2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_jbd2.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/ext4.h>
55 static struct proc_dir_entry *ext4_proc_root;
56 static struct kset *ext4_kset;
57 struct ext4_lazy_init *ext4_li_info;
58 struct mutex ext4_li_mtx;
59 struct ext4_features *ext4_feat;
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 struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
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 = {
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 = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
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)
666 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
669 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
671 struct block_device *bdev;
674 bdev = sbi->journal_bdev;
676 ret = ext4_blkdev_put(bdev);
677 sbi->journal_bdev = NULL;
682 static inline struct inode *orphan_list_entry(struct list_head *l)
684 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
687 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
691 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
692 le32_to_cpu(sbi->s_es->s_last_orphan));
694 printk(KERN_ERR "sb_info orphan list:\n");
695 list_for_each(l, &sbi->s_orphan) {
696 struct inode *inode = orphan_list_entry(l);
698 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
699 inode->i_sb->s_id, inode->i_ino, inode,
700 inode->i_mode, inode->i_nlink,
705 static void ext4_put_super(struct super_block *sb)
707 struct ext4_sb_info *sbi = EXT4_SB(sb);
708 struct ext4_super_block *es = sbi->s_es;
711 ext4_unregister_li_request(sb);
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);
719 ext4_commit_super(sb, 1);
721 if (sbi->s_journal) {
722 err = jbd2_journal_destroy(sbi->s_journal);
723 sbi->s_journal = NULL;
725 ext4_abort(sb, "Couldn't clean up the journal");
728 del_timer(&sbi->s_err_report);
729 ext4_release_system_zone(sb);
731 ext4_ext_release(sb);
732 ext4_xattr_put_super(sb);
734 if (!(sb->s_flags & MS_RDONLY)) {
735 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
736 es->s_state = cpu_to_le16(sbi->s_mount_state);
737 ext4_commit_super(sb, 1);
740 remove_proc_entry(sb->s_id, ext4_proc_root);
742 kobject_del(&sbi->s_kobj);
744 for (i = 0; i < sbi->s_gdb_count; i++)
745 brelse(sbi->s_group_desc[i]);
746 kfree(sbi->s_group_desc);
747 if (is_vmalloc_addr(sbi->s_flex_groups))
748 vfree(sbi->s_flex_groups);
750 kfree(sbi->s_flex_groups);
751 percpu_counter_destroy(&sbi->s_freeblocks_counter);
752 percpu_counter_destroy(&sbi->s_freeinodes_counter);
753 percpu_counter_destroy(&sbi->s_dirs_counter);
754 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
757 for (i = 0; i < MAXQUOTAS; i++)
758 kfree(sbi->s_qf_names[i]);
761 /* Debugging code just in case the in-memory inode orphan list
762 * isn't empty. The on-disk one can be non-empty if we've
763 * detected an error and taken the fs readonly, but the
764 * in-memory list had better be clean by this point. */
765 if (!list_empty(&sbi->s_orphan))
766 dump_orphan_list(sb, sbi);
767 J_ASSERT(list_empty(&sbi->s_orphan));
769 invalidate_bdev(sb->s_bdev);
770 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
772 * Invalidate the journal device's buffers. We don't want them
773 * floating about in memory - the physical journal device may
774 * hotswapped, and it breaks the `ro-after' testing code.
776 sync_blockdev(sbi->journal_bdev);
777 invalidate_bdev(sbi->journal_bdev);
778 ext4_blkdev_remove(sbi);
780 sb->s_fs_info = NULL;
782 * Now that we are completely done shutting down the
783 * superblock, we need to actually destroy the kobject.
786 kobject_put(&sbi->s_kobj);
787 wait_for_completion(&sbi->s_kobj_unregister);
788 kfree(sbi->s_blockgroup_lock);
792 static struct kmem_cache *ext4_inode_cachep;
795 * Called inside transaction, so use GFP_NOFS
797 static struct inode *ext4_alloc_inode(struct super_block *sb)
799 struct ext4_inode_info *ei;
801 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
805 ei->vfs_inode.i_version = 1;
806 ei->vfs_inode.i_data.writeback_index = 0;
807 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
808 INIT_LIST_HEAD(&ei->i_prealloc_list);
809 spin_lock_init(&ei->i_prealloc_lock);
811 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
812 * therefore it can be null here. Don't check it, just initialize
815 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
816 ei->i_reserved_data_blocks = 0;
817 ei->i_reserved_meta_blocks = 0;
818 ei->i_allocated_meta_blocks = 0;
819 ei->i_da_metadata_calc_len = 0;
820 ei->i_delalloc_reserved_flag = 0;
821 spin_lock_init(&(ei->i_block_reservation_lock));
823 ei->i_reserved_quota = 0;
825 INIT_LIST_HEAD(&ei->i_completed_io_list);
826 spin_lock_init(&ei->i_completed_io_lock);
827 ei->cur_aio_dio = NULL;
829 ei->i_datasync_tid = 0;
830 atomic_set(&ei->i_ioend_count, 0);
832 return &ei->vfs_inode;
835 static int ext4_drop_inode(struct inode *inode)
837 int drop = generic_drop_inode(inode);
839 trace_ext4_drop_inode(inode, drop);
843 static void ext4_destroy_inode(struct inode *inode)
845 ext4_ioend_wait(inode);
846 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
847 ext4_msg(inode->i_sb, KERN_ERR,
848 "Inode %lu (%p): orphan list check failed!",
849 inode->i_ino, EXT4_I(inode));
850 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
851 EXT4_I(inode), sizeof(struct ext4_inode_info),
855 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
858 static void init_once(void *foo)
860 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
862 INIT_LIST_HEAD(&ei->i_orphan);
863 #ifdef CONFIG_EXT4_FS_XATTR
864 init_rwsem(&ei->xattr_sem);
866 init_rwsem(&ei->i_data_sem);
867 inode_init_once(&ei->vfs_inode);
870 static int init_inodecache(void)
872 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
873 sizeof(struct ext4_inode_info),
874 0, (SLAB_RECLAIM_ACCOUNT|
877 if (ext4_inode_cachep == NULL)
882 static void destroy_inodecache(void)
884 kmem_cache_destroy(ext4_inode_cachep);
887 void ext4_clear_inode(struct inode *inode)
889 invalidate_inode_buffers(inode);
890 end_writeback(inode);
892 ext4_discard_preallocations(inode);
893 if (EXT4_JOURNAL(inode))
894 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
895 &EXT4_I(inode)->jinode);
898 static inline void ext4_show_quota_options(struct seq_file *seq,
899 struct super_block *sb)
901 #if defined(CONFIG_QUOTA)
902 struct ext4_sb_info *sbi = EXT4_SB(sb);
904 if (sbi->s_jquota_fmt) {
907 switch (sbi->s_jquota_fmt) {
918 seq_printf(seq, ",jqfmt=%s", fmtname);
921 if (sbi->s_qf_names[USRQUOTA])
922 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
924 if (sbi->s_qf_names[GRPQUOTA])
925 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
927 if (test_opt(sb, USRQUOTA))
928 seq_puts(seq, ",usrquota");
930 if (test_opt(sb, GRPQUOTA))
931 seq_puts(seq, ",grpquota");
937 * - it's set to a non-default value OR
938 * - if the per-sb default is different from the global default
940 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
943 unsigned long def_mount_opts;
944 struct super_block *sb = vfs->mnt_sb;
945 struct ext4_sb_info *sbi = EXT4_SB(sb);
946 struct ext4_super_block *es = sbi->s_es;
948 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
949 def_errors = le16_to_cpu(es->s_errors);
951 if (sbi->s_sb_block != 1)
952 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
953 if (test_opt(sb, MINIX_DF))
954 seq_puts(seq, ",minixdf");
955 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
956 seq_puts(seq, ",grpid");
957 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
958 seq_puts(seq, ",nogrpid");
959 if (sbi->s_resuid != EXT4_DEF_RESUID ||
960 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
961 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
963 if (sbi->s_resgid != EXT4_DEF_RESGID ||
964 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
965 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
967 if (test_opt(sb, ERRORS_RO)) {
968 if (def_errors == EXT4_ERRORS_PANIC ||
969 def_errors == EXT4_ERRORS_CONTINUE) {
970 seq_puts(seq, ",errors=remount-ro");
973 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
974 seq_puts(seq, ",errors=continue");
975 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
976 seq_puts(seq, ",errors=panic");
977 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
978 seq_puts(seq, ",nouid32");
979 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
980 seq_puts(seq, ",debug");
981 if (test_opt(sb, OLDALLOC))
982 seq_puts(seq, ",oldalloc");
983 #ifdef CONFIG_EXT4_FS_XATTR
984 if (test_opt(sb, XATTR_USER) &&
985 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
986 seq_puts(seq, ",user_xattr");
987 if (!test_opt(sb, XATTR_USER) &&
988 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
989 seq_puts(seq, ",nouser_xattr");
992 #ifdef CONFIG_EXT4_FS_POSIX_ACL
993 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
994 seq_puts(seq, ",acl");
995 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
996 seq_puts(seq, ",noacl");
998 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
999 seq_printf(seq, ",commit=%u",
1000 (unsigned) (sbi->s_commit_interval / HZ));
1002 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1003 seq_printf(seq, ",min_batch_time=%u",
1004 (unsigned) sbi->s_min_batch_time);
1006 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1007 seq_printf(seq, ",max_batch_time=%u",
1008 (unsigned) sbi->s_min_batch_time);
1012 * We're changing the default of barrier mount option, so
1013 * let's always display its mount state so it's clear what its
1016 seq_puts(seq, ",barrier=");
1017 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1018 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1019 seq_puts(seq, ",journal_async_commit");
1020 else if (test_opt(sb, JOURNAL_CHECKSUM))
1021 seq_puts(seq, ",journal_checksum");
1022 if (test_opt(sb, I_VERSION))
1023 seq_puts(seq, ",i_version");
1024 if (!test_opt(sb, DELALLOC) &&
1025 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1026 seq_puts(seq, ",nodelalloc");
1028 if (test_opt(sb, MBLK_IO_SUBMIT))
1029 seq_puts(seq, ",mblk_io_submit");
1031 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1033 * journal mode get enabled in different ways
1034 * So just print the value even if we didn't specify it
1036 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1037 seq_puts(seq, ",data=journal");
1038 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1039 seq_puts(seq, ",data=ordered");
1040 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1041 seq_puts(seq, ",data=writeback");
1043 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1044 seq_printf(seq, ",inode_readahead_blks=%u",
1045 sbi->s_inode_readahead_blks);
1047 if (test_opt(sb, DATA_ERR_ABORT))
1048 seq_puts(seq, ",data_err=abort");
1050 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1051 seq_puts(seq, ",noauto_da_alloc");
1053 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1054 seq_puts(seq, ",discard");
1056 if (test_opt(sb, NOLOAD))
1057 seq_puts(seq, ",norecovery");
1059 if (test_opt(sb, DIOREAD_NOLOCK))
1060 seq_puts(seq, ",dioread_nolock");
1062 if (test_opt(sb, BLOCK_VALIDITY) &&
1063 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1064 seq_puts(seq, ",block_validity");
1066 if (!test_opt(sb, INIT_INODE_TABLE))
1067 seq_puts(seq, ",noinit_inode_table");
1068 else if (sbi->s_li_wait_mult)
1069 seq_printf(seq, ",init_inode_table=%u",
1070 (unsigned) sbi->s_li_wait_mult);
1072 ext4_show_quota_options(seq, sb);
1077 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1078 u64 ino, u32 generation)
1080 struct inode *inode;
1082 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1083 return ERR_PTR(-ESTALE);
1084 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1085 return ERR_PTR(-ESTALE);
1087 /* iget isn't really right if the inode is currently unallocated!!
1089 * ext4_read_inode will return a bad_inode if the inode had been
1090 * deleted, so we should be safe.
1092 * Currently we don't know the generation for parent directory, so
1093 * a generation of 0 means "accept any"
1095 inode = ext4_iget(sb, ino);
1097 return ERR_CAST(inode);
1098 if (generation && inode->i_generation != generation) {
1100 return ERR_PTR(-ESTALE);
1106 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1107 int fh_len, int fh_type)
1109 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1110 ext4_nfs_get_inode);
1113 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1114 int fh_len, int fh_type)
1116 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1117 ext4_nfs_get_inode);
1121 * Try to release metadata pages (indirect blocks, directories) which are
1122 * mapped via the block device. Since these pages could have journal heads
1123 * which would prevent try_to_free_buffers() from freeing them, we must use
1124 * jbd2 layer's try_to_free_buffers() function to release them.
