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, ...)
391 struct va_format vaf;
397 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
398 sb->s_id, function, line, current->comm, &vaf);
401 ext4_handle_error(sb);
404 void ext4_error_inode(struct inode *inode, const char *function,
405 unsigned int line, ext4_fsblk_t block,
406 const char *fmt, ...)
409 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
411 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
412 es->s_last_error_block = cpu_to_le64(block);
413 save_error_info(inode->i_sb, function, line);
415 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
416 inode->i_sb->s_id, function, line, inode->i_ino);
418 printk("block %llu: ", block);
419 printk("comm %s: ", current->comm);
424 ext4_handle_error(inode->i_sb);
427 void ext4_error_file(struct file *file, const char *function,
428 unsigned int line, const char *fmt, ...)
431 struct ext4_super_block *es;
432 struct inode *inode = file->f_dentry->d_inode;
433 char pathname[80], *path;
435 es = EXT4_SB(inode->i_sb)->s_es;
436 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
437 save_error_info(inode->i_sb, function, line);
439 path = d_path(&(file->f_path), pathname, sizeof(pathname));
443 "EXT4-fs error (device %s): %s:%d: inode #%lu "
444 "(comm %s path %s): ",
445 inode->i_sb->s_id, function, line, inode->i_ino,
446 current->comm, path);
451 ext4_handle_error(inode->i_sb);
454 static const char *ext4_decode_error(struct super_block *sb, int errno,
461 errstr = "IO failure";
464 errstr = "Out of memory";
467 if (!sb || (EXT4_SB(sb)->s_journal &&
468 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
469 errstr = "Journal has aborted";
471 errstr = "Readonly filesystem";
474 /* If the caller passed in an extra buffer for unknown
475 * errors, textualise them now. Else we just return
478 /* Check for truncated error codes... */
479 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
488 /* __ext4_std_error decodes expected errors from journaling functions
489 * automatically and invokes the appropriate error response. */
491 void __ext4_std_error(struct super_block *sb, const char *function,
492 unsigned int line, int errno)
497 /* Special case: if the error is EROFS, and we're not already
498 * inside a transaction, then there's really no point in logging
500 if (errno == -EROFS && journal_current_handle() == NULL &&
501 (sb->s_flags & MS_RDONLY))
504 errstr = ext4_decode_error(sb, errno, nbuf);
505 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
506 sb->s_id, function, line, errstr);
507 save_error_info(sb, function, line);
509 ext4_handle_error(sb);
513 * ext4_abort is a much stronger failure handler than ext4_error. The
514 * abort function may be used to deal with unrecoverable failures such
515 * as journal IO errors or ENOMEM at a critical moment in log management.
517 * We unconditionally force the filesystem into an ABORT|READONLY state,
518 * unless the error response on the fs has been set to panic in which
519 * case we take the easy way out and panic immediately.
522 void __ext4_abort(struct super_block *sb, const char *function,
523 unsigned int line, const char *fmt, ...)
527 save_error_info(sb, function, line);
529 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 if ((sb->s_flags & MS_RDONLY) == 0) {
536 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
537 sb->s_flags |= MS_RDONLY;
538 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
539 if (EXT4_SB(sb)->s_journal)
540 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
541 save_error_info(sb, function, line);
543 if (test_opt(sb, ERRORS_PANIC))
544 panic("EXT4-fs panic from previous error\n");
547 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
549 struct va_format vaf;
555 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
559 void __ext4_warning(struct super_block *sb, const char *function,
560 unsigned int line, const char *fmt, ...)
562 struct va_format vaf;
568 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
569 sb->s_id, function, line, &vaf);
573 void __ext4_grp_locked_error(const char *function, unsigned int line,
574 struct super_block *sb, ext4_group_t grp,
575 unsigned long ino, ext4_fsblk_t block,
576 const char *fmt, ...)
580 struct va_format vaf;
582 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
584 es->s_last_error_ino = cpu_to_le32(ino);
585 es->s_last_error_block = cpu_to_le64(block);
586 __save_error_info(sb, function, line);
592 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
593 sb->s_id, function, line, grp);
595 printk(KERN_CONT "inode %lu: ", ino);
597 printk(KERN_CONT "block %llu:", (unsigned long long) block);
598 printk(KERN_CONT "%pV\n", &vaf);
601 if (test_opt(sb, ERRORS_CONT)) {
602 ext4_commit_super(sb, 0);
606 ext4_unlock_group(sb, grp);
607 ext4_handle_error(sb);
609 * We only get here in the ERRORS_RO case; relocking the group
610 * may be dangerous, but nothing bad will happen since the
611 * filesystem will have already been marked read/only and the
612 * journal has been aborted. We return 1 as a hint to callers
613 * who might what to use the return value from
614 * ext4_grp_locked_error() to distinguish beween the
615 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
616 * aggressively from the ext4 function in question, with a
617 * more appropriate error code.
619 ext4_lock_group(sb, grp);
623 void ext4_update_dynamic_rev(struct super_block *sb)
625 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
627 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
631 "updating to rev %d because of new feature flag, "
632 "running e2fsck is recommended",
635 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
636 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
637 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
638 /* leave es->s_feature_*compat flags alone */
639 /* es->s_uuid will be set by e2fsck if empty */
642 * The rest of the superblock fields should be zero, and if not it
643 * means they are likely already in use, so leave them alone. We
644 * can leave it up to e2fsck to clean up any inconsistencies there.
649 * Open the external journal device
651 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
653 struct block_device *bdev;
654 char b[BDEVNAME_SIZE];
656 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
662 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
663 __bdevname(dev, b), PTR_ERR(bdev));
668 * Release the journal device
670 static int ext4_blkdev_put(struct block_device *bdev)
673 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
676 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
678 struct block_device *bdev;
681 bdev = sbi->journal_bdev;
683 ret = ext4_blkdev_put(bdev);
684 sbi->journal_bdev = NULL;
689 static inline struct inode *orphan_list_entry(struct list_head *l)
691 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
694 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
698 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
699 le32_to_cpu(sbi->s_es->s_last_orphan));
701 printk(KERN_ERR "sb_info orphan list:\n");
702 list_for_each(l, &sbi->s_orphan) {
703 struct inode *inode = orphan_list_entry(l);
705 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
706 inode->i_sb->s_id, inode->i_ino, inode,
707 inode->i_mode, inode->i_nlink,
712 static void ext4_put_super(struct super_block *sb)
714 struct ext4_sb_info *sbi = EXT4_SB(sb);
715 struct ext4_super_block *es = sbi->s_es;
718 ext4_unregister_li_request(sb);
719 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
721 flush_workqueue(sbi->dio_unwritten_wq);
722 destroy_workqueue(sbi->dio_unwritten_wq);
726 ext4_commit_super(sb, 1);
728 if (sbi->s_journal) {
729 err = jbd2_journal_destroy(sbi->s_journal);
730 sbi->s_journal = NULL;
732 ext4_abort(sb, "Couldn't clean up the journal");
735 del_timer(&sbi->s_err_report);
736 ext4_release_system_zone(sb);
738 ext4_ext_release(sb);
739 ext4_xattr_put_super(sb);
741 if (!(sb->s_flags & MS_RDONLY)) {
742 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
743 es->s_state = cpu_to_le16(sbi->s_mount_state);
744 ext4_commit_super(sb, 1);
747 remove_proc_entry(sb->s_id, ext4_proc_root);
749 kobject_del(&sbi->s_kobj);
751 for (i = 0; i < sbi->s_gdb_count; i++)
752 brelse(sbi->s_group_desc[i]);
753 kfree(sbi->s_group_desc);
754 if (is_vmalloc_addr(sbi->s_flex_groups))
755 vfree(sbi->s_flex_groups);
757 kfree(sbi->s_flex_groups);
758 percpu_counter_destroy(&sbi->s_freeblocks_counter);
759 percpu_counter_destroy(&sbi->s_freeinodes_counter);
760 percpu_counter_destroy(&sbi->s_dirs_counter);
761 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
764 for (i = 0; i < MAXQUOTAS; i++)
765 kfree(sbi->s_qf_names[i]);
768 /* Debugging code just in case the in-memory inode orphan list
769 * isn't empty. The on-disk one can be non-empty if we've
770 * detected an error and taken the fs readonly, but the
771 * in-memory list had better be clean by this point. */
772 if (!list_empty(&sbi->s_orphan))
773 dump_orphan_list(sb, sbi);
774 J_ASSERT(list_empty(&sbi->s_orphan));
776 invalidate_bdev(sb->s_bdev);
777 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
779 * Invalidate the journal device's buffers. We don't want them
780 * floating about in memory - the physical journal device may
781 * hotswapped, and it breaks the `ro-after' testing code.
783 sync_blockdev(sbi->journal_bdev);
784 invalidate_bdev(sbi->journal_bdev);
785 ext4_blkdev_remove(sbi);
787 sb->s_fs_info = NULL;
789 * Now that we are completely done shutting down the
790 * superblock, we need to actually destroy the kobject.
793 kobject_put(&sbi->s_kobj);
794 wait_for_completion(&sbi->s_kobj_unregister);
795 kfree(sbi->s_blockgroup_lock);
799 static struct kmem_cache *ext4_inode_cachep;
802 * Called inside transaction, so use GFP_NOFS
804 static struct inode *ext4_alloc_inode(struct super_block *sb)
806 struct ext4_inode_info *ei;
808 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
812 ei->vfs_inode.i_version = 1;
813 ei->vfs_inode.i_data.writeback_index = 0;
814 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
815 INIT_LIST_HEAD(&ei->i_prealloc_list);
816 spin_lock_init(&ei->i_prealloc_lock);
818 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
819 * therefore it can be null here. Don't check it, just initialize
822 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
823 ei->i_reserved_data_blocks = 0;
824 ei->i_reserved_meta_blocks = 0;
825 ei->i_allocated_meta_blocks = 0;
826 ei->i_da_metadata_calc_len = 0;
827 ei->i_delalloc_reserved_flag = 0;
828 spin_lock_init(&(ei->i_block_reservation_lock));
830 ei->i_reserved_quota = 0;
832 INIT_LIST_HEAD(&ei->i_completed_io_list);
833 spin_lock_init(&ei->i_completed_io_lock);
834 ei->cur_aio_dio = NULL;
836 ei->i_datasync_tid = 0;
837 atomic_set(&ei->i_ioend_count, 0);
839 return &ei->vfs_inode;
842 static int ext4_drop_inode(struct inode *inode)
844 int drop = generic_drop_inode(inode);
846 trace_ext4_drop_inode(inode, drop);
850 static void ext4_destroy_inode(struct inode *inode)
852 ext4_ioend_wait(inode);
853 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
854 ext4_msg(inode->i_sb, KERN_ERR,
855 "Inode %lu (%p): orphan list check failed!",
856 inode->i_ino, EXT4_I(inode));
857 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
858 EXT4_I(inode), sizeof(struct ext4_inode_info),
862 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
865 static void init_once(void *foo)
867 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
869 INIT_LIST_HEAD(&ei->i_orphan);
870 #ifdef CONFIG_EXT4_FS_XATTR
871 init_rwsem(&ei->xattr_sem);
873 init_rwsem(&ei->i_data_sem);
874 inode_init_once(&ei->vfs_inode);
877 static int init_inodecache(void)
879 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
880 sizeof(struct ext4_inode_info),
881 0, (SLAB_RECLAIM_ACCOUNT|
884 if (ext4_inode_cachep == NULL)
889 static void destroy_inodecache(void)
891 kmem_cache_destroy(ext4_inode_cachep);
894 void ext4_clear_inode(struct inode *inode)
896 invalidate_inode_buffers(inode);
897 end_writeback(inode);
899 ext4_discard_preallocations(inode);
900 if (EXT4_JOURNAL(inode))
901 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
902 &EXT4_I(inode)->jinode);
905 static inline void ext4_show_quota_options(struct seq_file *seq,
906 struct super_block *sb)
908 #if defined(CONFIG_QUOTA)
909 struct ext4_sb_info *sbi = EXT4_SB(sb);
911 if (sbi->s_jquota_fmt) {
914 switch (sbi->s_jquota_fmt) {
925 seq_printf(seq, ",jqfmt=%s", fmtname);
928 if (sbi->s_qf_names[USRQUOTA])
929 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
931 if (sbi->s_qf_names[GRPQUOTA])
932 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
934 if (test_opt(sb, USRQUOTA))
935 seq_puts(seq, ",usrquota");
937 if (test_opt(sb, GRPQUOTA))
938 seq_puts(seq, ",grpquota");
944 * - it's set to a non-default value OR
945 * - if the per-sb default is different from the global default
947 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
950 unsigned long def_mount_opts;
951 struct super_block *sb = vfs->mnt_sb;
952 struct ext4_sb_info *sbi = EXT4_SB(sb);
953 struct ext4_super_block *es = sbi->s_es;
955 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
956 def_errors = le16_to_cpu(es->s_errors);
958 if (sbi->s_sb_block != 1)
959 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
960 if (test_opt(sb, MINIX_DF))
961 seq_puts(seq, ",minixdf");
962 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
963 seq_puts(seq, ",grpid");
964 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
965 seq_puts(seq, ",nogrpid");
966 if (sbi->s_resuid != EXT4_DEF_RESUID ||
967 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
968 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
970 if (sbi->s_resgid != EXT4_DEF_RESGID ||
971 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
972 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
974 if (test_opt(sb, ERRORS_RO)) {
975 if (def_errors == EXT4_ERRORS_PANIC ||
976 def_errors == EXT4_ERRORS_CONTINUE) {
977 seq_puts(seq, ",errors=remount-ro");
980 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
981 seq_puts(seq, ",errors=continue");
982 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
983 seq_puts(seq, ",errors=panic");
984 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
985 seq_puts(seq, ",nouid32");
986 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
987 seq_puts(seq, ",debug");
988 if (test_opt(sb, OLDALLOC))
989 seq_puts(seq, ",oldalloc");
990 #ifdef CONFIG_EXT4_FS_XATTR
991 if (test_opt(sb, XATTR_USER) &&
992 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
993 seq_puts(seq, ",user_xattr");
994 if (!test_opt(sb, XATTR_USER) &&
995 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
996 seq_puts(seq, ",nouser_xattr");
999 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1000 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1001 seq_puts(seq, ",acl");
1002 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1003 seq_puts(seq, ",noacl");
1005 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1006 seq_printf(seq, ",commit=%u",
1007 (unsigned) (sbi->s_commit_interval / HZ));
1009 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1010 seq_printf(seq, ",min_batch_time=%u",
1011 (unsigned) sbi->s_min_batch_time);
1013 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1014 seq_printf(seq, ",max_batch_time=%u",
1015 (unsigned) sbi->s_min_batch_time);
1019 * We're changing the default of barrier mount option, so
1020 * let's always display its mount state so it's clear what its
1023 seq_puts(seq, ",barrier=");
1024 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1025 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1026 seq_puts(seq, ",journal_async_commit");
1027 else if (test_opt(sb, JOURNAL_CHECKSUM))
1028 seq_puts(seq, ",journal_checksum");
1029 if (test_opt(sb, I_VERSION))
1030 seq_puts(seq, ",i_version");
1031 if (!test_opt(sb, DELALLOC) &&
1032 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1033 seq_puts(seq, ",nodelalloc");
1035 if (test_opt(sb, MBLK_IO_SUBMIT))
1036 seq_puts(seq, ",mblk_io_submit");
1038 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1040 * journal mode get enabled in different ways
1041 * So just print the value even if we didn't specify it
1043 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1044 seq_puts(seq, ",data=journal");
1045 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1046 seq_puts(seq, ",data=ordered");
1047 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1048 seq_puts(seq, ",data=writeback");
1050 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1051 seq_printf(seq, ",inode_readahead_blks=%u",
1052 sbi->s_inode_readahead_blks);
1054 if (test_opt(sb, DATA_ERR_ABORT))
1055 seq_puts(seq, ",data_err=abort");
1057 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1058 seq_puts(seq, ",noauto_da_alloc");
1060 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1061 seq_puts(seq, ",discard");
1063 if (test_opt(sb, NOLOAD))
1064 seq_puts(seq, ",norecovery");
1066 if (test_opt(sb, DIOREAD_NOLOCK))
1067 seq_puts(seq, ",dioread_nolock");
1069 if (test_opt(sb, BLOCK_VALIDITY) &&
1070 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1071 seq_puts(seq, ",block_validity");
1073 if (!test_opt(sb, INIT_INODE_TABLE))
1074 seq_puts(seq, ",noinit_inode_table");
1075 else if (sbi->s_li_wait_mult)
1076 seq_printf(seq, ",init_inode_table=%u",
1077 (unsigned) sbi->s_li_wait_mult);
1079 ext4_show_quota_options(seq, sb);
1084 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1085 u64 ino, u32 generation)
1087 struct inode *inode;
1089 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1090 return ERR_PTR(-ESTALE);
1091 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1092 return ERR_PTR(-ESTALE);
1094 /* iget isn't really right if the inode is currently unallocated!!
