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 va_format vaf;
410 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
412 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
413 es->s_last_error_block = cpu_to_le64(block);
414 save_error_info(inode->i_sb, function, line);
418 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
419 inode->i_sb->s_id, function, line, inode->i_ino);
421 printk(KERN_CONT "block %llu: ", block);
422 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
425 ext4_handle_error(inode->i_sb);
428 void ext4_error_file(struct file *file, const char *function,
429 unsigned int line, ext4_fsblk_t block,
430 const char *fmt, ...)
433 struct va_format vaf;
434 struct ext4_super_block *es;
435 struct inode *inode = file->f_dentry->d_inode;
436 char pathname[80], *path;
438 es = EXT4_SB(inode->i_sb)->s_es;
439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 save_error_info(inode->i_sb, function, line);
441 path = d_path(&(file->f_path), pathname, sizeof(pathname));
445 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
446 inode->i_sb->s_id, function, line, inode->i_ino);
448 printk(KERN_CONT "block %llu: ", block);
452 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
455 ext4_handle_error(inode->i_sb);
458 static const char *ext4_decode_error(struct super_block *sb, int errno,
465 errstr = "IO failure";
468 errstr = "Out of memory";
471 if (!sb || (EXT4_SB(sb)->s_journal &&
472 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
473 errstr = "Journal has aborted";
475 errstr = "Readonly filesystem";
478 /* If the caller passed in an extra buffer for unknown
479 * errors, textualise them now. Else we just return
482 /* Check for truncated error codes... */
483 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
492 /* __ext4_std_error decodes expected errors from journaling functions
493 * automatically and invokes the appropriate error response. */
495 void __ext4_std_error(struct super_block *sb, const char *function,
496 unsigned int line, int errno)
501 /* Special case: if the error is EROFS, and we're not already
502 * inside a transaction, then there's really no point in logging
504 if (errno == -EROFS && journal_current_handle() == NULL &&
505 (sb->s_flags & MS_RDONLY))
508 errstr = ext4_decode_error(sb, errno, nbuf);
509 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
510 sb->s_id, function, line, errstr);
511 save_error_info(sb, function, line);
513 ext4_handle_error(sb);
517 * ext4_abort is a much stronger failure handler than ext4_error. The
518 * abort function may be used to deal with unrecoverable failures such
519 * as journal IO errors or ENOMEM at a critical moment in log management.
521 * We unconditionally force the filesystem into an ABORT|READONLY state,
522 * unless the error response on the fs has been set to panic in which
523 * case we take the easy way out and panic immediately.
526 void __ext4_abort(struct super_block *sb, const char *function,
527 unsigned int line, const char *fmt, ...)
531 save_error_info(sb, function, line);
533 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
539 if ((sb->s_flags & MS_RDONLY) == 0) {
540 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
541 sb->s_flags |= MS_RDONLY;
542 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
543 if (EXT4_SB(sb)->s_journal)
544 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
545 save_error_info(sb, function, line);
547 if (test_opt(sb, ERRORS_PANIC))
548 panic("EXT4-fs panic from previous error\n");
551 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
553 struct va_format vaf;
559 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
563 void __ext4_warning(struct super_block *sb, const char *function,
564 unsigned int line, const char *fmt, ...)
566 struct va_format vaf;
572 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
573 sb->s_id, function, line, &vaf);
577 void __ext4_grp_locked_error(const char *function, unsigned int line,
578 struct super_block *sb, ext4_group_t grp,
579 unsigned long ino, ext4_fsblk_t block,
580 const char *fmt, ...)
584 struct va_format vaf;
586 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
588 es->s_last_error_ino = cpu_to_le32(ino);
589 es->s_last_error_block = cpu_to_le64(block);
590 __save_error_info(sb, function, line);
596 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
597 sb->s_id, function, line, grp);
599 printk(KERN_CONT "inode %lu: ", ino);
601 printk(KERN_CONT "block %llu:", (unsigned long long) block);
602 printk(KERN_CONT "%pV\n", &vaf);
605 if (test_opt(sb, ERRORS_CONT)) {
606 ext4_commit_super(sb, 0);
610 ext4_unlock_group(sb, grp);
611 ext4_handle_error(sb);
613 * We only get here in the ERRORS_RO case; relocking the group
614 * may be dangerous, but nothing bad will happen since the
615 * filesystem will have already been marked read/only and the
616 * journal has been aborted. We return 1 as a hint to callers
617 * who might what to use the return value from
618 * ext4_grp_locked_error() to distinguish beween the
619 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
620 * aggressively from the ext4 function in question, with a
621 * more appropriate error code.