1126 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1129 journal_t *journal = EXT4_SB(sb)->s_journal;
1131 WARN_ON(PageChecked(page));
1132 if (!page_has_buffers(page))
1135 return jbd2_journal_try_to_free_buffers(journal, page,
1136 wait & ~__GFP_WAIT);
1137 return try_to_free_buffers(page);
1141 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1142 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1144 static int ext4_write_dquot(struct dquot *dquot);
1145 static int ext4_acquire_dquot(struct dquot *dquot);
1146 static int ext4_release_dquot(struct dquot *dquot);
1147 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1148 static int ext4_write_info(struct super_block *sb, int type);
1149 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1151 static int ext4_quota_off(struct super_block *sb, int type);
1152 static int ext4_quota_on_mount(struct super_block *sb, int type);
1153 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1154 size_t len, loff_t off);
1155 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1156 const char *data, size_t len, loff_t off);
1158 static const struct dquot_operations ext4_quota_operations = {
1160 .get_reserved_space = ext4_get_reserved_space,
1162 .write_dquot = ext4_write_dquot,
1163 .acquire_dquot = ext4_acquire_dquot,
1164 .release_dquot = ext4_release_dquot,
1165 .mark_dirty = ext4_mark_dquot_dirty,
1166 .write_info = ext4_write_info,
1167 .alloc_dquot = dquot_alloc,
1168 .destroy_dquot = dquot_destroy,
1171 static const struct quotactl_ops ext4_qctl_operations = {
1172 .quota_on = ext4_quota_on,
1173 .quota_off = ext4_quota_off,
1174 .quota_sync = dquot_quota_sync,
1175 .get_info = dquot_get_dqinfo,
1176 .set_info = dquot_set_dqinfo,
1177 .get_dqblk = dquot_get_dqblk,
1178 .set_dqblk = dquot_set_dqblk
1182 static const struct super_operations ext4_sops = {
1183 .alloc_inode = ext4_alloc_inode,
1184 .destroy_inode = ext4_destroy_inode,
1185 .write_inode = ext4_write_inode,
1186 .dirty_inode = ext4_dirty_inode,
1187 .drop_inode = ext4_drop_inode,
1188 .evict_inode = ext4_evict_inode,
1189 .put_super = ext4_put_super,
1190 .sync_fs = ext4_sync_fs,
1191 .freeze_fs = ext4_freeze,
1192 .unfreeze_fs = ext4_unfreeze,
1193 .statfs = ext4_statfs,
1194 .remount_fs = ext4_remount,
1195 .show_options = ext4_show_options,
1197 .quota_read = ext4_quota_read,
1198 .quota_write = ext4_quota_write,
1200 .bdev_try_to_free_page = bdev_try_to_free_page,
1203 static const struct super_operations ext4_nojournal_sops = {
1204 .alloc_inode = ext4_alloc_inode,
1205 .destroy_inode = ext4_destroy_inode,
1206 .write_inode = ext4_write_inode,
1207 .dirty_inode = ext4_dirty_inode,
1208 .drop_inode = ext4_drop_inode,
1209 .evict_inode = ext4_evict_inode,
1210 .write_super = ext4_write_super,
1211 .put_super = ext4_put_super,
1212 .statfs = ext4_statfs,
1213 .remount_fs = ext4_remount,
1214 .show_options = ext4_show_options,
1216 .quota_read = ext4_quota_read,
1217 .quota_write = ext4_quota_write,
1219 .bdev_try_to_free_page = bdev_try_to_free_page,
1222 static const struct export_operations ext4_export_ops = {
1223 .fh_to_dentry = ext4_fh_to_dentry,
1224 .fh_to_parent = ext4_fh_to_parent,
1225 .get_parent = ext4_get_parent,
1229 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1230 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1231 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1232 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1233 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1234 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1235 Opt_journal_update, Opt_journal_dev,
1236 Opt_journal_checksum, Opt_journal_async_commit,
1237 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1238 Opt_data_err_abort, Opt_data_err_ignore,
1239 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1240 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1241 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1242 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1243 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1244 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1245 Opt_inode_readahead_blks, Opt_journal_ioprio,
1246 Opt_dioread_nolock, Opt_dioread_lock,
1247 Opt_discard, Opt_nodiscard,
1248 Opt_init_inode_table, Opt_noinit_inode_table,
1251 static const match_table_t tokens = {
1252 {Opt_bsd_df, "bsddf"},
1253 {Opt_minix_df, "minixdf"},
1254 {Opt_grpid, "grpid"},
1255 {Opt_grpid, "bsdgroups"},
1256 {Opt_nogrpid, "nogrpid"},
1257 {Opt_nogrpid, "sysvgroups"},
1258 {Opt_resgid, "resgid=%u"},
1259 {Opt_resuid, "resuid=%u"},
1261 {Opt_err_cont, "errors=continue"},
1262 {Opt_err_panic, "errors=panic"},
1263 {Opt_err_ro, "errors=remount-ro"},
1264 {Opt_nouid32, "nouid32"},
1265 {Opt_debug, "debug"},
1266 {Opt_oldalloc, "oldalloc"},
1267 {Opt_orlov, "orlov"},
1268 {Opt_user_xattr, "user_xattr"},
1269 {Opt_nouser_xattr, "nouser_xattr"},
1271 {Opt_noacl, "noacl"},
1272 {Opt_noload, "noload"},
1273 {Opt_noload, "norecovery"},
1276 {Opt_commit, "commit=%u"},
1277 {Opt_min_batch_time, "min_batch_time=%u"},
1278 {Opt_max_batch_time, "max_batch_time=%u"},
1279 {Opt_journal_update, "journal=update"},
1280 {Opt_journal_dev, "journal_dev=%u"},
1281 {Opt_journal_checksum, "journal_checksum"},
1282 {Opt_journal_async_commit, "journal_async_commit"},
1283 {Opt_abort, "abort"},
1284 {Opt_data_journal, "data=journal"},
1285 {Opt_data_ordered, "data=ordered"},
1286 {Opt_data_writeback, "data=writeback"},
1287 {Opt_data_err_abort, "data_err=abort"},
1288 {Opt_data_err_ignore, "data_err=ignore"},
1289 {Opt_offusrjquota, "usrjquota="},
1290 {Opt_usrjquota, "usrjquota=%s"},
1291 {Opt_offgrpjquota, "grpjquota="},
1292 {Opt_grpjquota, "grpjquota=%s"},
1293 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1294 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1295 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1296 {Opt_grpquota, "grpquota"},
1297 {Opt_noquota, "noquota"},
1298 {Opt_quota, "quota"},
1299 {Opt_usrquota, "usrquota"},
1300 {Opt_barrier, "barrier=%u"},
1301 {Opt_barrier, "barrier"},
1302 {Opt_nobarrier, "nobarrier"},
1303 {Opt_i_version, "i_version"},
1304 {Opt_stripe, "stripe=%u"},
1305 {Opt_resize, "resize"},
1306 {Opt_delalloc, "delalloc"},
1307 {Opt_nodelalloc, "nodelalloc"},
1308 {Opt_mblk_io_submit, "mblk_io_submit"},
1309 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1310 {Opt_block_validity, "block_validity"},
1311 {Opt_noblock_validity, "noblock_validity"},
1312 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1313 {Opt_journal_ioprio, "journal_ioprio=%u"},
1314 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1315 {Opt_auto_da_alloc, "auto_da_alloc"},
1316 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1317 {Opt_dioread_nolock, "dioread_nolock"},
1318 {Opt_dioread_lock, "dioread_lock"},
1319 {Opt_discard, "discard"},
1320 {Opt_nodiscard, "nodiscard"},
1321 {Opt_init_inode_table, "init_itable=%u"},
1322 {Opt_init_inode_table, "init_itable"},
1323 {Opt_noinit_inode_table, "noinit_itable"},
1327 static ext4_fsblk_t get_sb_block(void **data)
1329 ext4_fsblk_t sb_block;
1330 char *options = (char *) *data;
1332 if (!options || strncmp(options, "sb=", 3) != 0)
1333 return 1; /* Default location */
1336 /* TODO: use simple_strtoll with >32bit ext4 */
1337 sb_block = simple_strtoul(options, &options, 0);
1338 if (*options && *options != ',') {
1339 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1343 if (*options == ',')
1345 *data = (void *) options;
1350 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1351 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1352 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1355 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1357 struct ext4_sb_info *sbi = EXT4_SB(sb);
1360 if (sb_any_quota_loaded(sb) &&
1361 !sbi->s_qf_names[qtype]) {
1362 ext4_msg(sb, KERN_ERR,
1363 "Cannot change journaled "
1364 "quota options when quota turned on");
1367 qname = match_strdup(args);
1369 ext4_msg(sb, KERN_ERR,
1370 "Not enough memory for storing quotafile name");
1373 if (sbi->s_qf_names[qtype] &&
1374 strcmp(sbi->s_qf_names[qtype], qname)) {
1375 ext4_msg(sb, KERN_ERR,
1376 "%s quota file already specified", QTYPE2NAME(qtype));
1380 sbi->s_qf_names[qtype] = qname;
1381 if (strchr(sbi->s_qf_names[qtype], '/')) {
1382 ext4_msg(sb, KERN_ERR,
1383 "quotafile must be on filesystem root");
1384 kfree(sbi->s_qf_names[qtype]);
1385 sbi->s_qf_names[qtype] = NULL;
1388 set_opt(sbi->s_mount_opt, QUOTA);
1392 static int clear_qf_name(struct super_block *sb, int qtype)
1395 struct ext4_sb_info *sbi = EXT4_SB(sb);
1397 if (sb_any_quota_loaded(sb) &&
1398 sbi->s_qf_names[qtype]) {
1399 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1400 " when quota turned on");
1404 * The space will be released later when all options are confirmed
1407 sbi->s_qf_names[qtype] = NULL;
1412 static int parse_options(char *options, struct super_block *sb,
1413 unsigned long *journal_devnum,
1414 unsigned int *journal_ioprio,
1415 ext4_fsblk_t *n_blocks_count, int is_remount)
1417 struct ext4_sb_info *sbi = EXT4_SB(sb);
1419 substring_t args[MAX_OPT_ARGS];
1429 while ((p = strsep(&options, ",")) != NULL) {
1435 * Initialize args struct so we know whether arg was
1436 * found; some options take optional arguments.