1096 * ext4_read_inode will return a bad_inode if the inode had been
1097 * deleted, so we should be safe.
1099 * Currently we don't know the generation for parent directory, so
1100 * a generation of 0 means "accept any"
1102 inode = ext4_iget(sb, ino);
1104 return ERR_CAST(inode);
1105 if (generation && inode->i_generation != generation) {
1107 return ERR_PTR(-ESTALE);
1113 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1114 int fh_len, int fh_type)
1116 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1117 ext4_nfs_get_inode);
1120 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1121 int fh_len, int fh_type)
1123 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1124 ext4_nfs_get_inode);
1128 * Try to release metadata pages (indirect blocks, directories) which are
1129 * mapped via the block device. Since these pages could have journal heads
1130 * which would prevent try_to_free_buffers() from freeing them, we must use
1131 * jbd2 layer's try_to_free_buffers() function to release them.
1133 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1136 journal_t *journal = EXT4_SB(sb)->s_journal;
1138 WARN_ON(PageChecked(page));
1139 if (!page_has_buffers(page))
1142 return jbd2_journal_try_to_free_buffers(journal, page,
1143 wait & ~__GFP_WAIT);
1144 return try_to_free_buffers(page);
1148 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1149 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1151 static int ext4_write_dquot(struct dquot *dquot);
1152 static int ext4_acquire_dquot(struct dquot *dquot);
1153 static int ext4_release_dquot(struct dquot *dquot);
1154 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1155 static int ext4_write_info(struct super_block *sb, int type);
1156 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1158 static int ext4_quota_off(struct super_block *sb, int type);
1159 static int ext4_quota_on_mount(struct super_block *sb, int type);
1160 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1161 size_t len, loff_t off);
1162 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1163 const char *data, size_t len, loff_t off);
1165 static const struct dquot_operations ext4_quota_operations = {
1167 .get_reserved_space = ext4_get_reserved_space,
1169 .write_dquot = ext4_write_dquot,
1170 .acquire_dquot = ext4_acquire_dquot,
1171 .release_dquot = ext4_release_dquot,
1172 .mark_dirty = ext4_mark_dquot_dirty,
1173 .write_info = ext4_write_info,
1174 .alloc_dquot = dquot_alloc,
1175 .destroy_dquot = dquot_destroy,
1178 static const struct quotactl_ops ext4_qctl_operations = {
1179 .quota_on = ext4_quota_on,
1180 .quota_off = ext4_quota_off,
1181 .quota_sync = dquot_quota_sync,
1182 .get_info = dquot_get_dqinfo,
1183 .set_info = dquot_set_dqinfo,
1184 .get_dqblk = dquot_get_dqblk,
1185 .set_dqblk = dquot_set_dqblk
1189 static const struct super_operations ext4_sops = {
1190 .alloc_inode = ext4_alloc_inode,
1191 .destroy_inode = ext4_destroy_inode,
1192 .write_inode = ext4_write_inode,
1193 .dirty_inode = ext4_dirty_inode,
1194 .drop_inode = ext4_drop_inode,
1195 .evict_inode = ext4_evict_inode,
1196 .put_super = ext4_put_super,
1197 .sync_fs = ext4_sync_fs,
1198 .freeze_fs = ext4_freeze,
1199 .unfreeze_fs = ext4_unfreeze,
1200 .statfs = ext4_statfs,
1201 .remount_fs = ext4_remount,
1202 .show_options = ext4_show_options,
1204 .quota_read = ext4_quota_read,
1205 .quota_write = ext4_quota_write,
1207 .bdev_try_to_free_page = bdev_try_to_free_page,
1210 static const struct super_operations ext4_nojournal_sops = {
1211 .alloc_inode = ext4_alloc_inode,
1212 .destroy_inode = ext4_destroy_inode,
1213 .write_inode = ext4_write_inode,
1214 .dirty_inode = ext4_dirty_inode,
1215 .drop_inode = ext4_drop_inode,
1216 .evict_inode = ext4_evict_inode,
1217 .write_super = ext4_write_super,
1218 .put_super = ext4_put_super,
1219 .statfs = ext4_statfs,
1220 .remount_fs = ext4_remount,
1221 .show_options = ext4_show_options,
1223 .quota_read = ext4_quota_read,
1224 .quota_write = ext4_quota_write,
1226 .bdev_try_to_free_page = bdev_try_to_free_page,
1229 static const struct export_operations ext4_export_ops = {
1230 .fh_to_dentry = ext4_fh_to_dentry,
1231 .fh_to_parent = ext4_fh_to_parent,
1232 .get_parent = ext4_get_parent,
1236 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1237 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1238 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1239 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1240 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1241 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1242 Opt_journal_update, Opt_journal_dev,
1243 Opt_journal_checksum, Opt_journal_async_commit,
1244 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1245 Opt_data_err_abort, Opt_data_err_ignore,
1246 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1247 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1248 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1249 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1250 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1251 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1252 Opt_inode_readahead_blks, Opt_journal_ioprio,
1253 Opt_dioread_nolock, Opt_dioread_lock,
1254 Opt_discard, Opt_nodiscard,
1255 Opt_init_inode_table, Opt_noinit_inode_table,
1258 static const match_table_t tokens = {
1259 {Opt_bsd_df, "bsddf"},
1260 {Opt_minix_df, "minixdf"},
1261 {Opt_grpid, "grpid"},
1262 {Opt_grpid, "bsdgroups"},
1263 {Opt_nogrpid, "nogrpid"},
1264 {Opt_nogrpid, "sysvgroups"},
1265 {Opt_resgid, "resgid=%u"},
1266 {Opt_resuid, "resuid=%u"},
1268 {Opt_err_cont, "errors=continue"},
1269 {Opt_err_panic, "errors=panic"},
1270 {Opt_err_ro, "errors=remount-ro"},
1271 {Opt_nouid32, "nouid32"},
1272 {Opt_debug, "debug"},
1273 {Opt_oldalloc, "oldalloc"},
1274 {Opt_orlov, "orlov"},
1275 {Opt_user_xattr, "user_xattr"},
1276 {Opt_nouser_xattr, "nouser_xattr"},
1278 {Opt_noacl, "noacl"},
1279 {Opt_noload, "noload"},
1280 {Opt_noload, "norecovery"},
1283 {Opt_commit, "commit=%u"},
1284 {Opt_min_batch_time, "min_batch_time=%u"},
1285 {Opt_max_batch_time, "max_batch_time=%u"},
1286 {Opt_journal_update, "journal=update"},
1287 {Opt_journal_dev, "journal_dev=%u"},
1288 {Opt_journal_checksum, "journal_checksum"},
1289 {Opt_journal_async_commit, "journal_async_commit"},
1290 {Opt_abort, "abort"},
1291 {Opt_data_journal, "data=journal"},
1292 {Opt_data_ordered, "data=ordered"},
1293 {Opt_data_writeback, "data=writeback"},
1294 {Opt_data_err_abort, "data_err=abort"},
1295 {Opt_data_err_ignore, "data_err=ignore"},
1296 {Opt_offusrjquota, "usrjquota="},
1297 {Opt_usrjquota, "usrjquota=%s"},
1298 {Opt_offgrpjquota, "grpjquota="},
1299 {Opt_grpjquota, "grpjquota=%s"},
1300 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1301 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1302 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1303 {Opt_grpquota, "grpquota"},
1304 {Opt_noquota, "noquota"},
1305 {Opt_quota, "quota"},
1306 {Opt_usrquota, "usrquota"},
1307 {Opt_barrier, "barrier=%u"},
1308 {Opt_barrier, "barrier"},
1309 {Opt_nobarrier, "nobarrier"},
1310 {Opt_i_version, "i_version"},
1311 {Opt_stripe, "stripe=%u"},
1312 {Opt_resize, "resize"},
1313 {Opt_delalloc, "delalloc"},
1314 {Opt_nodelalloc, "nodelalloc"},
1315 {Opt_mblk_io_submit, "mblk_io_submit"},
1316 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1317 {Opt_block_validity, "block_validity"},
1318 {Opt_noblock_validity, "noblock_validity"},
1319 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1320 {Opt_journal_ioprio, "journal_ioprio=%u"},
1321 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1322 {Opt_auto_da_alloc, "auto_da_alloc"},
1323 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1324 {Opt_dioread_nolock, "dioread_nolock"},
1325 {Opt_dioread_lock, "dioread_lock"},
1326 {Opt_discard, "discard"},
1327 {Opt_nodiscard, "nodiscard"},
1328 {Opt_init_inode_table, "init_itable=%u"},
1329 {Opt_init_inode_table, "init_itable"},
1330 {Opt_noinit_inode_table, "noinit_itable"},
1334 static ext4_fsblk_t get_sb_block(void **data)
1336 ext4_fsblk_t sb_block;
1337 char *options = (char *) *data;
1339 if (!options || strncmp(options, "sb=", 3) != 0)
1340 return 1; /* Default location */
1343 /* TODO: use simple_strtoll with >32bit ext4 */
1344 sb_block = simple_strtoul(options, &options, 0);
1345 if (*options && *options != ',') {
1346 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1350 if (*options == ',')
1352 *data = (void *) options;
1357 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1358 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1359 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1362 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1364 struct ext4_sb_info *sbi = EXT4_SB(sb);
1367 if (sb_any_quota_loaded(sb) &&
1368 !sbi->s_qf_names[qtype]) {
1369 ext4_msg(sb, KERN_ERR,
1370 "Cannot change journaled "
1371 "quota options when quota turned on");
1374 qname = match_strdup(args);
1376 ext4_msg(sb, KERN_ERR,
1377 "Not enough memory for storing quotafile name");
1380 if (sbi->s_qf_names[qtype] &&
1381 strcmp(sbi->s_qf_names[qtype], qname)) {
1382 ext4_msg(sb, KERN_ERR,
1383 "%s quota file already specified", QTYPE2NAME(qtype));
1387 sbi->s_qf_names[qtype] = qname;
1388 if (strchr(sbi->s_qf_names[qtype], '/')) {
1389 ext4_msg(sb, KERN_ERR,
1390 "quotafile must be on filesystem root");
1391 kfree(sbi->s_qf_names[qtype]);
1392 sbi->s_qf_names[qtype] = NULL;
1399 static int clear_qf_name(struct super_block *sb, int qtype)
1402 struct ext4_sb_info *sbi = EXT4_SB(sb);
1404 if (sb_any_quota_loaded(sb) &&
1405 sbi->s_qf_names[qtype]) {
1406 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1407 " when quota turned on");
1411 * The space will be released later when all options are confirmed
1414 sbi->s_qf_names[qtype] = NULL;
1419 static int parse_options(char *options, struct super_block *sb,
1420 unsigned long *journal_devnum,
1421 unsigned int *journal_ioprio,
1422 ext4_fsblk_t *n_blocks_count, int is_remount)
1424 struct ext4_sb_info *sbi = EXT4_SB(sb);
1426 substring_t args[MAX_OPT_ARGS];
1436 while ((p = strsep(&options, ",")) != NULL) {
1442 * Initialize args struct so we know whether arg was
1443 * found; some options take optional arguments.