623 ext4_lock_group(sb, grp);
627 void ext4_update_dynamic_rev(struct super_block *sb)
629 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
631 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
635 "updating to rev %d because of new feature flag, "
636 "running e2fsck is recommended",
639 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
640 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
641 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
642 /* leave es->s_feature_*compat flags alone */
643 /* es->s_uuid will be set by e2fsck if empty */
646 * The rest of the superblock fields should be zero, and if not it
647 * means they are likely already in use, so leave them alone. We
648 * can leave it up to e2fsck to clean up any inconsistencies there.
653 * Open the external journal device
655 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
657 struct block_device *bdev;
658 char b[BDEVNAME_SIZE];
660 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
666 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
667 __bdevname(dev, b), PTR_ERR(bdev));
672 * Release the journal device
674 static int ext4_blkdev_put(struct block_device *bdev)
677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
680 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
682 struct block_device *bdev;
685 bdev = sbi->journal_bdev;
687 ret = ext4_blkdev_put(bdev);
688 sbi->journal_bdev = NULL;
693 static inline struct inode *orphan_list_entry(struct list_head *l)
695 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
698 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
702 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
703 le32_to_cpu(sbi->s_es->s_last_orphan));
705 printk(KERN_ERR "sb_info orphan list:\n");
706 list_for_each(l, &sbi->s_orphan) {
707 struct inode *inode = orphan_list_entry(l);
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode->i_sb->s_id, inode->i_ino, inode,
711 inode->i_mode, inode->i_nlink,
716 static void ext4_put_super(struct super_block *sb)
718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_super_block *es = sbi->s_es;
722 ext4_unregister_li_request(sb);
723 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
725 flush_workqueue(sbi->dio_unwritten_wq);
726 destroy_workqueue(sbi->dio_unwritten_wq);
730 ext4_commit_super(sb, 1);
732 if (sbi->s_journal) {
733 err = jbd2_journal_destroy(sbi->s_journal);
734 sbi->s_journal = NULL;
736 ext4_abort(sb, "Couldn't clean up the journal");
739 del_timer(&sbi->s_err_report);
740 ext4_release_system_zone(sb);
742 ext4_ext_release(sb);
743 ext4_xattr_put_super(sb);
745 if (!(sb->s_flags & MS_RDONLY)) {
746 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
747 es->s_state = cpu_to_le16(sbi->s_mount_state);
748 ext4_commit_super(sb, 1);
751 remove_proc_entry(sb->s_id, ext4_proc_root);
753 kobject_del(&sbi->s_kobj);
755 for (i = 0; i < sbi->s_gdb_count; i++)
756 brelse(sbi->s_group_desc[i]);
757 kfree(sbi->s_group_desc);
758 if (is_vmalloc_addr(sbi->s_flex_groups))
759 vfree(sbi->s_flex_groups);
761 kfree(sbi->s_flex_groups);
762 percpu_counter_destroy(&sbi->s_freeblocks_counter);
763 percpu_counter_destroy(&sbi->s_freeinodes_counter);
764 percpu_counter_destroy(&sbi->s_dirs_counter);
765 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
768 for (i = 0; i < MAXQUOTAS; i++)
769 kfree(sbi->s_qf_names[i]);
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi->s_orphan))
777 dump_orphan_list(sb, sbi);
778 J_ASSERT(list_empty(&sbi->s_orphan));
780 invalidate_bdev(sb->s_bdev);
781 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
787 sync_blockdev(sbi->journal_bdev);
788 invalidate_bdev(sbi->journal_bdev);
789 ext4_blkdev_remove(sbi);
791 sb->s_fs_info = NULL;
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
797 kobject_put(&sbi->s_kobj);
798 wait_for_completion(&sbi->s_kobj_unregister);
799 kfree(sbi->s_blockgroup_lock);
803 static struct kmem_cache *ext4_inode_cachep;
806 * Called inside transaction, so use GFP_NOFS
808 static struct inode *ext4_alloc_inode(struct super_block *sb)
810 struct ext4_inode_info *ei;
812 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
816 ei->vfs_inode.i_version = 1;
817 ei->vfs_inode.i_data.writeback_index = 0;
818 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
