2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 struct ext4_lazy_init *ext4_li_info;
59 struct mutex ext4_li_mtx;
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static const char *ext4_decode_error(struct super_block *sb, int errno,
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static void ext4_write_super(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data, struct vfsmount *mnt);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
81 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
82 static struct file_system_type ext3_fs_type = {
85 .get_sb = ext4_get_sb,
86 .kill_sb = kill_block_super,
87 .fs_flags = FS_REQUIRES_DEV,
89 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91 #define IS_EXT3_SB(sb) (0)
94 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
95 struct ext4_group_desc *bg)
97 return le32_to_cpu(bg->bg_block_bitmap_lo) |
98 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
99 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
102 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
103 struct ext4_group_desc *bg)
105 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
107 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
110 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
111 struct ext4_group_desc *bg)
113 return le32_to_cpu(bg->bg_inode_table_lo) |
114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
115 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
118 __u32 ext4_free_blks_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
121 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
123 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
126 __u32 ext4_free_inodes_count(struct super_block *sb,
127 struct ext4_group_desc *bg)
129 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
130 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
131 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
134 __u32 ext4_used_dirs_count(struct super_block *sb,
135 struct ext4_group_desc *bg)
137 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
138 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
139 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
142 __u32 ext4_itable_unused_count(struct super_block *sb,
143 struct ext4_group_desc *bg)
145 return le16_to_cpu(bg->bg_itable_unused_lo) |
146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
147 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
150 void ext4_block_bitmap_set(struct super_block *sb,
151 struct ext4_group_desc *bg, ext4_fsblk_t blk)
153 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
158 void ext4_inode_bitmap_set(struct super_block *sb,
159 struct ext4_group_desc *bg, ext4_fsblk_t blk)
161 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
166 void ext4_inode_table_set(struct super_block *sb,
167 struct ext4_group_desc *bg, ext4_fsblk_t blk)
169 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
174 void ext4_free_blks_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
177 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
182 void ext4_free_inodes_set(struct super_block *sb,
183 struct ext4_group_desc *bg, __u32 count)
185 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
186 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
187 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
190 void ext4_used_dirs_set(struct super_block *sb,
191 struct ext4_group_desc *bg, __u32 count)
193 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
194 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
195 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
198 void ext4_itable_unused_set(struct super_block *sb,
199 struct ext4_group_desc *bg, __u32 count)
201 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
207 /* Just increment the non-pointer handle value */
208 static handle_t *ext4_get_nojournal(void)
210 handle_t *handle = current->journal_info;
211 unsigned long ref_cnt = (unsigned long)handle;
213 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
216 handle = (handle_t *)ref_cnt;
218 current->journal_info = handle;
223 /* Decrement the non-pointer handle value */
224 static void ext4_put_nojournal(handle_t *handle)
226 unsigned long ref_cnt = (unsigned long)handle;
228 BUG_ON(ref_cnt == 0);
231 handle = (handle_t *)ref_cnt;
233 current->journal_info = handle;
237 * Wrappers for jbd2_journal_start/end.
239 * The only special thing we need to do here is to make sure that all
240 * journal_end calls result in the superblock being marked dirty, so
241 * that sync() will call the filesystem's write_super callback if
244 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
248 if (sb->s_flags & MS_RDONLY)
249 return ERR_PTR(-EROFS);
251 vfs_check_frozen(sb, SB_FREEZE_TRANS);
252 /* Special case here: if the journal has aborted behind our
253 * backs (eg. EIO in the commit thread), then we still need to
254 * take the FS itself readonly cleanly. */
255 journal = EXT4_SB(sb)->s_journal;
257 if (is_journal_aborted(journal)) {
258 ext4_abort(sb, "Detected aborted journal");
259 return ERR_PTR(-EROFS);
261 return jbd2_journal_start(journal, nblocks);
263 return ext4_get_nojournal();
267 * The only special thing we need to do here is to make sure that all
268 * jbd2_journal_stop calls result in the superblock being marked dirty, so
269 * that sync() will call the filesystem's write_super callback if
272 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
274 struct super_block *sb;
278 if (!ext4_handle_valid(handle)) {
279 ext4_put_nojournal(handle);
282 sb = handle->h_transaction->t_journal->j_private;
284 rc = jbd2_journal_stop(handle);
289 __ext4_std_error(sb, where, line, err);
293 void ext4_journal_abort_handle(const char *caller, unsigned int line,
294 const char *err_fn, struct buffer_head *bh,
295 handle_t *handle, int err)
298 const char *errstr = ext4_decode_error(NULL, err, nbuf);
300 BUG_ON(!ext4_handle_valid(handle));
303 BUFFER_TRACE(bh, "abort");
308 if (is_handle_aborted(handle))
311 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
312 caller, line, errstr, err_fn);
314 jbd2_journal_abort_handle(handle);
317 static void __save_error_info(struct super_block *sb, const char *func,
320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
324 es->s_last_error_time = cpu_to_le32(get_seconds());
325 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
326 es->s_last_error_line = cpu_to_le32(line);
327 if (!es->s_first_error_time) {
328 es->s_first_error_time = es->s_last_error_time;
329 strncpy(es->s_first_error_func, func,
330 sizeof(es->s_first_error_func));
331 es->s_first_error_line = cpu_to_le32(line);
332 es->s_first_error_ino = es->s_last_error_ino;
333 es->s_first_error_block = es->s_last_error_block;
336 * Start the daily error reporting function if it hasn't been
339 if (!es->s_error_count)
340 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
341 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
344 static void save_error_info(struct super_block *sb, const char *func,
347 __save_error_info(sb, func, line);
348 ext4_commit_super(sb, 1);
352 /* Deal with the reporting of failure conditions on a filesystem such as
353 * inconsistencies detected or read IO failures.
355 * On ext2, we can store the error state of the filesystem in the
356 * superblock. That is not possible on ext4, because we may have other
357 * write ordering constraints on the superblock which prevent us from
358 * writing it out straight away; and given that the journal is about to
359 * be aborted, we can't rely on the current, or future, transactions to
360 * write out the superblock safely.
362 * We'll just use the jbd2_journal_abort() error code to record an error in
363 * the journal instead. On recovery, the journal will complain about
364 * that error until we've noted it down and cleared it.
367 static void ext4_handle_error(struct super_block *sb)
369 if (sb->s_flags & MS_RDONLY)
372 if (!test_opt(sb, ERRORS_CONT)) {
373 journal_t *journal = EXT4_SB(sb)->s_journal;
375 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
377 jbd2_journal_abort(journal, -EIO);
379 if (test_opt(sb, ERRORS_RO)) {
380 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
381 sb->s_flags |= MS_RDONLY;
383 if (test_opt(sb, ERRORS_PANIC))
384 panic("EXT4-fs (device %s): panic forced after error\n",
388 void __ext4_error(struct super_block *sb, const char *function,
389 unsigned int line, const char *fmt, ...)
394 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
395 sb->s_id, function, line, current->comm);
400 ext4_handle_error(sb);
403 void ext4_error_inode(struct inode *inode, const char *function,
404 unsigned int line, ext4_fsblk_t block,
405 const char *fmt, ...)
408 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
410 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
411 es->s_last_error_block = cpu_to_le64(block);
412 save_error_info(inode->i_sb, function, line);
414 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
415 inode->i_sb->s_id, function, line, inode->i_ino);
417 printk("block %llu: ", block);
418 printk("comm %s: ", current->comm);
423 ext4_handle_error(inode->i_sb);
426 void ext4_error_file(struct file *file, const char *function,
427 unsigned int line, const char *fmt, ...)
430 struct ext4_super_block *es;
431 struct inode *inode = file->f_dentry->d_inode;
432 char pathname[80], *path;
434 es = EXT4_SB(inode->i_sb)->s_es;
435 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
436 save_error_info(inode->i_sb, function, line);
438 path = d_path(&(file->f_path), pathname, sizeof(pathname));
442 "EXT4-fs error (device %s): %s:%d: inode #%lu "
443 "(comm %s path %s): ",
444 inode->i_sb->s_id, function, line, inode->i_ino,
445 current->comm, path);
450 ext4_handle_error(inode->i_sb);
453 static const char *ext4_decode_error(struct super_block *sb, int errno,
460 errstr = "IO failure";
463 errstr = "Out of memory";
466 if (!sb || (EXT4_SB(sb)->s_journal &&
467 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
468 errstr = "Journal has aborted";
470 errstr = "Readonly filesystem";
473 /* If the caller passed in an extra buffer for unknown
474 * errors, textualise them now. Else we just return
477 /* Check for truncated error codes... */
478 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
487 /* __ext4_std_error decodes expected errors from journaling functions
488 * automatically and invokes the appropriate error response. */
490 void __ext4_std_error(struct super_block *sb, const char *function,
491 unsigned int line, int errno)
496 /* Special case: if the error is EROFS, and we're not already
497 * inside a transaction, then there's really no point in logging
499 if (errno == -EROFS && journal_current_handle() == NULL &&
500 (sb->s_flags & MS_RDONLY))
503 errstr = ext4_decode_error(sb, errno, nbuf);
504 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
505 sb->s_id, function, line, errstr);
506 save_error_info(sb, function, line);
508 ext4_handle_error(sb);
512 * ext4_abort is a much stronger failure handler than ext4_error. The
513 * abort function may be used to deal with unrecoverable failures such
514 * as journal IO errors or ENOMEM at a critical moment in log management.
516 * We unconditionally force the filesystem into an ABORT|READONLY state,
517 * unless the error response on the fs has been set to panic in which
518 * case we take the easy way out and panic immediately.
521 void __ext4_abort(struct super_block *sb, const char *function,
522 unsigned int line, const char *fmt, ...)
526 save_error_info(sb, function, line);
528 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
534 if ((sb->s_flags & MS_RDONLY) == 0) {
535 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
536 sb->s_flags |= MS_RDONLY;
537 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
538 if (EXT4_SB(sb)->s_journal)
539 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
540 save_error_info(sb, function, line);
542 if (test_opt(sb, ERRORS_PANIC))
543 panic("EXT4-fs panic from previous error\n");
546 void ext4_msg (struct super_block * sb, const char *prefix,
547 const char *fmt, ...)
552 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
558 void __ext4_warning(struct super_block *sb, const char *function,
559 unsigned int line, const char *fmt, ...)
564 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
565 sb->s_id, function, line);
571 void __ext4_grp_locked_error(const char *function, unsigned int line,
572 struct super_block *sb, ext4_group_t grp,
573 unsigned long ino, ext4_fsblk_t block,
574 const char *fmt, ...)
579 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
581 es->s_last_error_ino = cpu_to_le32(ino);
582 es->s_last_error_block = cpu_to_le64(block);
583 __save_error_info(sb, function, line);
585 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
586 sb->s_id, function, line, grp);
588 printk("inode %lu: ", ino);
590 printk("block %llu:", (unsigned long long) block);
595 if (test_opt(sb, ERRORS_CONT)) {
596 ext4_commit_super(sb, 0);
600 ext4_unlock_group(sb, grp);
601 ext4_handle_error(sb);
603 * We only get here in the ERRORS_RO case; relocking the group
604 * may be dangerous, but nothing bad will happen since the
605 * filesystem will have already been marked read/only and the
606 * journal has been aborted. We return 1 as a hint to callers
607 * who might what to use the return value from
608 * ext4_grp_locked_error() to distinguish beween the
609 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
610 * aggressively from the ext4 function in question, with a
611 * more appropriate error code.
613 ext4_lock_group(sb, grp);
617 void ext4_update_dynamic_rev(struct super_block *sb)
619 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
621 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
625 "updating to rev %d because of new feature flag, "
626 "running e2fsck is recommended",
629 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
630 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
631 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
632 /* leave es->s_feature_*compat flags alone */
633 /* es->s_uuid will be set by e2fsck if empty */
636 * The rest of the superblock fields should be zero, and if not it
637 * means they are likely already in use, so leave them alone. We
638 * can leave it up to e2fsck to clean up any inconsistencies there.
643 * Open the external journal device
645 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
647 struct block_device *bdev;
648 char b[BDEVNAME_SIZE];
650 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
656 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
657 __bdevname(dev, b), PTR_ERR(bdev));
662 * Release the journal device
664 static int ext4_blkdev_put(struct block_device *bdev)
667 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
670 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
672 struct block_device *bdev;
675 bdev = sbi->journal_bdev;
677 ret = ext4_blkdev_put(bdev);
678 sbi->journal_bdev = NULL;
683 static inline struct inode *orphan_list_entry(struct list_head *l)
685 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
688 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
692 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
693 le32_to_cpu(sbi->s_es->s_last_orphan));
695 printk(KERN_ERR "sb_info orphan list:\n");
696 list_for_each(l, &sbi->s_orphan) {
697 struct inode *inode = orphan_list_entry(l);
699 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
700 inode->i_sb->s_id, inode->i_ino, inode,
701 inode->i_mode, inode->i_nlink,
706 static void ext4_put_super(struct super_block *sb)
708 struct ext4_sb_info *sbi = EXT4_SB(sb);
709 struct ext4_super_block *es = sbi->s_es;
712 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
714 flush_workqueue(sbi->dio_unwritten_wq);
715 destroy_workqueue(sbi->dio_unwritten_wq);
720 ext4_commit_super(sb, 1);
722 if (sbi->s_journal) {
723 err = jbd2_journal_destroy(sbi->s_journal);
724 sbi->s_journal = NULL;
726 ext4_abort(sb, "Couldn't clean up the journal");
729 del_timer(&sbi->s_err_report);
730 ext4_unregister_li_request(sb);
731 ext4_release_system_zone(sb);
733 ext4_ext_release(sb);
734 ext4_xattr_put_super(sb);
736 if (!(sb->s_flags & MS_RDONLY)) {
737 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
738 es->s_state = cpu_to_le16(sbi->s_mount_state);
739 ext4_commit_super(sb, 1);
742 remove_proc_entry(sb->s_id, ext4_proc_root);
744 kobject_del(&sbi->s_kobj);
746 for (i = 0; i < sbi->s_gdb_count; i++)
747 brelse(sbi->s_group_desc[i]);
748 kfree(sbi->s_group_desc);
749 if (is_vmalloc_addr(sbi->s_flex_groups))
750 vfree(sbi->s_flex_groups);
752 kfree(sbi->s_flex_groups);
753 percpu_counter_destroy(&sbi->s_freeblocks_counter);
754 percpu_counter_destroy(&sbi->s_freeinodes_counter);
755 percpu_counter_destroy(&sbi->s_dirs_counter);
756 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
759 for (i = 0; i < MAXQUOTAS; i++)
760 kfree(sbi->s_qf_names[i]);
763 /* Debugging code just in case the in-memory inode orphan list
764 * isn't empty. The on-disk one can be non-empty if we've
765 * detected an error and taken the fs readonly, but the
766 * in-memory list had better be clean by this point. */
767 if (!list_empty(&sbi->s_orphan))
768 dump_orphan_list(sb, sbi);
769 J_ASSERT(list_empty(&sbi->s_orphan));
771 invalidate_bdev(sb->s_bdev);
772 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
774 * Invalidate the journal device's buffers. We don't want them
775 * floating about in memory - the physical journal device may
776 * hotswapped, and it breaks the `ro-after' testing code.