1445 args[0].to = args[0].from = 0;
1446 token = match_token(p, tokens, args);
1449 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1450 clear_opt(sb, MINIX_DF);
1453 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1454 set_opt(sb, MINIX_DF);
1458 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1463 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1464 clear_opt(sb, GRPID);
1468 if (match_int(&args[0], &option))
1470 sbi->s_resuid = option;
1473 if (match_int(&args[0], &option))
1475 sbi->s_resgid = option;
1478 /* handled by get_sb_block() instead of here */
1479 /* *sb_block = match_int(&args[0]); */
1482 clear_opt(sb, ERRORS_CONT);
1483 clear_opt(sb, ERRORS_RO);
1484 set_opt(sb, ERRORS_PANIC);
1487 clear_opt(sb, ERRORS_CONT);
1488 clear_opt(sb, ERRORS_PANIC);
1489 set_opt(sb, ERRORS_RO);
1492 clear_opt(sb, ERRORS_RO);
1493 clear_opt(sb, ERRORS_PANIC);
1494 set_opt(sb, ERRORS_CONT);
1497 set_opt(sb, NO_UID32);
1503 set_opt(sb, OLDALLOC);
1506 clear_opt(sb, OLDALLOC);
1508 #ifdef CONFIG_EXT4_FS_XATTR
1509 case Opt_user_xattr:
1510 set_opt(sb, XATTR_USER);
1512 case Opt_nouser_xattr:
1513 clear_opt(sb, XATTR_USER);
1516 case Opt_user_xattr:
1517 case Opt_nouser_xattr:
1518 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1521 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1523 set_opt(sb, POSIX_ACL);
1526 clear_opt(sb, POSIX_ACL);
1531 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1534 case Opt_journal_update:
1536 /* Eventually we will want to be able to create
1537 a journal file here. For now, only allow the
1538 user to specify an existing inode to be the
1541 ext4_msg(sb, KERN_ERR,
1542 "Cannot specify journal on remount");
1545 set_opt(sb, UPDATE_JOURNAL);
1547 case Opt_journal_dev:
1549 ext4_msg(sb, KERN_ERR,
1550 "Cannot specify journal on remount");
1553 if (match_int(&args[0], &option))
1555 *journal_devnum = option;
1557 case Opt_journal_checksum:
1558 set_opt(sb, JOURNAL_CHECKSUM);
1560 case Opt_journal_async_commit:
1561 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1562 set_opt(sb, JOURNAL_CHECKSUM);
1565 set_opt(sb, NOLOAD);
1568 if (match_int(&args[0], &option))
1573 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1574 sbi->s_commit_interval = HZ * option;
1576 case Opt_max_batch_time:
1577 if (match_int(&args[0], &option))
1582 option = EXT4_DEF_MAX_BATCH_TIME;
1583 sbi->s_max_batch_time = option;
1585 case Opt_min_batch_time:
1586 if (match_int(&args[0], &option))
1590 sbi->s_min_batch_time = option;
1592 case Opt_data_journal:
1593 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1595 case Opt_data_ordered:
1596 data_opt = EXT4_MOUNT_ORDERED_DATA;
1598 case Opt_data_writeback:
1599 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1602 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1603 ext4_msg(sb, KERN_ERR,
1604 "Cannot change data mode on remount");
1608 clear_opt(sb, DATA_FLAGS);
1609 sbi->s_mount_opt |= data_opt;
1612 case Opt_data_err_abort:
1613 set_opt(sb, DATA_ERR_ABORT);
1615 case Opt_data_err_ignore:
1616 clear_opt(sb, DATA_ERR_ABORT);
1620 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1624 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1627 case Opt_offusrjquota:
1628 if (!clear_qf_name(sb, USRQUOTA))
1631 case Opt_offgrpjquota:
1632 if (!clear_qf_name(sb, GRPQUOTA))
1636 case Opt_jqfmt_vfsold:
1637 qfmt = QFMT_VFS_OLD;
1639 case Opt_jqfmt_vfsv0:
1642 case Opt_jqfmt_vfsv1:
1645 if (sb_any_quota_loaded(sb) &&
1646 sbi->s_jquota_fmt != qfmt) {
1647 ext4_msg(sb, KERN_ERR, "Cannot change "
1648 "journaled quota options when "
1652 sbi->s_jquota_fmt = qfmt;
1657 set_opt(sb, USRQUOTA);
1661 set_opt(sb, GRPQUOTA);
1664 if (sb_any_quota_loaded(sb)) {
1665 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1666 "options when quota turned on");
1669 clear_opt(sb, QUOTA);
1670 clear_opt(sb, USRQUOTA);
1671 clear_opt(sb, GRPQUOTA);
1677 ext4_msg(sb, KERN_ERR,
1678 "quota options not supported");
1682 case Opt_offusrjquota:
1683 case Opt_offgrpjquota:
1684 case Opt_jqfmt_vfsold:
1685 case Opt_jqfmt_vfsv0:
1686 case Opt_jqfmt_vfsv1:
1687 ext4_msg(sb, KERN_ERR,
1688 "journaled quota options not supported");
1694 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1697 clear_opt(sb, BARRIER);
1701 if (match_int(&args[0], &option))
1704 option = 1; /* No argument, default to 1 */
1706 set_opt(sb, BARRIER);
1708 clear_opt(sb, BARRIER);
1714 ext4_msg(sb, KERN_ERR,
1715 "resize option only available "
1719 if (match_int(&args[0], &option) != 0)
1721 *n_blocks_count = option;
1724 ext4_msg(sb, KERN_WARNING,
1725 "Ignoring deprecated nobh option");
1728 ext4_msg(sb, KERN_WARNING,
1729 "Ignoring deprecated bh option");
1732 set_opt(sb, I_VERSION);
1733 sb->s_flags |= MS_I_VERSION;
1735 case Opt_nodelalloc:
1736 clear_opt(sb, DELALLOC);
1738 case Opt_mblk_io_submit:
1739 set_opt(sb, MBLK_IO_SUBMIT);
1741 case Opt_nomblk_io_submit:
1742 clear_opt(sb, MBLK_IO_SUBMIT);
1745 if (match_int(&args[0], &option))
1749 sbi->s_stripe = option;
1752 set_opt(sb, DELALLOC);
1754 case Opt_block_validity:
1755 set_opt(sb, BLOCK_VALIDITY);
1757 case Opt_noblock_validity:
1758 clear_opt(sb, BLOCK_VALIDITY);
1760 case Opt_inode_readahead_blks:
1761 if (match_int(&args[0], &option))
1763 if (option < 0 || option > (1 << 30))
1765 if (!is_power_of_2(option)) {
1766 ext4_msg(sb, KERN_ERR,
1767 "EXT4-fs: inode_readahead_blks"
1768 " must be a power of 2");
1771 sbi->s_inode_readahead_blks = option;
1773 case Opt_journal_ioprio:
1774 if (match_int(&args[0], &option))
1776 if (option < 0 || option > 7)
1778 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1781 case Opt_noauto_da_alloc:
1782 set_opt(sb, NO_AUTO_DA_ALLOC);
1784 case Opt_auto_da_alloc:
1786 if (match_int(&args[0], &option))
1789 option = 1; /* No argument, default to 1 */
1791 clear_opt(sb, NO_AUTO_DA_ALLOC);
1793 set_opt(sb,NO_AUTO_DA_ALLOC);
1796 set_opt(sb, DISCARD);
1799 clear_opt(sb, DISCARD);
1801 case Opt_dioread_nolock:
1802 set_opt(sb, DIOREAD_NOLOCK);
1804 case Opt_dioread_lock:
1805 clear_opt(sb, DIOREAD_NOLOCK);
1807 case Opt_init_inode_table:
1808 set_opt(sb, INIT_INODE_TABLE);
1810 if (match_int(&args[0], &option))
1813 option = EXT4_DEF_LI_WAIT_MULT;
1816 sbi->s_li_wait_mult = option;
1818 case Opt_noinit_inode_table:
1819 clear_opt(sb, INIT_INODE_TABLE);
1822 ext4_msg(sb, KERN_ERR,
1823 "Unrecognized mount option \"%s\" "
1824 "or missing value", p);
1829 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1830 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1831 clear_opt(sb, USRQUOTA);
1833 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1834 clear_opt(sb, GRPQUOTA);
1836 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1837 ext4_msg(sb, KERN_ERR, "old and new quota "
1842 if (!sbi->s_jquota_fmt) {
1843 ext4_msg(sb, KERN_ERR, "journaled quota format "
1848 if (sbi->s_jquota_fmt) {
1849 ext4_msg(sb, KERN_ERR, "journaled quota format "
1850 "specified with no journaling "
1859 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1862 struct ext4_sb_info *sbi = EXT4_SB(sb);
1865 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1866 ext4_msg(sb, KERN_ERR, "revision level too high, "
1867 "forcing read-only mode");
1872 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1873 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1874 "running e2fsck is recommended");
1875 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1876 ext4_msg(sb, KERN_WARNING,
1877 "warning: mounting fs with errors, "
1878 "running e2fsck is recommended");
1879 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1880 le16_to_cpu(es->s_mnt_count) >=
1881 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: maximal mount count reached, "
1884 "running e2fsck is recommended");
1885 else if (le32_to_cpu(es->s_checkinterval) &&
1886 (le32_to_cpu(es->s_lastcheck) +
1887 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: checktime reached, "
1890 "running e2fsck is recommended");
1891 if (!sbi->s_journal)
1892 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1893 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1894 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1895 le16_add_cpu(&es->s_mnt_count, 1);
1896 es->s_mtime = cpu_to_le32(get_seconds());
1897 ext4_update_dynamic_rev(sb);
1899 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1901 ext4_commit_super(sb, 1);
1902 if (test_opt(sb, DEBUG))
1903 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1904 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1906 sbi->s_groups_count,
1907 EXT4_BLOCKS_PER_GROUP(sb),
1908 EXT4_INODES_PER_GROUP(sb),
1909 sbi->s_mount_opt, sbi->s_mount_opt2);
1914 static int ext4_fill_flex_info(struct super_block *sb)
1916 struct ext4_sb_info *sbi = EXT4_SB(sb);
1917 struct ext4_group_desc *gdp = NULL;
1918 ext4_group_t flex_group_count;
1919 ext4_group_t flex_group;
1920 int groups_per_flex = 0;
1924 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1925 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1927 if (groups_per_flex < 2) {
1928 sbi->s_log_groups_per_flex = 0;
1932 /* We allocate both existing and potentially added groups */
1933 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1934 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1935 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1936 size = flex_group_count * sizeof(struct flex_groups);
1937 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1938 if (sbi->s_flex_groups == NULL) {
1939 sbi->s_flex_groups = vzalloc(size);
1940 if (sbi->s_flex_groups == NULL) {
1941 ext4_msg(sb, KERN_ERR,
1942 "not enough memory for %u flex groups",
1948 for (i = 0; i < sbi->s_groups_count; i++) {
1949 gdp = ext4_get_group_desc(sb, i, NULL);
1951 flex_group = ext4_flex_group(sbi, i);
1952 atomic_add(ext4_free_inodes_count(sb, gdp),
1953 &sbi->s_flex_groups[flex_group].free_inodes);
1954 atomic_add(ext4_free_blks_count(sb, gdp),
1955 &sbi->s_flex_groups[flex_group].free_blocks);
1956 atomic_add(ext4_used_dirs_count(sb, gdp),
1957 &sbi->s_flex_groups[flex_group].used_dirs);
1965 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1966 struct ext4_group_desc *gdp)
1970 if (sbi->s_es->s_feature_ro_compat &
1971 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1972 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1973 __le32 le_group = cpu_to_le32(block_group);
1975 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1976 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1977 crc = crc16(crc, (__u8 *)gdp, offset);
1978 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1979 /* for checksum of struct ext4_group_desc do the rest...*/
1980 if ((sbi->s_es->s_feature_incompat &
1981 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1982 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1983 crc = crc16(crc, (__u8 *)gdp + offset,
1984 le16_to_cpu(sbi->s_es->s_desc_size) -
1988 return cpu_to_le16(crc);
1991 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1992 struct ext4_group_desc *gdp)
1994 if ((sbi->s_es->s_feature_ro_compat &
1995 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1996 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2002 /* Called at mount-time, super-block is locked */
2003 static int ext4_check_descriptors(struct super_block *sb,
2004 ext4_group_t *first_not_zeroed)
2006 struct ext4_sb_info *sbi = EXT4_SB(sb);
2007 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2008 ext4_fsblk_t last_block;
2009 ext4_fsblk_t block_bitmap;
2010 ext4_fsblk_t inode_bitmap;
2011 ext4_fsblk_t inode_table;
2012 int flexbg_flag = 0;
2013 ext4_group_t i, grp = sbi->s_groups_count;
2015 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2018 ext4_debug("Checking group descriptors");
2020 for (i = 0; i < sbi->s_groups_count; i++) {
2021 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2023 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2024 last_block = ext4_blocks_count(sbi->s_es) - 1;
2026 last_block = first_block +
2027 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2029 if ((grp == sbi->s_groups_count) &&
2030 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2033 block_bitmap = ext4_block_bitmap(sb, gdp);
2034 if (block_bitmap < first_block || block_bitmap > last_block) {
2035 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2036 "Block bitmap for group %u not in group "
2037 "(block %llu)!", i, block_bitmap);
2040 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2041 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2042 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2043 "Inode bitmap for group %u not in group "
2044 "(block %llu)!", i, inode_bitmap);
2047 inode_table = ext4_inode_table(sb, gdp);
2048 if (inode_table < first_block ||
2049 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2050 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2051 "Inode table for group %u not in group "
2052 "(block %llu)!", i, inode_table);
2055 ext4_lock_group(sb, i);
2056 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2057 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2058 "Checksum for group %u failed (%u!=%u)",
2059 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2060 gdp)), le16_to_cpu(gdp->bg_checksum));
2061 if (!(sb->s_flags & MS_RDONLY)) {
2062 ext4_unlock_group(sb, i);
2066 ext4_unlock_group(sb, i);
2068 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2070 if (NULL != first_not_zeroed)
2071 *first_not_zeroed = grp;
2073 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2074 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2078 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2079 * the superblock) which were deleted from all directories, but held open by
2080 * a process at the time of a crash. We walk the list and try to delete these
2081 * inodes at recovery time (only with a read-write filesystem).