1438 args[0].to = args[0].from = 0;
1439 token = match_token(p, tokens, args);
1442 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1443 clear_opt(sbi->s_mount_opt, MINIX_DF);
1446 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1447 set_opt(sbi->s_mount_opt, MINIX_DF);
1451 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1452 set_opt(sbi->s_mount_opt, GRPID);
1456 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1457 clear_opt(sbi->s_mount_opt, GRPID);
1461 if (match_int(&args[0], &option))
1463 sbi->s_resuid = option;
1466 if (match_int(&args[0], &option))
1468 sbi->s_resgid = option;
1471 /* handled by get_sb_block() instead of here */
1472 /* *sb_block = match_int(&args[0]); */
1475 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1476 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1477 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1480 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1481 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1482 set_opt(sbi->s_mount_opt, ERRORS_RO);
1485 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1486 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1487 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1490 set_opt(sbi->s_mount_opt, NO_UID32);
1493 set_opt(sbi->s_mount_opt, DEBUG);
1496 set_opt(sbi->s_mount_opt, OLDALLOC);
1499 clear_opt(sbi->s_mount_opt, OLDALLOC);
1501 #ifdef CONFIG_EXT4_FS_XATTR
1502 case Opt_user_xattr:
1503 set_opt(sbi->s_mount_opt, XATTR_USER);
1505 case Opt_nouser_xattr:
1506 clear_opt(sbi->s_mount_opt, XATTR_USER);
1509 case Opt_user_xattr:
1510 case Opt_nouser_xattr:
1511 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1514 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1516 set_opt(sbi->s_mount_opt, POSIX_ACL);
1519 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1524 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1527 case Opt_journal_update:
1529 /* Eventually we will want to be able to create
1530 a journal file here. For now, only allow the
1531 user to specify an existing inode to be the
1534 ext4_msg(sb, KERN_ERR,
1535 "Cannot specify journal on remount");
1538 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1540 case Opt_journal_dev:
1542 ext4_msg(sb, KERN_ERR,
1543 "Cannot specify journal on remount");
1546 if (match_int(&args[0], &option))
1548 *journal_devnum = option;
1550 case Opt_journal_checksum:
1551 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1553 case Opt_journal_async_commit:
1554 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1555 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1558 set_opt(sbi->s_mount_opt, NOLOAD);
1561 if (match_int(&args[0], &option))
1566 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1567 sbi->s_commit_interval = HZ * option;
1569 case Opt_max_batch_time:
1570 if (match_int(&args[0], &option))
1575 option = EXT4_DEF_MAX_BATCH_TIME;
1576 sbi->s_max_batch_time = option;
1578 case Opt_min_batch_time:
1579 if (match_int(&args[0], &option))
1583 sbi->s_min_batch_time = option;
1585 case Opt_data_journal:
1586 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1588 case Opt_data_ordered:
1589 data_opt = EXT4_MOUNT_ORDERED_DATA;
1591 case Opt_data_writeback:
1592 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1595 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1596 ext4_msg(sb, KERN_ERR,
1597 "Cannot change data mode on remount");
1601 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1602 sbi->s_mount_opt |= data_opt;
1605 case Opt_data_err_abort:
1606 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1608 case Opt_data_err_ignore:
1609 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1613 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1617 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1620 case Opt_offusrjquota:
1621 if (!clear_qf_name(sb, USRQUOTA))
1624 case Opt_offgrpjquota:
1625 if (!clear_qf_name(sb, GRPQUOTA))
1629 case Opt_jqfmt_vfsold:
1630 qfmt = QFMT_VFS_OLD;
1632 case Opt_jqfmt_vfsv0:
1635 case Opt_jqfmt_vfsv1:
1638 if (sb_any_quota_loaded(sb) &&
1639 sbi->s_jquota_fmt != qfmt) {
1640 ext4_msg(sb, KERN_ERR, "Cannot change "
1641 "journaled quota options when "
1645 sbi->s_jquota_fmt = qfmt;
1649 set_opt(sbi->s_mount_opt, QUOTA);
1650 set_opt(sbi->s_mount_opt, USRQUOTA);
1653 set_opt(sbi->s_mount_opt, QUOTA);
1654 set_opt(sbi->s_mount_opt, GRPQUOTA);
1657 if (sb_any_quota_loaded(sb)) {
1658 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1659 "options when quota turned on");
1662 clear_opt(sbi->s_mount_opt, QUOTA);
1663 clear_opt(sbi->s_mount_opt, USRQUOTA);
1664 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1670 ext4_msg(sb, KERN_ERR,
1671 "quota options not supported");
1675 case Opt_offusrjquota:
1676 case Opt_offgrpjquota:
1677 case Opt_jqfmt_vfsold:
1678 case Opt_jqfmt_vfsv0:
1679 case Opt_jqfmt_vfsv1:
1680 ext4_msg(sb, KERN_ERR,
1681 "journaled quota options not supported");
1687 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1690 clear_opt(sbi->s_mount_opt, BARRIER);
1694 if (match_int(&args[0], &option))
1697 option = 1; /* No argument, default to 1 */
1699 set_opt(sbi->s_mount_opt, BARRIER);
1701 clear_opt(sbi->s_mount_opt, BARRIER);
1707 ext4_msg(sb, KERN_ERR,
1708 "resize option only available "
1712 if (match_int(&args[0], &option) != 0)
1714 *n_blocks_count = option;
1717 ext4_msg(sb, KERN_WARNING,
1718 "Ignoring deprecated nobh option");
1721 ext4_msg(sb, KERN_WARNING,
1722 "Ignoring deprecated bh option");
1725 set_opt(sbi->s_mount_opt, I_VERSION);
1726 sb->s_flags |= MS_I_VERSION;
1728 case Opt_nodelalloc:
1729 clear_opt(sbi->s_mount_opt, DELALLOC);
1731 case Opt_mblk_io_submit:
1732 set_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT);
1734 case Opt_nomblk_io_submit:
1735 clear_opt(sbi->s_mount_opt, MBLK_IO_SUBMIT);
1738 if (match_int(&args[0], &option))
1742 sbi->s_stripe = option;
1745 set_opt(sbi->s_mount_opt, DELALLOC);
1747 case Opt_block_validity:
1748 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1750 case Opt_noblock_validity:
1751 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1753 case Opt_inode_readahead_blks:
1754 if (match_int(&args[0], &option))
1756 if (option < 0 || option > (1 << 30))
1758 if (!is_power_of_2(option)) {
1759 ext4_msg(sb, KERN_ERR,
1760 "EXT4-fs: inode_readahead_blks"
1761 " must be a power of 2");
1764 sbi->s_inode_readahead_blks = option;
1766 case Opt_journal_ioprio:
1767 if (match_int(&args[0], &option))
1769 if (option < 0 || option > 7)
1771 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1774 case Opt_noauto_da_alloc:
1775 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1777 case Opt_auto_da_alloc:
1779 if (match_int(&args[0], &option))
1782 option = 1; /* No argument, default to 1 */
1784 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1786 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1789 set_opt(sbi->s_mount_opt, DISCARD);
1792 clear_opt(sbi->s_mount_opt, DISCARD);
1794 case Opt_dioread_nolock:
1795 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1797 case Opt_dioread_lock:
1798 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1800 case Opt_init_inode_table:
1801 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1803 if (match_int(&args[0], &option))
1806 option = EXT4_DEF_LI_WAIT_MULT;
1809 sbi->s_li_wait_mult = option;
1811 case Opt_noinit_inode_table:
1812 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1815 ext4_msg(sb, KERN_ERR,
1816 "Unrecognized mount option \"%s\" "
1817 "or missing value", p);
1822 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1823 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1824 clear_opt(sbi->s_mount_opt, USRQUOTA);
1826 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1827 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1829 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1830 ext4_msg(sb, KERN_ERR, "old and new quota "
1835 if (!sbi->s_jquota_fmt) {
1836 ext4_msg(sb, KERN_ERR, "journaled quota format "
1841 if (sbi->s_jquota_fmt) {
1842 ext4_msg(sb, KERN_ERR, "journaled quota format "
1843 "specified with no journaling "
1852 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1855 struct ext4_sb_info *sbi = EXT4_SB(sb);
1858 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1859 ext4_msg(sb, KERN_ERR, "revision level too high, "
1860 "forcing read-only mode");
1865 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1866 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1867 "running e2fsck is recommended");
1868 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1869 ext4_msg(sb, KERN_WARNING,
1870 "warning: mounting fs with errors, "
1871 "running e2fsck is recommended");
1872 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1873 le16_to_cpu(es->s_mnt_count) >=
1874 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1875 ext4_msg(sb, KERN_WARNING,
1876 "warning: maximal mount count reached, "
1877 "running e2fsck is recommended");
1878 else if (le32_to_cpu(es->s_checkinterval) &&
1879 (le32_to_cpu(es->s_lastcheck) +
1880 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1881 ext4_msg(sb, KERN_WARNING,
1882 "warning: checktime reached, "
1883 "running e2fsck is recommended");
1884 if (!sbi->s_journal)
1885 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1886 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1887 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1888 le16_add_cpu(&es->s_mnt_count, 1);
1889 es->s_mtime = cpu_to_le32(get_seconds());
1890 ext4_update_dynamic_rev(sb);
1892 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1894 ext4_commit_super(sb, 1);
1895 if (test_opt(sb, DEBUG))
1896 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1897 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1899 sbi->s_groups_count,
1900 EXT4_BLOCKS_PER_GROUP(sb),
1901 EXT4_INODES_PER_GROUP(sb),
1907 static int ext4_fill_flex_info(struct super_block *sb)
1909 struct ext4_sb_info *sbi = EXT4_SB(sb);
1910 struct ext4_group_desc *gdp = NULL;
1911 ext4_group_t flex_group_count;
1912 ext4_group_t flex_group;
1913 int groups_per_flex = 0;
1917 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1918 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1920 if (groups_per_flex < 2) {
1921 sbi->s_log_groups_per_flex = 0;
1925 /* We allocate both existing and potentially added groups */
1926 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1927 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1928 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1929 size = flex_group_count * sizeof(struct flex_groups);
1930 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1931 if (sbi->s_flex_groups == NULL) {
1932 sbi->s_flex_groups = vmalloc(size);
1933 if (sbi->s_flex_groups)
1934 memset(sbi->s_flex_groups, 0, size);
1936 if (sbi->s_flex_groups == NULL) {
1937 ext4_msg(sb, KERN_ERR, "not enough memory for "
1938 "%u flex groups", flex_group_count);
1942 for (i = 0; i < sbi->s_groups_count; i++) {
1943 gdp = ext4_get_group_desc(sb, i, NULL);
1945 flex_group = ext4_flex_group(sbi, i);
1946 atomic_add(ext4_free_inodes_count(sb, gdp),
1947 &sbi->s_flex_groups[flex_group].free_inodes);
1948 atomic_add(ext4_free_blks_count(sb, gdp),
1949 &sbi->s_flex_groups[flex_group].free_blocks);
1950 atomic_add(ext4_used_dirs_count(sb, gdp),
1951 &sbi->s_flex_groups[flex_group].used_dirs);
1959 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1960 struct ext4_group_desc *gdp)
1964 if (sbi->s_es->s_feature_ro_compat &
1965 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1966 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1967 __le32 le_group = cpu_to_le32(block_group);
1969 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1970 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1971 crc = crc16(crc, (__u8 *)gdp, offset);
1972 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1973 /* for checksum of struct ext4_group_desc do the rest...*/
1974 if ((sbi->s_es->s_feature_incompat &
1975 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1976 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1977 crc = crc16(crc, (__u8 *)gdp + offset,
1978 le16_to_cpu(sbi->s_es->s_desc_size) -
1982 return cpu_to_le16(crc);
1985 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1986 struct ext4_group_desc *gdp)
1988 if ((sbi->s_es->s_feature_ro_compat &
1989 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1990 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1996 /* Called at mount-time, super-block is locked */
1997 static int ext4_check_descriptors(struct super_block *sb,
1998 ext4_group_t *first_not_zeroed)
2000 struct ext4_sb_info *sbi = EXT4_SB(sb);
2001 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2002 ext4_fsblk_t last_block;
2003 ext4_fsblk_t block_bitmap;
2004 ext4_fsblk_t inode_bitmap;
2005 ext4_fsblk_t inode_table;
2006 int flexbg_flag = 0;
2007 ext4_group_t i, grp = sbi->s_groups_count;
2009 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2012 ext4_debug("Checking group descriptors");
2014 for (i = 0; i < sbi->s_groups_count; i++) {
2015 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2017 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2018 last_block = ext4_blocks_count(sbi->s_es) - 1;
2020 last_block = first_block +
2021 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2023 if ((grp == sbi->s_groups_count) &&
2024 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2027 block_bitmap = ext4_block_bitmap(sb, gdp);
2028 if (block_bitmap < first_block || block_bitmap > last_block) {
2029 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2030 "Block bitmap for group %u not in group "
2031 "(block %llu)!", i, block_bitmap);
2034 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2035 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2036 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2037 "Inode bitmap for group %u not in group "
2038 "(block %llu)!", i, inode_bitmap);
2041 inode_table = ext4_inode_table(sb, gdp);
2042 if (inode_table < first_block ||
2043 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2044 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2045 "Inode table for group %u not in group "
2046 "(block %llu)!", i, inode_table);
2049 ext4_lock_group(sb, i);
2050 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2052 "Checksum for group %u failed (%u!=%u)",
2053 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2054 gdp)), le16_to_cpu(gdp->bg_checksum));
2055 if (!(sb->s_flags & MS_RDONLY)) {
2056 ext4_unlock_group(sb, i);
2060 ext4_unlock_group(sb, i);
2062 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2064 if (NULL != first_not_zeroed)
2065 *first_not_zeroed = grp;
2067 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2068 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2072 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2073 * the superblock) which were deleted from all directories, but held open by
2074 * a process at the time of a crash. We walk the list and try to delete these
2075 * inodes at recovery time (only with a read-write filesystem).
2077 * In order to keep the orphan inode chain consistent during traversal (in
2078 * case of crash during recovery), we link each inode into the superblock
2079 * orphan list_head and handle it the same way as an inode deletion during
2080 * normal operation (which journals the operations for us).
2082 * We only do an iget() and an iput() on each inode, which is very safe if we
2083 * accidentally point at an in-use or already deleted inode. The worst that
2084 * can happen in this case is that we get a "bit already cleared" message from
2085 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2086 * e2fsck was run on this filesystem, and it must have already done the orphan
2087 * inode cleanup for us, so we can safely abort without any further action.