778 sync_blockdev(sbi->journal_bdev);
779 invalidate_bdev(sbi->journal_bdev);
780 ext4_blkdev_remove(sbi);
782 sb->s_fs_info = NULL;
784 * Now that we are completely done shutting down the
785 * superblock, we need to actually destroy the kobject.
789 kobject_put(&sbi->s_kobj);
790 wait_for_completion(&sbi->s_kobj_unregister);
791 kfree(sbi->s_blockgroup_lock);
795 static struct kmem_cache *ext4_inode_cachep;
798 * Called inside transaction, so use GFP_NOFS
800 static struct inode *ext4_alloc_inode(struct super_block *sb)
802 struct ext4_inode_info *ei;
804 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
808 ei->vfs_inode.i_version = 1;
809 ei->vfs_inode.i_data.writeback_index = 0;
810 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
811 INIT_LIST_HEAD(&ei->i_prealloc_list);
812 spin_lock_init(&ei->i_prealloc_lock);
814 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
815 * therefore it can be null here. Don't check it, just initialize
818 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
819 ei->i_reserved_data_blocks = 0;
820 ei->i_reserved_meta_blocks = 0;
821 ei->i_allocated_meta_blocks = 0;
822 ei->i_da_metadata_calc_len = 0;
823 ei->i_delalloc_reserved_flag = 0;
824 spin_lock_init(&(ei->i_block_reservation_lock));
826 ei->i_reserved_quota = 0;
828 INIT_LIST_HEAD(&ei->i_completed_io_list);
829 spin_lock_init(&ei->i_completed_io_lock);
830 ei->cur_aio_dio = NULL;
832 ei->i_datasync_tid = 0;
834 return &ei->vfs_inode;
837 static void ext4_destroy_inode(struct inode *inode)
839 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
840 ext4_msg(inode->i_sb, KERN_ERR,
841 "Inode %lu (%p): orphan list check failed!",
842 inode->i_ino, EXT4_I(inode));
843 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
844 EXT4_I(inode), sizeof(struct ext4_inode_info),
848 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
851 static void init_once(void *foo)
853 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
855 INIT_LIST_HEAD(&ei->i_orphan);
856 #ifdef CONFIG_EXT4_FS_XATTR
857 init_rwsem(&ei->xattr_sem);
859 init_rwsem(&ei->i_data_sem);
860 inode_init_once(&ei->vfs_inode);
863 static int init_inodecache(void)
865 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
866 sizeof(struct ext4_inode_info),
867 0, (SLAB_RECLAIM_ACCOUNT|
870 if (ext4_inode_cachep == NULL)
875 static void destroy_inodecache(void)
877 kmem_cache_destroy(ext4_inode_cachep);
880 void ext4_clear_inode(struct inode *inode)
882 invalidate_inode_buffers(inode);
883 end_writeback(inode);
885 ext4_discard_preallocations(inode);
886 if (EXT4_JOURNAL(inode))
887 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
888 &EXT4_I(inode)->jinode);
891 static inline void ext4_show_quota_options(struct seq_file *seq,
892 struct super_block *sb)
894 #if defined(CONFIG_QUOTA)
895 struct ext4_sb_info *sbi = EXT4_SB(sb);
897 if (sbi->s_jquota_fmt) {
900 switch (sbi->s_jquota_fmt) {
911 seq_printf(seq, ",jqfmt=%s", fmtname);
914 if (sbi->s_qf_names[USRQUOTA])
915 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
917 if (sbi->s_qf_names[GRPQUOTA])
918 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
920 if (test_opt(sb, USRQUOTA))
921 seq_puts(seq, ",usrquota");
923 if (test_opt(sb, GRPQUOTA))
924 seq_puts(seq, ",grpquota");
930 * - it's set to a non-default value OR
931 * - if the per-sb default is different from the global default
933 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
936 unsigned long def_mount_opts;
937 struct super_block *sb = vfs->mnt_sb;
938 struct ext4_sb_info *sbi = EXT4_SB(sb);
939 struct ext4_super_block *es = sbi->s_es;
941 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
942 def_errors = le16_to_cpu(es->s_errors);
944 if (sbi->s_sb_block != 1)
945 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
946 if (test_opt(sb, MINIX_DF))
947 seq_puts(seq, ",minixdf");
948 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
949 seq_puts(seq, ",grpid");
950 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
951 seq_puts(seq, ",nogrpid");
952 if (sbi->s_resuid != EXT4_DEF_RESUID ||
953 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
954 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
956 if (sbi->s_resgid != EXT4_DEF_RESGID ||
957 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
958 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
960 if (test_opt(sb, ERRORS_RO)) {
961 if (def_errors == EXT4_ERRORS_PANIC ||
962 def_errors == EXT4_ERRORS_CONTINUE) {
963 seq_puts(seq, ",errors=remount-ro");
966 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
967 seq_puts(seq, ",errors=continue");
968 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
969 seq_puts(seq, ",errors=panic");
970 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
971 seq_puts(seq, ",nouid32");
972 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
973 seq_puts(seq, ",debug");
974 if (test_opt(sb, OLDALLOC))
975 seq_puts(seq, ",oldalloc");
976 #ifdef CONFIG_EXT4_FS_XATTR
977 if (test_opt(sb, XATTR_USER) &&
978 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
979 seq_puts(seq, ",user_xattr");
980 if (!test_opt(sb, XATTR_USER) &&
981 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
982 seq_puts(seq, ",nouser_xattr");
985 #ifdef CONFIG_EXT4_FS_POSIX_ACL
986 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
987 seq_puts(seq, ",acl");
988 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
989 seq_puts(seq, ",noacl");
991 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
992 seq_printf(seq, ",commit=%u",
993 (unsigned) (sbi->s_commit_interval / HZ));
995 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
996 seq_printf(seq, ",min_batch_time=%u",
997 (unsigned) sbi->s_min_batch_time);
999 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1000 seq_printf(seq, ",max_batch_time=%u",
1001 (unsigned) sbi->s_min_batch_time);
1005 * We're changing the default of barrier mount option, so
1006 * let's always display its mount state so it's clear what its
1009 seq_puts(seq, ",barrier=");
1010 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1011 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1012 seq_puts(seq, ",journal_async_commit");
1013 else if (test_opt(sb, JOURNAL_CHECKSUM))
1014 seq_puts(seq, ",journal_checksum");
1015 if (test_opt(sb, I_VERSION))
1016 seq_puts(seq, ",i_version");
1017 if (!test_opt(sb, DELALLOC) &&
1018 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1019 seq_puts(seq, ",nodelalloc");
1022 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1024 * journal mode get enabled in different ways
1025 * So just print the value even if we didn't specify it
1027 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1028 seq_puts(seq, ",data=journal");
1029 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1030 seq_puts(seq, ",data=ordered");
1031 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1032 seq_puts(seq, ",data=writeback");
1034 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1035 seq_printf(seq, ",inode_readahead_blks=%u",
1036 sbi->s_inode_readahead_blks);
1038 if (test_opt(sb, DATA_ERR_ABORT))
1039 seq_puts(seq, ",data_err=abort");
1041 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1042 seq_puts(seq, ",noauto_da_alloc");
1044 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1045 seq_puts(seq, ",discard");
1047 if (test_opt(sb, NOLOAD))
1048 seq_puts(seq, ",norecovery");
1050 if (test_opt(sb, DIOREAD_NOLOCK))
1051 seq_puts(seq, ",dioread_nolock");
1053 if (test_opt(sb, BLOCK_VALIDITY) &&
1054 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1055 seq_puts(seq, ",block_validity");
1057 if (!test_opt(sb, INIT_INODE_TABLE))
1058 seq_puts(seq, ",noinit_inode_table");
1059 else if (sbi->s_li_wait_mult)
1060 seq_printf(seq, ",init_inode_table=%u",
1061 (unsigned) sbi->s_li_wait_mult);
1063 ext4_show_quota_options(seq, sb);
1068 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1069 u64 ino, u32 generation)
1071 struct inode *inode;
1073 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1074 return ERR_PTR(-ESTALE);
1075 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1076 return ERR_PTR(-ESTALE);
1078 /* iget isn't really right if the inode is currently unallocated!!
1080 * ext4_read_inode will return a bad_inode if the inode had been
1081 * deleted, so we should be safe.
1083 * Currently we don't know the generation for parent directory, so
1084 * a generation of 0 means "accept any"
1086 inode = ext4_iget(sb, ino);
1088 return ERR_CAST(inode);
1089 if (generation && inode->i_generation != generation) {
1091 return ERR_PTR(-ESTALE);
1097 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1098 int fh_len, int fh_type)
1100 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1101 ext4_nfs_get_inode);
1104 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1105 int fh_len, int fh_type)
1107 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1108 ext4_nfs_get_inode);
1112 * Try to release metadata pages (indirect blocks, directories) which are
1113 * mapped via the block device. Since these pages could have journal heads
1114 * which would prevent try_to_free_buffers() from freeing them, we must use
1115 * jbd2 layer's try_to_free_buffers() function to release them.