2083 * In order to keep the orphan inode chain consistent during traversal (in
2084 * case of crash during recovery), we link each inode into the superblock
2085 * orphan list_head and handle it the same way as an inode deletion during
2086 * normal operation (which journals the operations for us).
2088 * We only do an iget() and an iput() on each inode, which is very safe if we
2089 * accidentally point at an in-use or already deleted inode. The worst that
2090 * can happen in this case is that we get a "bit already cleared" message from
2091 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2092 * e2fsck was run on this filesystem, and it must have already done the orphan
2093 * inode cleanup for us, so we can safely abort without any further action.
2095 static void ext4_orphan_cleanup(struct super_block *sb,
2096 struct ext4_super_block *es)
2098 unsigned int s_flags = sb->s_flags;
2099 int nr_orphans = 0, nr_truncates = 0;
2103 if (!es->s_last_orphan) {
2104 jbd_debug(4, "no orphan inodes to clean up\n");
2108 if (bdev_read_only(sb->s_bdev)) {
2109 ext4_msg(sb, KERN_ERR, "write access "
2110 "unavailable, skipping orphan cleanup");
2114 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2115 if (es->s_last_orphan)
2116 jbd_debug(1, "Errors on filesystem, "
2117 "clearing orphan list.\n");
2118 es->s_last_orphan = 0;
2119 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2123 if (s_flags & MS_RDONLY) {
2124 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2125 sb->s_flags &= ~MS_RDONLY;
2128 /* Needed for iput() to work correctly and not trash data */
2129 sb->s_flags |= MS_ACTIVE;
2130 /* Turn on quotas so that they are updated correctly */
2131 for (i = 0; i < MAXQUOTAS; i++) {
2132 if (EXT4_SB(sb)->s_qf_names[i]) {
2133 int ret = ext4_quota_on_mount(sb, i);
2135 ext4_msg(sb, KERN_ERR,
2136 "Cannot turn on journaled "
2137 "quota: error %d", ret);
2142 while (es->s_last_orphan) {
2143 struct inode *inode;
2145 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2146 if (IS_ERR(inode)) {
2147 es->s_last_orphan = 0;
2151 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2152 dquot_initialize(inode);
2153 if (inode->i_nlink) {
2154 ext4_msg(sb, KERN_DEBUG,
2155 "%s: truncating inode %lu to %lld bytes",
2156 __func__, inode->i_ino, inode->i_size);
2157 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2158 inode->i_ino, inode->i_size);
2159 ext4_truncate(inode);
2162 ext4_msg(sb, KERN_DEBUG,
2163 "%s: deleting unreferenced inode %lu",
2164 __func__, inode->i_ino);
2165 jbd_debug(2, "deleting unreferenced inode %lu\n",
2169 iput(inode); /* The delete magic happens here! */
2172 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2175 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2176 PLURAL(nr_orphans));
2178 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2179 PLURAL(nr_truncates));
2181 /* Turn quotas off */
2182 for (i = 0; i < MAXQUOTAS; i++) {
2183 if (sb_dqopt(sb)->files[i])
2184 dquot_quota_off(sb, i);
2187 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2191 * Maximal extent format file size.
2192 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2193 * extent format containers, within a sector_t, and within i_blocks
2194 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2195 * so that won't be a limiting factor.
2197 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2199 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2202 loff_t upper_limit = MAX_LFS_FILESIZE;
2204 /* small i_blocks in vfs inode? */
2205 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2207 * CONFIG_LBDAF is not enabled implies the inode
2208 * i_block represent total blocks in 512 bytes
2209 * 32 == size of vfs inode i_blocks * 8
2211 upper_limit = (1LL << 32) - 1;
2213 /* total blocks in file system block size */
2214 upper_limit >>= (blkbits - 9);
2215 upper_limit <<= blkbits;
2218 /* 32-bit extent-start container, ee_block */
2223 /* Sanity check against vm- & vfs- imposed limits */
2224 if (res > upper_limit)
2231 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2232 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2233 * We need to be 1 filesystem block less than the 2^48 sector limit.
2235 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2237 loff_t res = EXT4_NDIR_BLOCKS;
2240 /* This is calculated to be the largest file size for a dense, block
2241 * mapped file such that the file's total number of 512-byte sectors,
2242 * including data and all indirect blocks, does not exceed (2^48 - 1).
2244 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2245 * number of 512-byte sectors of the file.
2248 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2250 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2251 * the inode i_block field represents total file blocks in
2252 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2254 upper_limit = (1LL << 32) - 1;
2256 /* total blocks in file system block size */
2257 upper_limit >>= (bits - 9);
2261 * We use 48 bit ext4_inode i_blocks
2262 * With EXT4_HUGE_FILE_FL set the i_blocks
2263 * represent total number of blocks in
2264 * file system block size
2266 upper_limit = (1LL << 48) - 1;
2270 /* indirect blocks */
2272 /* double indirect blocks */
2273 meta_blocks += 1 + (1LL << (bits-2));
2274 /* tripple indirect blocks */
2275 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2277 upper_limit -= meta_blocks;
2278 upper_limit <<= bits;
2280 res += 1LL << (bits-2);
2281 res += 1LL << (2*(bits-2));
2282 res += 1LL << (3*(bits-2));
2284 if (res > upper_limit)
2287 if (res > MAX_LFS_FILESIZE)
2288 res = MAX_LFS_FILESIZE;
2293 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2294 ext4_fsblk_t logical_sb_block, int nr)
2296 struct ext4_sb_info *sbi = EXT4_SB(sb);
2297 ext4_group_t bg, first_meta_bg;
2300 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2302 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2304 return logical_sb_block + nr + 1;
2305 bg = sbi->s_desc_per_block * nr;
2306 if (ext4_bg_has_super(sb, bg))
2309 return (has_super + ext4_group_first_block_no(sb, bg));
2313 * ext4_get_stripe_size: Get the stripe size.
2314 * @sbi: In memory super block info
2316 * If we have specified it via mount option, then
2317 * use the mount option value. If the value specified at mount time is
2318 * greater than the blocks per group use the super block value.
2319 * If the super block value is greater than blocks per group return 0.
2320 * Allocator needs it be less than blocks per group.
2323 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2325 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2326 unsigned long stripe_width =
2327 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2329 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2330 return sbi->s_stripe;
2332 if (stripe_width <= sbi->s_blocks_per_group)
2333 return stripe_width;
2335 if (stride <= sbi->s_blocks_per_group)
2344 struct attribute attr;
2345 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2346 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2347 const char *, size_t);
2351 static int parse_strtoul(const char *buf,
2352 unsigned long max, unsigned long *value)
2356 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2357 endp = skip_spaces(endp);
2358 if (*endp || *value > max)
2364 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2365 struct ext4_sb_info *sbi,
2368 return snprintf(buf, PAGE_SIZE, "%llu\n",
2369 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2372 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2373 struct ext4_sb_info *sbi, char *buf)
2375 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2377 if (!sb->s_bdev->bd_part)
2378 return snprintf(buf, PAGE_SIZE, "0\n");
2379 return snprintf(buf, PAGE_SIZE, "%lu\n",
2380 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2381 sbi->s_sectors_written_start) >> 1);
2384 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2385 struct ext4_sb_info *sbi, char *buf)
2387 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2389 if (!sb->s_bdev->bd_part)
2390 return snprintf(buf, PAGE_SIZE, "0\n");
2391 return snprintf(buf, PAGE_SIZE, "%llu\n",
2392 (unsigned long long)(sbi->s_kbytes_written +
2393 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2394 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2397 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2398 struct ext4_sb_info *sbi,
2399 const char *buf, size_t count)
2403 if (parse_strtoul(buf, 0x40000000, &t))
2406 if (!is_power_of_2(t))
2409 sbi->s_inode_readahead_blks = t;
2413 static ssize_t sbi_ui_show(struct ext4_attr *a,
2414 struct ext4_sb_info *sbi, char *buf)
2416 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2418 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2421 static ssize_t sbi_ui_store(struct ext4_attr *a,
2422 struct ext4_sb_info *sbi,
2423 const char *buf, size_t count)
2425 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2428 if (parse_strtoul(buf, 0xffffffff, &t))
2434 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2435 static struct ext4_attr ext4_attr_##_name = { \
2436 .attr = {.name = __stringify(_name), .mode = _mode }, \
2439 .offset = offsetof(struct ext4_sb_info, _elname), \
2441 #define EXT4_ATTR(name, mode, show, store) \
2442 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2444 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2445 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2446 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2447 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2448 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2449 #define ATTR_LIST(name) &ext4_attr_##name.attr
2451 EXT4_RO_ATTR(delayed_allocation_blocks);
2452 EXT4_RO_ATTR(session_write_kbytes);
2453 EXT4_RO_ATTR(lifetime_write_kbytes);
2454 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2455 inode_readahead_blks_store, s_inode_readahead_blks);
2456 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2457 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2458 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2459 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2460 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2461 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2462 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2463 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2465 static struct attribute *ext4_attrs[] = {
2466 ATTR_LIST(delayed_allocation_blocks),
2467 ATTR_LIST(session_write_kbytes),
2468 ATTR_LIST(lifetime_write_kbytes),
2469 ATTR_LIST(inode_readahead_blks),
2470 ATTR_LIST(inode_goal),
2471 ATTR_LIST(mb_stats),
2472 ATTR_LIST(mb_max_to_scan),
2473 ATTR_LIST(mb_min_to_scan),
2474 ATTR_LIST(mb_order2_req),
2475 ATTR_LIST(mb_stream_req),
2476 ATTR_LIST(mb_group_prealloc),
2477 ATTR_LIST(max_writeback_mb_bump),
2481 /* Features this copy of ext4 supports */
2482 EXT4_INFO_ATTR(lazy_itable_init);
2483 EXT4_INFO_ATTR(batched_discard);
2485 static struct attribute *ext4_feat_attrs[] = {
2486 ATTR_LIST(lazy_itable_init),
2487 ATTR_LIST(batched_discard),
2491 static ssize_t ext4_attr_show(struct kobject *kobj,
2492 struct attribute *attr, char *buf)
2494 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2496 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2498 return a->show ? a->show(a, sbi, buf) : 0;
2501 static ssize_t ext4_attr_store(struct kobject *kobj,
2502 struct attribute *attr,
2503 const char *buf, size_t len)
2505 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2507 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2509 return a->store ? a->store(a, sbi, buf, len) : 0;
2512 static void ext4_sb_release(struct kobject *kobj)
2514 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2516 complete(&sbi->s_kobj_unregister);
2519 static const struct sysfs_ops ext4_attr_ops = {
2520 .show = ext4_attr_show,
2521 .store = ext4_attr_store,
2524 static struct kobj_type ext4_ktype = {
2525 .default_attrs = ext4_attrs,
2526 .sysfs_ops = &ext4_attr_ops,
2527 .release = ext4_sb_release,
2530 static void ext4_feat_release(struct kobject *kobj)
2532 complete(&ext4_feat->f_kobj_unregister);
2535 static struct kobj_type ext4_feat_ktype = {
2536 .default_attrs = ext4_feat_attrs,
2537 .sysfs_ops = &ext4_attr_ops,
2538 .release = ext4_feat_release,
2542 * Check whether this filesystem can be mounted based on
2543 * the features present and the RDONLY/RDWR mount requested.