2089 static void ext4_orphan_cleanup(struct super_block *sb,
2090 struct ext4_super_block *es)
2092 unsigned int s_flags = sb->s_flags;
2093 int nr_orphans = 0, nr_truncates = 0;
2097 if (!es->s_last_orphan) {
2098 jbd_debug(4, "no orphan inodes to clean up\n");
2102 if (bdev_read_only(sb->s_bdev)) {
2103 ext4_msg(sb, KERN_ERR, "write access "
2104 "unavailable, skipping orphan cleanup");
2108 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2109 if (es->s_last_orphan)
2110 jbd_debug(1, "Errors on filesystem, "
2111 "clearing orphan list.\n");
2112 es->s_last_orphan = 0;
2113 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2117 if (s_flags & MS_RDONLY) {
2118 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2119 sb->s_flags &= ~MS_RDONLY;
2122 /* Needed for iput() to work correctly and not trash data */
2123 sb->s_flags |= MS_ACTIVE;
2124 /* Turn on quotas so that they are updated correctly */
2125 for (i = 0; i < MAXQUOTAS; i++) {
2126 if (EXT4_SB(sb)->s_qf_names[i]) {
2127 int ret = ext4_quota_on_mount(sb, i);
2129 ext4_msg(sb, KERN_ERR,
2130 "Cannot turn on journaled "
2131 "quota: error %d", ret);
2136 while (es->s_last_orphan) {
2137 struct inode *inode;
2139 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2140 if (IS_ERR(inode)) {
2141 es->s_last_orphan = 0;
2145 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2146 dquot_initialize(inode);
2147 if (inode->i_nlink) {
2148 ext4_msg(sb, KERN_DEBUG,
2149 "%s: truncating inode %lu to %lld bytes",
2150 __func__, inode->i_ino, inode->i_size);
2151 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2152 inode->i_ino, inode->i_size);
2153 ext4_truncate(inode);
2156 ext4_msg(sb, KERN_DEBUG,
2157 "%s: deleting unreferenced inode %lu",
2158 __func__, inode->i_ino);
2159 jbd_debug(2, "deleting unreferenced inode %lu\n",
2163 iput(inode); /* The delete magic happens here! */
2166 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2169 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2170 PLURAL(nr_orphans));
2172 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2173 PLURAL(nr_truncates));
2175 /* Turn quotas off */
2176 for (i = 0; i < MAXQUOTAS; i++) {
2177 if (sb_dqopt(sb)->files[i])
2178 dquot_quota_off(sb, i);
2181 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2185 * Maximal extent format file size.
2186 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2187 * extent format containers, within a sector_t, and within i_blocks
2188 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2189 * so that won't be a limiting factor.
2191 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2193 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2196 loff_t upper_limit = MAX_LFS_FILESIZE;
2198 /* small i_blocks in vfs inode? */
2199 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2201 * CONFIG_LBDAF is not enabled implies the inode
2202 * i_block represent total blocks in 512 bytes
2203 * 32 == size of vfs inode i_blocks * 8
2205 upper_limit = (1LL << 32) - 1;
2207 /* total blocks in file system block size */
2208 upper_limit >>= (blkbits - 9);
2209 upper_limit <<= blkbits;
2212 /* 32-bit extent-start container, ee_block */
2217 /* Sanity check against vm- & vfs- imposed limits */
2218 if (res > upper_limit)
2225 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2226 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2227 * We need to be 1 filesystem block less than the 2^48 sector limit.
2229 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2231 loff_t res = EXT4_NDIR_BLOCKS;
2234 /* This is calculated to be the largest file size for a dense, block
2235 * mapped file such that the file's total number of 512-byte sectors,
2236 * including data and all indirect blocks, does not exceed (2^48 - 1).
2238 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2239 * number of 512-byte sectors of the file.
2242 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2244 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2245 * the inode i_block field represents total file blocks in
2246 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2248 upper_limit = (1LL << 32) - 1;
2250 /* total blocks in file system block size */
2251 upper_limit >>= (bits - 9);
2255 * We use 48 bit ext4_inode i_blocks
2256 * With EXT4_HUGE_FILE_FL set the i_blocks
2257 * represent total number of blocks in
2258 * file system block size
2260 upper_limit = (1LL << 48) - 1;
2264 /* indirect blocks */
2266 /* double indirect blocks */
2267 meta_blocks += 1 + (1LL << (bits-2));
2268 /* tripple indirect blocks */
2269 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2271 upper_limit -= meta_blocks;
2272 upper_limit <<= bits;
2274 res += 1LL << (bits-2);
2275 res += 1LL << (2*(bits-2));
2276 res += 1LL << (3*(bits-2));
2278 if (res > upper_limit)
2281 if (res > MAX_LFS_FILESIZE)
2282 res = MAX_LFS_FILESIZE;
2287 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2288 ext4_fsblk_t logical_sb_block, int nr)
2290 struct ext4_sb_info *sbi = EXT4_SB(sb);
2291 ext4_group_t bg, first_meta_bg;
2294 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2296 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2298 return logical_sb_block + nr + 1;
2299 bg = sbi->s_desc_per_block * nr;
2300 if (ext4_bg_has_super(sb, bg))
2303 return (has_super + ext4_group_first_block_no(sb, bg));
2307 * ext4_get_stripe_size: Get the stripe size.
2308 * @sbi: In memory super block info
2310 * If we have specified it via mount option, then
2311 * use the mount option value. If the value specified at mount time is
2312 * greater than the blocks per group use the super block value.
2313 * If the super block value is greater than blocks per group return 0.
2314 * Allocator needs it be less than blocks per group.
2317 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2319 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2320 unsigned long stripe_width =
2321 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2323 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2324 return sbi->s_stripe;
2326 if (stripe_width <= sbi->s_blocks_per_group)
2327 return stripe_width;
2329 if (stride <= sbi->s_blocks_per_group)
2338 struct attribute attr;
2339 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2340 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2341 const char *, size_t);
2345 static int parse_strtoul(const char *buf,
2346 unsigned long max, unsigned long *value)
2350 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2351 endp = skip_spaces(endp);
2352 if (*endp || *value > max)
2358 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2359 struct ext4_sb_info *sbi,
2362 return snprintf(buf, PAGE_SIZE, "%llu\n",
2363 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2366 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2367 struct ext4_sb_info *sbi, char *buf)
2369 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2371 if (!sb->s_bdev->bd_part)
2372 return snprintf(buf, PAGE_SIZE, "0\n");
2373 return snprintf(buf, PAGE_SIZE, "%lu\n",
2374 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2375 sbi->s_sectors_written_start) >> 1);
2378 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2379 struct ext4_sb_info *sbi, char *buf)
2381 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2383 if (!sb->s_bdev->bd_part)
2384 return snprintf(buf, PAGE_SIZE, "0\n");
2385 return snprintf(buf, PAGE_SIZE, "%llu\n",
2386 (unsigned long long)(sbi->s_kbytes_written +
2387 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2388 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2391 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2392 struct ext4_sb_info *sbi,
2393 const char *buf, size_t count)
2397 if (parse_strtoul(buf, 0x40000000, &t))
2400 if (!is_power_of_2(t))
2403 sbi->s_inode_readahead_blks = t;
2407 static ssize_t sbi_ui_show(struct ext4_attr *a,
2408 struct ext4_sb_info *sbi, char *buf)
2410 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2412 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2415 static ssize_t sbi_ui_store(struct ext4_attr *a,
2416 struct ext4_sb_info *sbi,
2417 const char *buf, size_t count)
2419 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2422 if (parse_strtoul(buf, 0xffffffff, &t))
2428 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2429 static struct ext4_attr ext4_attr_##_name = { \
2430 .attr = {.name = __stringify(_name), .mode = _mode }, \
2433 .offset = offsetof(struct ext4_sb_info, _elname), \
2435 #define EXT4_ATTR(name, mode, show, store) \
2436 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2438 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2439 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2440 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2441 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2442 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2443 #define ATTR_LIST(name) &ext4_attr_##name.attr
2445 EXT4_RO_ATTR(delayed_allocation_blocks);
2446 EXT4_RO_ATTR(session_write_kbytes);
2447 EXT4_RO_ATTR(lifetime_write_kbytes);
2448 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2449 inode_readahead_blks_store, s_inode_readahead_blks);
2450 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2451 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2452 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2453 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2454 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2455 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2456 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2457 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2459 static struct attribute *ext4_attrs[] = {
2460 ATTR_LIST(delayed_allocation_blocks),
2461 ATTR_LIST(session_write_kbytes),
2462 ATTR_LIST(lifetime_write_kbytes),
2463 ATTR_LIST(inode_readahead_blks),
2464 ATTR_LIST(inode_goal),
2465 ATTR_LIST(mb_stats),
2466 ATTR_LIST(mb_max_to_scan),
2467 ATTR_LIST(mb_min_to_scan),
2468 ATTR_LIST(mb_order2_req),
2469 ATTR_LIST(mb_stream_req),
2470 ATTR_LIST(mb_group_prealloc),
2471 ATTR_LIST(max_writeback_mb_bump),
2475 /* Features this copy of ext4 supports */
2476 EXT4_INFO_ATTR(lazy_itable_init);
2477 EXT4_INFO_ATTR(batched_discard);
2479 static struct attribute *ext4_feat_attrs[] = {
2480 ATTR_LIST(lazy_itable_init),
2481 ATTR_LIST(batched_discard),
2485 static ssize_t ext4_attr_show(struct kobject *kobj,
2486 struct attribute *attr, char *buf)
2488 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2490 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2492 return a->show ? a->show(a, sbi, buf) : 0;
2495 static ssize_t ext4_attr_store(struct kobject *kobj,
2496 struct attribute *attr,
2497 const char *buf, size_t len)
2499 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2501 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2503 return a->store ? a->store(a, sbi, buf, len) : 0;
2506 static void ext4_sb_release(struct kobject *kobj)
2508 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2510 complete(&sbi->s_kobj_unregister);
2513 static const struct sysfs_ops ext4_attr_ops = {
2514 .show = ext4_attr_show,
2515 .store = ext4_attr_store,
2518 static struct kobj_type ext4_ktype = {
2519 .default_attrs = ext4_attrs,
2520 .sysfs_ops = &ext4_attr_ops,
2521 .release = ext4_sb_release,
2524 static void ext4_feat_release(struct kobject *kobj)
2526 complete(&ext4_feat->f_kobj_unregister);
2529 static struct kobj_type ext4_feat_ktype = {
2530 .default_attrs = ext4_feat_attrs,
2531 .sysfs_ops = &ext4_attr_ops,
2532 .release = ext4_feat_release,
2536 * Check whether this filesystem can be mounted based on
2537 * the features present and the RDONLY/RDWR mount requested.
2538 * Returns 1 if this filesystem can be mounted as requested,
2539 * 0 if it cannot be.
