount.
3220 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3222 struct ext4_sb_info *sbi = EXT4_SB(sb);
3224 journal->j_commit_interval = sbi->s_commit_interval;
3225 journal->j_min_batch_time = sbi->s_min_batch_time;
3226 journal->j_max_batch_time = sbi->s_max_batch_time;
3228 write_lock(&journal->j_state_lock);
3229 if (test_opt(sb, BARRIER))
3230 journal->j_flags |= JBD2_BARRIER;
3232 journal->j_flags &= ~JBD2_BARRIER;
3233 if (test_opt(sb, DATA_ERR_ABORT))
3234 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3236 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3237 write_unlock(&journal->j_state_lock);
3240 static journal_t *ext4_get_journal(struct super_block *sb,
3241 unsigned int journal_inum)
3243 struct inode *journal_inode;
3246 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3248 /* First, test for the existence of a valid inode on disk. Bad
3249 * things happen if we iget() an unused inode, as the subsequent
3250 * iput() will try to delete it. */
3252 journal_inode = ext4_iget(sb, journal_inum);
3253 if (IS_ERR(journal_inode)) {
3254 ext4_msg(sb, KERN_ERR, "no journal found");
3257 if (!journal_inode->i_nlink) {
3258 make_bad_inode(journal_inode);
3259 iput(journal_inode);
3260 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3264 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3265 journal_inode, journal_inode->i_size);
3266 if (!S_ISREG(journal_inode->i_mode)) {
3267 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3268 iput(journal_inode);
3272 journal = jbd2_journal_init_inode(journal_inode);
3274 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3275 iput(journal_inode);
3278 journal->j_private = sb;
3279 ext4_init_journal_params(sb, journal);
3283 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3286 struct buffer_head *bh;
3290 int hblock, blocksize;
3291 ext4_fsblk_t sb_block;
3292 unsigned long offset;
3293 struct ext4_super_block *es;
3294 struct block_device *bdev;
3296 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3298 bdev = ext4_blkdev_get(j_dev, sb);
3302 if (bd_claim(bdev, sb)) {
3303 ext4_msg(sb, KERN_ERR,
3304 "failed to claim external journal device");
3305 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3309 blocksize = sb->s_blocksize;
3310 hblock = bdev_logical_block_size(bdev);
3311 if (blocksize < hblock) {
3312 ext4_msg(sb, KERN_ERR,
3313 "blocksize too small for journal device");
3317 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3318 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3319 set_blocksize(bdev, blocksize);
3320 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3321 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3322 "external journal");
3326 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3327 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3328 !(le32_to_cpu(es->s_feature_incompat) &
3329 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3330 ext4_msg(sb, KERN_ERR, "external journal has "
3336 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3337 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3342 len = ext4_blocks_count(es);
3343 start = sb_block + 1;
3344 brelse(bh); /* we're done with the superblock */
3346 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3347 start, len, blocksize);
3349 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3352 journal->j_private = sb;
3353 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3354 wait_on_buffer(journal->j_sb_buffer);
3355 if (!buffer_uptodate(journal->j_sb_buffer)) {
3356 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3359 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3360 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3361 "user (unsupported) - %d",
3362 be32_to_cpu(journal->j_superblock->s_nr_users));
3365 EXT4_SB(sb)->journal_bdev = bdev;
3366 ext4_init_journal_params(sb, journal);
3370 jbd2_journal_destroy(journal);
3372 ext4_blkdev_put(bdev);
3376 static int ext4_load_journal(struct super_block *sb,
3377 struct ext4_super_block *es,
3378 unsigned long journal_devnum)
3381 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3384 int really_read_only;
3386 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3388 if (journal_devnum &&
3389 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3390 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3391 "numbers have changed");
3392 journal_dev = new_decode_dev(journal_devnum);
3394 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3396 really_read_only = bdev_read_only(sb->s_bdev);
3399 * Are we loading a blank journal or performing recovery after a
3400 * crash? For recovery, we need to check in advance whether we
3401 * can get read-write access to the device.