2544 * Returns 1 if this filesystem can be mounted as requested,
2545 * 0 if it cannot be.
2547 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2549 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2550 ext4_msg(sb, KERN_ERR,
2551 "Couldn't mount because of "
2552 "unsupported optional features (%x)",
2553 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2554 ~EXT4_FEATURE_INCOMPAT_SUPP));
2561 /* Check that feature set is OK for a read-write mount */
2562 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2563 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2564 "unsupported optional features (%x)",
2565 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2566 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2570 * Large file size enabled file system can only be mounted
2571 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2573 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2574 if (sizeof(blkcnt_t) < sizeof(u64)) {
2575 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2576 "cannot be mounted RDWR without "
2585 * This function is called once a day if we have errors logged
2586 * on the file system
2588 static void print_daily_error_info(unsigned long arg)
2590 struct super_block *sb = (struct super_block *) arg;
2591 struct ext4_sb_info *sbi;
2592 struct ext4_super_block *es;
2597 if (es->s_error_count)
2598 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2599 le32_to_cpu(es->s_error_count));
2600 if (es->s_first_error_time) {
2601 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2602 sb->s_id, le32_to_cpu(es->s_first_error_time),
2603 (int) sizeof(es->s_first_error_func),
2604 es->s_first_error_func,
2605 le32_to_cpu(es->s_first_error_line));
2606 if (es->s_first_error_ino)
2607 printk(": inode %u",
2608 le32_to_cpu(es->s_first_error_ino));
2609 if (es->s_first_error_block)
2610 printk(": block %llu", (unsigned long long)
2611 le64_to_cpu(es->s_first_error_block));
2614 if (es->s_last_error_time) {
2615 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2616 sb->s_id, le32_to_cpu(es->s_last_error_time),
2617 (int) sizeof(es->s_last_error_func),
2618 es->s_last_error_func,
2619 le32_to_cpu(es->s_last_error_line));
2620 if (es->s_last_error_ino)
2621 printk(": inode %u",
2622 le32_to_cpu(es->s_last_error_ino));
2623 if (es->s_last_error_block)
2624 printk(": block %llu", (unsigned long long)
2625 le64_to_cpu(es->s_last_error_block));
2628 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2631 static void ext4_lazyinode_timeout(unsigned long data)
2633 struct task_struct *p = (struct task_struct *)data;
2637 /* Find next suitable group and run ext4_init_inode_table */
2638 static int ext4_run_li_request(struct ext4_li_request *elr)
2640 struct ext4_group_desc *gdp = NULL;
2641 ext4_group_t group, ngroups;
2642 struct super_block *sb;
2643 unsigned long timeout = 0;
2647 ngroups = EXT4_SB(sb)->s_groups_count;
2649 for (group = elr->lr_next_group; group < ngroups; group++) {
2650 gdp = ext4_get_group_desc(sb, group, NULL);
2656 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2660 if (group == ngroups)
2665 ret = ext4_init_inode_table(sb, group,
2666 elr->lr_timeout ? 0 : 1);
2667 if (elr->lr_timeout == 0) {
2668 timeout = jiffies - timeout;
2669 if (elr->lr_sbi->s_li_wait_mult)
2670 timeout *= elr->lr_sbi->s_li_wait_mult;
2673 elr->lr_timeout = timeout;
2675 elr->lr_next_sched = jiffies + elr->lr_timeout;
2676 elr->lr_next_group = group + 1;
2683 * Remove lr_request from the list_request and free the
2684 * request tructure. Should be called with li_list_mtx held
2686 static void ext4_remove_li_request(struct ext4_li_request *elr)
2688 struct ext4_sb_info *sbi;
2695 list_del(&elr->lr_request);
2696 sbi->s_li_request = NULL;
2700 static void ext4_unregister_li_request(struct super_block *sb)
2702 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2707 mutex_lock(&ext4_li_info->li_list_mtx);
2708 ext4_remove_li_request(elr);
2709 mutex_unlock(&ext4_li_info->li_list_mtx);
2713 * This is the function where ext4lazyinit thread lives. It walks
2714 * through the request list searching for next scheduled filesystem.
2715 * When such a fs is found, run the lazy initialization request
2716 * (ext4_rn_li_request) and keep track of the time spend in this
2717 * function. Based on that time we compute next schedule time of
2718 * the request. When walking through the list is complete, compute
2719 * next waking time and put itself into sleep.
2721 static int ext4_lazyinit_thread(void *arg)
2723 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2724 struct list_head *pos, *n;
2725 struct ext4_li_request *elr;
2726 unsigned long next_wakeup;
2729 BUG_ON(NULL == eli);
2731 eli->li_timer.data = (unsigned long)current;
2732 eli->li_timer.function = ext4_lazyinode_timeout;
2734 eli->li_task = current;
2735 wake_up(&eli->li_wait_task);
2739 next_wakeup = MAX_JIFFY_OFFSET;
2741 mutex_lock(&eli->li_list_mtx);
2742 if (list_empty(&eli->li_request_list)) {
2743 mutex_unlock(&eli->li_list_mtx);
2747 list_for_each_safe(pos, n, &eli->li_request_list) {
2748 elr = list_entry(pos, struct ext4_li_request,
2751 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2752 if (ext4_run_li_request(elr) != 0) {
2753 /* error, remove the lazy_init job */
2754 ext4_remove_li_request(elr);
2759 if (time_before(elr->lr_next_sched, next_wakeup))
2760 next_wakeup = elr->lr_next_sched;
2762 mutex_unlock(&eli->li_list_mtx);
2764 if (freezing(current))
2767 if ((time_after_eq(jiffies, next_wakeup)) ||
2768 (MAX_JIFFY_OFFSET == next_wakeup)) {
2773 eli->li_timer.expires = next_wakeup;
2774 add_timer(&eli->li_timer);
2775 prepare_to_wait(&eli->li_wait_daemon, &wait,
2776 TASK_INTERRUPTIBLE);
2777 if (time_before(jiffies, next_wakeup))
2779 finish_wait(&eli->li_wait_daemon, &wait);
2784 * It looks like the request list is empty, but we need
2785 * to check it under the li_list_mtx lock, to prevent any
2786 * additions into it, and of course we should lock ext4_li_mtx
2787 * to atomically free the list and ext4_li_info, because at
2788 * this point another ext4 filesystem could be registering
2791 mutex_lock(&ext4_li_mtx);
2792 mutex_lock(&eli->li_list_mtx);
2793 if (!list_empty(&eli->li_request_list)) {
2794 mutex_unlock(&eli->li_list_mtx);
2795 mutex_unlock(&ext4_li_mtx);
2798 mutex_unlock(&eli->li_list_mtx);
2799 del_timer_sync(&ext4_li_info->li_timer);
2800 eli->li_task = NULL;
2801 wake_up(&eli->li_wait_task);
2803 kfree(ext4_li_info);
2804 ext4_li_info = NULL;
2805 mutex_unlock(&ext4_li_mtx);
2810 static void ext4_clear_request_list(void)
2812 struct list_head *pos, *n;
2813 struct ext4_li_request *elr;
2815 mutex_lock(&ext4_li_info->li_list_mtx);
2816 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2817 elr = list_entry(pos, struct ext4_li_request,
2819 ext4_remove_li_request(elr);
2821 mutex_unlock(&ext4_li_info->li_list_mtx);
2824 static int ext4_run_lazyinit_thread(void)
2826 struct task_struct *t;
2828 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2830 int err = PTR_ERR(t);
2831 ext4_clear_request_list();
2832 del_timer_sync(&ext4_li_info->li_timer);
2833 kfree(ext4_li_info);
2834 ext4_li_info = NULL;
2835 printk(KERN_CRIT "EXT4: error %d creating inode table "
2836 "initialization thread\n",
2840 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2842 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2847 * Check whether it make sense to run itable init. thread or not.
2848 * If there is at least one uninitialized inode table, return
2849 * corresponding group number, else the loop goes through all
2850 * groups and return total number of groups.
2852 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2854 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2855 struct ext4_group_desc *gdp = NULL;
2857 for (group = 0; group < ngroups; group++) {
2858 gdp = ext4_get_group_desc(sb, group, NULL);
2862 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2869 static int ext4_li_info_new(void)
2871 struct ext4_lazy_init *eli = NULL;
2873 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2877 eli->li_task = NULL;
2878 INIT_LIST_HEAD(&eli->li_request_list);
2879 mutex_init(&eli->li_list_mtx);
2881 init_waitqueue_head(&eli->li_wait_daemon);
2882 init_waitqueue_head(&eli->li_wait_task);
2883 init_timer(&eli->li_timer);
2884 eli->li_state |= EXT4_LAZYINIT_QUIT;
2891 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2894 struct ext4_sb_info *sbi = EXT4_SB(sb);
2895 struct ext4_li_request *elr;
2898 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2904 elr->lr_next_group = start;
2907 * Randomize first schedule time of the request to
2908 * spread the inode table initialization requests
2911 get_random_bytes(&rnd, sizeof(rnd));
2912 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2913 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2918 static int ext4_register_li_request(struct super_block *sb,
2919 ext4_group_t first_not_zeroed)
2921 struct ext4_sb_info *sbi = EXT4_SB(sb);
2922 struct ext4_li_request *elr;
2923 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2926 if (sbi->s_li_request != NULL)
2929 if (first_not_zeroed == ngroups ||
2930 (sb->s_flags & MS_RDONLY) ||
2931 !test_opt(sb, INIT_INODE_TABLE)) {
2932 sbi->s_li_request = NULL;
2936 if (first_not_zeroed == ngroups) {
2937 sbi->s_li_request = NULL;
2941 elr = ext4_li_request_new(sb, first_not_zeroed);
2945 mutex_lock(&ext4_li_mtx);
2947 if (NULL == ext4_li_info) {
2948 ret = ext4_li_info_new();
2953 mutex_lock(&ext4_li_info->li_list_mtx);
2954 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2955 mutex_unlock(&ext4_li_info->li_list_mtx);
2957 sbi->s_li_request = elr;
2959 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2960 ret = ext4_run_lazyinit_thread();
2965 mutex_unlock(&ext4_li_mtx);
2972 * We do not need to lock anything since this is called on
2975 static void ext4_destroy_lazyinit_thread(void)
2978 * If thread exited earlier
2979 * there's nothing to be done.
2984 ext4_clear_request_list();
2986 while (ext4_li_info->li_task) {
2987 wake_up(&ext4_li_info->li_wait_daemon);
2988 wait_event(ext4_li_info->li_wait_task,
2989 ext4_li_info->li_task == NULL);
2993 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2994 __releases(kernel_lock)
2995 __acquires(kernel_lock)
2997 char *orig_data = kstrdup(data, GFP_KERNEL);
2998 struct buffer_head *bh;
2999 struct ext4_super_block *es = NULL;
3000 struct ext4_sb_info *sbi;
3002 ext4_fsblk_t sb_block = get_sb_block(&data);
3003 ext4_fsblk_t logical_sb_block;
3004 unsigned long offset = 0;
3005 unsigned long journal_devnum = 0;
3006 unsigned long def_mount_opts;
3012 unsigned int db_count;
3014 int needs_recovery, has_huge_files;
3017 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3018 ext4_group_t first_not_zeroed;
3020 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3024 sbi->s_blockgroup_lock =
3025 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3026 if (!sbi->s_blockgroup_lock) {
3030 sb->s_fs_info = sbi;
3031 sbi->s_mount_opt = 0;
3032 sbi->s_resuid = EXT4_DEF_RESUID;
3033 sbi->s_resgid = EXT4_DEF_RESGID;
3034 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3035 sbi->s_sb_block = sb_block;
3036 if (sb->s_bdev->bd_part)
3037 sbi->s_sectors_written_start =
3038 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3040 /* Cleanup superblock name */
3041 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3045 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3047 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3052 * The ext4 superblock will not be buffer aligned for other than 1kB
3053 * block sizes. We need to calculate the offset from buffer start.