2541 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2543 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2544 ext4_msg(sb, KERN_ERR,
2545 "Couldn't mount because of "
2546 "unsupported optional features (%x)",
2547 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2548 ~EXT4_FEATURE_INCOMPAT_SUPP));
2555 /* Check that feature set is OK for a read-write mount */
2556 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2557 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2558 "unsupported optional features (%x)",
2559 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2560 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2564 * Large file size enabled file system can only be mounted
2565 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2567 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2568 if (sizeof(blkcnt_t) < sizeof(u64)) {
2569 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2570 "cannot be mounted RDWR without "
2579 * This function is called once a day if we have errors logged
2580 * on the file system
2582 static void print_daily_error_info(unsigned long arg)
2584 struct super_block *sb = (struct super_block *) arg;
2585 struct ext4_sb_info *sbi;
2586 struct ext4_super_block *es;
2591 if (es->s_error_count)
2592 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2593 le32_to_cpu(es->s_error_count));
2594 if (es->s_first_error_time) {
2595 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2596 sb->s_id, le32_to_cpu(es->s_first_error_time),
2597 (int) sizeof(es->s_first_error_func),
2598 es->s_first_error_func,
2599 le32_to_cpu(es->s_first_error_line));
2600 if (es->s_first_error_ino)
2601 printk(": inode %u",
2602 le32_to_cpu(es->s_first_error_ino));
2603 if (es->s_first_error_block)
2604 printk(": block %llu", (unsigned long long)
2605 le64_to_cpu(es->s_first_error_block));
2608 if (es->s_last_error_time) {
2609 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2610 sb->s_id, le32_to_cpu(es->s_last_error_time),
2611 (int) sizeof(es->s_last_error_func),
2612 es->s_last_error_func,
2613 le32_to_cpu(es->s_last_error_line));
2614 if (es->s_last_error_ino)
2615 printk(": inode %u",
2616 le32_to_cpu(es->s_last_error_ino));
2617 if (es->s_last_error_block)
2618 printk(": block %llu", (unsigned long long)
2619 le64_to_cpu(es->s_last_error_block));
2622 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2625 static void ext4_lazyinode_timeout(unsigned long data)
2627 struct task_struct *p = (struct task_struct *)data;
2631 /* Find next suitable group and run ext4_init_inode_table */
2632 static int ext4_run_li_request(struct ext4_li_request *elr)
2634 struct ext4_group_desc *gdp = NULL;
2635 ext4_group_t group, ngroups;
2636 struct super_block *sb;
2637 unsigned long timeout = 0;
2641 ngroups = EXT4_SB(sb)->s_groups_count;
2643 for (group = elr->lr_next_group; group < ngroups; group++) {
2644 gdp = ext4_get_group_desc(sb, group, NULL);
2650 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2654 if (group == ngroups)
2659 ret = ext4_init_inode_table(sb, group,
2660 elr->lr_timeout ? 0 : 1);
2661 if (elr->lr_timeout == 0) {
2662 timeout = jiffies - timeout;
2663 if (elr->lr_sbi->s_li_wait_mult)
2664 timeout *= elr->lr_sbi->s_li_wait_mult;
2667 elr->lr_timeout = timeout;
2669 elr->lr_next_sched = jiffies + elr->lr_timeout;
2670 elr->lr_next_group = group + 1;
2677 * Remove lr_request from the list_request and free the
2678 * request tructure. Should be called with li_list_mtx held
2680 static void ext4_remove_li_request(struct ext4_li_request *elr)
2682 struct ext4_sb_info *sbi;
2689 list_del(&elr->lr_request);
2690 sbi->s_li_request = NULL;
2694 static void ext4_unregister_li_request(struct super_block *sb)
2696 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2701 mutex_lock(&ext4_li_info->li_list_mtx);
2702 ext4_remove_li_request(elr);
2703 mutex_unlock(&ext4_li_info->li_list_mtx);
2707 * This is the function where ext4lazyinit thread lives. It walks
2708 * through the request list searching for next scheduled filesystem.
2709 * When such a fs is found, run the lazy initialization request
2710 * (ext4_rn_li_request) and keep track of the time spend in this
2711 * function. Based on that time we compute next schedule time of
2712 * the request. When walking through the list is complete, compute
2713 * next waking time and put itself into sleep.
2715 static int ext4_lazyinit_thread(void *arg)
2717 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2718 struct list_head *pos, *n;
2719 struct ext4_li_request *elr;
2720 unsigned long next_wakeup;
2723 BUG_ON(NULL == eli);
2725 eli->li_timer.data = (unsigned long)current;
2726 eli->li_timer.function = ext4_lazyinode_timeout;
2728 eli->li_task = current;
2729 wake_up(&eli->li_wait_task);
2733 next_wakeup = MAX_JIFFY_OFFSET;
2735 mutex_lock(&eli->li_list_mtx);
2736 if (list_empty(&eli->li_request_list)) {
2737 mutex_unlock(&eli->li_list_mtx);
2741 list_for_each_safe(pos, n, &eli->li_request_list) {
2742 elr = list_entry(pos, struct ext4_li_request,
2745 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2746 if (ext4_run_li_request(elr) != 0) {
2747 /* error, remove the lazy_init job */
2748 ext4_remove_li_request(elr);
2753 if (time_before(elr->lr_next_sched, next_wakeup))
2754 next_wakeup = elr->lr_next_sched;
2756 mutex_unlock(&eli->li_list_mtx);
2758 if (freezing(current))
2761 if ((time_after_eq(jiffies, next_wakeup)) ||
2762 (MAX_JIFFY_OFFSET == next_wakeup)) {
2767 eli->li_timer.expires = next_wakeup;
2768 add_timer(&eli->li_timer);
2769 prepare_to_wait(&eli->li_wait_daemon, &wait,
2770 TASK_INTERRUPTIBLE);
2771 if (time_before(jiffies, next_wakeup))
2773 finish_wait(&eli->li_wait_daemon, &wait);
2778 * It looks like the request list is empty, but we need
2779 * to check it under the li_list_mtx lock, to prevent any
2780 * additions into it, and of course we should lock ext4_li_mtx
2781 * to atomically free the list and ext4_li_info, because at
2782 * this point another ext4 filesystem could be registering
2785 mutex_lock(&ext4_li_mtx);
2786 mutex_lock(&eli->li_list_mtx);
2787 if (!list_empty(&eli->li_request_list)) {
2788 mutex_unlock(&eli->li_list_mtx);
2789 mutex_unlock(&ext4_li_mtx);
2792 mutex_unlock(&eli->li_list_mtx);
2793 del_timer_sync(&ext4_li_info->li_timer);
2794 eli->li_task = NULL;
2795 wake_up(&eli->li_wait_task);
2797 kfree(ext4_li_info);
2798 ext4_li_info = NULL;
2799 mutex_unlock(&ext4_li_mtx);
2804 static void ext4_clear_request_list(void)
2806 struct list_head *pos, *n;
2807 struct ext4_li_request *elr;
2809 mutex_lock(&ext4_li_info->li_list_mtx);
2810 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2811 elr = list_entry(pos, struct ext4_li_request,
2813 ext4_remove_li_request(elr);
2815 mutex_unlock(&ext4_li_info->li_list_mtx);
2818 static int ext4_run_lazyinit_thread(void)
2820 struct task_struct *t;
2822 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2824 int err = PTR_ERR(t);
2825 ext4_clear_request_list();
2826 del_timer_sync(&ext4_li_info->li_timer);
2827 kfree(ext4_li_info);
2828 ext4_li_info = NULL;
2829 printk(KERN_CRIT "EXT4: error %d creating inode table "
2830 "initialization thread\n",
2834 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2836 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2841 * Check whether it make sense to run itable init. thread or not.
2842 * If there is at least one uninitialized inode table, return
2843 * corresponding group number, else the loop goes through all
2844 * groups and return total number of groups.
2846 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2848 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2849 struct ext4_group_desc *gdp = NULL;
2851 for (group = 0; group < ngroups; group++) {
2852 gdp = ext4_get_group_desc(sb, group, NULL);
2856 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2863 static int ext4_li_info_new(void)
2865 struct ext4_lazy_init *eli = NULL;
2867 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2871 eli->li_task = NULL;
2872 INIT_LIST_HEAD(&eli->li_request_list);
2873 mutex_init(&eli->li_list_mtx);
2875 init_waitqueue_head(&eli->li_wait_daemon);
2876 init_waitqueue_head(&eli->li_wait_task);
2877 init_timer(&eli->li_timer);
2878 eli->li_state |= EXT4_LAZYINIT_QUIT;
2885 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2888 struct ext4_sb_info *sbi = EXT4_SB(sb);
2889 struct ext4_li_request *elr;
2892 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2898 elr->lr_next_group = start;
2901 * Randomize first schedule time of the request to
2902 * spread the inode table initialization requests
2905 get_random_bytes(&rnd, sizeof(rnd));
2906 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2907 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2912 static int ext4_register_li_request(struct super_block *sb,
2913 ext4_group_t first_not_zeroed)
2915 struct ext4_sb_info *sbi = EXT4_SB(sb);
2916 struct ext4_li_request *elr;
2917 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2920 if (sbi->s_li_request != NULL)
2923 if (first_not_zeroed == ngroups ||
2924 (sb->s_flags & MS_RDONLY) ||
2925 !test_opt(sb, INIT_INODE_TABLE)) {
2926 sbi->s_li_request = NULL;
2930 if (first_not_zeroed == ngroups) {
2931 sbi->s_li_request = NULL;
2935 elr = ext4_li_request_new(sb, first_not_zeroed);
2939 mutex_lock(&ext4_li_mtx);
2941 if (NULL == ext4_li_info) {
2942 ret = ext4_li_info_new();
2947 mutex_lock(&ext4_li_info->li_list_mtx);
2948 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2949 mutex_unlock(&ext4_li_info->li_list_mtx);
2951 sbi->s_li_request = elr;
2953 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2954 ret = ext4_run_lazyinit_thread();
2959 mutex_unlock(&ext4_li_mtx);
2966 * We do not need to lock anything since this is called on
2969 static void ext4_destroy_lazyinit_thread(void)
2972 * If thread exited earlier
2973 * there's nothing to be done.
2978 ext4_clear_request_list();
2980 while (ext4_li_info->li_task) {
2981 wake_up(&ext4_li_info->li_wait_daemon);
2982 wait_event(ext4_li_info->li_wait_task,
2983 ext4_li_info->li_task == NULL);
2987 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2988 __releases(kernel_lock)
2989 __acquires(kernel_lock)
2991 char *orig_data = kstrdup(data, GFP_KERNEL);
2992 struct buffer_head *bh;
2993 struct ext4_super_block *es = NULL;
2994 struct ext4_sb_info *sbi;
2996 ext4_fsblk_t sb_block = get_sb_block(&data);
2997 ext4_fsblk_t logical_sb_block;
2998 unsigned long offset = 0;
2999 unsigned long journal_devnum = 0;
3000 unsigned long def_mount_opts;
3006 unsigned int db_count;
3008 int needs_recovery, has_huge_files;
3011 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3012 ext4_group_t first_not_zeroed;
3014 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3018 sbi->s_blockgroup_lock =
3019 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3020 if (!sbi->s_blockgroup_lock) {
3024 sb->s_fs_info = sbi;
3025 sbi->s_mount_opt = 0;
3026 sbi->s_resuid = EXT4_DEF_RESUID;
3027 sbi->s_resgid = EXT4_DEF_RESGID;
3028 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3029 sbi->s_sb_block = sb_block;
3030 if (sb->s_bdev->bd_part)
3031 sbi->s_sectors_written_start =
3032 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3034 /* Cleanup superblock name */
3035 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3039 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3041 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3046 * The ext4 superblock will not be buffer aligned for other than 1kB
3047 * block sizes. We need to calculate the offset from buffer start.
3049 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3050 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3051 offset = do_div(logical_sb_block, blocksize);
3053 logical_sb_block = sb_block;
3056 if (!(bh = sb_bread(sb, logical_sb_block))) {
3057 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3061 * Note: s_es must be initialized as soon as possible because
3062 * some ext4 macro-instructions depend on its value
3064 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3066 sb->s_magic = le16_to_cpu(es->s_magic);
3067 if (sb->s_magic != EXT4_SUPER_MAGIC)
3069 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3071 /* Set defaults before we parse the mount options */
3072 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3073 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
3074 if (def_mount_opts & EXT4_DEFM_DEBUG)
3075 set_opt(sbi->s_mount_opt, DEBUG);
3076 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3077 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3079 set_opt(sbi->s_mount_opt, GRPID);
3081 if (def_mount_opts & EXT4_DEFM_UID16)
3082 set_opt(sbi->s_mount_opt, NO_UID32);
3083 #ifdef CONFIG_EXT4_FS_XATTR
3084 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3085 set_opt(sbi->s_mount_opt, XATTR_USER);
3087 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3088 if (def_mount_opts & EXT4_DEFM_ACL)
3089 set_opt(sbi->s_mount_opt, POSIX_ACL);
3091 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3092 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3093 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3094 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3095 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3096 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3098 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3099 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
3100 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3101 set_opt(sbi->s_mount_opt, ERRORS_CONT);
3103 set_opt(sbi->s_mount_opt, ERRORS_RO);
3104 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3105 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3106 if (def_mount_opts & EXT4_DEFM_DISCARD)
3107 set_opt(sbi->s_mount_opt, DISCARD);
3109 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3110 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3111 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3112 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3113 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3115 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3116 set_opt(sbi->s_mount_opt, BARRIER);
3119 * enable delayed allocation by default
3120 * Use -o nodelalloc to turn it off
3122 if (!IS_EXT3_SB(sb) &&
3123 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3124 set_opt(sbi->s_mount_opt, DELALLOC);
3126 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3127 &journal_devnum, &journal_ioprio, NULL, 0)) {
3128 ext4_msg(sb, KERN_WARNING,
3129 "failed to parse options in superblock: %s",
3130 sbi->s_es->s_mount_opts);
3132 if (!parse_options((char *) data, sb, &journal_devnum,
3133 &journal_ioprio, NULL, 0))
3136 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3137 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3139 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3140 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3141 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3142 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3143 ext4_msg(sb, KERN_WARNING,
3144 "feature flags set on rev 0 fs, "
3145 "running e2fsck is recommended");
3148 * Check feature flags regardless of the revision level, since we
3149 * previously didn't change the revision level when setting the flags,
3150 * so there is a chance incompat flags are set on a rev 0 filesystem.