3403 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3404 if (sb->s_flags & MS_RDONLY) {
3405 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3406 "required on readonly filesystem");
3407 if (really_read_only) {
3408 ext4_msg(sb, KERN_ERR, "write access "
3409 "unavailable, cannot proceed");
3412 ext4_msg(sb, KERN_INFO, "write access will "
3413 "be enabled during recovery");
3417 if (journal_inum && journal_dev) {
3418 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3419 "and inode journals!");
3424 if (!(journal = ext4_get_journal(sb, journal_inum)))
3427 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3431 if (!(journal->j_flags & JBD2_BARRIER))
3432 ext4_msg(sb, KERN_INFO, "barriers disabled");
3434 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3435 err = jbd2_journal_update_format(journal);
3437 ext4_msg(sb, KERN_ERR, "error updating journal");
3438 jbd2_journal_destroy(journal);
3443 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3444 err = jbd2_journal_wipe(journal, !really_read_only);
3446 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3448 memcpy(save, ((char *) es) +
3449 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3450 err = jbd2_journal_load(journal);
3452 memcpy(((char *) es) + EXT4_S_ERR_START,
3453 save, EXT4_S_ERR_LEN);
3458 ext4_msg(sb, KERN_ERR, "error loading journal");
3459 jbd2_journal_destroy(journal);
3463 EXT4_SB(sb)->s_journal = journal;
3464 ext4_clear_journal_err(sb, es);
3466 if (journal_devnum &&
3467 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3468 es->s_journal_dev = cpu_to_le32(journal_devnum);
3470 /* Make sure we flush the recovery flag to disk. */
3471 ext4_commit_super(sb, 1);
3477 static int ext4_commit_super(struct super_block *sb, int sync)
3479 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3480 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3485 if (buffer_write_io_error(sbh)) {
3487 * Oh, dear. A previous attempt to write the
3488 * superblock failed. This could happen because the
3489 * USB device was yanked out. Or it could happen to
3490 * be a transient write error and maybe the block will
3491 * be remapped. Nothing we can do but to retry the
3492 * write and hope for the best.
3494 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3495 "superblock detected");
3496 clear_buffer_write_io_error(sbh);
3497 set_buffer_uptodate(sbh);
3500 * If the file system is mounted read-only, don't update the
3501 * superblock write time. This avoids updating the superblock
3502 * write time when we are mounting the root file system
3503 * read/only but we need to replay the journal; at that point,
3504 * for people who are east of GMT and who make their clock
3505 * tick in localtime for Windows bug-for-bug compatibility,
3506 * the clock is set in the future, and this will cause e2fsck
3507 * to complain and force a full file system check.
3509 if (!(sb->s_flags & MS_RDONLY))
3510 es->s_wtime = cpu_to_le32(get_seconds());
3511 if (sb->s_bdev->bd_part)
3512 es->s_kbytes_written =
3513 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3514 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3515 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3517 es->s_kbytes_written =
3518 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3519 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3520 &EXT4_SB(sb)->s_freeblocks_counter));
3521 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3522 &EXT4_SB(sb)->s_freeinodes_counter));
3524 BUFFER_TRACE(sbh, "marking dirty");
3525 mark_buffer_dirty(sbh);
3527 error = sync_dirty_buffer(sbh);
3531 error = buffer_write_io_error(sbh);
3533 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3535 clear_buffer_write_io_error(sbh);
3536 set_buffer_uptodate(sbh);
3543 * Have we just finished recovery? If so, and if we are mounting (or
3544 * remounting) the filesystem readonly, then we will end up with a
3545 * consistent fs on disk. Record that fact.
3547 static void ext4_mark_recovery_complete(struct super_block *sb,
3548 struct ext4_super_block *es)
3550 journal_t *journal = EXT4_SB(sb)->s_journal;
3552 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3553 BUG_ON(journal != NULL);
3556 jbd2_journal_lock_updates(journal);
3557 if (jbd2_journal_flush(journal) < 0)
3560 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3561 sb->s_flags & MS_RDONLY) {
3562 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3563 ext4_commit_super(sb, 1);
3567 jbd2_journal_unlock_updates(journal);
3571 * If we are mounting (or read-write remounting) a filesystem whose journal
3572 * has recorded an error from a previous lifetime, move that error to the
3573 * main filesystem now.