3055 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3056 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3057 offset = do_div(logical_sb_block, blocksize);
3059 logical_sb_block = sb_block;
3062 if (!(bh = sb_bread(sb, logical_sb_block))) {
3063 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3067 * Note: s_es must be initialized as soon as possible because
3068 * some ext4 macro-instructions depend on its value
3070 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3072 sb->s_magic = le16_to_cpu(es->s_magic);
3073 if (sb->s_magic != EXT4_SUPER_MAGIC)
3075 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3077 /* Set defaults before we parse the mount options */
3078 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3079 set_opt(sb, INIT_INODE_TABLE);
3080 if (def_mount_opts & EXT4_DEFM_DEBUG)
3082 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3083 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3087 if (def_mount_opts & EXT4_DEFM_UID16)
3088 set_opt(sb, NO_UID32);
3089 #ifdef CONFIG_EXT4_FS_XATTR
3090 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3091 set_opt(sb, XATTR_USER);
3093 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3094 if (def_mount_opts & EXT4_DEFM_ACL)
3095 set_opt(sb, POSIX_ACL);
3097 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3098 set_opt(sb, JOURNAL_DATA);
3099 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3100 set_opt(sb, ORDERED_DATA);
3101 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3102 set_opt(sb, WRITEBACK_DATA);
3104 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3105 set_opt(sb, ERRORS_PANIC);
3106 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3107 set_opt(sb, ERRORS_CONT);
3109 set_opt(sb, ERRORS_RO);
3110 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3111 set_opt(sb, BLOCK_VALIDITY);
3112 if (def_mount_opts & EXT4_DEFM_DISCARD)
3113 set_opt(sb, DISCARD);
3115 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3116 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3117 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3118 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3119 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3121 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3122 set_opt(sb, BARRIER);
3125 * enable delayed allocation by default
3126 * Use -o nodelalloc to turn it off
3128 if (!IS_EXT3_SB(sb) &&
3129 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3130 set_opt(sb, DELALLOC);
3132 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3133 &journal_devnum, &journal_ioprio, NULL, 0)) {
3134 ext4_msg(sb, KERN_WARNING,
3135 "failed to parse options in superblock: %s",
3136 sbi->s_es->s_mount_opts);
3138 if (!parse_options((char *) data, sb, &journal_devnum,
3139 &journal_ioprio, NULL, 0))
3142 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3143 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3145 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3146 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3147 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3148 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3149 ext4_msg(sb, KERN_WARNING,
3150 "feature flags set on rev 0 fs, "
3151 "running e2fsck is recommended");
3154 * Check feature flags regardless of the revision level, since we
3155 * previously didn't change the revision level when setting the flags,
3156 * so there is a chance incompat flags are set on a rev 0 filesystem.
3158 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3161 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3163 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3164 blocksize > EXT4_MAX_BLOCK_SIZE) {
3165 ext4_msg(sb, KERN_ERR,
3166 "Unsupported filesystem blocksize %d", blocksize);
3170 if (sb->s_blocksize != blocksize) {
3171 /* Validate the filesystem blocksize */
3172 if (!sb_set_blocksize(sb, blocksize)) {
3173 ext4_msg(sb, KERN_ERR, "bad block size %d",
3179 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3180 offset = do_div(logical_sb_block, blocksize);
3181 bh = sb_bread(sb, logical_sb_block);
3183 ext4_msg(sb, KERN_ERR,
3184 "Can't read superblock on 2nd try");
3187 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3189 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3190 ext4_msg(sb, KERN_ERR,
3191 "Magic mismatch, very weird!");
3196 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3197 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3198 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3200 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3202 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3203 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3204 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3206 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3207 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3208 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3209 (!is_power_of_2(sbi->s_inode_size)) ||
3210 (sbi->s_inode_size > blocksize)) {
3211 ext4_msg(sb, KERN_ERR,
3212 "unsupported inode size: %d",
3216 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3217 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3220 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3222 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3223 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3224 !is_power_of_2(sbi->s_desc_size)) {
3225 ext4_msg(sb, KERN_ERR,
3226 "unsupported descriptor size %lu",
3231 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3233 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3234 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3235 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3238 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3239 if (sbi->s_inodes_per_block == 0)
3241 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3242 sbi->s_inodes_per_block;
3243 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3245 sbi->s_mount_state = le16_to_cpu(es->s_state);
3246 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3247 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3249 for (i = 0; i < 4; i++)
3250 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3251 sbi->s_def_hash_version = es->s_def_hash_version;
3252 i = le32_to_cpu(es->s_flags);
3253 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3254 sbi->s_hash_unsigned = 3;
3255 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3256 #ifdef __CHAR_UNSIGNED__
3257 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3258 sbi->s_hash_unsigned = 3;
3260 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3265 if (sbi->s_blocks_per_group > blocksize * 8) {
3266 ext4_msg(sb, KERN_ERR,
3267 "#blocks per group too big: %lu",
3268 sbi->s_blocks_per_group);
3271 if (sbi->s_inodes_per_group > blocksize * 8) {
3272 ext4_msg(sb, KERN_ERR,
3273 "#inodes per group too big: %lu",
3274 sbi->s_inodes_per_group);
3279 * Test whether we have more sectors than will fit in sector_t,
3280 * and whether the max offset is addressable by the page cache.
3282 err = generic_check_addressable(sb->s_blocksize_bits,
3283 ext4_blocks_count(es));
3285 ext4_msg(sb, KERN_ERR, "filesystem"
3286 " too large to mount safely on this system");
3287 if (sizeof(sector_t) < 8)
3288 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3293 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3296 /* check blocks count against device size */
3297 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3298 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3299 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3300 "exceeds size of device (%llu blocks)",
3301 ext4_blocks_count(es), blocks_count);
3306 * It makes no sense for the first data block to be beyond the end
3307 * of the filesystem.
3309 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3310 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3311 "block %u is beyond end of filesystem (%llu)",
3312 le32_to_cpu(es->s_first_data_block),
3313 ext4_blocks_count(es));
3316 blocks_count = (ext4_blocks_count(es) -
3317 le32_to_cpu(es->s_first_data_block) +
3318 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3319 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3320 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3321 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3322 "(block count %llu, first data block %u, "
3323 "blocks per group %lu)", sbi->s_groups_count,
3324 ext4_blocks_count(es),
3325 le32_to_cpu(es->s_first_data_block),
3326 EXT4_BLOCKS_PER_GROUP(sb));
3329 sbi->s_groups_count = blocks_count;
3330 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3331 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3332 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3333 EXT4_DESC_PER_BLOCK(sb);
3334 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3336 if (sbi->s_group_desc == NULL) {
3337 ext4_msg(sb, KERN_ERR, "not enough memory");
3341 #ifdef CONFIG_PROC_FS
3343 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3346 bgl_lock_init(sbi->s_blockgroup_lock);
3348 for (i = 0; i < db_count; i++) {
3349 block = descriptor_loc(sb, logical_sb_block, i);
3350 sbi->s_group_desc[i] = sb_bread(sb, block);
3351 if (!sbi->s_group_desc[i]) {
3352 ext4_msg(sb, KERN_ERR,
3353 "can't read group descriptor %d", i);
3358 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3359 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3362 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3363 if (!ext4_fill_flex_info(sb)) {
3364 ext4_msg(sb, KERN_ERR,
3365 "unable to initialize "
3366 "flex_bg meta info!");
3370 sbi->s_gdb_count = db_count;
3371 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3372 spin_lock_init(&sbi->s_next_gen_lock);
3374 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3375 ext4_count_free_blocks(sb));
3377 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3378 ext4_count_free_inodes(sb));
3381 err = percpu_counter_init(&sbi->s_dirs_counter,
3382 ext4_count_dirs(sb));
3385 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3388 ext4_msg(sb, KERN_ERR, "insufficient memory");
3392 sbi->s_stripe = ext4_get_stripe_size(sbi);
3393 sbi->s_max_writeback_mb_bump = 128;
3396 * set up enough so that it can read an inode
3398 if (!test_opt(sb, NOLOAD) &&
3399 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3400 sb->s_op = &ext4_sops;
3402 sb->s_op = &ext4_nojournal_sops;
3403 sb->s_export_op = &ext4_export_ops;
3404 sb->s_xattr = ext4_xattr_handlers;
3406 sb->s_qcop = &ext4_qctl_operations;
3407 sb->dq_op = &ext4_quota_operations;
3409 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3410 mutex_init(&sbi->s_orphan_lock);
3411 mutex_init(&sbi->s_resize_lock);
3415 needs_recovery = (es->s_last_orphan != 0 ||
3416 EXT4_HAS_INCOMPAT_FEATURE(sb,
3417 EXT4_FEATURE_INCOMPAT_RECOVER));
3420 * The first inode we look at is the journal inode. Don't try
3421 * root first: it may be modified in the journal!
3423 if (!test_opt(sb, NOLOAD) &&
3424 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3425 if (ext4_load_journal(sb, es, journal_devnum))
3427 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3428 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3429 ext4_msg(sb, KERN_ERR, "required journal recovery "
3430 "suppressed and not mounted read-only");
3431 goto failed_mount_wq;
3433 clear_opt(sb, DATA_FLAGS);
3434 set_opt(sb, WRITEBACK_DATA);
3435 sbi->s_journal = NULL;
3440 if (ext4_blocks_count(es) > 0xffffffffULL &&
3441 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3442 JBD2_FEATURE_INCOMPAT_64BIT)) {
3443 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3444 goto failed_mount_wq;
3447 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3448 jbd2_journal_set_features(sbi->s_journal,
3449 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3450 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3451 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3452 jbd2_journal_set_features(sbi->s_journal,
3453 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3454 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3455 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3457 jbd2_journal_clear_features(sbi->s_journal,
3458 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3462 /* We have now updated the journal if required, so we can
3463 * validate the data journaling mode. */
3464 switch (test_opt(sb, DATA_FLAGS)) {
3466 /* No mode set, assume a default based on the journal
3467 * capabilities: ORDERED_DATA if the journal can
3468 * cope, else JOURNAL_DATA
3470 if (jbd2_journal_check_available_features
3471 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3472 set_opt(sb, ORDERED_DATA);
3474 set_opt(sb, JOURNAL_DATA);
3477 case EXT4_MOUNT_ORDERED_DATA:
3478 case EXT4_MOUNT_WRITEBACK_DATA:
3479 if (!jbd2_journal_check_available_features
3480 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3481 ext4_msg(sb, KERN_ERR, "Journal does not support "
3482 "requested data journaling mode");
3483 goto failed_mount_wq;
3488 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3491 * The journal may have updated the bg summary counts, so we
3492 * need to update the global counters.
3494 percpu_counter_set(&sbi->s_freeblocks_counter,
3495 ext4_count_free_blocks(sb));
3496 percpu_counter_set(&sbi->s_freeinodes_counter,
3497 ext4_count_free_inodes(sb));
3498 percpu_counter_set(&sbi->s_dirs_counter,
3499 ext4_count_dirs(sb));
3500 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3503 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3504 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3505 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3506 goto failed_mount_wq;
3510 * The jbd2_journal_load will have done any necessary log recovery,
3511 * so we can safely mount the rest of the filesystem now.
3514 root = ext4_iget(sb, EXT4_ROOT_INO);
3516 ext4_msg(sb, KERN_ERR, "get root inode failed");
3517 ret = PTR_ERR(root);
3520 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3522 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3525 sb->s_root = d_alloc_root(root);
3527 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3533 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3535 /* determine the minimum size of new large inodes, if present */
3536 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3537 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3538 EXT4_GOOD_OLD_INODE_SIZE;
3539 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3540 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3541 if (sbi->s_want_extra_isize <
3542 le16_to_cpu(es->s_want_extra_isize))
3543 sbi->s_want_extra_isize =
3544 le16_to_cpu(es->s_want_extra_isize);
3545 if (sbi->s_want_extra_isize <
3546 le16_to_cpu(es->s_min_extra_isize))
3547 sbi->s_want_extra_isize =
3548 le16_to_cpu(es->s_min_extra_isize);
3551 /* Check if enough inode space is available */
3552 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3553 sbi->s_inode_size) {
3554 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3555 EXT4_GOOD_OLD_INODE_SIZE;
3556 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3560 if (test_opt(sb, DELALLOC) &&
3561 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3562 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3563 "requested data journaling mode");
3564 clear_opt(sb, DELALLOC);
3566 if (test_opt(sb, DIOREAD_NOLOCK)) {
3567 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3568 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3569 "option - requested data journaling mode");
3570 clear_opt(sb, DIOREAD_NOLOCK);
3572 if (sb->s_blocksize < PAGE_SIZE) {
3573 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3574 "option - block size is too small");
3575 clear_opt(sb, DIOREAD_NOLOCK);
3579 err = ext4_setup_system_zone(sb);
3581 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3587 err = ext4_mb_init(sb, needs_recovery);
3589 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3594 err = ext4_register_li_request(sb, first_not_zeroed);
3598 sbi->s_kobj.kset = ext4_kset;
3599 init_completion(&sbi->s_kobj_unregister);
3600 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3603 ext4_mb_release(sb);
3604 ext4_ext_release(sb);
3608 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3609 ext4_orphan_cleanup(sb, es);
3610 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3611 if (needs_recovery) {
3612 ext4_msg(sb, KERN_INFO, "recovery complete");
3613 ext4_mark_recovery_complete(sb, es);
3615 if (EXT4_SB(sb)->s_journal) {
3616 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3617 descr = " journalled data mode";
3618 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3619 descr = " ordered data mode";
3621 descr = " writeback data mode";
3623 descr = "out journal";
3625 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3626 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3627 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3629 init_timer(&sbi->s_err_report);
3630 sbi->s_err_report.function = print_daily_error_info;
3631 sbi->s_err_report.data = (unsigned long) sb;
3632 if (es->s_error_count)
3633 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3640 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3644 ext4_msg(sb, KERN_ERR, "mount failed");
3645 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3647 ext4_release_system_zone(sb);
3648 if (sbi->s_journal) {
3649 jbd2_journal_destroy(sbi->s_journal);
3650 sbi->s_journal = NULL;
3653 if (sbi->s_flex_groups) {
3654 if (is_vmalloc_addr(sbi->s_flex_groups))
3655 vfree(sbi->s_flex_groups);
3657 kfree(sbi->s_flex_groups);
3659 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3660 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3661 percpu_counter_destroy(&sbi->s_dirs_counter);
3662 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3664 for (i = 0; i < db_count; i++)
3665 brelse(sbi->s_group_desc[i]);
3666 kfree(sbi->s_group_desc);
3669 remove_proc_entry(sb->s_id, ext4_proc_root);
3672 for (i = 0; i < MAXQUOTAS; i++)
3673 kfree(sbi->s_qf_names[i]);
3675 ext4_blkdev_remove(sbi);
3678 sb->s_fs_info = NULL;
3679 kfree(sbi->s_blockgroup_lock);
3687 * Setup any per-fs journal parameters now. We'll do this both on
3688 * initial mount, once the journal has been initialised but before we've
3689 * done any recovery; and again on any subsequent remount.