1117 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1120 journal_t *journal = EXT4_SB(sb)->s_journal;
1122 WARN_ON(PageChecked(page));
1123 if (!page_has_buffers(page))
1126 return jbd2_journal_try_to_free_buffers(journal, page,
1127 wait & ~__GFP_WAIT);
1128 return try_to_free_buffers(page);
1132 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1133 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1135 static int ext4_write_dquot(struct dquot *dquot);
1136 static int ext4_acquire_dquot(struct dquot *dquot);
1137 static int ext4_release_dquot(struct dquot *dquot);
1138 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1139 static int ext4_write_info(struct super_block *sb, int type);
1140 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1142 static int ext4_quota_off(struct super_block *sb, int type);
1143 static int ext4_quota_on_mount(struct super_block *sb, int type);
1144 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1145 size_t len, loff_t off);
1146 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1147 const char *data, size_t len, loff_t off);
1149 static const struct dquot_operations ext4_quota_operations = {
1151 .get_reserved_space = ext4_get_reserved_space,
1153 .write_dquot = ext4_write_dquot,
1154 .acquire_dquot = ext4_acquire_dquot,
1155 .release_dquot = ext4_release_dquot,
1156 .mark_dirty = ext4_mark_dquot_dirty,
1157 .write_info = ext4_write_info,
1158 .alloc_dquot = dquot_alloc,
1159 .destroy_dquot = dquot_destroy,
1162 static const struct quotactl_ops ext4_qctl_operations = {
1163 .quota_on = ext4_quota_on,
1164 .quota_off = ext4_quota_off,
1165 .quota_sync = dquot_quota_sync,
1166 .get_info = dquot_get_dqinfo,
1167 .set_info = dquot_set_dqinfo,
1168 .get_dqblk = dquot_get_dqblk,
1169 .set_dqblk = dquot_set_dqblk
1173 static const struct super_operations ext4_sops = {
1174 .alloc_inode = ext4_alloc_inode,
1175 .destroy_inode = ext4_destroy_inode,
1176 .write_inode = ext4_write_inode,
1177 .dirty_inode = ext4_dirty_inode,
1178 .evict_inode = ext4_evict_inode,
1179 .put_super = ext4_put_super,
1180 .sync_fs = ext4_sync_fs,
1181 .freeze_fs = ext4_freeze,
1182 .unfreeze_fs = ext4_unfreeze,
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 super_operations ext4_nojournal_sops = {
1194 .alloc_inode = ext4_alloc_inode,
1195 .destroy_inode = ext4_destroy_inode,
1196 .write_inode = ext4_write_inode,
1197 .dirty_inode = ext4_dirty_inode,
1198 .evict_inode = ext4_evict_inode,
1199 .write_super = ext4_write_super,
1200 .put_super = ext4_put_super,
1201 .statfs = ext4_statfs,
1202 .remount_fs = ext4_remount,
1203 .show_options = ext4_show_options,
1205 .quota_read = ext4_quota_read,
1206 .quota_write = ext4_quota_write,
1208 .bdev_try_to_free_page = bdev_try_to_free_page,
1211 static const struct export_operations ext4_export_ops = {
1212 .fh_to_dentry = ext4_fh_to_dentry,
1213 .fh_to_parent = ext4_fh_to_parent,
1214 .get_parent = ext4_get_parent,
1218 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1219 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1220 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1221 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1222 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1223 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1224 Opt_journal_update, Opt_journal_dev,
1225 Opt_journal_checksum, Opt_journal_async_commit,
1226 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1227 Opt_data_err_abort, Opt_data_err_ignore,
1228 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1229 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1230 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1231 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1232 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1233 Opt_block_validity, Opt_noblock_validity,
1234 Opt_inode_readahead_blks, Opt_journal_ioprio,
1235 Opt_dioread_nolock, Opt_dioread_lock,
1236 Opt_discard, Opt_nodiscard,
1237 Opt_init_inode_table, Opt_noinit_inode_table,
1240 static const match_table_t tokens = {
1241 {Opt_bsd_df, "bsddf"},
1242 {Opt_minix_df, "minixdf"},
1243 {Opt_grpid, "grpid"},
1244 {Opt_grpid, "bsdgroups"},
1245 {Opt_nogrpid, "nogrpid"},
1246 {Opt_nogrpid, "sysvgroups"},
1247 {Opt_resgid, "resgid=%u"},
1248 {Opt_resuid, "resuid=%u"},
1250 {Opt_err_cont, "errors=continue"},
1251 {Opt_err_panic, "errors=panic"},
1252 {Opt_err_ro, "errors=remount-ro"},
1253 {Opt_nouid32, "nouid32"},
1254 {Opt_debug, "debug"},
1255 {Opt_oldalloc, "oldalloc"},
1256 {Opt_orlov, "orlov"},
1257 {Opt_user_xattr, "user_xattr"},
1258 {Opt_nouser_xattr, "nouser_xattr"},
1260 {Opt_noacl, "noacl"},
1261 {Opt_noload, "noload"},
1262 {Opt_noload, "norecovery"},
1265 {Opt_commit, "commit=%u"},
1266 {Opt_min_batch_time, "min_batch_time=%u"},
1267 {Opt_max_batch_time, "max_batch_time=%u"},
1268 {Opt_journal_update, "journal=update"},
1269 {Opt_journal_dev, "journal_dev=%u"},
1270 {Opt_journal_checksum, "journal_checksum"},
1271 {Opt_journal_async_commit, "journal_async_commit"},
1272 {Opt_abort, "abort"},
1273 {Opt_data_journal, "data=journal"},
1274 {Opt_data_ordered, "data=ordered"},
1275 {Opt_data_writeback, "data=writeback"},
1276 {Opt_data_err_abort, "data_err=abort"},
1277 {Opt_data_err_ignore, "data_err=ignore"},
1278 {Opt_offusrjquota, "usrjquota="},
1279 {Opt_usrjquota, "usrjquota=%s"},
1280 {Opt_offgrpjquota, "grpjquota="},
1281 {Opt_grpjquota, "grpjquota=%s"},
1282 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1283 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1284 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1285 {Opt_grpquota, "grpquota"},
1286 {Opt_noquota, "noquota"},
1287 {Opt_quota, "quota"},
1288 {Opt_usrquota, "usrquota"},
1289 {Opt_barrier, "barrier=%u"},
1290 {Opt_barrier, "barrier"},
1291 {Opt_nobarrier, "nobarrier"},
1292 {Opt_i_version, "i_version"},
1293 {Opt_stripe, "stripe=%u"},
1294 {Opt_resize, "resize"},
1295 {Opt_delalloc, "delalloc"},
1296 {Opt_nodelalloc, "nodelalloc"},
1297 {Opt_block_validity, "block_validity"},
1298 {Opt_noblock_validity, "noblock_validity"},
1299 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1300 {Opt_journal_ioprio, "journal_ioprio=%u"},
1301 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1302 {Opt_auto_da_alloc, "auto_da_alloc"},
1303 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1304 {Opt_dioread_nolock, "dioread_nolock"},
1305 {Opt_dioread_lock, "dioread_lock"},
1306 {Opt_discard, "discard"},
1307 {Opt_nodiscard, "nodiscard"},
1308 {Opt_init_inode_table, "init_itable=%u"},
1309 {Opt_init_inode_table, "init_itable"},
1310 {Opt_noinit_inode_table, "noinit_itable"},
1314 static ext4_fsblk_t get_sb_block(void **data)
1316 ext4_fsblk_t sb_block;
1317 char *options = (char *) *data;
1319 if (!options || strncmp(options, "sb=", 3) != 0)
1320 return 1; /* Default location */
1323 /* TODO: use simple_strtoll with >32bit ext4 */
1324 sb_block = simple_strtoul(options, &options, 0);
1325 if (*options && *options != ',') {
1326 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1330 if (*options == ',')
1332 *data = (void *) options;
1337 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1338 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1339 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1342 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1344 struct ext4_sb_info *sbi = EXT4_SB(sb);
1347 if (sb_any_quota_loaded(sb) &&
1348 !sbi->s_qf_names[qtype]) {
1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change journaled "
1351 "quota options when quota turned on");
1354 qname = match_strdup(args);
1356 ext4_msg(sb, KERN_ERR,
1357 "Not enough memory for storing quotafile name");
1360 if (sbi->s_qf_names[qtype] &&
1361 strcmp(sbi->s_qf_names[qtype], qname)) {
1362 ext4_msg(sb, KERN_ERR,
1363 "%s quota file already specified", QTYPE2NAME(qtype));
1367 sbi->s_qf_names[qtype] = qname;
1368 if (strchr(sbi->s_qf_names[qtype], '/')) {
1369 ext4_msg(sb, KERN_ERR,
1370 "quotafile must be on filesystem root");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
1375 set_opt(sbi->s_mount_opt, QUOTA);
1379 static int clear_qf_name(struct super_block *sb, int qtype)
1382 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384 if (sb_any_quota_loaded(sb) &&
1385 sbi->s_qf_names[qtype]) {
1386 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1387 " when quota turned on");
1391 * The space will be released later when all options are confirmed
1394 sbi->s_qf_names[qtype] = NULL;
1399 static int parse_options(char *options, struct super_block *sb,
1400 unsigned long *journal_devnum,
1401 unsigned int *journal_ioprio,
1402 ext4_fsblk_t *n_blocks_count, int is_remount)
1404 struct ext4_sb_info *sbi = EXT4_SB(sb);
1406 substring_t args[MAX_OPT_ARGS];
1416 while ((p = strsep(&options, ",")) != NULL) {
1422 * Initialize args struct so we know whether arg was
1423 * found; some options take optional arguments.
1425 args[0].to = args[0].from = 0;
1426 token = match_token(p, tokens, args);
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1430 clear_opt(sbi->s_mount_opt, MINIX_DF);
1433 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1434 set_opt(sbi->s_mount_opt, MINIX_DF);
1438 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1439 set_opt(sbi->s_mount_opt, GRPID);
1443 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1444 clear_opt(sbi->s_mount_opt, GRPID);
1448 if (match_int(&args[0], &option))
1450 sbi->s_resuid = option;
1453 if (match_int(&args[0], &option))
1455 sbi->s_resgid = option;
1458 /* handled by get_sb_block() instead of here */
1459 /* *sb_block = match_int(&args[0]); */
1462 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1463 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1464 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1467 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1468 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1469 set_opt(sbi->s_mount_opt, ERRORS_RO);
1472 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1473 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1474 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1477 set_opt(sbi->s_mount_opt, NO_UID32);
1480 set_opt(sbi->s_mount_opt, DEBUG);
1483 set_opt(sbi->s_mount_opt, OLDALLOC);
1486 clear_opt(sbi->s_mount_opt, OLDALLOC);
1488 #ifdef CONFIG_EXT4_FS_XATTR
1489 case Opt_user_xattr:
1490 set_opt(sbi->s_mount_opt, XATTR_USER);
1492 case Opt_nouser_xattr:
1493 clear_opt(sbi->s_mount_opt, XATTR_USER);
1496 case Opt_user_xattr:
1497 case Opt_nouser_xattr:
1498 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1501 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1503 set_opt(sbi->s_mount_opt, POSIX_ACL);
1506 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1511 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1514 case Opt_journal_update:
1516 /* Eventually we will want to be able to create
1517 a journal file here. For now, only allow the
1518 user to specify an existing inode to be the
1521 ext4_msg(sb, KERN_ERR,
1522 "Cannot specify journal on remount");
1525 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1527 case Opt_journal_dev:
1529 ext4_msg(sb, KERN_ERR,
1530 "Cannot specify journal on remount");
1533 if (match_int(&args[0], &option))
1535 *journal_devnum = option;
1537 case Opt_journal_checksum:
1538 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1540 case Opt_journal_async_commit:
1541 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1542 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1545 set_opt(sbi->s_mount_opt, NOLOAD);
1548 if (match_int(&args[0], &option))
1553 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1554 sbi->s_commit_interval = HZ * option;
1556 case Opt_max_batch_time:
1557 if (match_int(&args[0], &option))
1562 option = EXT4_DEF_MAX_BATCH_TIME;
1563 sbi->s_max_batch_time = option;
1565 case Opt_min_batch_time:
1566 if (match_int(&args[0], &option))
1570 sbi->s_min_batch_time = option;
1572 case Opt_data_journal:
1573 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1575 case Opt_data_ordered:
1576 data_opt = EXT4_MOUNT_ORDERED_DATA;
1578 case Opt_data_writeback:
1579 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1582 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1583 ext4_msg(sb, KERN_ERR,
1584 "Cannot change data mode on remount");
1588 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1589 sbi->s_mount_opt |= data_opt;
1592 case Opt_data_err_abort:
1593 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1595 case Opt_data_err_ignore:
1596 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1600 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1604 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1607 case Opt_offusrjquota:
1608 if (!clear_qf_name(sb, USRQUOTA))
1611 case Opt_offgrpjquota:
1612 if (!clear_qf_name(sb, GRPQUOTA))
1616 case Opt_jqfmt_vfsold:
1617 qfmt = QFMT_VFS_OLD;
1619 case Opt_jqfmt_vfsv0:
1622 case Opt_jqfmt_vfsv1:
1625 if (sb_any_quota_loaded(sb) &&
1626 sbi->s_jquota_fmt != qfmt) {
1627 ext4_msg(sb, KERN_ERR, "Cannot change "
1628 "journaled quota options when "
1632 sbi->s_jquota_fmt = qfmt;
1636 set_opt(sbi->s_mount_opt, QUOTA);
1637 set_opt(sbi->s_mount_opt, USRQUOTA);
1640 set_opt(sbi->s_mount_opt, QUOTA);
1641 set_opt(sbi->s_mount_opt, GRPQUOTA);
1644 if (sb_any_quota_loaded(sb)) {
1645 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1646 "options when quota turned on");
1649 clear_opt(sbi->s_mount_opt, QUOTA);
1650 clear_opt(sbi->s_mount_opt, USRQUOTA);
1651 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1657 ext4_msg(sb, KERN_ERR,
1658 "quota options not supported");
1662 case Opt_offusrjquota:
1663 case Opt_offgrpjquota:
1664 case Opt_jqfmt_vfsold:
1665 case Opt_jqfmt_vfsv0:
1666 case Opt_jqfmt_vfsv1:
1667 ext4_msg(sb, KERN_ERR,
1668 "journaled quota options not supported");
1674 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1677 clear_opt(sbi->s_mount_opt, BARRIER);
1681 if (match_int(&args[0], &option))
1684 option = 1; /* No argument, default to 1 */
1686 set_opt(sbi->s_mount_opt, BARRIER);
1688 clear_opt(sbi->s_mount_opt, BARRIER);
1694 ext4_msg(sb, KERN_ERR,
1695 "resize option only available "
1699 if (match_int(&args[0], &option) != 0)
1701 *n_blocks_count = option;
1704 ext4_msg(sb, KERN_WARNING,
1705 "Ignoring deprecated nobh option");
1708 ext4_msg(sb, KERN_WARNING,
1709 "Ignoring deprecated bh option");
1712 set_opt(sbi->s_mount_opt, I_VERSION);
1713 sb->s_flags |= MS_I_VERSION;
1715 case Opt_nodelalloc:
1716 clear_opt(sbi->s_mount_opt, DELALLOC);
1719 if (match_int(&args[0], &option))
1723 sbi->s_stripe = option;
1726 set_opt(sbi->s_mount_opt, DELALLOC);
1728 case Opt_block_validity:
1729 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1731 case Opt_noblock_validity:
1732 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1734 case Opt_inode_readahead_blks:
1735 if (match_int(&args[0], &option))
1737 if (option < 0 || option > (1 << 30))
1739 if (!is_power_of_2(option)) {
1740 ext4_msg(sb, KERN_ERR,
1741 "EXT4-fs: inode_readahead_blks"
1742 " must be a power of 2");
1745 sbi->s_inode_readahead_blks = option;
1747 case Opt_journal_ioprio:
1748 if (match_int(&args[0], &option))