3575 static void ext4_clear_journal_err(struct super_block *sb,
3576 struct ext4_super_block *es)
3582 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3584 journal = EXT4_SB(sb)->s_journal;
3587 * Now check for any error status which may have been recorded in the
3588 * journal by a prior ext4_error() or ext4_abort()
3591 j_errno = jbd2_journal_errno(journal);
3595 errstr = ext4_decode_error(sb, j_errno, nbuf);
3596 ext4_warning(sb, "Filesystem error recorded "
3597 "from previous mount: %s", errstr);
3598 ext4_warning(sb, "Marking fs in need of filesystem check.");
3600 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3601 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3602 ext4_commit_super(sb, 1);
3604 jbd2_journal_clear_err(journal);
3609 * Force the running and committing transactions to commit,
3610 * and wait on the commit.
3612 int ext4_force_commit(struct super_block *sb)
3617 if (sb->s_flags & MS_RDONLY)
3620 journal = EXT4_SB(sb)->s_journal;
3622 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3623 ret = ext4_journal_force_commit(journal);
3629 static void ext4_write_super(struct super_block *sb)
3632 ext4_commit_super(sb, 1);
3636 static int ext4_sync_fs(struct super_block *sb, int wait)
3640 struct ext4_sb_info *sbi = EXT4_SB(sb);
3642 trace_ext4_sync_fs(sb, wait);
3643 flush_workqueue(sbi->dio_unwritten_wq);
3644 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3646 jbd2_log_wait_commit(sbi->s_journal, target);
3652 * LVM calls this function before a (read-only) snapshot is created. This
3653 * gives us a chance to flush the journal completely and mark the fs clean.
3655 static int ext4_freeze(struct super_block *sb)
3660 if (sb->s_flags & MS_RDONLY)
3663 journal = EXT4_SB(sb)->s_journal;
3665 /* Now we set up the journal barrier. */
3666 jbd2_journal_lock_updates(journal);
3669 * Don't clear the needs_recovery flag if we failed to flush
3672 error = jbd2_journal_flush(journal);
3676 /* Journal blocked and flushed, clear needs_recovery flag. */
3677 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3678 error = ext4_commit_super(sb, 1);
3680 /* we rely on s_frozen to stop further updates */
3681 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3686 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3687 * flag here, even though the filesystem is not technically dirty yet.
3689 static int ext4_unfreeze(struct super_block *sb)
3691 if (sb->s_flags & MS_RDONLY)
3695 /* Reset the needs_recovery flag before the fs is unlocked. */
3696 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3697 ext4_commit_super(sb, 1);
3702 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3704 struct ext4_super_block *es;
3705 struct ext4_sb_info *sbi = EXT4_SB(sb);
3706 ext4_fsblk_t n_blocks_count = 0;
3707 unsigned long old_sb_flags;
3708 struct ext4_mount_options old_opts;
3709 int enable_quota = 0;
3711 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3716 char *orig_data = kstrdup(data, GFP_KERNEL);
3718 /* Store the original options */
3720 old_sb_flags = sb->s_flags;
3721 old_opts.s_mount_opt = sbi->s_mount_opt;
3722 old_opts.s_resuid = sbi->s_resuid;
3723 old_opts.s_resgid = sbi->s_resgid;
3724 old_opts.s_commit_interval = sbi->s_commit_interval;
3725 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3726 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3728 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3729 for (i = 0; i < MAXQUOTAS; i++)
3730 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3732 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3733 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3736 * Allow the "check" option to be passed as a remount option.
3738 if (!parse_options(data, sb, NULL, &journal_ioprio,
3739 &n_blocks_count, 1)) {
3744 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3745 ext4_abort(sb, "Abort forced by user");
3747 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3748 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3752 if (sbi->s_journal) {
3753 ext4_init_journal_params(sb, sbi->s_journal);
3754 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3757 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3758 n_blocks_count > ext4_blocks_count(es)) {
3759 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3764 if (*flags & MS_RDONLY) {
3765 err = dquot_suspend(sb, -1);
3770 * First of all, the unconditional stuff we have to do
3771 * to disable replay of the journal when we next remount
3773 sb->s_flags |= MS_RDONLY;
3776 * OK, test if we are remounting a valid rw partition
3777 * readonly, and if so set the rdonly flag and then
3778 * mark the partition as valid again.