3691 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3693 struct ext4_sb_info *sbi = EXT4_SB(sb);
3695 journal->j_commit_interval = sbi->s_commit_interval;
3696 journal->j_min_batch_time = sbi->s_min_batch_time;
3697 journal->j_max_batch_time = sbi->s_max_batch_time;
3699 write_lock(&journal->j_state_lock);
3700 if (test_opt(sb, BARRIER))
3701 journal->j_flags |= JBD2_BARRIER;
3703 journal->j_flags &= ~JBD2_BARRIER;
3704 if (test_opt(sb, DATA_ERR_ABORT))
3705 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3707 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3708 write_unlock(&journal->j_state_lock);
3711 static journal_t *ext4_get_journal(struct super_block *sb,
3712 unsigned int journal_inum)
3714 struct inode *journal_inode;
3717 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3719 /* First, test for the existence of a valid inode on disk. Bad
3720 * things happen if we iget() an unused inode, as the subsequent
3721 * iput() will try to delete it. */
3723 journal_inode = ext4_iget(sb, journal_inum);
3724 if (IS_ERR(journal_inode)) {
3725 ext4_msg(sb, KERN_ERR, "no journal found");
3728 if (!journal_inode->i_nlink) {
3729 make_bad_inode(journal_inode);
3730 iput(journal_inode);
3731 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3735 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3736 journal_inode, journal_inode->i_size);
3737 if (!S_ISREG(journal_inode->i_mode)) {
3738 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3739 iput(journal_inode);
3743 journal = jbd2_journal_init_inode(journal_inode);
3745 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3746 iput(journal_inode);
3749 journal->j_private = sb;
3750 ext4_init_journal_params(sb, journal);
3754 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3757 struct buffer_head *bh;
3761 int hblock, blocksize;
3762 ext4_fsblk_t sb_block;
3763 unsigned long offset;
3764 struct ext4_super_block *es;
3765 struct block_device *bdev;
3767 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3769 bdev = ext4_blkdev_get(j_dev, sb);
3773 if (bd_claim(bdev, sb)) {
3774 ext4_msg(sb, KERN_ERR,
3775 "failed to claim external journal device");
3776 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3780 blocksize = sb->s_blocksize;
3781 hblock = bdev_logical_block_size(bdev);
3782 if (blocksize < hblock) {
3783 ext4_msg(sb, KERN_ERR,
3784 "blocksize too small for journal device");
3788 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3789 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3790 set_blocksize(bdev, blocksize);
3791 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3792 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3793 "external journal");
3797 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3798 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3799 !(le32_to_cpu(es->s_feature_incompat) &
3800 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3801 ext4_msg(sb, KERN_ERR, "external journal has "
3807 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3808 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3813 len = ext4_blocks_count(es);
3814 start = sb_block + 1;
3815 brelse(bh); /* we're done with the superblock */
3817 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3818 start, len, blocksize);
3820 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3823 journal->j_private = sb;
3824 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3825 wait_on_buffer(journal->j_sb_buffer);
3826 if (!buffer_uptodate(journal->j_sb_buffer)) {
3827 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3830 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3831 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3832 "user (unsupported) - %d",
3833 be32_to_cpu(journal->j_superblock->s_nr_users));
3836 EXT4_SB(sb)->journal_bdev = bdev;
3837 ext4_init_journal_params(sb, journal);
3841 jbd2_journal_destroy(journal);
3843 ext4_blkdev_put(bdev);
3847 static int ext4_load_journal(struct super_block *sb,
3848 struct ext4_super_block *es,
3849 unsigned long journal_devnum)
3852 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3855 int really_read_only;
3857 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3859 if (journal_devnum &&
3860 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3861 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3862 "numbers have changed");
3863 journal_dev = new_decode_dev(journal_devnum);
3865 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3867 really_read_only = bdev_read_only(sb->s_bdev);
3870 * Are we loading a blank journal or performing recovery after a
3871 * crash? For recovery, we need to check in advance whether we
3872 * can get read-write access to the device.
3874 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3875 if (sb->s_flags & MS_RDONLY) {
3876 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3877 "required on readonly filesystem");
3878 if (really_read_only) {
3879 ext4_msg(sb, KERN_ERR, "write access "
3880 "unavailable, cannot proceed");
3883 ext4_msg(sb, KERN_INFO, "write access will "
3884 "be enabled during recovery");
3888 if (journal_inum && journal_dev) {
3889 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3890 "and inode journals!");
3895 if (!(journal = ext4_get_journal(sb, journal_inum)))
3898 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3902 if (!(journal->j_flags & JBD2_BARRIER))
3903 ext4_msg(sb, KERN_INFO, "barriers disabled");
3905 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3906 err = jbd2_journal_update_format(journal);
3908 ext4_msg(sb, KERN_ERR, "error updating journal");
3909 jbd2_journal_destroy(journal);
3914 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3915 err = jbd2_journal_wipe(journal, !really_read_only);
3917 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3919 memcpy(save, ((char *) es) +
3920 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3921 err = jbd2_journal_load(journal);
3923 memcpy(((char *) es) + EXT4_S_ERR_START,
3924 save, EXT4_S_ERR_LEN);
3929 ext4_msg(sb, KERN_ERR, "error loading journal");
3930 jbd2_journal_destroy(journal);
3934 EXT4_SB(sb)->s_journal = journal;
3935 ext4_clear_journal_err(sb, es);
3937 if (!really_read_only && journal_devnum &&
3938 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3939 es->s_journal_dev = cpu_to_le32(journal_devnum);
3941 /* Make sure we flush the recovery flag to disk. */
3942 ext4_commit_super(sb, 1);
3948 static int ext4_commit_super(struct super_block *sb, int sync)
3950 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3951 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3956 if (buffer_write_io_error(sbh)) {
3958 * Oh, dear. A previous attempt to write the
3959 * superblock failed. This could happen because the
3960 * USB device was yanked out. Or it could happen to
3961 * be a transient write error and maybe the block will
3962 * be remapped. Nothing we can do but to retry the
3963 * write and hope for the best.
3965 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3966 "superblock detected");
3967 clear_buffer_write_io_error(sbh);
3968 set_buffer_uptodate(sbh);
3971 * If the file system is mounted read-only, don't update the
3972 * superblock write time. This avoids updating the superblock
3973 * write time when we are mounting the root file system
3974 * read/only but we need to replay the journal; at that point,
3975 * for people who are east of GMT and who make their clock
3976 * tick in localtime for Windows bug-for-bug compatibility,
3977 * the clock is set in the future, and this will cause e2fsck
3978 * to complain and force a full file system check.
3980 if (!(sb->s_flags & MS_RDONLY))
3981 es->s_wtime = cpu_to_le32(get_seconds());
3982 if (sb->s_bdev->bd_part)
3983 es->s_kbytes_written =
3984 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3985 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3986 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3988 es->s_kbytes_written =
3989 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3990 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3991 &EXT4_SB(sb)->s_freeblocks_counter));
3992 es->s_free_inodes_count =
3993 cpu_to_le32(percpu_counter_sum_positive(
3994 &EXT4_SB(sb)->s_freeinodes_counter));
3996 BUFFER_TRACE(sbh, "marking dirty");
3997 mark_buffer_dirty(sbh);
3999 error = sync_dirty_buffer(sbh);
4003 error = buffer_write_io_error(sbh);
4005 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4007 clear_buffer_write_io_error(sbh);
4008 set_buffer_uptodate(sbh);
4015 * Have we just finished recovery? If so, and if we are mounting (or
4016 * remounting) the filesystem readonly, then we will end up with a
4017 * consistent fs on disk. Record that fact.
4019 static void ext4_mark_recovery_complete(struct super_block *sb,
4020 struct ext4_super_block *es)
4022 journal_t *journal = EXT4_SB(sb)->s_journal;
4024 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4025 BUG_ON(journal != NULL);
4028 jbd2_journal_lock_updates(journal);
4029 if (jbd2_journal_flush(journal) < 0)
4032 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4033 sb->s_flags & MS_RDONLY) {
4034 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4035 ext4_commit_super(sb, 1);
4039 jbd2_journal_unlock_updates(journal);
4043 * If we are mounting (or read-write remounting) a filesystem whose journal
4044 * has recorded an error from a previous lifetime, move that error to the
4045 * main filesystem now.
4047 static void ext4_clear_journal_err(struct super_block *sb,
4048 struct ext4_super_block *es)
4054 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4056 journal = EXT4_SB(sb)->s_journal;
4059 * Now check for any error status which may have been recorded in the
4060 * journal by a prior ext4_error() or ext4_abort()
4063 j_errno = jbd2_journal_errno(journal);
4067 errstr = ext4_decode_error(sb, j_errno, nbuf);
4068 ext4_warning(sb, "Filesystem error recorded "
4069 "from previous mount: %s", errstr);
4070 ext4_warning(sb, "Marking fs in need of filesystem check.");
4072 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4073 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4074 ext4_commit_super(sb, 1);
4076 jbd2_journal_clear_err(journal);
4081 * Force the running and committing transactions to commit,
4082 * and wait on the commit.
4084 int ext4_force_commit(struct super_block *sb)
4089 if (sb->s_flags & MS_RDONLY)
4092 journal = EXT4_SB(sb)->s_journal;
4094 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4095 ret = ext4_journal_force_commit(journal);
4101 static void ext4_write_super(struct super_block *sb)
4104 ext4_commit_super(sb, 1);
4108 static int ext4_sync_fs(struct super_block *sb, int wait)
4112 struct ext4_sb_info *sbi = EXT4_SB(sb);
4114 trace_ext4_sync_fs(sb, wait);
4115 flush_workqueue(sbi->dio_unwritten_wq);
4116 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4118 jbd2_log_wait_commit(sbi->s_journal, target);
4124 * LVM calls this function before a (read-only) snapshot is created. This
4125 * gives us a chance to flush the journal completely and mark the fs clean.
4127 static int ext4_freeze(struct super_block *sb)
4132 if (sb->s_flags & MS_RDONLY)
4135 journal = EXT4_SB(sb)->s_journal;
4137 /* Now we set up the journal barrier. */
4138 jbd2_journal_lock_updates(journal);
4141 * Don't clear the needs_recovery flag if we failed to flush
4144 error = jbd2_journal_flush(journal);
4148 /* Journal blocked and flushed, clear needs_recovery flag. */
4149 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4150 error = ext4_commit_super(sb, 1);
4152 /* we rely on s_frozen to stop further updates */
4153 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4158 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4159 * flag here, even though the filesystem is not technically dirty yet.
4161 static int ext4_unfreeze(struct super_block *sb)
4163 if (sb->s_flags & MS_RDONLY)
4167 /* Reset the needs_recovery flag before the fs is unlocked. */
4168 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4169 ext4_commit_super(sb, 1);
4175 * Structure to save mount options for ext4_remount's benefit
4177 struct ext4_mount_options {
4178 unsigned long s_mount_opt;
4179 unsigned long s_mount_opt2;
4182 unsigned long s_commit_interval;
4183 u32 s_min_batch_time, s_max_batch_time;
4186 char *s_qf_names[MAXQUOTAS];
4190 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4192 struct ext4_super_block *es;
4193 struct ext4_sb_info *sbi = EXT4_SB(sb);
4194 ext4_fsblk_t n_blocks_count = 0;
4195 unsigned long old_sb_flags;
4196 struct ext4_mount_options old_opts;
4197 int enable_quota = 0;
4199 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4204 char *orig_data = kstrdup(data, GFP_KERNEL);
4206 /* Store the original options */
4208 old_sb_flags = sb->s_flags;
4209 old_opts.s_mount_opt = sbi->s_mount_opt;
4210 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4211 old_opts.s_resuid = sbi->s_resuid;
4212 old_opts.s_resgid = sbi->s_resgid;
4213 old_opts.s_commit_interval = sbi->s_commit_interval;
4214 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4215 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4217 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4218 for (i = 0; i < MAXQUOTAS; i++)
4219 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4221 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4222 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4225 * Allow the "check" option to be passed as a remount option.
4227 if (!parse_options(data, sb, NULL, &journal_ioprio,
4228 &n_blocks_count, 1)) {
4233 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4234 ext4_abort(sb, "Abort forced by user");
4236 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4237 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4241 if (sbi->s_journal) {
4242 ext4_init_journal_params(sb, sbi->s_journal);
4243 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4246 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4247 n_blocks_count > ext4_blocks_count(es)) {
4248 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4253 if (*flags & MS_RDONLY) {
4254 err = dquot_suspend(sb, -1);
4259 * First of all, the unconditional stuff we have to do
4260 * to disable replay of the journal when we next remount
4262 sb->s_flags |= MS_RDONLY;
4265 * OK, test if we are remounting a valid rw partition
4266 * readonly, and if so set the rdonly flag and then
4267 * mark the partition as valid again.