1750 if (option < 0 || option > 7)
1752 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1755 case Opt_noauto_da_alloc:
1756 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1758 case Opt_auto_da_alloc:
1760 if (match_int(&args[0], &option))
1763 option = 1; /* No argument, default to 1 */
1765 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1767 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1770 set_opt(sbi->s_mount_opt, DISCARD);
1773 clear_opt(sbi->s_mount_opt, DISCARD);
1775 case Opt_dioread_nolock:
1776 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1778 case Opt_dioread_lock:
1779 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1781 case Opt_init_inode_table:
1782 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1784 if (match_int(&args[0], &option))
1787 option = EXT4_DEF_LI_WAIT_MULT;
1790 sbi->s_li_wait_mult = option;
1792 case Opt_noinit_inode_table:
1793 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1796 ext4_msg(sb, KERN_ERR,
1797 "Unrecognized mount option \"%s\" "
1798 "or missing value", p);
1803 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1804 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1805 clear_opt(sbi->s_mount_opt, USRQUOTA);
1807 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1808 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1810 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1811 ext4_msg(sb, KERN_ERR, "old and new quota "
1816 if (!sbi->s_jquota_fmt) {
1817 ext4_msg(sb, KERN_ERR, "journaled quota format "
1822 if (sbi->s_jquota_fmt) {
1823 ext4_msg(sb, KERN_ERR, "journaled quota format "
1824 "specified with no journaling "
1833 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1836 struct ext4_sb_info *sbi = EXT4_SB(sb);
1839 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1840 ext4_msg(sb, KERN_ERR, "revision level too high, "
1841 "forcing read-only mode");
1846 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1847 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1848 "running e2fsck is recommended");
1849 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1850 ext4_msg(sb, KERN_WARNING,
1851 "warning: mounting fs with errors, "
1852 "running e2fsck is recommended");
1853 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1854 le16_to_cpu(es->s_mnt_count) >=
1855 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1856 ext4_msg(sb, KERN_WARNING,
1857 "warning: maximal mount count reached, "
1858 "running e2fsck is recommended");
1859 else if (le32_to_cpu(es->s_checkinterval) &&
1860 (le32_to_cpu(es->s_lastcheck) +
1861 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1862 ext4_msg(sb, KERN_WARNING,
1863 "warning: checktime reached, "
1864 "running e2fsck is recommended");
1865 if (!sbi->s_journal)
1866 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1867 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1868 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1869 le16_add_cpu(&es->s_mnt_count, 1);
1870 es->s_mtime = cpu_to_le32(get_seconds());
1871 ext4_update_dynamic_rev(sb);
1873 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1875 ext4_commit_super(sb, 1);
1876 if (test_opt(sb, DEBUG))
1877 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1878 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1880 sbi->s_groups_count,
1881 EXT4_BLOCKS_PER_GROUP(sb),
1882 EXT4_INODES_PER_GROUP(sb),
1888 static int ext4_fill_flex_info(struct super_block *sb)
1890 struct ext4_sb_info *sbi = EXT4_SB(sb);
1891 struct ext4_group_desc *gdp = NULL;
1892 ext4_group_t flex_group_count;
1893 ext4_group_t flex_group;
1894 int groups_per_flex = 0;
1898 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1899 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1901 if (groups_per_flex < 2) {
1902 sbi->s_log_groups_per_flex = 0;
1906 /* We allocate both existing and potentially added groups */
1907 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1908 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1909 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1910 size = flex_group_count * sizeof(struct flex_groups);
1911 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1912 if (sbi->s_flex_groups == NULL) {
1913 sbi->s_flex_groups = vmalloc(size);
1914 if (sbi->s_flex_groups)
1915 memset(sbi->s_flex_groups, 0, size);
1917 if (sbi->s_flex_groups == NULL) {
1918 ext4_msg(sb, KERN_ERR, "not enough memory for "
1919 "%u flex groups", flex_group_count);
1923 for (i = 0; i < sbi->s_groups_count; i++) {
1924 gdp = ext4_get_group_desc(sb, i, NULL);
1926 flex_group = ext4_flex_group(sbi, i);
1927 atomic_add(ext4_free_inodes_count(sb, gdp),
1928 &sbi->s_flex_groups[flex_group].free_inodes);
1929 atomic_add(ext4_free_blks_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].free_blocks);
1931 atomic_add(ext4_used_dirs_count(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].used_dirs);
1940 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1941 struct ext4_group_desc *gdp)
1945 if (sbi->s_es->s_feature_ro_compat &
1946 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1947 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1948 __le32 le_group = cpu_to_le32(block_group);
1950 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1951 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1952 crc = crc16(crc, (__u8 *)gdp, offset);
1953 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1954 /* for checksum of struct ext4_group_desc do the rest...*/
1955 if ((sbi->s_es->s_feature_incompat &
1956 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1957 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1958 crc = crc16(crc, (__u8 *)gdp + offset,
1959 le16_to_cpu(sbi->s_es->s_desc_size) -
1963 return cpu_to_le16(crc);
1966 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1967 struct ext4_group_desc *gdp)
1969 if ((sbi->s_es->s_feature_ro_compat &
1970 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1971 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1977 /* Called at mount-time, super-block is locked */
1978 static int ext4_check_descriptors(struct super_block *sb,
1979 ext4_group_t *first_not_zeroed)
1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
1982 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1983 ext4_fsblk_t last_block;
1984 ext4_fsblk_t block_bitmap;
1985 ext4_fsblk_t inode_bitmap;
1986 ext4_fsblk_t inode_table;
1987 int flexbg_flag = 0;
1988 ext4_group_t i, grp = sbi->s_groups_count;
1990 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1993 ext4_debug("Checking group descriptors");
1995 for (i = 0; i < sbi->s_groups_count; i++) {
1996 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1998 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1999 last_block = ext4_blocks_count(sbi->s_es) - 1;
2001 last_block = first_block +
2002 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2004 if ((grp == sbi->s_groups_count) &&
2005 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2008 block_bitmap = ext4_block_bitmap(sb, gdp);
2009 if (block_bitmap < first_block || block_bitmap > last_block) {
2010 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2011 "Block bitmap for group %u not in group "
2012 "(block %llu)!", i, block_bitmap);
2015 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2016 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2017 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2018 "Inode bitmap for group %u not in group "
2019 "(block %llu)!", i, inode_bitmap);
2022 inode_table = ext4_inode_table(sb, gdp);
2023 if (inode_table < first_block ||
2024 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2025 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2026 "Inode table for group %u not in group "
2027 "(block %llu)!", i, inode_table);
2030 ext4_lock_group(sb, i);
2031 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2032 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2033 "Checksum for group %u failed (%u!=%u)",
2034 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2035 gdp)), le16_to_cpu(gdp->bg_checksum));
2036 if (!(sb->s_flags & MS_RDONLY)) {
2037 ext4_unlock_group(sb, i);
2041 ext4_unlock_group(sb, i);
2043 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2045 if (NULL != first_not_zeroed)
2046 *first_not_zeroed = grp;
2048 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2049 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2053 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2054 * the superblock) which were deleted from all directories, but held open by
2055 * a process at the time of a crash. We walk the list and try to delete these
2056 * inodes at recovery time (only with a read-write filesystem).
2058 * In order to keep the orphan inode chain consistent during traversal (in
2059 * case of crash during recovery), we link each inode into the superblock
2060 * orphan list_head and handle it the same way as an inode deletion during
2061 * normal operation (which journals the operations for us).
2063 * We only do an iget() and an iput() on each inode, which is very safe if we
2064 * accidentally point at an in-use or already deleted inode. The worst that
2065 * can happen in this case is that we get a "bit already cleared" message from
2066 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2067 * e2fsck was run on this filesystem, and it must have already done the orphan
2068 * inode cleanup for us, so we can safely abort without any further action.
2070 static void ext4_orphan_cleanup(struct super_block *sb,
2071 struct ext4_super_block *es)
2073 unsigned int s_flags = sb->s_flags;
2074 int nr_orphans = 0, nr_truncates = 0;
2078 if (!es->s_last_orphan) {
2079 jbd_debug(4, "no orphan inodes to clean up\n");
2083 if (bdev_read_only(sb->s_bdev)) {
2084 ext4_msg(sb, KERN_ERR, "write access "
2085 "unavailable, skipping orphan cleanup");
2089 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2090 if (es->s_last_orphan)
2091 jbd_debug(1, "Errors on filesystem, "
2092 "clearing orphan list.\n");
2093 es->s_last_orphan = 0;
2094 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2098 if (s_flags & MS_RDONLY) {
2099 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2100 sb->s_flags &= ~MS_RDONLY;
2103 /* Needed for iput() to work correctly and not trash data */
2104 sb->s_flags |= MS_ACTIVE;
2105 /* Turn on quotas so that they are updated correctly */
2106 for (i = 0; i < MAXQUOTAS; i++) {
2107 if (EXT4_SB(sb)->s_qf_names[i]) {
2108 int ret = ext4_quota_on_mount(sb, i);
2110 ext4_msg(sb, KERN_ERR,
2111 "Cannot turn on journaled "
2112 "quota: error %d", ret);
2117 while (es->s_last_orphan) {
2118 struct inode *inode;
2120 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2121 if (IS_ERR(inode)) {
2122 es->s_last_orphan = 0;
2126 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2127 dquot_initialize(inode);
2128 if (inode->i_nlink) {
2129 ext4_msg(sb, KERN_DEBUG,
2130 "%s: truncating inode %lu to %lld bytes",
2131 __func__, inode->i_ino, inode->i_size);
2132 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2133 inode->i_ino, inode->i_size);
2134 ext4_truncate(inode);
2137 ext4_msg(sb, KERN_DEBUG,
2138 "%s: deleting unreferenced inode %lu",
2139 __func__, inode->i_ino);
2140 jbd_debug(2, "deleting unreferenced inode %lu\n",
2144 iput(inode); /* The delete magic happens here! */
2147 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2150 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2151 PLURAL(nr_orphans));
2153 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2154 PLURAL(nr_truncates));
2156 /* Turn quotas off */
2157 for (i = 0; i < MAXQUOTAS; i++) {
2158 if (sb_dqopt(sb)->files[i])
2159 dquot_quota_off(sb, i);
2162 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2166 * Maximal extent format file size.
2167 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2168 * extent format containers, within a sector_t, and within i_blocks
2169 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2170 * so that won't be a limiting factor.
2172 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2174 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2177 loff_t upper_limit = MAX_LFS_FILESIZE;
2179 /* small i_blocks in vfs inode? */
2180 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2182 * CONFIG_LBDAF is not enabled implies the inode
2183 * i_block represent total blocks in 512 bytes
2184 * 32 == size of vfs inode i_blocks * 8
2186 upper_limit = (1LL << 32) - 1;
2188 /* total blocks in file system block size */
2189 upper_limit >>= (blkbits - 9);
2190 upper_limit <<= blkbits;
2193 /* 32-bit extent-start container, ee_block */
2198 /* Sanity check against vm- & vfs- imposed limits */
2199 if (res > upper_limit)
2206 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2207 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2208 * We need to be 1 filesystem block less than the 2^48 sector limit.
2210 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2212 loff_t res = EXT4_NDIR_BLOCKS;
2215 /* This is calculated to be the largest file size for a dense, block
2216 * mapped file such that the file's total number of 512-byte sectors,
2217 * including data and all indirect blocks, does not exceed (2^48 - 1).
2219 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2220 * number of 512-byte sectors of the file.
2223 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2225 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2226 * the inode i_block field represents total file blocks in
2227 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2229 upper_limit = (1LL << 32) - 1;
2231 /* total blocks in file system block size */
2232 upper_limit >>= (bits - 9);
2236 * We use 48 bit ext4_inode i_blocks
2237 * With EXT4_HUGE_FILE_FL set the i_blocks
2238 * represent total number of blocks in
2239 * file system block size
2241 upper_limit = (1LL << 48) - 1;
2245 /* indirect blocks */
2247 /* double indirect blocks */
2248 meta_blocks += 1 + (1LL << (bits-2));
2249 /* tripple indirect blocks */
2250 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2252 upper_limit -= meta_blocks;
2253 upper_limit <<= bits;
2255 res += 1LL << (bits-2);
2256 res += 1LL << (2*(bits-2));
2257 res += 1LL << (3*(bits-2));
2259 if (res > upper_limit)
2262 if (res > MAX_LFS_FILESIZE)
2263 res = MAX_LFS_FILESIZE;
2268 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2269 ext4_fsblk_t logical_sb_block, int nr)
2271 struct ext4_sb_info *sbi = EXT4_SB(sb);
2272 ext4_group_t bg, first_meta_bg;
2275 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2277 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2279 return logical_sb_block + nr + 1;
2280 bg = sbi->s_desc_per_block * nr;
2281 if (ext4_bg_has_super(sb, bg))
2284 return (has_super + ext4_group_first_block_no(sb, bg));
2288 * ext4_get_stripe_size: Get the stripe size.
2289 * @sbi: In memory super block info
2291 * If we have specified it via mount option, then
2292 * use the mount option value. If the value specified at mount time is
2293 * greater than the blocks per group use the super block value.
2294 * If the super block value is greater than blocks per group return 0.
2295 * Allocator needs it be less than blocks per group.