3780 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3781 (sbi->s_mount_state & EXT4_VALID_FS))
3782 es->s_state = cpu_to_le16(sbi->s_mount_state);
3785 ext4_mark_recovery_complete(sb, es);
3787 /* Make sure we can mount this feature set readwrite */
3788 if (!ext4_feature_set_ok(sb, 0)) {
3793 * Make sure the group descriptor checksums
3794 * are sane. If they aren't, refuse to remount r/w.
3796 for (g = 0; g < sbi->s_groups_count; g++) {
3797 struct ext4_group_desc *gdp =
3798 ext4_get_group_desc(sb, g, NULL);
3800 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3801 ext4_msg(sb, KERN_ERR,
3802 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3803 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3804 le16_to_cpu(gdp->bg_checksum));
3811 * If we have an unprocessed orphan list hanging
3812 * around from a previously readonly bdev mount,
3813 * require a full umount/remount for now.
3815 if (es->s_last_orphan) {
3816 ext4_msg(sb, KERN_WARNING, "Couldn't "
3817 "remount RDWR because of unprocessed "
3818 "orphan inode list. Please "
3819 "umount/remount instead");
3825 * Mounting a RDONLY partition read-write, so reread
3826 * and store the current valid flag. (It may have
3827 * been changed by e2fsck since we originally mounted
3831 ext4_clear_journal_err(sb, es);
3832 sbi->s_mount_state = le16_to_cpu(es->s_state);
3833 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3835 if (!ext4_setup_super(sb, es, 0))
3836 sb->s_flags &= ~MS_RDONLY;
3840 ext4_setup_system_zone(sb);
3841 if (sbi->s_journal == NULL)
3842 ext4_commit_super(sb, 1);
3845 /* Release old quota file names */
3846 for (i = 0; i < MAXQUOTAS; i++)
3847 if (old_opts.s_qf_names[i] &&
3848 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3849 kfree(old_opts.s_qf_names[i]);
3853 dquot_resume(sb, -1);
3855 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3860 sb->s_flags = old_sb_flags;
3861 sbi->s_mount_opt = old_opts.s_mount_opt;
3862 sbi->s_resuid = old_opts.s_resuid;
3863 sbi->s_resgid = old_opts.s_resgid;
3864 sbi->s_commit_interval = old_opts.s_commit_interval;
3865 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3866 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3868 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3869 for (i = 0; i < MAXQUOTAS; i++) {
3870 if (sbi->s_qf_names[i] &&
3871 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3872 kfree(sbi->s_qf_names[i]);
3873 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3881 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3883 struct super_block *sb = dentry->d_sb;
3884 struct ext4_sb_info *sbi = EXT4_SB(sb);
3885 struct ext4_super_block *es = sbi->s_es;
3888 if (test_opt(sb, MINIX_DF)) {
3889 sbi->s_overhead_last = 0;
3890 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3891 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3892 ext4_fsblk_t overhead = 0;
3895 * Compute the overhead (FS structures). This is constant
3896 * for a given filesystem unless the number of block groups
3897 * changes so we cache the previous value until it does.
3901 * All of the blocks before first_data_block are
3904 overhead = le32_to_cpu(es->s_first_data_block);
3907 * Add the overhead attributed to the superblock and
3908 * block group descriptors. If the sparse superblocks
3909 * feature is turned on, then not all groups have this.
3911 for (i = 0; i < ngroups; i++) {
3912 overhead += ext4_bg_has_super(sb, i) +
3913 ext4_bg_num_gdb(sb, i);
3918 * Every block group has an inode bitmap, a block
3919 * bitmap, and an inode table.