3152 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3155 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3157 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3158 blocksize > EXT4_MAX_BLOCK_SIZE) {
3159 ext4_msg(sb, KERN_ERR,
3160 "Unsupported filesystem blocksize %d", blocksize);
3164 if (sb->s_blocksize != blocksize) {
3165 /* Validate the filesystem blocksize */
3166 if (!sb_set_blocksize(sb, blocksize)) {
3167 ext4_msg(sb, KERN_ERR, "bad block size %d",
3173 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3174 offset = do_div(logical_sb_block, blocksize);
3175 bh = sb_bread(sb, logical_sb_block);
3177 ext4_msg(sb, KERN_ERR,
3178 "Can't read superblock on 2nd try");
3181 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3183 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3184 ext4_msg(sb, KERN_ERR,
3185 "Magic mismatch, very weird!");
3190 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3191 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3192 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3194 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3196 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3197 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3198 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3200 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3201 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3202 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3203 (!is_power_of_2(sbi->s_inode_size)) ||
3204 (sbi->s_inode_size > blocksize)) {
3205 ext4_msg(sb, KERN_ERR,
3206 "unsupported inode size: %d",
3210 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3211 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3214 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3215 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3216 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3217 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3218 !is_power_of_2(sbi->s_desc_size)) {
3219 ext4_msg(sb, KERN_ERR,
3220 "unsupported descriptor size %lu",
3225 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3227 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3228 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3229 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3232 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3233 if (sbi->s_inodes_per_block == 0)
3235 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3236 sbi->s_inodes_per_block;
3237 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3239 sbi->s_mount_state = le16_to_cpu(es->s_state);
3240 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3241 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3243 for (i = 0; i < 4; i++)
3244 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3245 sbi->s_def_hash_version = es->s_def_hash_version;
3246 i = le32_to_cpu(es->s_flags);
3247 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3248 sbi->s_hash_unsigned = 3;
3249 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3250 #ifdef __CHAR_UNSIGNED__
3251 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3252 sbi->s_hash_unsigned = 3;
3254 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3259 if (sbi->s_blocks_per_group > blocksize * 8) {
3260 ext4_msg(sb, KERN_ERR,
3261 "#blocks per group too big: %lu",
3262 sbi->s_blocks_per_group);
3265 if (sbi->s_inodes_per_group > blocksize * 8) {
3266 ext4_msg(sb, KERN_ERR,
3267 "#inodes per group too big: %lu",
3268 sbi->s_inodes_per_group);
3273 * Test whether we have more sectors than will fit in sector_t,
3274 * and whether the max offset is addressable by the page cache.
3276 err = generic_check_addressable(sb->s_blocksize_bits,
3277 ext4_blocks_count(es));
3279 ext4_msg(sb, KERN_ERR, "filesystem"
3280 " too large to mount safely on this system");
3281 if (sizeof(sector_t) < 8)
3282 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3287 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3290 /* check blocks count against device size */
3291 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3292 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3293 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3294 "exceeds size of device (%llu blocks)",
3295 ext4_blocks_count(es), blocks_count);
3300 * It makes no sense for the first data block to be beyond the end
3301 * of the filesystem.
3303 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3304 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3305 "block %u is beyond end of filesystem (%llu)",
3306 le32_to_cpu(es->s_first_data_block),
3307 ext4_blocks_count(es));
3310 blocks_count = (ext4_blocks_count(es) -
3311 le32_to_cpu(es->s_first_data_block) +
3312 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3313 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3314 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3315 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3316 "(block count %llu, first data block %u, "
3317 "blocks per group %lu)", sbi->s_groups_count,
3318 ext4_blocks_count(es),
3319 le32_to_cpu(es->s_first_data_block),
3320 EXT4_BLOCKS_PER_GROUP(sb));
3323 sbi->s_groups_count = blocks_count;
3324 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3325 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3326 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3327 EXT4_DESC_PER_BLOCK(sb);
3328 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3330 if (sbi->s_group_desc == NULL) {
3331 ext4_msg(sb, KERN_ERR, "not enough memory");
3335 #ifdef CONFIG_PROC_FS
3337 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3340 bgl_lock_init(sbi->s_blockgroup_lock);
3342 for (i = 0; i < db_count; i++) {
3343 block = descriptor_loc(sb, logical_sb_block, i);
3344 sbi->s_group_desc[i] = sb_bread(sb, block);
3345 if (!sbi->s_group_desc[i]) {
3346 ext4_msg(sb, KERN_ERR,
3347 "can't read group descriptor %d", i);
3352 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3353 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3356 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3357 if (!ext4_fill_flex_info(sb)) {
3358 ext4_msg(sb, KERN_ERR,
3359 "unable to initialize "
3360 "flex_bg meta info!");
3364 sbi->s_gdb_count = db_count;
3365 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3366 spin_lock_init(&sbi->s_next_gen_lock);
3368 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3369 ext4_count_free_blocks(sb));
3371 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3372 ext4_count_free_inodes(sb));
3375 err = percpu_counter_init(&sbi->s_dirs_counter,
3376 ext4_count_dirs(sb));
3379 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3382 ext4_msg(sb, KERN_ERR, "insufficient memory");
3386 sbi->s_stripe = ext4_get_stripe_size(sbi);
3387 sbi->s_max_writeback_mb_bump = 128;
3390 * set up enough so that it can read an inode
3392 if (!test_opt(sb, NOLOAD) &&
3393 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3394 sb->s_op = &ext4_sops;
3396 sb->s_op = &ext4_nojournal_sops;
3397 sb->s_export_op = &ext4_export_ops;
3398 sb->s_xattr = ext4_xattr_handlers;
3400 sb->s_qcop = &ext4_qctl_operations;
3401 sb->dq_op = &ext4_quota_operations;
3403 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3404 mutex_init(&sbi->s_orphan_lock);
3405 mutex_init(&sbi->s_resize_lock);
3409 needs_recovery = (es->s_last_orphan != 0 ||
3410 EXT4_HAS_INCOMPAT_FEATURE(sb,
3411 EXT4_FEATURE_INCOMPAT_RECOVER));
3414 * The first inode we look at is the journal inode. Don't try
3415 * root first: it may be modified in the journal!
3417 if (!test_opt(sb, NOLOAD) &&
3418 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3419 if (ext4_load_journal(sb, es, journal_devnum))
3421 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3422 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3423 ext4_msg(sb, KERN_ERR, "required journal recovery "
3424 "suppressed and not mounted read-only");
3425 goto failed_mount_wq;
3427 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
3428 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3429 sbi->s_journal = NULL;
3434 if (ext4_blocks_count(es) > 0xffffffffULL &&
3435 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3436 JBD2_FEATURE_INCOMPAT_64BIT)) {
3437 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3438 goto failed_mount_wq;
3441 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3442 jbd2_journal_set_features(sbi->s_journal,
3443 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3444 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3445 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3446 jbd2_journal_set_features(sbi->s_journal,
3447 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3448 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3449 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3451 jbd2_journal_clear_features(sbi->s_journal,
3452 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3453 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3456 /* We have now updated the journal if required, so we can
3457 * validate the data journaling mode. */
3458 switch (test_opt(sb, DATA_FLAGS)) {
3460 /* No mode set, assume a default based on the journal
3461 * capabilities: ORDERED_DATA if the journal can
3462 * cope, else JOURNAL_DATA
3464 if (jbd2_journal_check_available_features
3465 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3466 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3468 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3471 case EXT4_MOUNT_ORDERED_DATA:
3472 case EXT4_MOUNT_WRITEBACK_DATA:
3473 if (!jbd2_journal_check_available_features
3474 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3475 ext4_msg(sb, KERN_ERR, "Journal does not support "
3476 "requested data journaling mode");
3477 goto failed_mount_wq;
3482 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3485 * The journal may have updated the bg summary counts, so we
3486 * need to update the global counters.
3488 percpu_counter_set(&sbi->s_freeblocks_counter,
3489 ext4_count_free_blocks(sb));
3490 percpu_counter_set(&sbi->s_freeinodes_counter,
3491 ext4_count_free_inodes(sb));
3492 percpu_counter_set(&sbi->s_dirs_counter,
3493 ext4_count_dirs(sb));
3494 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3497 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3498 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3499 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3500 goto failed_mount_wq;
3504 * The jbd2_journal_load will have done any necessary log recovery,
3505 * so we can safely mount the rest of the filesystem now.
3508 root = ext4_iget(sb, EXT4_ROOT_INO);
3510 ext4_msg(sb, KERN_ERR, "get root inode failed");
3511 ret = PTR_ERR(root);
3514 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3516 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3519 sb->s_root = d_alloc_root(root);
3521 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3527 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3529 /* determine the minimum size of new large inodes, if present */
3530 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3531 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3532 EXT4_GOOD_OLD_INODE_SIZE;
3533 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3534 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3535 if (sbi->s_want_extra_isize <
3536 le16_to_cpu(es->s_want_extra_isize))
3537 sbi->s_want_extra_isize =
3538 le16_to_cpu(es->s_want_extra_isize);
3539 if (sbi->s_want_extra_isize <
3540 le16_to_cpu(es->s_min_extra_isize))
3541 sbi->s_want_extra_isize =
3542 le16_to_cpu(es->s_min_extra_isize);
3545 /* Check if enough inode space is available */
3546 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3547 sbi->s_inode_size) {
3548 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3549 EXT4_GOOD_OLD_INODE_SIZE;
3550 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3554 if (test_opt(sb, DELALLOC) &&
3555 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3556 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3557 "requested data journaling mode");
3558 clear_opt(sbi->s_mount_opt, DELALLOC);
3560 if (test_opt(sb, DIOREAD_NOLOCK)) {
3561 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3562 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3563 "option - requested data journaling mode");
3564 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3566 if (sb->s_blocksize < PAGE_SIZE) {
3567 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3568 "option - block size is too small");
3569 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3573 err = ext4_setup_system_zone(sb);
3575 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3581 err = ext4_mb_init(sb, needs_recovery);
3583 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3588 err = ext4_register_li_request(sb, first_not_zeroed);
3592 sbi->s_kobj.kset = ext4_kset;
3593 init_completion(&sbi->s_kobj_unregister);
3594 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3597 ext4_mb_release(sb);
3598 ext4_ext_release(sb);
3602 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3603 ext4_orphan_cleanup(sb, es);
3604 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3605 if (needs_recovery) {
3606 ext4_msg(sb, KERN_INFO, "recovery complete");
3607 ext4_mark_recovery_complete(sb, es);
3609 if (EXT4_SB(sb)->s_journal) {
3610 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3611 descr = " journalled data mode";
3612 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3613 descr = " ordered data mode";
3615 descr = " writeback data mode";
3617 descr = "out journal";
3619 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3620 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3621 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3623 init_timer(&sbi->s_err_report);
3624 sbi->s_err_report.function = print_daily_error_info;
3625 sbi->s_err_report.data = (unsigned long) sb;
3626 if (es->s_error_count)
3627 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3634 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3638 ext4_msg(sb, KERN_ERR, "mount failed");
3639 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3641 ext4_release_system_zone(sb);
3642 if (sbi->s_journal) {
3643 jbd2_journal_destroy(sbi->s_journal);
3644 sbi->s_journal = NULL;
3647 if (sbi->s_flex_groups) {
3648 if (is_vmalloc_addr(sbi->s_flex_groups))
3649 vfree(sbi->s_flex_groups);
3651 kfree(sbi->s_flex_groups);
3653 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3654 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3655 percpu_counter_destroy(&sbi->s_dirs_counter);
3656 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3658 for (i = 0; i < db_count; i++)
3659 brelse(sbi->s_group_desc[i]);
3660 kfree(sbi->s_group_desc);
3663 remove_proc_entry(sb->s_id, ext4_proc_root);
3666 for (i = 0; i < MAXQUOTAS; i++)
3667 kfree(sbi->s_qf_names[i]);
3669 ext4_blkdev_remove(sbi);
3672 sb->s_fs_info = NULL;
3673 kfree(sbi->s_blockgroup_lock);
3681 * Setup any per-fs journal parameters now. We'll do this both on
3682 * initial mount, once the journal has been initialised but before we've
3683 * done any recovery; and again on any subsequent remount.