2298 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2300 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2301 unsigned long stripe_width =
2302 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2304 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2305 return sbi->s_stripe;
2307 if (stripe_width <= sbi->s_blocks_per_group)
2308 return stripe_width;
2310 if (stride <= sbi->s_blocks_per_group)
2319 struct attribute attr;
2320 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2321 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2322 const char *, size_t);
2326 static int parse_strtoul(const char *buf,
2327 unsigned long max, unsigned long *value)
2331 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2332 endp = skip_spaces(endp);
2333 if (*endp || *value > max)
2339 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2340 struct ext4_sb_info *sbi,
2343 return snprintf(buf, PAGE_SIZE, "%llu\n",
2344 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2347 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2348 struct ext4_sb_info *sbi, char *buf)
2350 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2352 if (!sb->s_bdev->bd_part)
2353 return snprintf(buf, PAGE_SIZE, "0\n");
2354 return snprintf(buf, PAGE_SIZE, "%lu\n",
2355 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2356 sbi->s_sectors_written_start) >> 1);
2359 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2360 struct ext4_sb_info *sbi, char *buf)
2362 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2364 if (!sb->s_bdev->bd_part)
2365 return snprintf(buf, PAGE_SIZE, "0\n");
2366 return snprintf(buf, PAGE_SIZE, "%llu\n",
2367 (unsigned long long)(sbi->s_kbytes_written +
2368 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2369 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2372 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2373 struct ext4_sb_info *sbi,
2374 const char *buf, size_t count)
2378 if (parse_strtoul(buf, 0x40000000, &t))
2381 if (!is_power_of_2(t))
2384 sbi->s_inode_readahead_blks = t;
2388 static ssize_t sbi_ui_show(struct ext4_attr *a,
2389 struct ext4_sb_info *sbi, char *buf)
2391 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2393 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2396 static ssize_t sbi_ui_store(struct ext4_attr *a,
2397 struct ext4_sb_info *sbi,
2398 const char *buf, size_t count)
2400 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2403 if (parse_strtoul(buf, 0xffffffff, &t))
2409 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2410 static struct ext4_attr ext4_attr_##_name = { \
2411 .attr = {.name = __stringify(_name), .mode = _mode }, \
2414 .offset = offsetof(struct ext4_sb_info, _elname), \
2416 #define EXT4_ATTR(name, mode, show, store) \
2417 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2419 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2420 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2421 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2422 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2423 #define ATTR_LIST(name) &ext4_attr_##name.attr
2425 EXT4_RO_ATTR(delayed_allocation_blocks);
2426 EXT4_RO_ATTR(session_write_kbytes);
2427 EXT4_RO_ATTR(lifetime_write_kbytes);
2428 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2429 inode_readahead_blks_store, s_inode_readahead_blks);
2430 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2431 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2432 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2433 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2434 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2435 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2436 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2437 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2439 static struct attribute *ext4_attrs[] = {
2440 ATTR_LIST(delayed_allocation_blocks),
2441 ATTR_LIST(session_write_kbytes),
2442 ATTR_LIST(lifetime_write_kbytes),
2443 ATTR_LIST(inode_readahead_blks),
2444 ATTR_LIST(inode_goal),
2445 ATTR_LIST(mb_stats),
2446 ATTR_LIST(mb_max_to_scan),
2447 ATTR_LIST(mb_min_to_scan),
2448 ATTR_LIST(mb_order2_req),
2449 ATTR_LIST(mb_stream_req),
2450 ATTR_LIST(mb_group_prealloc),
2451 ATTR_LIST(max_writeback_mb_bump),
2455 static ssize_t ext4_attr_show(struct kobject *kobj,
2456 struct attribute *attr, char *buf)
2458 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2460 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2462 return a->show ? a->show(a, sbi, buf) : 0;
2465 static ssize_t ext4_attr_store(struct kobject *kobj,
2466 struct attribute *attr,
2467 const char *buf, size_t len)
2469 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2471 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2473 return a->store ? a->store(a, sbi, buf, len) : 0;
2476 static void ext4_sb_release(struct kobject *kobj)
2478 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2480 complete(&sbi->s_kobj_unregister);
2484 static const struct sysfs_ops ext4_attr_ops = {
2485 .show = ext4_attr_show,
2486 .store = ext4_attr_store,
2489 static struct kobj_type ext4_ktype = {
2490 .default_attrs = ext4_attrs,
2491 .sysfs_ops = &ext4_attr_ops,
2492 .release = ext4_sb_release,
2496 * Check whether this filesystem can be mounted based on
2497 * the features present and the RDONLY/RDWR mount requested.
2498 * Returns 1 if this filesystem can be mounted as requested,
2499 * 0 if it cannot be.
2501 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2503 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2504 ext4_msg(sb, KERN_ERR,
2505 "Couldn't mount because of "
2506 "unsupported optional features (%x)",
2507 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2508 ~EXT4_FEATURE_INCOMPAT_SUPP));
2515 /* Check that feature set is OK for a read-write mount */
2516 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2517 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2518 "unsupported optional features (%x)",
2519 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2520 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2524 * Large file size enabled file system can only be mounted
2525 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2527 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2528 if (sizeof(blkcnt_t) < sizeof(u64)) {
2529 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2530 "cannot be mounted RDWR without "
2539 * This function is called once a day if we have errors logged
2540 * on the file system
2542 static void print_daily_error_info(unsigned long arg)
2544 struct super_block *sb = (struct super_block *) arg;
2545 struct ext4_sb_info *sbi;
2546 struct ext4_super_block *es;
2551 if (es->s_error_count)
2552 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2553 le32_to_cpu(es->s_error_count));
2554 if (es->s_first_error_time) {
2555 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2556 sb->s_id, le32_to_cpu(es->s_first_error_time),
2557 (int) sizeof(es->s_first_error_func),
2558 es->s_first_error_func,
2559 le32_to_cpu(es->s_first_error_line));
2560 if (es->s_first_error_ino)
2561 printk(": inode %u",
2562 le32_to_cpu(es->s_first_error_ino));
2563 if (es->s_first_error_block)
2564 printk(": block %llu", (unsigned long long)
2565 le64_to_cpu(es->s_first_error_block));
2568 if (es->s_last_error_time) {
2569 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2570 sb->s_id, le32_to_cpu(es->s_last_error_time),
2571 (int) sizeof(es->s_last_error_func),
2572 es->s_last_error_func,
2573 le32_to_cpu(es->s_last_error_line));
2574 if (es->s_last_error_ino)
2575 printk(": inode %u",
2576 le32_to_cpu(es->s_last_error_ino));
2577 if (es->s_last_error_block)
2578 printk(": block %llu", (unsigned long long)
2579 le64_to_cpu(es->s_last_error_block));
2582 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2585 static void ext4_lazyinode_timeout(unsigned long data)
2587 struct task_struct *p = (struct task_struct *)data;
2591 /* Find next suitable group and run ext4_init_inode_table */
2592 static int ext4_run_li_request(struct ext4_li_request *elr)
2594 struct ext4_group_desc *gdp = NULL;
2595 ext4_group_t group, ngroups;
2596 struct super_block *sb;
2597 unsigned long timeout = 0;
2601 ngroups = EXT4_SB(sb)->s_groups_count;
2603 for (group = elr->lr_next_group; group < ngroups; group++) {
2604 gdp = ext4_get_group_desc(sb, group, NULL);
2610 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2614 if (group == ngroups)
2619 ret = ext4_init_inode_table(sb, group,
2620 elr->lr_timeout ? 0 : 1);
2621 if (elr->lr_timeout == 0) {
2622 timeout = jiffies - timeout;
2623 if (elr->lr_sbi->s_li_wait_mult)
2624 timeout *= elr->lr_sbi->s_li_wait_mult;
2627 elr->lr_timeout = timeout;
2629 elr->lr_next_sched = jiffies + elr->lr_timeout;
2630 elr->lr_next_group = group + 1;
2637 * Remove lr_request from the list_request and free the
2638 * request tructure. Should be called with li_list_mtx held
2640 static void ext4_remove_li_request(struct ext4_li_request *elr)
2642 struct ext4_sb_info *sbi;
2649 list_del(&elr->lr_request);
2650 sbi->s_li_request = NULL;
2654 static void ext4_unregister_li_request(struct super_block *sb)
2656 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2661 mutex_lock(&ext4_li_info->li_list_mtx);
2662 ext4_remove_li_request(elr);
2663 mutex_unlock(&ext4_li_info->li_list_mtx);
2667 * This is the function where ext4lazyinit thread lives. It walks
2668 * through the request list searching for next scheduled filesystem.
2669 * When such a fs is found, run the lazy initialization request
2670 * (ext4_rn_li_request) and keep track of the time spend in this
2671 * function. Based on that time we compute next schedule time of
2672 * the request. When walking through the list is complete, compute
2673 * next waking time and put itself into sleep.
2675 static int ext4_lazyinit_thread(void *arg)
2677 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2678 struct list_head *pos, *n;
2679 struct ext4_li_request *elr;
2680 unsigned long next_wakeup;
2684 BUG_ON(NULL == eli);
2686 eli->li_timer.data = (unsigned long)current;
2687 eli->li_timer.function = ext4_lazyinode_timeout;
2689 eli->li_task = current;
2690 wake_up(&eli->li_wait_task);
2694 next_wakeup = MAX_JIFFY_OFFSET;
2696 mutex_lock(&eli->li_list_mtx);
2697 if (list_empty(&eli->li_request_list)) {
2698 mutex_unlock(&eli->li_list_mtx);
2702 list_for_each_safe(pos, n, &eli->li_request_list) {
2703 elr = list_entry(pos, struct ext4_li_request,
2706 if (time_after_eq(jiffies, elr->lr_next_sched))
2707 ret = ext4_run_li_request(elr);
2711 ext4_remove_li_request(elr);
2715 if (time_before(elr->lr_next_sched, next_wakeup))
2716 next_wakeup = elr->lr_next_sched;
2718 mutex_unlock(&eli->li_list_mtx);
2720 if (freezing(current))
2723 if (time_after_eq(jiffies, next_wakeup)) {
2728 eli->li_timer.expires = next_wakeup;
2729 add_timer(&eli->li_timer);
2730 prepare_to_wait(&eli->li_wait_daemon, &wait,
2731 TASK_INTERRUPTIBLE);
2732 if (time_before(jiffies, next_wakeup))
2734 finish_wait(&eli->li_wait_daemon, &wait);
2739 * It looks like the request list is empty, but we need
2740 * to check it under the li_list_mtx lock, to prevent any
2741 * additions into it, and of course we should lock ext4_li_mtx
2742 * to atomically free the list and ext4_li_info, because at
2743 * this point another ext4 filesystem could be registering
2746 mutex_lock(&ext4_li_mtx);
2747 mutex_lock(&eli->li_list_mtx);
2748 if (!list_empty(&eli->li_request_list)) {
2749 mutex_unlock(&eli->li_list_mtx);
2750 mutex_unlock(&ext4_li_mtx);
2753 mutex_unlock(&eli->li_list_mtx);
2754 del_timer_sync(&ext4_li_info->li_timer);
2755 eli->li_task = NULL;
2756 wake_up(&eli->li_wait_task);
2758 kfree(ext4_li_info);
2759 ext4_li_info = NULL;
2760 mutex_unlock(&ext4_li_mtx);
2765 static void ext4_clear_request_list(void)
2767 struct list_head *pos, *n;
2768 struct ext4_li_request *elr;
2770 mutex_lock(&ext4_li_info->li_list_mtx);
2771 if (list_empty(&ext4_li_info->li_request_list))
2774 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2775 elr = list_entry(pos, struct ext4_li_request,
2777 ext4_remove_li_request(elr);
2779 mutex_unlock(&ext4_li_info->li_list_mtx);
2782 static int ext4_run_lazyinit_thread(void)
2784 struct task_struct *t;
2786 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2788 int err = PTR_ERR(t);
2789 ext4_clear_request_list();
2790 del_timer_sync(&ext4_li_info->li_timer);
2791 kfree(ext4_li_info);
2792 ext4_li_info = NULL;
2793 printk(KERN_CRIT "EXT4: error %d creating inode table "
2794 "initialization thread\n",
2798 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2800 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2805 * Check whether it make sense to run itable init. thread or not.
2806 * If there is at least one uninitialized inode table, return
2807 * corresponding group number, else the loop goes through all
2808 * groups and return total number of groups.
2810 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2812 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2813 struct ext4_group_desc *gdp = NULL;
2815 for (group = 0; group < ngroups; group++) {
2816 gdp = ext4_get_group_desc(sb, group, NULL);
2820 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2827 static int ext4_li_info_new(void)
2829 struct ext4_lazy_init *eli = NULL;
2831 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2835 eli->li_task = NULL;
2836 INIT_LIST_HEAD(&eli->li_request_list);
2837 mutex_init(&eli->li_list_mtx);
2839 init_waitqueue_head(&eli->li_wait_daemon);
2840 init_waitqueue_head(&eli->li_wait_task);
2841 init_timer(&eli->li_timer);
2842 eli->li_state |= EXT4_LAZYINIT_QUIT;
2849 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2852 struct ext4_sb_info *sbi = EXT4_SB(sb);
2853 struct ext4_li_request *elr;
2856 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2862 elr->lr_next_group = start;
2865 * Randomize first schedule time of the request to
2866 * spread the inode table initialization requests
2869 get_random_bytes(&rnd, sizeof(rnd));
2870 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2871 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2876 static int ext4_register_li_request(struct super_block *sb,
2877 ext4_group_t first_not_zeroed)
2879 struct ext4_sb_info *sbi = EXT4_SB(sb);
2880 struct ext4_li_request *elr;
2881 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2884 if (sbi->s_li_request != NULL)
2887 if (first_not_zeroed == ngroups ||
2888 (sb->s_flags & MS_RDONLY) ||
2889 !test_opt(sb, INIT_INODE_TABLE)) {
2890 sbi->s_li_request = NULL;
2894 if (first_not_zeroed == ngroups) {
2895 sbi->s_li_request = NULL;
2899 elr = ext4_li_request_new(sb, first_not_zeroed);
2905 mutex_lock(&ext4_li_mtx);
2907 if (NULL == ext4_li_info) {
2908 ret = ext4_li_info_new();
2913 mutex_lock(&ext4_li_info->li_list_mtx);
2914 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2915 mutex_unlock(&ext4_li_info->li_list_mtx);
2917 sbi->s_li_request = elr;
2919 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2920 ret = ext4_run_lazyinit_thread();
2925 mutex_unlock(&ext4_li_mtx);
2929 mutex_unlock(&ext4_li_mtx);
2936 * We do not need to lock anything since this is called on
2939 static void ext4_destroy_lazyinit_thread(void)
2942 * If thread exited earlier
2943 * there's nothing to be done.