3921 overhead += ngroups * (2 + sbi->s_itb_per_group);
3922 sbi->s_overhead_last = overhead;
3924 sbi->s_blocks_last = ext4_blocks_count(es);
3927 buf->f_type = EXT4_SUPER_MAGIC;
3928 buf->f_bsize = sb->s_blocksize;
3929 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3930 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3931 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3932 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3933 if (buf->f_bfree < ext4_r_blocks_count(es))
3935 buf->f_files = le32_to_cpu(es->s_inodes_count);
3936 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3937 buf->f_namelen = EXT4_NAME_LEN;
3938 fsid = le64_to_cpup((void *)es->s_uuid) ^
3939 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3940 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3941 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3946 /* Helper function for writing quotas on sync - we need to start transaction
3947 * before quota file is locked for write. Otherwise the are possible deadlocks:
3948 * Process 1 Process 2
3949 * ext4_create() quota_sync()
3950 * jbd2_journal_start() write_dquot()
3951 * dquot_initialize() down(dqio_mutex)
3952 * down(dqio_mutex) jbd2_journal_start()
3958 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3960 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3963 static int ext4_write_dquot(struct dquot *dquot)
3967 struct inode *inode;
3969 inode = dquot_to_inode(dquot);
3970 handle = ext4_journal_start(inode,
3971 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3973 return PTR_ERR(handle);
3974 ret = dquot_commit(dquot);
3975 err = ext4_journal_stop(handle);
3981 static int ext4_acquire_dquot(struct dquot *dquot)
3986 handle = ext4_journal_start(dquot_to_inode(dquot),
3987 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3989 return PTR_ERR(handle);
3990 ret = dquot_acquire(dquot);
3991 err = ext4_journal_stop(handle);
3997 static int ext4_release_dquot(struct dquot *dquot)
4002 handle = ext4_journal_start(dquot_to_inode(dquot),
4003 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4004 if (IS_ERR(handle)) {
4005 /* Release dquot anyway to avoid endless cycle in dqput() */
4006 dquot_release(dquot);
4007 return PTR_ERR(handle);
4009 ret = dquot_release(dquot);
4010 err = ext4_journal_stop(handle);
4016 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4018 /* Are we journaling quotas? */
4019 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4020 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4021 dquot_mark_dquot_dirty(dquot);
4022 return ext4_write_dquot(dquot);
4024 return dquot_mark_dquot_dirty(dquot);
4028 static int ext4_write_info(struct super_block *sb, int type)
4033 /* Data block + inode block */
4034 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4036 return PTR_ERR(handle);
4037 ret = dquot_commit_info(sb, type);
4038 err = ext4_journal_stop(handle);
4045 * Turn on quotas during mount time - we need to find
4046 * the quota file and such...
4048 static int ext4_quota_on_mount(struct super_block *sb, int type)
4050 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4051 EXT4_SB(sb)->s_jquota_fmt, type);
4055 * Standard function to be called on quota_on
4057 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4063 if (!test_opt(sb, QUOTA))
4066 err = kern_path(name, LOOKUP_FOLLOW, &path);
4070 /* Quotafile not on the same filesystem? */
4071 if (path.mnt->mnt_sb != sb) {
4075 /* Journaling quota? */
4076 if (EXT4_SB(sb)->s_qf_names[type]) {
4077 /* Quotafile not in fs root? */
4078 if (path.dentry->d_parent != sb->s_root)
4079 ext4_msg(sb, KERN_WARNING,
4080 "Quota file not on filesystem root. "
4081 "Journaled quota will not work");
4085 * When we journal data on quota file, we have to flush journal to see
4086 * all updates to the file when we bypass pagecache...
4088 if (EXT4_SB(sb)->s_journal &&
4089 ext4_should_journal_data(path.dentry->d_inode)) {
4091 * We don't need to lock updates but journal_flush() could
4092 * otherwise be livelocked...
4094 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4095 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4096 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4103 err = dquot_quota_on_path(sb, type, format_id, &path);
4108 static int ext4_quota_off(struct super_block *sb, int type)
4110 /* Force all delayed allocation blocks to be allocated */
4111 if (test_opt(sb, DELALLOC)) {
4112 down_read(&sb->s_umount);
4113 sync_filesystem(sb);
4114 up_read(&sb->s_umount);
4117 return dquot_quota_off(sb, type);
4120 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4121 * acquiring the locks... As quota files are never truncated and quota code
4122 * itself serializes the operations (and noone else should touch the files)
4123 * we don't have to be afraid of races */
4124 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4125 size_t len, loff_t off)
4127 struct inode *inode = sb_dqopt(sb)->files[type];
4128 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4130 int offset = off & (sb->s_blocksize - 1);
4133 struct buffer_head *bh;
4134 loff_t i_size = i_size_read(inode);
4138 if (off+len > i_size)
4141 while (toread > 0) {
4142 tocopy = sb->s_blocksize - offset < toread ?