3685 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3687 struct ext4_sb_info *sbi = EXT4_SB(sb);
3689 journal->j_commit_interval = sbi->s_commit_interval;
3690 journal->j_min_batch_time = sbi->s_min_batch_time;
3691 journal->j_max_batch_time = sbi->s_max_batch_time;
3693 write_lock(&journal->j_state_lock);
3694 if (test_opt(sb, BARRIER))
3695 journal->j_flags |= JBD2_BARRIER;
3697 journal->j_flags &= ~JBD2_BARRIER;
3698 if (test_opt(sb, DATA_ERR_ABORT))
3699 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3701 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3702 write_unlock(&journal->j_state_lock);
3705 static journal_t *ext4_get_journal(struct super_block *sb,
3706 unsigned int journal_inum)
3708 struct inode *journal_inode;
3711 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3713 /* First, test for the existence of a valid inode on disk. Bad
3714 * things happen if we iget() an unused inode, as the subsequent
3715 * iput() will try to delete it. */
3717 journal_inode = ext4_iget(sb, journal_inum);
3718 if (IS_ERR(journal_inode)) {
3719 ext4_msg(sb, KERN_ERR, "no journal found");
3722 if (!journal_inode->i_nlink) {
3723 make_bad_inode(journal_inode);
3724 iput(journal_inode);
3725 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3729 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3730 journal_inode, journal_inode->i_size);
3731 if (!S_ISREG(journal_inode->i_mode)) {
3732 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3733 iput(journal_inode);
3737 journal = jbd2_journal_init_inode(journal_inode);
3739 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3740 iput(journal_inode);
3743 journal->j_private = sb;
3744 ext4_init_journal_params(sb, journal);
3748 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3751 struct buffer_head *bh;
3755 int hblock, blocksize;
3756 ext4_fsblk_t sb_block;
3757 unsigned long offset;
3758 struct ext4_super_block *es;
3759 struct block_device *bdev;
3761 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3763 bdev = ext4_blkdev_get(j_dev, sb);
3767 blocksize = sb->s_blocksize;
3768 hblock = bdev_logical_block_size(bdev);
3769 if (blocksize < hblock) {
3770 ext4_msg(sb, KERN_ERR,
3771 "blocksize too small for journal device");
3775 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3776 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3777 set_blocksize(bdev, blocksize);
3778 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3779 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3780 "external journal");
3784 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3785 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3786 !(le32_to_cpu(es->s_feature_incompat) &
3787 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3788 ext4_msg(sb, KERN_ERR, "external journal has "
3794 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3795 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3800 len = ext4_blocks_count(es);
3801 start = sb_block + 1;
3802 brelse(bh); /* we're done with the superblock */
3804 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3805 start, len, blocksize);
3807 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3810 journal->j_private = sb;
3811 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3812 wait_on_buffer(journal->j_sb_buffer);
3813 if (!buffer_uptodate(journal->j_sb_buffer)) {
3814 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3817 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3818 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3819 "user (unsupported) - %d",
3820 be32_to_cpu(journal->j_superblock->s_nr_users));
3823 EXT4_SB(sb)->journal_bdev = bdev;
3824 ext4_init_journal_params(sb, journal);
3828 jbd2_journal_destroy(journal);
3830 ext4_blkdev_put(bdev);
3834 static int ext4_load_journal(struct super_block *sb,
3835 struct ext4_super_block *es,
3836 unsigned long journal_devnum)
3839 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3842 int really_read_only;
3844 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3846 if (journal_devnum &&
3847 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3848 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3849 "numbers have changed");
3850 journal_dev = new_decode_dev(journal_devnum);
3852 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3854 really_read_only = bdev_read_only(sb->s_bdev);
3857 * Are we loading a blank journal or performing recovery after a
3858 * crash? For recovery, we need to check in advance whether we
3859 * can get read-write access to the device.
3861 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3862 if (sb->s_flags & MS_RDONLY) {
3863 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3864 "required on readonly filesystem");
3865 if (really_read_only) {
3866 ext4_msg(sb, KERN_ERR, "write access "
3867 "unavailable, cannot proceed");
3870 ext4_msg(sb, KERN_INFO, "write access will "
3871 "be enabled during recovery");
3875 if (journal_inum && journal_dev) {
3876 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3877 "and inode journals!");
3882 if (!(journal = ext4_get_journal(sb, journal_inum)))
3885 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3889 if (!(journal->j_flags & JBD2_BARRIER))
3890 ext4_msg(sb, KERN_INFO, "barriers disabled");
3892 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3893 err = jbd2_journal_update_format(journal);
3895 ext4_msg(sb, KERN_ERR, "error updating journal");
3896 jbd2_journal_destroy(journal);
3901 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3902 err = jbd2_journal_wipe(journal, !really_read_only);
3904 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3906 memcpy(save, ((char *) es) +
3907 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3908 err = jbd2_journal_load(journal);
3910 memcpy(((char *) es) + EXT4_S_ERR_START,
3911 save, EXT4_S_ERR_LEN);
3916 ext4_msg(sb, KERN_ERR, "error loading journal");
3917 jbd2_journal_destroy(journal);
3921 EXT4_SB(sb)->s_journal = journal;
3922 ext4_clear_journal_err(sb, es);
3924 if (!really_read_only && journal_devnum &&
3925 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3926 es->s_journal_dev = cpu_to_le32(journal_devnum);
3928 /* Make sure we flush the recovery flag to disk. */
3929 ext4_commit_super(sb, 1);
3935 static int ext4_commit_super(struct super_block *sb, int sync)
3937 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3938 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3943 if (buffer_write_io_error(sbh)) {
3945 * Oh, dear. A previous attempt to write the
3946 * superblock failed. This could happen because the
3947 * USB device was yanked out. Or it could happen to
3948 * be a transient write error and maybe the block will
3949 * be remapped. Nothing we can do but to retry the
3950 * write and hope for the best.
3952 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3953 "superblock detected");
3954 clear_buffer_write_io_error(sbh);
3955 set_buffer_uptodate(sbh);
3958 * If the file system is mounted read-only, don't update the
3959 * superblock write time. This avoids updating the superblock
3960 * write time when we are mounting the root file system
3961 * read/only but we need to replay the journal; at that point,
3962 * for people who are east of GMT and who make their clock
3963 * tick in localtime for Windows bug-for-bug compatibility,
3964 * the clock is set in the future, and this will cause e2fsck
3965 * to complain and force a full file system check.
3967 if (!(sb->s_flags & MS_RDONLY))
3968 es->s_wtime = cpu_to_le32(get_seconds());
3969 if (sb->s_bdev->bd_part)
3970 es->s_kbytes_written =
3971 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3972 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3973 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3975 es->s_kbytes_written =
3976 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3977 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3978 &EXT4_SB(sb)->s_freeblocks_counter));
3979 es->s_free_inodes_count =
3980 cpu_to_le32(percpu_counter_sum_positive(
3981 &EXT4_SB(sb)->s_freeinodes_counter));
3983 BUFFER_TRACE(sbh, "marking dirty");
3984 mark_buffer_dirty(sbh);
3986 error = sync_dirty_buffer(sbh);
3990 error = buffer_write_io_error(sbh);
3992 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3994 clear_buffer_write_io_error(sbh);
3995 set_buffer_uptodate(sbh);
4002 * Have we just finished recovery? If so, and if we are mounting (or
4003 * remounting) the filesystem readonly, then we will end up with a
4004 * consistent fs on disk. Record that fact.
4006 static void ext4_mark_recovery_complete(struct super_block *sb,
4007 struct ext4_super_block *es)
4009 journal_t *journal = EXT4_SB(sb)->s_journal;
4011 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4012 BUG_ON(journal != NULL);
4015 jbd2_journal_lock_updates(journal);
4016 if (jbd2_journal_flush(journal) < 0)
4019 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4020 sb->s_flags & MS_RDONLY) {
4021 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4022 ext4_commit_super(sb, 1);
4026 jbd2_journal_unlock_updates(journal);
4030 * If we are mounting (or read-write remounting) a filesystem whose journal
4031 * has recorded an error from a previous lifetime, move that error to the
4032 * main filesystem now.
4034 static void ext4_clear_journal_err(struct super_block *sb,
4035 struct ext4_super_block *es)
4041 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4043 journal = EXT4_SB(sb)->s_journal;
4046 * Now check for any error status which may have been recorded in the
4047 * journal by a prior ext4_error() or ext4_abort()
4050 j_errno = jbd2_journal_errno(journal);
4054 errstr = ext4_decode_error(sb, j_errno, nbuf);
4055 ext4_warning(sb, "Filesystem error recorded "
4056 "from previous mount: %s", errstr);
4057 ext4_warning(sb, "Marking fs in need of filesystem check.");
4059 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4060 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4061 ext4_commit_super(sb, 1);
4063 jbd2_journal_clear_err(journal);
4068 * Force the running and committing transactions to commit,
4069 * and wait on the commit.
4071 int ext4_force_commit(struct super_block *sb)
4076 if (sb->s_flags & MS_RDONLY)
4079 journal = EXT4_SB(sb)->s_journal;
4081 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4082 ret = ext4_journal_force_commit(journal);
4088 static void ext4_write_super(struct super_block *sb)
4091 ext4_commit_super(sb, 1);
4095 static int ext4_sync_fs(struct super_block *sb, int wait)
4099 struct ext4_sb_info *sbi = EXT4_SB(sb);
4101 trace_ext4_sync_fs(sb, wait);
4102 flush_workqueue(sbi->dio_unwritten_wq);
4103 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4105 jbd2_log_wait_commit(sbi->s_journal, target);
4111 * LVM calls this function before a (read-only) snapshot is created. This
4112 * gives us a chance to flush the journal completely and mark the fs clean.
4114 static int ext4_freeze(struct super_block *sb)
4119 if (sb->s_flags & MS_RDONLY)
4122 journal = EXT4_SB(sb)->s_journal;
4124 /* Now we set up the journal barrier. */
4125 jbd2_journal_lock_updates(journal);
4128 * Don't clear the needs_recovery flag if we failed to flush
4131 error = jbd2_journal_flush(journal);
4135 /* Journal blocked and flushed, clear needs_recovery flag. */
4136 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4137 error = ext4_commit_super(sb, 1);
4139 /* we rely on s_frozen to stop further updates */
4140 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4145 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4146 * flag here, even though the filesystem is not technically dirty yet.
4148 static int ext4_unfreeze(struct super_block *sb)
4150 if (sb->s_flags & MS_RDONLY)
4154 /* Reset the needs_recovery flag before the fs is unlocked. */
4155 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4156 ext4_commit_super(sb, 1);
4161 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4163 struct ext4_super_block *es;
4164 struct ext4_sb_info *sbi = EXT4_SB(sb);
4165 ext4_fsblk_t n_blocks_count = 0;
4166 unsigned long old_sb_flags;
4167 struct ext4_mount_options old_opts;
4168 int enable_quota = 0;
4170 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4175 char *orig_data = kstrdup(data, GFP_KERNEL);
4177 /* Store the original options */
4179 old_sb_flags = sb->s_flags;
4180 old_opts.s_mount_opt = sbi->s_mount_opt;
4181 old_opts.s_resuid = sbi->s_resuid;
4182 old_opts.s_resgid = sbi->s_resgid;
4183 old_opts.s_commit_interval = sbi->s_commit_interval;
4184 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4185 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4187 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4188 for (i = 0; i < MAXQUOTAS; i++)
4189 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4191 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4192 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4195 * Allow the "check" option to be passed as a remount option.
4197 if (!parse_options(data, sb, NULL, &journal_ioprio,
4198 &n_blocks_count, 1)) {
4203 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4204 ext4_abort(sb, "Abort forced by user");
4206 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4207 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4211 if (sbi->s_journal) {
4212 ext4_init_journal_params(sb, sbi->s_journal);
4213 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4216 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4217 n_blocks_count > ext4_blocks_count(es)) {
4218 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4223 if (*flags & MS_RDONLY) {
4224 err = dquot_suspend(sb, -1);
4229 * First of all, the unconditional stuff we have to do
4230 * to disable replay of the journal when we next remount
4232 sb->s_flags |= MS_RDONLY;
4235 * OK, test if we are remounting a valid rw partition
4236 * readonly, and if so set the rdonly flag and then
4237 * mark the partition as valid again.
4239 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4240 (sbi->s_mount_state & EXT4_VALID_FS))
4241 es->s_state = cpu_to_le16(sbi->s_mount_state);
4244 ext4_mark_recovery_complete(sb, es);
4246 /* Make sure we can mount this feature set readwrite */
4247 if (!ext4_feature_set_ok(sb, 0)) {
4252 * Make sure the group descriptor checksums
4253 * are sane. If they aren't, refuse to remount r/w.