2948 ext4_clear_request_list();
2950 while (ext4_li_info->li_task) {
2951 wake_up(&ext4_li_info->li_wait_daemon);
2952 wait_event(ext4_li_info->li_wait_task,
2953 ext4_li_info->li_task == NULL);
2957 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2958 __releases(kernel_lock)
2959 __acquires(kernel_lock)
2961 char *orig_data = kstrdup(data, GFP_KERNEL);
2962 struct buffer_head *bh;
2963 struct ext4_super_block *es = NULL;
2964 struct ext4_sb_info *sbi;
2966 ext4_fsblk_t sb_block = get_sb_block(&data);
2967 ext4_fsblk_t logical_sb_block;
2968 unsigned long offset = 0;
2969 unsigned long journal_devnum = 0;
2970 unsigned long def_mount_opts;
2976 unsigned int db_count;
2978 int needs_recovery, has_huge_files;
2981 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2982 ext4_group_t first_not_zeroed;
2984 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2988 sbi->s_blockgroup_lock =
2989 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2990 if (!sbi->s_blockgroup_lock) {
2994 sb->s_fs_info = sbi;
2995 sbi->s_mount_opt = 0;
2996 sbi->s_resuid = EXT4_DEF_RESUID;
2997 sbi->s_resgid = EXT4_DEF_RESGID;
2998 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2999 sbi->s_sb_block = sb_block;
3000 if (sb->s_bdev->bd_part)
3001 sbi->s_sectors_written_start =
3002 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3006 /* Cleanup superblock name */
3007 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3011 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3013 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3018 * The ext4 superblock will not be buffer aligned for other than 1kB
3019 * block sizes. We need to calculate the offset from buffer start.
3021 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3022 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3023 offset = do_div(logical_sb_block, blocksize);
3025 logical_sb_block = sb_block;
3028 if (!(bh = sb_bread(sb, logical_sb_block))) {
3029 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3033 * Note: s_es must be initialized as soon as possible because
3034 * some ext4 macro-instructions depend on its value
3036 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3038 sb->s_magic = le16_to_cpu(es->s_magic);
3039 if (sb->s_magic != EXT4_SUPER_MAGIC)
3041 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3043 /* Set defaults before we parse the mount options */
3044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3045 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
3046 if (def_mount_opts & EXT4_DEFM_DEBUG)
3047 set_opt(sbi->s_mount_opt, DEBUG);
3048 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3049 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3051 set_opt(sbi->s_mount_opt, GRPID);
3053 if (def_mount_opts & EXT4_DEFM_UID16)
3054 set_opt(sbi->s_mount_opt, NO_UID32);
3055 #ifdef CONFIG_EXT4_FS_XATTR
3056 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3057 set_opt(sbi->s_mount_opt, XATTR_USER);
3059 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3060 if (def_mount_opts & EXT4_DEFM_ACL)
3061 set_opt(sbi->s_mount_opt, POSIX_ACL);
3063 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3064 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3065 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3066 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3067 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3068 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3070 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3071 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
3072 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3073 set_opt(sbi->s_mount_opt, ERRORS_CONT);
3075 set_opt(sbi->s_mount_opt, ERRORS_RO);
3076 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3077 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3078 if (def_mount_opts & EXT4_DEFM_DISCARD)
3079 set_opt(sbi->s_mount_opt, DISCARD);
3081 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3082 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3083 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3084 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3085 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3087 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3088 set_opt(sbi->s_mount_opt, BARRIER);
3091 * enable delayed allocation by default
3092 * Use -o nodelalloc to turn it off
3094 if (!IS_EXT3_SB(sb) &&
3095 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3096 set_opt(sbi->s_mount_opt, DELALLOC);
3098 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3099 &journal_devnum, &journal_ioprio, NULL, 0)) {
3100 ext4_msg(sb, KERN_WARNING,
3101 "failed to parse options in superblock: %s",
3102 sbi->s_es->s_mount_opts);
3104 if (!parse_options((char *) data, sb, &journal_devnum,
3105 &journal_ioprio, NULL, 0))
3108 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3109 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3111 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3112 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3113 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3114 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3115 ext4_msg(sb, KERN_WARNING,
3116 "feature flags set on rev 0 fs, "
3117 "running e2fsck is recommended");
3120 * Check feature flags regardless of the revision level, since we
3121 * previously didn't change the revision level when setting the flags,
3122 * so there is a chance incompat flags are set on a rev 0 filesystem.
3124 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3127 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3129 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3130 blocksize > EXT4_MAX_BLOCK_SIZE) {
3131 ext4_msg(sb, KERN_ERR,
3132 "Unsupported filesystem blocksize %d", blocksize);
3136 if (sb->s_blocksize != blocksize) {
3137 /* Validate the filesystem blocksize */
3138 if (!sb_set_blocksize(sb, blocksize)) {
3139 ext4_msg(sb, KERN_ERR, "bad block size %d",
3145 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3146 offset = do_div(logical_sb_block, blocksize);
3147 bh = sb_bread(sb, logical_sb_block);
3149 ext4_msg(sb, KERN_ERR,
3150 "Can't read superblock on 2nd try");
3153 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3155 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3156 ext4_msg(sb, KERN_ERR,
3157 "Magic mismatch, very weird!");
3162 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3163 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3164 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3166 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3168 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3169 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3170 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3172 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3173 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3174 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3175 (!is_power_of_2(sbi->s_inode_size)) ||
3176 (sbi->s_inode_size > blocksize)) {
3177 ext4_msg(sb, KERN_ERR,
3178 "unsupported inode size: %d",
3182 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3183 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3186 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3187 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3188 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3189 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3190 !is_power_of_2(sbi->s_desc_size)) {
3191 ext4_msg(sb, KERN_ERR,
3192 "unsupported descriptor size %lu",
3197 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3199 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3200 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3201 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3204 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3205 if (sbi->s_inodes_per_block == 0)
3207 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3208 sbi->s_inodes_per_block;
3209 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3211 sbi->s_mount_state = le16_to_cpu(es->s_state);
3212 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3213 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3215 for (i = 0; i < 4; i++)
3216 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3217 sbi->s_def_hash_version = es->s_def_hash_version;
3218 i = le32_to_cpu(es->s_flags);
3219 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3220 sbi->s_hash_unsigned = 3;
3221 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3222 #ifdef __CHAR_UNSIGNED__
3223 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3224 sbi->s_hash_unsigned = 3;
3226 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3231 if (sbi->s_blocks_per_group > blocksize * 8) {
3232 ext4_msg(sb, KERN_ERR,
3233 "#blocks per group too big: %lu",
3234 sbi->s_blocks_per_group);
3237 if (sbi->s_inodes_per_group > blocksize * 8) {
3238 ext4_msg(sb, KERN_ERR,
3239 "#inodes per group too big: %lu",
3240 sbi->s_inodes_per_group);
3245 * Test whether we have more sectors than will fit in sector_t,
3246 * and whether the max offset is addressable by the page cache.
3248 if ((ext4_blocks_count(es) >
3249 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
3250 (ext4_blocks_count(es) >
3251 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
3252 ext4_msg(sb, KERN_ERR, "filesystem"
3253 " too large to mount safely on this system");
3254 if (sizeof(sector_t) < 8)
3255 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3260 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3263 /* check blocks count against device size */
3264 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3265 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3266 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3267 "exceeds size of device (%llu blocks)",
3268 ext4_blocks_count(es), blocks_count);
3273 * It makes no sense for the first data block to be beyond the end
3274 * of the filesystem.
3276 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3277 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3278 "block %u is beyond end of filesystem (%llu)",
3279 le32_to_cpu(es->s_first_data_block),
3280 ext4_blocks_count(es));
3283 blocks_count = (ext4_blocks_count(es) -
3284 le32_to_cpu(es->s_first_data_block) +
3285 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3286 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3287 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3288 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3289 "(block count %llu, first data block %u, "
3290 "blocks per group %lu)", sbi->s_groups_count,
3291 ext4_blocks_count(es),
3292 le32_to_cpu(es->s_first_data_block),
3293 EXT4_BLOCKS_PER_GROUP(sb));
3296 sbi->s_groups_count = blocks_count;
3297 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3298 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3299 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3300 EXT4_DESC_PER_BLOCK(sb);
3301 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3303 if (sbi->s_group_desc == NULL) {
3304 ext4_msg(sb, KERN_ERR, "not enough memory");
3308 #ifdef CONFIG_PROC_FS
3310 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3313 bgl_lock_init(sbi->s_blockgroup_lock);
3315 for (i = 0; i < db_count; i++) {
3316 block = descriptor_loc(sb, logical_sb_block, i);
3317 sbi->s_group_desc[i] = sb_bread(sb, block);
3318 if (!sbi->s_group_desc[i]) {
3319 ext4_msg(sb, KERN_ERR,
3320 "can't read group descriptor %d", i);
3325 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3326 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3329 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3330 if (!ext4_fill_flex_info(sb)) {
3331 ext4_msg(sb, KERN_ERR,
3332 "unable to initialize "
3333 "flex_bg meta info!");
3337 sbi->s_gdb_count = db_count;
3338 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3339 spin_lock_init(&sbi->s_next_gen_lock);
3341 sbi->s_stripe = ext4_get_stripe_size(sbi);
3342 sbi->s_max_writeback_mb_bump = 128;
3345 * set up enough so that it can read an inode
3347 if (!test_opt(sb, NOLOAD) &&
3348 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3349 sb->s_op = &ext4_sops;
3351 sb->s_op = &ext4_nojournal_sops;
3352 sb->s_export_op = &ext4_export_ops;
3353 sb->s_xattr = ext4_xattr_handlers;
3355 sb->s_qcop = &ext4_qctl_operations;
3356 sb->dq_op = &ext4_quota_operations;
3358 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3359 mutex_init(&sbi->s_orphan_lock);
3360 mutex_init(&sbi->s_resize_lock);
3364 needs_recovery = (es->s_last_orphan != 0 ||
3365 EXT4_HAS_INCOMPAT_FEATURE(sb,
3366 EXT4_FEATURE_INCOMPAT_RECOVER));
3369 * The first inode we look at is the journal inode. Don't try
3370 * root first: it may be modified in the journal!
3372 if (!test_opt(sb, NOLOAD) &&
3373 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3374 if (ext4_load_journal(sb, es, journal_devnum))
3376 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3377 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3378 ext4_msg(sb, KERN_ERR, "required journal recovery "
3379 "suppressed and not mounted read-only");
3380 goto failed_mount_wq;
3382 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
3383 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3384 sbi->s_journal = NULL;
3389 if (ext4_blocks_count(es) > 0xffffffffULL &&
3390 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3391 JBD2_FEATURE_INCOMPAT_64BIT)) {
3392 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3393 goto failed_mount_wq;
3396 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3397 jbd2_journal_set_features(sbi->s_journal,
3398 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3399 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3400 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3401 jbd2_journal_set_features(sbi->s_journal,
3402 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3403 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3404 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3406 jbd2_journal_clear_features(sbi->s_journal,
3407 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3408 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3411 /* We have now updated the journal if required, so we can
3412 * validate the data journaling mode. */
3413 switch (test_opt(sb, DATA_FLAGS)) {
3415 /* No mode set, assume a default based on the journal
3416 * capabilities: ORDERED_DATA if the journal can
3417 * cope, else JOURNAL_DATA
3419 if (jbd2_journal_check_available_features
3420 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3421 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3423 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3426 case EXT4_MOUNT_ORDERED_DATA:
3427 case EXT4_MOUNT_WRITEBACK_DATA:
3428 if (!jbd2_journal_check_available_features
3429 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3430 ext4_msg(sb, KERN_ERR, "Journal does not support "
3431 "requested data journaling mode");
3432 goto failed_mount_wq;
3437 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3440 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3441 ext4_count_free_blocks(sb));
3443 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3444 ext4_count_free_inodes(sb));
3446 err = percpu_counter_init(&sbi->s_dirs_counter,
3447 ext4_count_dirs(sb));
3449 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3451 ext4_msg(sb, KERN_ERR, "insufficient memory");
3452 goto failed_mount_wq;
3455 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3456 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3457 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3458 goto failed_mount_wq;
3462 * The jbd2_journal_load will have done any necessary log recovery,
3463 * so we can safely mount the rest of the filesystem now.