4143 sb->s_blocksize - offset : toread;
4144 bh = ext4_bread(NULL, inode, blk, 0, &err);
4147 if (!bh) /* A hole? */
4148 memset(data, 0, tocopy);
4150 memcpy(data, bh->b_data+offset, tocopy);
4160 /* Write to quotafile (we know the transaction is already started and has
4161 * enough credits) */
4162 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4163 const char *data, size_t len, loff_t off)
4165 struct inode *inode = sb_dqopt(sb)->files[type];
4166 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4168 int offset = off & (sb->s_blocksize - 1);
4169 struct buffer_head *bh;
4170 handle_t *handle = journal_current_handle();
4172 if (EXT4_SB(sb)->s_journal && !handle) {
4173 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4174 " cancelled because transaction is not started",
4175 (unsigned long long)off, (unsigned long long)len);
4179 * Since we account only one data block in transaction credits,
4180 * then it is impossible to cross a block boundary.
4182 if (sb->s_blocksize - offset < len) {
4183 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4184 " cancelled because not block aligned",
4185 (unsigned long long)off, (unsigned long long)len);
4189 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4190 bh = ext4_bread(handle, inode, blk, 1, &err);
4193 err = ext4_journal_get_write_access(handle, bh);
4199 memcpy(bh->b_data+offset, data, len);
4200 flush_dcache_page(bh->b_page);
4202 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4206 mutex_unlock(&inode->i_mutex);
4209 if (inode->i_size < off + len) {
4210 i_size_write(inode, off + len);
4211 EXT4_I(inode)->i_disksize = inode->i_size;
4213 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4214 ext4_mark_inode_dirty(handle, inode);
4215 mutex_unlock(&inode->i_mutex);
4221 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4222 const char *dev_name, void *data, struct vfsmount *mnt)
4224 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4227 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4228 static struct file_system_type ext2_fs_type = {
4229 .owner = THIS_MODULE,
4231 .get_sb = ext4_get_sb,
4232 .kill_sb = kill_block_super,
4233 .fs_flags = FS_REQUIRES_DEV,
4236 static inline void register_as_ext2(void)
4238 int err = register_filesystem(&ext2_fs_type);
4241 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4244 static inline void unregister_as_ext2(void)
4246 unregister_filesystem(&ext2_fs_type);
4248 MODULE_ALIAS("ext2");
4250 static inline void register_as_ext2(void) { }
4251 static inline void unregister_as_ext2(void) { }
4254 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4255 static inline void register_as_ext3(void)
4257 int err = register_filesystem(&ext3_fs_type);
4260 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4263 static inline void unregister_as_ext3(void)
4265 unregister_filesystem(&ext3_fs_type);
4267 MODULE_ALIAS("ext3");
4269 static inline void register_as_ext3(void) { }
4270 static inline void unregister_as_ext3(void) { }
4273 static struct file_system_type ext4_fs_type = {
4274 .owner = THIS_MODULE,
4276 .get_sb = ext4_get_sb,
4277 .kill_sb = kill_block_super,
4278 .fs_flags = FS_REQUIRES_DEV,
4281 static int __init init_ext4_fs(void)
4285 ext4_check_flag_values();
4286 err = init_ext4_system_zone();
4289 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4292 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4293 err = init_ext4_mballoc();
4297 err = init_ext4_xattr();
4300 err = init_inodecache();
4305 err = register_filesystem(&ext4_fs_type);
4310 unregister_as_ext2();
4311 unregister_as_ext3();
4312 destroy_inodecache();
4316 exit_ext4_mballoc();
4318 remove_proc_entry("fs/ext4", NULL);
4319 kset_unregister(ext4_kset);
4321 exit_ext4_system_zone();
4325 static void __exit exit_ext4_fs(void)
4327 unregister_as_ext2();
4328 unregister_as_ext3();
4329 unregister_filesystem(&ext4_fs_type);
4330 destroy_inodecache();
4332 exit_ext4_mballoc();
4333 remove_proc_entry("fs/ext4", NULL);
4334 kset_unregister(ext4_kset);
4335 exit_ext4_system_zone();
4338 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4339 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4340 MODULE_LICENSE("GPL");
4341 module_init(init_ext4_fs)
4342 module_exit(exit_ext4_fs)