4255 for (g = 0; g < sbi->s_groups_count; g++) {
4256 struct ext4_group_desc *gdp =
4257 ext4_get_group_desc(sb, g, NULL);
4259 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4260 ext4_msg(sb, KERN_ERR,
4261 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4262 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4263 le16_to_cpu(gdp->bg_checksum));
4270 * If we have an unprocessed orphan list hanging
4271 * around from a previously readonly bdev mount,
4272 * require a full umount/remount for now.
4274 if (es->s_last_orphan) {
4275 ext4_msg(sb, KERN_WARNING, "Couldn't "
4276 "remount RDWR because of unprocessed "
4277 "orphan inode list. Please "
4278 "umount/remount instead");
4284 * Mounting a RDONLY partition read-write, so reread
4285 * and store the current valid flag. (It may have
4286 * been changed by e2fsck since we originally mounted
4290 ext4_clear_journal_err(sb, es);
4291 sbi->s_mount_state = le16_to_cpu(es->s_state);
4292 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4294 if (!ext4_setup_super(sb, es, 0))
4295 sb->s_flags &= ~MS_RDONLY;
4301 * Reinitialize lazy itable initialization thread based on
4304 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4305 ext4_unregister_li_request(sb);
4307 ext4_group_t first_not_zeroed;
4308 first_not_zeroed = ext4_has_uninit_itable(sb);
4309 ext4_register_li_request(sb, first_not_zeroed);
4312 ext4_setup_system_zone(sb);
4313 if (sbi->s_journal == NULL)
4314 ext4_commit_super(sb, 1);
4317 /* Release old quota file names */
4318 for (i = 0; i < MAXQUOTAS; i++)
4319 if (old_opts.s_qf_names[i] &&
4320 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4321 kfree(old_opts.s_qf_names[i]);
4325 dquot_resume(sb, -1);
4327 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4332 sb->s_flags = old_sb_flags;
4333 sbi->s_mount_opt = old_opts.s_mount_opt;
4334 sbi->s_resuid = old_opts.s_resuid;
4335 sbi->s_resgid = old_opts.s_resgid;
4336 sbi->s_commit_interval = old_opts.s_commit_interval;
4337 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4338 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4340 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4341 for (i = 0; i < MAXQUOTAS; i++) {
4342 if (sbi->s_qf_names[i] &&
4343 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4344 kfree(sbi->s_qf_names[i]);
4345 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4353 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4355 struct super_block *sb = dentry->d_sb;
4356 struct ext4_sb_info *sbi = EXT4_SB(sb);
4357 struct ext4_super_block *es = sbi->s_es;
4360 if (test_opt(sb, MINIX_DF)) {
4361 sbi->s_overhead_last = 0;
4362 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4363 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4364 ext4_fsblk_t overhead = 0;
4367 * Compute the overhead (FS structures). This is constant
4368 * for a given filesystem unless the number of block groups
4369 * changes so we cache the previous value until it does.
4373 * All of the blocks before first_data_block are
4376 overhead = le32_to_cpu(es->s_first_data_block);
4379 * Add the overhead attributed to the superblock and
4380 * block group descriptors. If the sparse superblocks
4381 * feature is turned on, then not all groups have this.
4383 for (i = 0; i < ngroups; i++) {
4384 overhead += ext4_bg_has_super(sb, i) +
4385 ext4_bg_num_gdb(sb, i);
4390 * Every block group has an inode bitmap, a block
4391 * bitmap, and an inode table.
4393 overhead += ngroups * (2 + sbi->s_itb_per_group);
4394 sbi->s_overhead_last = overhead;
4396 sbi->s_blocks_last = ext4_blocks_count(es);
4399 buf->f_type = EXT4_SUPER_MAGIC;
4400 buf->f_bsize = sb->s_blocksize;
4401 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4402 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4403 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4404 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4405 if (buf->f_bfree < ext4_r_blocks_count(es))
4407 buf->f_files = le32_to_cpu(es->s_inodes_count);
4408 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4409 buf->f_namelen = EXT4_NAME_LEN;
4410 fsid = le64_to_cpup((void *)es->s_uuid) ^
4411 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4412 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4413 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4418 /* Helper function for writing quotas on sync - we need to start transaction
4419 * before quota file is locked for write. Otherwise the are possible deadlocks:
4420 * Process 1 Process 2
4421 * ext4_create() quota_sync()
4422 * jbd2_journal_start() write_dquot()
4423 * dquot_initialize() down(dqio_mutex)
4424 * down(dqio_mutex) jbd2_journal_start()
4430 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4432 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4435 static int ext4_write_dquot(struct dquot *dquot)
4439 struct inode *inode;
4441 inode = dquot_to_inode(dquot);
4442 handle = ext4_journal_start(inode,
4443 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4445 return PTR_ERR(handle);
4446 ret = dquot_commit(dquot);
4447 err = ext4_journal_stop(handle);
4453 static int ext4_acquire_dquot(struct dquot *dquot)
4458 handle = ext4_journal_start(dquot_to_inode(dquot),
4459 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4461 return PTR_ERR(handle);
4462 ret = dquot_acquire(dquot);
4463 err = ext4_journal_stop(handle);
4469 static int ext4_release_dquot(struct dquot *dquot)
4474 handle = ext4_journal_start(dquot_to_inode(dquot),
4475 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4476 if (IS_ERR(handle)) {
4477 /* Release dquot anyway to avoid endless cycle in dqput() */
4478 dquot_release(dquot);
4479 return PTR_ERR(handle);
4481 ret = dquot_release(dquot);
4482 err = ext4_journal_stop(handle);
4488 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4490 /* Are we journaling quotas? */
4491 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4492 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4493 dquot_mark_dquot_dirty(dquot);
4494 return ext4_write_dquot(dquot);
4496 return dquot_mark_dquot_dirty(dquot);
4500 static int ext4_write_info(struct super_block *sb, int type)
4505 /* Data block + inode block */
4506 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4508 return PTR_ERR(handle);
4509 ret = dquot_commit_info(sb, type);
4510 err = ext4_journal_stop(handle);
4517 * Turn on quotas during mount time - we need to find
4518 * the quota file and such...
4520 static int ext4_quota_on_mount(struct super_block *sb, int type)
4522 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4523 EXT4_SB(sb)->s_jquota_fmt, type);
4527 * Standard function to be called on quota_on
4529 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4535 if (!test_opt(sb, QUOTA))
4538 err = kern_path(name, LOOKUP_FOLLOW, &path);
4542 /* Quotafile not on the same filesystem? */
4543 if (path.mnt->mnt_sb != sb) {
4547 /* Journaling quota? */
4548 if (EXT4_SB(sb)->s_qf_names[type]) {
4549 /* Quotafile not in fs root? */
4550 if (path.dentry->d_parent != sb->s_root)
4551 ext4_msg(sb, KERN_WARNING,
4552 "Quota file not on filesystem root. "
4553 "Journaled quota will not work");
4557 * When we journal data on quota file, we have to flush journal to see
4558 * all updates to the file when we bypass pagecache...
4560 if (EXT4_SB(sb)->s_journal &&
4561 ext4_should_journal_data(path.dentry->d_inode)) {
4563 * We don't need to lock updates but journal_flush() could
4564 * otherwise be livelocked...
4566 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4567 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4568 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4575 err = dquot_quota_on_path(sb, type, format_id, &path);
4580 static int ext4_quota_off(struct super_block *sb, int type)
4582 /* Force all delayed allocation blocks to be allocated.
4583 * Caller already holds s_umount sem */
4584 if (test_opt(sb, DELALLOC))
4585 sync_filesystem(sb);
4587 return dquot_quota_off(sb, type);
4590 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4591 * acquiring the locks... As quota files are never truncated and quota code
4592 * itself serializes the operations (and noone else should touch the files)
4593 * we don't have to be afraid of races */
4594 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4595 size_t len, loff_t off)
4597 struct inode *inode = sb_dqopt(sb)->files[type];
4598 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4600 int offset = off & (sb->s_blocksize - 1);
4603 struct buffer_head *bh;
4604 loff_t i_size = i_size_read(inode);
4608 if (off+len > i_size)
4611 while (toread > 0) {
4612 tocopy = sb->s_blocksize - offset < toread ?
4613 sb->s_blocksize - offset : toread;
4614 bh = ext4_bread(NULL, inode, blk, 0, &err);
4617 if (!bh) /* A hole? */
4618 memset(data, 0, tocopy);
4620 memcpy(data, bh->b_data+offset, tocopy);
4630 /* Write to quotafile (we know the transaction is already started and has
4631 * enough credits) */
4632 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4633 const char *data, size_t len, loff_t off)
4635 struct inode *inode = sb_dqopt(sb)->files[type];
4636 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4638 int offset = off & (sb->s_blocksize - 1);
4639 struct buffer_head *bh;
4640 handle_t *handle = journal_current_handle();
4642 if (EXT4_SB(sb)->s_journal && !handle) {
4643 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4644 " cancelled because transaction is not started",
4645 (unsigned long long)off, (unsigned long long)len);
4649 * Since we account only one data block in transaction credits,
4650 * then it is impossible to cross a block boundary.
4652 if (sb->s_blocksize - offset < len) {
4653 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4654 " cancelled because not block aligned",
4655 (unsigned long long)off, (unsigned long long)len);
4659 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4660 bh = ext4_bread(handle, inode, blk, 1, &err);
4663 err = ext4_journal_get_write_access(handle, bh);
4669 memcpy(bh->b_data+offset, data, len);
4670 flush_dcache_page(bh->b_page);
4672 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4676 mutex_unlock(&inode->i_mutex);
4679 if (inode->i_size < off + len) {
4680 i_size_write(inode, off + len);
4681 EXT4_I(inode)->i_disksize = inode->i_size;
4683 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4684 ext4_mark_inode_dirty(handle, inode);
4685 mutex_unlock(&inode->i_mutex);
4691 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4692 const char *dev_name, void *data)
4694 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4697 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4698 static struct file_system_type ext2_fs_type = {
4699 .owner = THIS_MODULE,
4701 .mount = ext4_mount,
4702 .kill_sb = kill_block_super,
4703 .fs_flags = FS_REQUIRES_DEV,
4706 static inline void register_as_ext2(void)
4708 int err = register_filesystem(&ext2_fs_type);
4711 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4714 static inline void unregister_as_ext2(void)
4716 unregister_filesystem(&ext2_fs_type);
4718 MODULE_ALIAS("ext2");
4720 static inline void register_as_ext2(void) { }
4721 static inline void unregister_as_ext2(void) { }
4724 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4725 static inline void register_as_ext3(void)
4727 int err = register_filesystem(&ext3_fs_type);
4730 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4733 static inline void unregister_as_ext3(void)
4735 unregister_filesystem(&ext3_fs_type);
4737 MODULE_ALIAS("ext3");
4739 static inline void register_as_ext3(void) { }
4740 static inline void unregister_as_ext3(void) { }
4743 static struct file_system_type ext4_fs_type = {
4744 .owner = THIS_MODULE,
4746 .mount = ext4_mount,
4747 .kill_sb = kill_block_super,
4748 .fs_flags = FS_REQUIRES_DEV,
4751 int __init ext4_init_feat_adverts(void)
4753 struct ext4_features *ef;
4756 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4760 ef->f_kobj.kset = ext4_kset;
4761 init_completion(&ef->f_kobj_unregister);
4762 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4775 static int __init ext4_init_fs(void)
4779 ext4_check_flag_values();
4780 err = ext4_init_pageio();
4783 err = ext4_init_system_zone();
4786 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4789 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4791 err = ext4_init_feat_adverts();
4793 err = ext4_init_mballoc();
4797 err = ext4_init_xattr();
4800 err = init_inodecache();
4805 err = register_filesystem(&ext4_fs_type);
4809 ext4_li_info = NULL;
4810 mutex_init(&ext4_li_mtx);
4813 unregister_as_ext2();
4814 unregister_as_ext3();
4815 destroy_inodecache();
4819 ext4_exit_mballoc();
4822 remove_proc_entry("fs/ext4", NULL);
4823 kset_unregister(ext4_kset);
4825 ext4_exit_system_zone();
4831 static void __exit ext4_exit_fs(void)
4833 ext4_destroy_lazyinit_thread();
4834 unregister_as_ext2();
4835 unregister_as_ext3();
4836 unregister_filesystem(&ext4_fs_type);
4837 destroy_inodecache();
4839 ext4_exit_mballoc();
4840 remove_proc_entry("fs/ext4", NULL);
4841 kset_unregister(ext4_kset);
4842 ext4_exit_system_zone();
4846 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4847 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4848 MODULE_LICENSE("GPL");
4849 module_init(ext4_init_fs)
4850 module_exit(ext4_exit_fs)