3466 root = ext4_iget(sb, EXT4_ROOT_INO);
3468 ext4_msg(sb, KERN_ERR, "get root inode failed");
3469 ret = PTR_ERR(root);
3472 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3474 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3477 sb->s_root = d_alloc_root(root);
3479 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3485 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3487 /* determine the minimum size of new large inodes, if present */
3488 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3489 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3490 EXT4_GOOD_OLD_INODE_SIZE;
3491 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3492 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3493 if (sbi->s_want_extra_isize <
3494 le16_to_cpu(es->s_want_extra_isize))
3495 sbi->s_want_extra_isize =
3496 le16_to_cpu(es->s_want_extra_isize);
3497 if (sbi->s_want_extra_isize <
3498 le16_to_cpu(es->s_min_extra_isize))
3499 sbi->s_want_extra_isize =
3500 le16_to_cpu(es->s_min_extra_isize);
3503 /* Check if enough inode space is available */
3504 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3505 sbi->s_inode_size) {
3506 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3507 EXT4_GOOD_OLD_INODE_SIZE;
3508 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3512 if (test_opt(sb, DELALLOC) &&
3513 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3514 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3515 "requested data journaling mode");
3516 clear_opt(sbi->s_mount_opt, DELALLOC);
3518 if (test_opt(sb, DIOREAD_NOLOCK)) {
3519 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3520 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3521 "option - requested data journaling mode");
3522 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3524 if (sb->s_blocksize < PAGE_SIZE) {
3525 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3526 "option - block size is too small");
3527 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3531 err = ext4_setup_system_zone(sb);
3533 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3539 err = ext4_mb_init(sb, needs_recovery);
3541 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3546 err = ext4_register_li_request(sb, first_not_zeroed);
3550 sbi->s_kobj.kset = ext4_kset;
3551 init_completion(&sbi->s_kobj_unregister);
3552 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3555 ext4_mb_release(sb);
3556 ext4_ext_release(sb);
3560 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3561 ext4_orphan_cleanup(sb, es);
3562 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3563 if (needs_recovery) {
3564 ext4_msg(sb, KERN_INFO, "recovery complete");
3565 ext4_mark_recovery_complete(sb, es);
3567 if (EXT4_SB(sb)->s_journal) {
3568 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3569 descr = " journalled data mode";
3570 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3571 descr = " ordered data mode";
3573 descr = " writeback data mode";
3575 descr = "out journal";
3577 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3578 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3579 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3581 init_timer(&sbi->s_err_report);
3582 sbi->s_err_report.function = print_daily_error_info;
3583 sbi->s_err_report.data = (unsigned long) sb;
3584 if (es->s_error_count)
3585 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3593 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3597 ext4_msg(sb, KERN_ERR, "mount failed");
3598 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3600 ext4_release_system_zone(sb);
3601 if (sbi->s_journal) {
3602 jbd2_journal_destroy(sbi->s_journal);
3603 sbi->s_journal = NULL;
3605 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3606 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3607 percpu_counter_destroy(&sbi->s_dirs_counter);
3608 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3610 if (sbi->s_flex_groups) {
3611 if (is_vmalloc_addr(sbi->s_flex_groups))
3612 vfree(sbi->s_flex_groups);
3614 kfree(sbi->s_flex_groups);
3617 for (i = 0; i < db_count; i++)
3618 brelse(sbi->s_group_desc[i]);
3619 kfree(sbi->s_group_desc);
3622 remove_proc_entry(sb->s_id, ext4_proc_root);
3625 for (i = 0; i < MAXQUOTAS; i++)
3626 kfree(sbi->s_qf_names[i]);
3628 ext4_blkdev_remove(sbi);
3631 sb->s_fs_info = NULL;
3632 kfree(sbi->s_blockgroup_lock);
3641 * Setup any per-fs journal parameters now. We'll do this both on
3642 * initial mount, once the journal has been initialised but before we've
3643 * done any recovery; and again on any subsequent remount.
3645 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3647 struct ext4_sb_info *sbi = EXT4_SB(sb);
3649 journal->j_commit_interval = sbi->s_commit_interval;
3650 journal->j_min_batch_time = sbi->s_min_batch_time;
3651 journal->j_max_batch_time = sbi->s_max_batch_time;
3653 write_lock(&journal->j_state_lock);
3654 if (test_opt(sb, BARRIER))
3655 journal->j_flags |= JBD2_BARRIER;
3657 journal->j_flags &= ~JBD2_BARRIER;
3658 if (test_opt(sb, DATA_ERR_ABORT))
3659 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3661 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3662 write_unlock(&journal->j_state_lock);
3665 static journal_t *ext4_get_journal(struct super_block *sb,
3666 unsigned int journal_inum)
3668 struct inode *journal_inode;
3671 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3673 /* First, test for the existence of a valid inode on disk. Bad
3674 * things happen if we iget() an unused inode, as the subsequent
3675 * iput() will try to delete it. */
3677 journal_inode = ext4_iget(sb, journal_inum);
3678 if (IS_ERR(journal_inode)) {
3679 ext4_msg(sb, KERN_ERR, "no journal found");
3682 if (!journal_inode->i_nlink) {
3683 make_bad_inode(journal_inode);
3684 iput(journal_inode);
3685 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3689 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3690 journal_inode, journal_inode->i_size);
3691 if (!S_ISREG(journal_inode->i_mode)) {
3692 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3693 iput(journal_inode);
3697 journal = jbd2_journal_init_inode(journal_inode);
3699 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3700 iput(journal_inode);
3703 journal->j_private = sb;
3704 ext4_init_journal_params(sb, journal);
3708 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3711 struct buffer_head *bh;
3715 int hblock, blocksize;
3716 ext4_fsblk_t sb_block;
3717 unsigned long offset;
3718 struct ext4_super_block *es;
3719 struct block_device *bdev;
3721 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3723 bdev = ext4_blkdev_get(j_dev, sb);
3727 if (bd_claim(bdev, sb)) {
3728 ext4_msg(sb, KERN_ERR,
3729 "failed to claim external journal device");
3730 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3734 blocksize = sb->s_blocksize;
3735 hblock = bdev_logical_block_size(bdev);
3736 if (blocksize < hblock) {
3737 ext4_msg(sb, KERN_ERR,
3738 "blocksize too small for journal device");
3742 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3743 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3744 set_blocksize(bdev, blocksize);
3745 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3746 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3747 "external journal");
3751 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3752 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3753 !(le32_to_cpu(es->s_feature_incompat) &
3754 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3755 ext4_msg(sb, KERN_ERR, "external journal has "
3761 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3762 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3767 len = ext4_blocks_count(es);
3768 start = sb_block + 1;
3769 brelse(bh); /* we're done with the superblock */
3771 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3772 start, len, blocksize);
3774 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3777 journal->j_private = sb;
3778 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3779 wait_on_buffer(journal->j_sb_buffer);
3780 if (!buffer_uptodate(journal->j_sb_buffer)) {
3781 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3784 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3785 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3786 "user (unsupported) - %d",
3787 be32_to_cpu(journal->j_superblock->s_nr_users));
3790 EXT4_SB(sb)->journal_bdev = bdev;
3791 ext4_init_journal_params(sb, journal);
3795 jbd2_journal_destroy(journal);
3797 ext4_blkdev_put(bdev);
3801 static int ext4_load_journal(struct super_block *sb,
3802 struct ext4_super_block *es,
3803 unsigned long journal_devnum)
3806 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3809 int really_read_only;
3811 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3813 if (journal_devnum &&
3814 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3815 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3816 "numbers have changed");
3817 journal_dev = new_decode_dev(journal_devnum);
3819 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3821 really_read_only = bdev_read_only(sb->s_bdev);
3824 * Are we loading a blank journal or performing recovery after a
3825 * crash? For recovery, we need to check in advance whether we
3826 * can get read-write access to the device.
3828 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3829 if (sb->s_flags & MS_RDONLY) {
3830 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3831 "required on readonly filesystem");
3832 if (really_read_only) {
3833 ext4_msg(sb, KERN_ERR, "write access "
3834 "unavailable, cannot proceed");
3837 ext4_msg(sb, KERN_INFO, "write access will "
3838 "be enabled during recovery");
3842 if (journal_inum && journal_dev) {
3843 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3844 "and inode journals!");
3849 if (!(journal = ext4_get_journal(sb, journal_inum)))
3852 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3856 if (!(journal->j_flags & JBD2_BARRIER))
3857 ext4_msg(sb, KERN_INFO, "barriers disabled");
3859 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3860 err = jbd2_journal_update_format(journal);
3862 ext4_msg(sb, KERN_ERR, "error updating journal");
3863 jbd2_journal_destroy(journal);
3868 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3869 err = jbd2_journal_wipe(journal, !really_read_only);
3871 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3873 memcpy(save, ((char *) es) +
3874 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3875 err = jbd2_journal_load(journal);
3877 memcpy(((char *) es) + EXT4_S_ERR_START,
3878 save, EXT4_S_ERR_LEN);
3883 ext4_msg(sb, KERN_ERR, "error loading journal");
3884 jbd2_journal_destroy(journal);
3888 EXT4_SB(sb)->s_journal = journal;
3889 ext4_clear_journal_err(sb, es);
3891 if (journal_devnum &&
3892 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3893 es->s_journal_dev = cpu_to_le32(journal_devnum);
3895 /* Make sure we flush the recovery flag to disk. */
3896 ext4_commit_super(sb, 1);
3902 static int ext4_commit_super(struct super_block *sb, int sync)
3904 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3905 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3910 if (buffer_write_io_error(sbh)) {
3912 * Oh, dear. A previous attempt to write the
3913 * superblock failed. This could happen because the
3914 * USB device was yanked out. Or it could happen to
3915 * be a transient write error and maybe the block will
3916 * be remapped. Nothing we can do but to retry the
3917 * write and hope for the best.
3919 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3920 "superblock detected");
3921 clear_buffer_write_io_error(sbh);
3922 set_buffer_uptodate(sbh);
3925 * If the file system is mounted read-only, don't update the
3926 * superblock write time. This avoids updating the superblock
3927 * write time when we are mounting the root file system
3928 * read/only but we need to replay the journal; at that point,
3929 * for people who are east of GMT and who make their clock
3930 * tick in localtime for Windows bug-for-bug compatibility,
3931 * the clock is set in the future, and this will cause e2fsck
3932 * to complain and force a full file system check.
3934 if (!(sb->s_flags & MS_RDONLY))
3935 es->s_wtime = cpu_to_le32(get_seconds());
3936 if (sb->s_bdev->bd_part)
3937 es->s_kbytes_written =
3938 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3939 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3940 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3942 es->s_kbytes_written =
3943 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3944 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3945 &EXT4_SB(sb)->s_freeblocks_counter));
3946 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3947 &EXT4_SB(sb)->s_freeinodes_counter));
3949 BUFFER_TRACE(sbh, "marking dirty");
3950 mark_buffer_dirty(sbh);
3952 error = sync_dirty_buffer(sbh);
3956 error = buffer_write_io_error(sbh);
3958 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3960 clear_buffer_write_io_error(sbh);
3961 set_buffer_uptodate(sbh);
3968 * Have we just finished recovery? If so, and if we are mounting (or
3969 * remounting) the filesystem readonly, then we will end up with a
3970 * consistent fs on disk. Record that fact.
3972 static void ext4_mark_recovery_complete(struct super_block *sb,
3973 struct ext4_super_block *es)
3975 journal_t *journal = EXT4_SB(sb)->s_journal;
3977 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3978 BUG_ON(journal != NULL);
3981 jbd2_journal_lock_updates(journal);
3982 if (jbd2_journal_flush(journal) < 0)
3985 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3986 sb->s_flags & MS_RDONLY) {
3987 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3988 ext4_commit_super(sb, 1);
3992 jbd2_journal_unlock_updates(journal);
3996 * If we are mounting (or read-write remounting) a filesystem whose journal
3997 * has recorded an error from a previous lifetime, move that error to the
3998 * main filesystem now.
4000 static void ext4_clear_journal_err(struct super_block *sb,
4001 struct ext4_super_block *es)
4007 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4009 journal = EXT4_SB(sb)->s_journal;
4012 * Now check for any error status which may have been recorded in the
4013 * journal by a prior ext4_error() or ext4_abort()
4016 j_errno = jbd2_journal_errno(journal);
4020 errstr = ext4_decode_error(sb, j_errno, nbuf);
4021 ext4_warning(sb, "Filesystem error recorded "
4022 "from previous mount: %s", errstr);
4023 ext4_warning(sb, "Marking fs in need of filesystem check.");
4025 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4026 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4027 ext4_commit_super(sb, 1);
4029 jbd2_journal_clear_err(journal);
4034 * Force the running and committing transactions to commit,
4035 * and wait on the commit.
4037 int ext4_force_commit(struct super_block *sb)
4042 if (sb->s_flags & MS_RDONLY)
4045 journal = EXT4_SB(sb)->s_journal;
4047 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4048 ret = ext4_journal_force_commit(journal);
4054 static void ext4_write_super(struct super_block *sb)
4057 ext4_commit_super(sb, 1);
4061 static int ext4_sync_fs(struct super_block *sb, int wait)
4065 struct ext4_sb_info *sbi = EXT4_SB(sb);
4067 trace_ext4_sync_fs(sb, wait);
4068 flush_workqueue(sbi->dio_unwritten_wq);
4069 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4071 jbd2_log_wait_commit(sbi->s_journal, target);
4077 * LVM calls this function before a (read-only) snapshot is created. This
4078 * gives us a chance to flush the journal completely and mark the fs clean.
4080 static int ext4_freeze(struct super_block *sb)
4085 if (sb->s_flags & MS_RDONLY)
4088 journal = EXT4_SB(sb)->s_journal;
4090 /* Now we set up the journal barrier. */
4091 jbd2_journal_lock_updates(journal);
4094 * Don't clear the needs_recovery flag if we failed to flush
4097 error = jbd2_journal_flush(journal);
4101 /* Journal blocked and flushed, clear needs_recovery flag. */
4102 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4103 error = ext4_commit_super(sb, 1);
4105 /* we rely on s_frozen to stop further updates */
4106 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4111 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4112 * flag here, even though the filesystem is not technically dirty yet.
4114 static int ext4_unfreeze(struct super_block *sb)
4116 if (sb->s_flags & MS_RDONLY)
4120 /* Reset the needs_recovery flag before the fs is unlocked. */
4121 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4122 ext4_commit_super(sb, 1);
4127 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4129 struct ext4_super_block *es;
4130 struct ext4_sb_info *sbi = EXT4_SB(sb);
4131 ext4_fsblk_t n_blocks_count = 0;
4132 unsigned long old_sb_flags;
4133 struct ext4_mount_options old_opts;
4134 int enable_quota = 0;
4136 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4141 char *orig_data = kstrdup(data, GFP_KERNEL);
4145 /* Store the original options */
4147 old_sb_flags = sb->s_flags;
4148 old_opts.s_mount_opt = sbi->s_mount_opt;
4149 old_opts.s_resuid = sbi->s_resuid;
4150 old_opts.s_resgid = sbi->s_resgid;
4151 old_opts.s_commit_interval = sbi->s_commit_interval;
4152 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4153 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4155 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4156 for (i = 0; i < MAXQUOTAS; i++)
4157 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4159 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4160 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4163 * Allow the "check" option to be passed as a remount option.