4269 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4270 (sbi->s_mount_state & EXT4_VALID_FS))
4271 es->s_state = cpu_to_le16(sbi->s_mount_state);
4274 ext4_mark_recovery_complete(sb, es);
4276 /* Make sure we can mount this feature set readwrite */
4277 if (!ext4_feature_set_ok(sb, 0)) {
4282 * Make sure the group descriptor checksums
4283 * are sane. If they aren't, refuse to remount r/w.
4285 for (g = 0; g < sbi->s_groups_count; g++) {
4286 struct ext4_group_desc *gdp =
4287 ext4_get_group_desc(sb, g, NULL);
4289 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4290 ext4_msg(sb, KERN_ERR,
4291 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4292 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4293 le16_to_cpu(gdp->bg_checksum));
4300 * If we have an unprocessed orphan list hanging
4301 * around from a previously readonly bdev mount,
4302 * require a full umount/remount for now.
4304 if (es->s_last_orphan) {
4305 ext4_msg(sb, KERN_WARNING, "Couldn't "
4306 "remount RDWR because of unprocessed "
4307 "orphan inode list. Please "
4308 "umount/remount instead");
4314 * Mounting a RDONLY partition read-write, so reread
4315 * and store the current valid flag. (It may have
4316 * been changed by e2fsck since we originally mounted
4320 ext4_clear_journal_err(sb, es);
4321 sbi->s_mount_state = le16_to_cpu(es->s_state);
4322 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4324 if (!ext4_setup_super(sb, es, 0))
4325 sb->s_flags &= ~MS_RDONLY;
4331 * Reinitialize lazy itable initialization thread based on
4334 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4335 ext4_unregister_li_request(sb);
4337 ext4_group_t first_not_zeroed;
4338 first_not_zeroed = ext4_has_uninit_itable(sb);
4339 ext4_register_li_request(sb, first_not_zeroed);
4342 ext4_setup_system_zone(sb);
4343 if (sbi->s_journal == NULL)
4344 ext4_commit_super(sb, 1);
4347 /* Release old quota file names */
4348 for (i = 0; i < MAXQUOTAS; i++)
4349 if (old_opts.s_qf_names[i] &&
4350 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4351 kfree(old_opts.s_qf_names[i]);
4355 dquot_resume(sb, -1);
4357 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4362 sb->s_flags = old_sb_flags;
4363 sbi->s_mount_opt = old_opts.s_mount_opt;
4364 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4365 sbi->s_resuid = old_opts.s_resuid;
4366 sbi->s_resgid = old_opts.s_resgid;
4367 sbi->s_commit_interval = old_opts.s_commit_interval;
4368 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4369 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4371 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4372 for (i = 0; i < MAXQUOTAS; i++) {
4373 if (sbi->s_qf_names[i] &&
4374 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4375 kfree(sbi->s_qf_names[i]);
4376 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4384 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4386 struct super_block *sb = dentry->d_sb;
4387 struct ext4_sb_info *sbi = EXT4_SB(sb);
4388 struct ext4_super_block *es = sbi->s_es;
4391 if (test_opt(sb, MINIX_DF)) {
4392 sbi->s_overhead_last = 0;
4393 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4394 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4395 ext4_fsblk_t overhead = 0;
4398 * Compute the overhead (FS structures). This is constant
4399 * for a given filesystem unless the number of block groups
4400 * changes so we cache the previous value until it does.
4404 * All of the blocks before first_data_block are
4407 overhead = le32_to_cpu(es->s_first_data_block);
4410 * Add the overhead attributed to the superblock and
4411 * block group descriptors. If the sparse superblocks
4412 * feature is turned on, then not all groups have this.
4414 for (i = 0; i < ngroups; i++) {
4415 overhead += ext4_bg_has_super(sb, i) +
4416 ext4_bg_num_gdb(sb, i);
4421 * Every block group has an inode bitmap, a block
4422 * bitmap, and an inode table.
4424 overhead += ngroups * (2 + sbi->s_itb_per_group);
4425 sbi->s_overhead_last = overhead;
4427 sbi->s_blocks_last = ext4_blocks_count(es);
4430 buf->f_type = EXT4_SUPER_MAGIC;
4431 buf->f_bsize = sb->s_blocksize;
4432 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4433 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4434 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4435 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4436 if (buf->f_bfree < ext4_r_blocks_count(es))
4438 buf->f_files = le32_to_cpu(es->s_inodes_count);
4439 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4440 buf->f_namelen = EXT4_NAME_LEN;
4441 fsid = le64_to_cpup((void *)es->s_uuid) ^
4442 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4443 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4444 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4449 /* Helper function for writing quotas on sync - we need to start transaction
4450 * before quota file is locked for write. Otherwise the are possible deadlocks:
4451 * Process 1 Process 2
4452 * ext4_create() quota_sync()
4453 * jbd2_journal_start() write_dquot()
4454 * dquot_initialize() down(dqio_mutex)
4455 * down(dqio_mutex) jbd2_journal_start()
4461 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4463 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4466 static int ext4_write_dquot(struct dquot *dquot)
4470 struct inode *inode;
4472 inode = dquot_to_inode(dquot);
4473 handle = ext4_journal_start(inode,
4474 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4476 return PTR_ERR(handle);
4477 ret = dquot_commit(dquot);
4478 err = ext4_journal_stop(handle);
4484 static int ext4_acquire_dquot(struct dquot *dquot)
4489 handle = ext4_journal_start(dquot_to_inode(dquot),
4490 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4492 return PTR_ERR(handle);
4493 ret = dquot_acquire(dquot);
4494 err = ext4_journal_stop(handle);
4500 static int ext4_release_dquot(struct dquot *dquot)
4505 handle = ext4_journal_start(dquot_to_inode(dquot),
4506 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4507 if (IS_ERR(handle)) {
4508 /* Release dquot anyway to avoid endless cycle in dqput() */
4509 dquot_release(dquot);
4510 return PTR_ERR(handle);
4512 ret = dquot_release(dquot);
4513 err = ext4_journal_stop(handle);
4519 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4521 /* Are we journaling quotas? */
4522 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4523 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4524 dquot_mark_dquot_dirty(dquot);
4525 return ext4_write_dquot(dquot);
4527 return dquot_mark_dquot_dirty(dquot);
4531 static int ext4_write_info(struct super_block *sb, int type)
4536 /* Data block + inode block */
4537 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4539 return PTR_ERR(handle);
4540 ret = dquot_commit_info(sb, type);
4541 err = ext4_journal_stop(handle);
4548 * Turn on quotas during mount time - we need to find
4549 * the quota file and such...
4551 static int ext4_quota_on_mount(struct super_block *sb, int type)
4553 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4554 EXT4_SB(sb)->s_jquota_fmt, type);
4558 * Standard function to be called on quota_on
4560 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4566 if (!test_opt(sb, QUOTA))
4569 err = kern_path(name, LOOKUP_FOLLOW, &path);
4573 /* Quotafile not on the same filesystem? */
4574 if (path.mnt->mnt_sb != sb) {
4578 /* Journaling quota? */
4579 if (EXT4_SB(sb)->s_qf_names[type]) {
4580 /* Quotafile not in fs root? */
4581 if (path.dentry->d_parent != sb->s_root)
4582 ext4_msg(sb, KERN_WARNING,
4583 "Quota file not on filesystem root. "
4584 "Journaled quota will not work");
4588 * When we journal data on quota file, we have to flush journal to see
4589 * all updates to the file when we bypass pagecache...
4591 if (EXT4_SB(sb)->s_journal &&
4592 ext4_should_journal_data(path.dentry->d_inode)) {
4594 * We don't need to lock updates but journal_flush() could
4595 * otherwise be livelocked...
4597 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4598 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4599 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4606 err = dquot_quota_on_path(sb, type, format_id, &path);
4611 static int ext4_quota_off(struct super_block *sb, int type)
4613 /* Force all delayed allocation blocks to be allocated.
4614 * Caller already holds s_umount sem */
4615 if (test_opt(sb, DELALLOC))
4616 sync_filesystem(sb);
4618 return dquot_quota_off(sb, type);
4621 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4622 * acquiring the locks... As quota files are never truncated and quota code
4623 * itself serializes the operations (and noone else should touch the files)
4624 * we don't have to be afraid of races */
4625 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4626 size_t len, loff_t off)
4628 struct inode *inode = sb_dqopt(sb)->files[type];
4629 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4631 int offset = off & (sb->s_blocksize - 1);
4634 struct buffer_head *bh;
4635 loff_t i_size = i_size_read(inode);
4639 if (off+len > i_size)
4642 while (toread > 0) {
4643 tocopy = sb->s_blocksize - offset < toread ?
4644 sb->s_blocksize - offset : toread;
4645 bh = ext4_bread(NULL, inode, blk, 0, &err);
4648 if (!bh) /* A hole? */
4649 memset(data, 0, tocopy);
4651 memcpy(data, bh->b_data+offset, tocopy);
4661 /* Write to quotafile (we know the transaction is already started and has
4662 * enough credits) */
4663 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4664 const char *data, size_t len, loff_t off)
4666 struct inode *inode = sb_dqopt(sb)->files[type];
4667 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4669 int offset = off & (sb->s_blocksize - 1);
4670 struct buffer_head *bh;
4671 handle_t *handle = journal_current_handle();
4673 if (EXT4_SB(sb)->s_journal && !handle) {
4674 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4675 " cancelled because transaction is not started",
4676 (unsigned long long)off, (unsigned long long)len);
4680 * Since we account only one data block in transaction credits,
4681 * then it is impossible to cross a block boundary.
4683 if (sb->s_blocksize - offset < len) {
4684 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4685 " cancelled because not block aligned",
4686 (unsigned long long)off, (unsigned long long)len);
4690 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4691 bh = ext4_bread(handle, inode, blk, 1, &err);
4694 err = ext4_journal_get_write_access(handle, bh);
4700 memcpy(bh->b_data+offset, data, len);
4701 flush_dcache_page(bh->b_page);
4703 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4707 mutex_unlock(&inode->i_mutex);
4710 if (inode->i_size < off + len) {
4711 i_size_write(inode, off + len);
4712 EXT4_I(inode)->i_disksize = inode->i_size;
4714 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4715 ext4_mark_inode_dirty(handle, inode);
4716 mutex_unlock(&inode->i_mutex);
4722 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4723 const char *dev_name, void *data)
4725 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4728 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4729 static struct file_system_type ext2_fs_type = {
4730 .owner = THIS_MODULE,
4732 .mount = ext4_mount,
4733 .kill_sb = kill_block_super,
4734 .fs_flags = FS_REQUIRES_DEV,
4737 static inline void register_as_ext2(void)
4739 int err = register_filesystem(&ext2_fs_type);
4742 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4745 static inline void unregister_as_ext2(void)
4747 unregister_filesystem(&ext2_fs_type);
4749 MODULE_ALIAS("ext2");
4751 static inline void register_as_ext2(void) { }
4752 static inline void unregister_as_ext2(void) { }
4755 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4756 static inline void register_as_ext3(void)
4758 int err = register_filesystem(&ext3_fs_type);
4761 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4764 static inline void unregister_as_ext3(void)
4766 unregister_filesystem(&ext3_fs_type);
4768 MODULE_ALIAS("ext3");
4770 static inline void register_as_ext3(void) { }
4771 static inline void unregister_as_ext3(void) { }
4774 static struct file_system_type ext4_fs_type = {
4775 .owner = THIS_MODULE,
4777 .mount = ext4_mount,
4778 .kill_sb = kill_block_super,
4779 .fs_flags = FS_REQUIRES_DEV,
4782 int __init ext4_init_feat_adverts(void)
4784 struct ext4_features *ef;
4787 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4791 ef->f_kobj.kset = ext4_kset;
4792 init_completion(&ef->f_kobj_unregister);
4793 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4806 static int __init ext4_init_fs(void)
4810 ext4_check_flag_values();
4811 err = ext4_init_pageio();
4814 err = ext4_init_system_zone();
4817 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4820 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4822 err = ext4_init_feat_adverts();
4824 err = ext4_init_mballoc();
4828 err = ext4_init_xattr();
4831 err = init_inodecache();
4836 err = register_filesystem(&ext4_fs_type);
4840 ext4_li_info = NULL;
4841 mutex_init(&ext4_li_mtx);
4844 unregister_as_ext2();
4845 unregister_as_ext3();
4846 destroy_inodecache();
4850 ext4_exit_mballoc();
4853 remove_proc_entry("fs/ext4", NULL);
4854 kset_unregister(ext4_kset);
4856 ext4_exit_system_zone();
4862 static void __exit ext4_exit_fs(void)
4864 ext4_destroy_lazyinit_thread();
4865 unregister_as_ext2();
4866 unregister_as_ext3();
4867 unregister_filesystem(&ext4_fs_type);
4868 destroy_inodecache();
4870 ext4_exit_mballoc();
4871 remove_proc_entry("fs/ext4", NULL);
4872 kset_unregister(ext4_kset);
4873 ext4_exit_system_zone();
4877 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4878 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4879 MODULE_LICENSE("GPL");
4880 module_init(ext4_init_fs)
4881 module_exit(ext4_exit_fs)