4165 if (!parse_options(data, sb, NULL, &journal_ioprio,
4166 &n_blocks_count, 1)) {
4171 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4172 ext4_abort(sb, "Abort forced by user");
4174 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4175 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4179 if (sbi->s_journal) {
4180 ext4_init_journal_params(sb, sbi->s_journal);
4181 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4184 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4185 n_blocks_count > ext4_blocks_count(es)) {
4186 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4191 if (*flags & MS_RDONLY) {
4192 err = dquot_suspend(sb, -1);
4197 * First of all, the unconditional stuff we have to do
4198 * to disable replay of the journal when we next remount
4200 sb->s_flags |= MS_RDONLY;
4203 * OK, test if we are remounting a valid rw partition
4204 * readonly, and if so set the rdonly flag and then
4205 * mark the partition as valid again.
4207 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4208 (sbi->s_mount_state & EXT4_VALID_FS))
4209 es->s_state = cpu_to_le16(sbi->s_mount_state);
4212 ext4_mark_recovery_complete(sb, es);
4214 /* Make sure we can mount this feature set readwrite */
4215 if (!ext4_feature_set_ok(sb, 0)) {
4220 * Make sure the group descriptor checksums
4221 * are sane. If they aren't, refuse to remount r/w.
4223 for (g = 0; g < sbi->s_groups_count; g++) {
4224 struct ext4_group_desc *gdp =
4225 ext4_get_group_desc(sb, g, NULL);
4227 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4228 ext4_msg(sb, KERN_ERR,
4229 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4230 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4231 le16_to_cpu(gdp->bg_checksum));
4238 * If we have an unprocessed orphan list hanging
4239 * around from a previously readonly bdev mount,
4240 * require a full umount/remount for now.
4242 if (es->s_last_orphan) {
4243 ext4_msg(sb, KERN_WARNING, "Couldn't "
4244 "remount RDWR because of unprocessed "
4245 "orphan inode list. Please "
4246 "umount/remount instead");
4252 * Mounting a RDONLY partition read-write, so reread
4253 * and store the current valid flag. (It may have
4254 * been changed by e2fsck since we originally mounted
4258 ext4_clear_journal_err(sb, es);
4259 sbi->s_mount_state = le16_to_cpu(es->s_state);
4260 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4262 if (!ext4_setup_super(sb, es, 0))
4263 sb->s_flags &= ~MS_RDONLY;
4269 * Reinitialize lazy itable initialization thread based on
4272 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4273 ext4_unregister_li_request(sb);
4275 ext4_group_t first_not_zeroed;
4276 first_not_zeroed = ext4_has_uninit_itable(sb);
4277 ext4_register_li_request(sb, first_not_zeroed);
4280 ext4_setup_system_zone(sb);
4281 if (sbi->s_journal == NULL)
4282 ext4_commit_super(sb, 1);
4285 /* Release old quota file names */
4286 for (i = 0; i < MAXQUOTAS; i++)
4287 if (old_opts.s_qf_names[i] &&
4288 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4289 kfree(old_opts.s_qf_names[i]);
4294 dquot_resume(sb, -1);
4296 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4301 sb->s_flags = old_sb_flags;
4302 sbi->s_mount_opt = old_opts.s_mount_opt;
4303 sbi->s_resuid = old_opts.s_resuid;
4304 sbi->s_resgid = old_opts.s_resgid;
4305 sbi->s_commit_interval = old_opts.s_commit_interval;
4306 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4307 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4309 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4310 for (i = 0; i < MAXQUOTAS; i++) {
4311 if (sbi->s_qf_names[i] &&
4312 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4313 kfree(sbi->s_qf_names[i]);
4314 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4323 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4325 struct super_block *sb = dentry->d_sb;
4326 struct ext4_sb_info *sbi = EXT4_SB(sb);
4327 struct ext4_super_block *es = sbi->s_es;
4330 if (test_opt(sb, MINIX_DF)) {
4331 sbi->s_overhead_last = 0;
4332 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4333 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4334 ext4_fsblk_t overhead = 0;
4337 * Compute the overhead (FS structures). This is constant
4338 * for a given filesystem unless the number of block groups
4339 * changes so we cache the previous value until it does.
4343 * All of the blocks before first_data_block are
4346 overhead = le32_to_cpu(es->s_first_data_block);
4349 * Add the overhead attributed to the superblock and
4350 * block group descriptors. If the sparse superblocks
4351 * feature is turned on, then not all groups have this.
4353 for (i = 0; i < ngroups; i++) {
4354 overhead += ext4_bg_has_super(sb, i) +
4355 ext4_bg_num_gdb(sb, i);
4360 * Every block group has an inode bitmap, a block
4361 * bitmap, and an inode table.
4363 overhead += ngroups * (2 + sbi->s_itb_per_group);
4364 sbi->s_overhead_last = overhead;
4366 sbi->s_blocks_last = ext4_blocks_count(es);
4369 buf->f_type = EXT4_SUPER_MAGIC;
4370 buf->f_bsize = sb->s_blocksize;
4371 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4372 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4373 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4374 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4375 if (buf->f_bfree < ext4_r_blocks_count(es))
4377 buf->f_files = le32_to_cpu(es->s_inodes_count);
4378 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4379 buf->f_namelen = EXT4_NAME_LEN;
4380 fsid = le64_to_cpup((void *)es->s_uuid) ^
4381 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4382 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4383 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4388 /* Helper function for writing quotas on sync - we need to start transaction
4389 * before quota file is locked for write. Otherwise the are possible deadlocks:
4390 * Process 1 Process 2
4391 * ext4_create() quota_sync()
4392 * jbd2_journal_start() write_dquot()
4393 * dquot_initialize() down(dqio_mutex)
4394 * down(dqio_mutex) jbd2_journal_start()
4400 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4402 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4405 static int ext4_write_dquot(struct dquot *dquot)
4409 struct inode *inode;
4411 inode = dquot_to_inode(dquot);
4412 handle = ext4_journal_start(inode,
4413 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4415 return PTR_ERR(handle);
4416 ret = dquot_commit(dquot);
4417 err = ext4_journal_stop(handle);
4423 static int ext4_acquire_dquot(struct dquot *dquot)
4428 handle = ext4_journal_start(dquot_to_inode(dquot),
4429 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4431 return PTR_ERR(handle);
4432 ret = dquot_acquire(dquot);
4433 err = ext4_journal_stop(handle);
4439 static int ext4_release_dquot(struct dquot *dquot)
4444 handle = ext4_journal_start(dquot_to_inode(dquot),
4445 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4446 if (IS_ERR(handle)) {
4447 /* Release dquot anyway to avoid endless cycle in dqput() */
4448 dquot_release(dquot);
4449 return PTR_ERR(handle);
4451 ret = dquot_release(dquot);
4452 err = ext4_journal_stop(handle);
4458 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4460 /* Are we journaling quotas? */
4461 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4462 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4463 dquot_mark_dquot_dirty(dquot);
4464 return ext4_write_dquot(dquot);
4466 return dquot_mark_dquot_dirty(dquot);
4470 static int ext4_write_info(struct super_block *sb, int type)
4475 /* Data block + inode block */
4476 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4478 return PTR_ERR(handle);
4479 ret = dquot_commit_info(sb, type);
4480 err = ext4_journal_stop(handle);
4487 * Turn on quotas during mount time - we need to find
4488 * the quota file and such...
4490 static int ext4_quota_on_mount(struct super_block *sb, int type)
4492 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4493 EXT4_SB(sb)->s_jquota_fmt, type);
4497 * Standard function to be called on quota_on
4499 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4505 if (!test_opt(sb, QUOTA))
4508 err = kern_path(name, LOOKUP_FOLLOW, &path);
4512 /* Quotafile not on the same filesystem? */
4513 if (path.mnt->mnt_sb != sb) {
4517 /* Journaling quota? */
4518 if (EXT4_SB(sb)->s_qf_names[type]) {
4519 /* Quotafile not in fs root? */
4520 if (path.dentry->d_parent != sb->s_root)
4521 ext4_msg(sb, KERN_WARNING,
4522 "Quota file not on filesystem root. "
4523 "Journaled quota will not work");
4527 * When we journal data on quota file, we have to flush journal to see
4528 * all updates to the file when we bypass pagecache...
4530 if (EXT4_SB(sb)->s_journal &&
4531 ext4_should_journal_data(path.dentry->d_inode)) {
4533 * We don't need to lock updates but journal_flush() could
4534 * otherwise be livelocked...
4536 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4537 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4538 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4545 err = dquot_quota_on_path(sb, type, format_id, &path);
4550 static int ext4_quota_off(struct super_block *sb, int type)
4552 /* Force all delayed allocation blocks to be allocated */
4553 if (test_opt(sb, DELALLOC)) {
4554 down_read(&sb->s_umount);
4555 sync_filesystem(sb);
4556 up_read(&sb->s_umount);
4559 return dquot_quota_off(sb, type);
4562 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4563 * acquiring the locks... As quota files are never truncated and quota code
4564 * itself serializes the operations (and noone else should touch the files)
4565 * we don't have to be afraid of races */
4566 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4567 size_t len, loff_t off)
4569 struct inode *inode = sb_dqopt(sb)->files[type];
4570 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4572 int offset = off & (sb->s_blocksize - 1);
4575 struct buffer_head *bh;
4576 loff_t i_size = i_size_read(inode);
4580 if (off+len > i_size)
4583 while (toread > 0) {
4584 tocopy = sb->s_blocksize - offset < toread ?
4585 sb->s_blocksize - offset : toread;
4586 bh = ext4_bread(NULL, inode, blk, 0, &err);
4589 if (!bh) /* A hole? */
4590 memset(data, 0, tocopy);
4592 memcpy(data, bh->b_data+offset, tocopy);
4602 /* Write to quotafile (we know the transaction is already started and has
4603 * enough credits) */
4604 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4605 const char *data, size_t len, loff_t off)
4607 struct inode *inode = sb_dqopt(sb)->files[type];
4608 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4610 int offset = off & (sb->s_blocksize - 1);
4611 struct buffer_head *bh;
4612 handle_t *handle = journal_current_handle();
4614 if (EXT4_SB(sb)->s_journal && !handle) {
4615 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4616 " cancelled because transaction is not started",
4617 (unsigned long long)off, (unsigned long long)len);
4621 * Since we account only one data block in transaction credits,
4622 * then it is impossible to cross a block boundary.
4624 if (sb->s_blocksize - offset < len) {
4625 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4626 " cancelled because not block aligned",
4627 (unsigned long long)off, (unsigned long long)len);
4631 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4632 bh = ext4_bread(handle, inode, blk, 1, &err);
4635 err = ext4_journal_get_write_access(handle, bh);
4641 memcpy(bh->b_data+offset, data, len);
4642 flush_dcache_page(bh->b_page);
4644 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4648 mutex_unlock(&inode->i_mutex);
4651 if (inode->i_size < off + len) {
4652 i_size_write(inode, off + len);
4653 EXT4_I(inode)->i_disksize = inode->i_size;
4655 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4656 ext4_mark_inode_dirty(handle, inode);
4657 mutex_unlock(&inode->i_mutex);
4663 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4664 const char *dev_name, void *data, struct vfsmount *mnt)
4666 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4669 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4670 static struct file_system_type ext2_fs_type = {
4671 .owner = THIS_MODULE,
4673 .get_sb = ext4_get_sb,
4674 .kill_sb = kill_block_super,
4675 .fs_flags = FS_REQUIRES_DEV,
4678 static inline void register_as_ext2(void)
4680 int err = register_filesystem(&ext2_fs_type);
4683 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4686 static inline void unregister_as_ext2(void)
4688 unregister_filesystem(&ext2_fs_type);
4690 MODULE_ALIAS("ext2");
4692 static inline void register_as_ext2(void) { }
4693 static inline void unregister_as_ext2(void) { }
4696 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4697 static inline void register_as_ext3(void)
4699 int err = register_filesystem(&ext3_fs_type);
4702 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4705 static inline void unregister_as_ext3(void)
4707 unregister_filesystem(&ext3_fs_type);
4709 MODULE_ALIAS("ext3");
4711 static inline void register_as_ext3(void) { }
4712 static inline void unregister_as_ext3(void) { }
4715 static struct file_system_type ext4_fs_type = {
4716 .owner = THIS_MODULE,
4718 .get_sb = ext4_get_sb,
4719 .kill_sb = kill_block_super,
4720 .fs_flags = FS_REQUIRES_DEV,
4723 static int __init init_ext4_fs(void)
4727 ext4_check_flag_values();
4728 err = init_ext4_system_zone();
4731 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4734 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4735 err = init_ext4_mballoc();
4739 err = init_ext4_xattr();
4742 err = init_inodecache();
4747 err = register_filesystem(&ext4_fs_type);
4751 ext4_li_info = NULL;
4752 mutex_init(&ext4_li_mtx);
4755 unregister_as_ext2();
4756 unregister_as_ext3();
4757 destroy_inodecache();
4761 exit_ext4_mballoc();
4763 remove_proc_entry("fs/ext4", NULL);
4764 kset_unregister(ext4_kset);
4766 exit_ext4_system_zone();
4770 static void __exit exit_ext4_fs(void)
4772 ext4_destroy_lazyinit_thread();
4773 unregister_as_ext2();
4774 unregister_as_ext3();
4775 unregister_filesystem(&ext4_fs_type);
4776 destroy_inodecache();
4778 exit_ext4_mballoc();
4779 remove_proc_entry("fs/ext4", NULL);
4780 kset_unregister(ext4_kset);
4781 exit_ext4_system_zone();
4784 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4785 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4786 MODULE_LICENSE("GPL");
4787 module_init(init_ext4_fs)
4788 module_exit(exit_ext4_fs)