819 INIT_LIST_HEAD(&ei->i_prealloc_list);
820 spin_lock_init(&ei->i_prealloc_lock);
822 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
823 * therefore it can be null here. Don't check it, just initialize
826 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
827 ei->i_reserved_data_blocks = 0;
828 ei->i_reserved_meta_blocks = 0;
829 ei->i_allocated_meta_blocks = 0;
830 ei->i_da_metadata_calc_len = 0;
831 ei->i_delalloc_reserved_flag = 0;
832 spin_lock_init(&(ei->i_block_reservation_lock));
834 ei->i_reserved_quota = 0;
836 INIT_LIST_HEAD(&ei->i_completed_io_list);
837 spin_lock_init(&ei->i_completed_io_lock);
838 ei->cur_aio_dio = NULL;
840 ei->i_datasync_tid = 0;
841 atomic_set(&ei->i_ioend_count, 0);
843 return &ei->vfs_inode;
846 static int ext4_drop_inode(struct inode *inode)
848 int drop = generic_drop_inode(inode);
850 trace_ext4_drop_inode(inode, drop);
854 static void ext4_destroy_inode(struct inode *inode)
856 ext4_ioend_wait(inode);
857 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
858 ext4_msg(inode->i_sb, KERN_ERR,
859 "Inode %lu (%p): orphan list check failed!",
860 inode->i_ino, EXT4_I(inode));
861 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
862 EXT4_I(inode), sizeof(struct ext4_inode_info),
866 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
869 static void init_once(void *foo)
871 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
873 INIT_LIST_HEAD(&ei->i_orphan);
874 #ifdef CONFIG_EXT4_FS_XATTR
875 init_rwsem(&ei->xattr_sem);
877 init_rwsem(&ei->i_data_sem);
878 inode_init_once(&ei->vfs_inode);
881 static int init_inodecache(void)
883 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
884 sizeof(struct ext4_inode_info),
885 0, (SLAB_RECLAIM_ACCOUNT|
888 if (ext4_inode_cachep == NULL)
893 static void destroy_inodecache(void)
895 kmem_cache_destroy(ext4_inode_cachep);
898 void ext4_clear_inode(struct inode *inode)
900 invalidate_inode_buffers(inode);
901 end_writeback(inode);
903 ext4_discard_preallocations(inode);
904 if (EXT4_JOURNAL(inode))
905 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
906 &EXT4_I(inode)->jinode);
909 static inline void ext4_show_quota_options(struct seq_file *seq,
910 struct super_block *sb)
912 #if defined(CONFIG_QUOTA)
913 struct ext4_sb_info *sbi = EXT4_SB(sb);
915 if (sbi->s_jquota_fmt) {
918 switch (sbi->s_jquota_fmt) {
929 seq_printf(seq, ",jqfmt=%s", fmtname);
932 if (sbi->s_qf_names[USRQUOTA])
933 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
935 if (sbi->s_qf_names[GRPQUOTA])
936 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
938 if (test_opt(sb, USRQUOTA))
939 seq_puts(seq, ",usrquota");
941 if (test_opt(sb, GRPQUOTA))
942 seq_puts(seq, ",grpquota");
948 * - it's set to a non-default value OR
949 * - if the per-sb default is different from the global default
951 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
954 unsigned long def_mount_opts;
955 struct super_block *sb = vfs->mnt_sb;
956 struct ext4_sb_info *sbi = EXT4_SB(sb);
957 struct ext4_super_block *es = sbi->s_es;
959 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
960 def_errors = le16_to_cpu(es->s_errors);
962 if (sbi->s_sb_block != 1)
963 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
964 if (test_opt(sb, MINIX_DF))
965 seq_puts(seq, ",minixdf");
966 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
967 seq_puts(seq, ",grpid");
968 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
969 seq_puts(seq, ",nogrpid");
970 if (sbi->s_resuid != EXT4_DEF_RESUID ||
971 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
972 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
974 if (sbi->s_resgid != EXT4_DEF_RESGID ||
975 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
976 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
978 if (test_opt(sb, ERRORS_RO)) {
979 if (def_errors == EXT4_ERRORS_PANIC ||
980 def_errors == EXT4_ERRORS_CONTINUE) {
981 seq_puts(seq, ",errors=remount-ro");
984 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
985 seq_puts(seq, ",errors=continue");
986 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
987 seq_puts(seq, ",errors=panic");
988 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
989 seq_puts(seq, ",nouid32");
990 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
991 seq_puts(seq, ",debug");
992 if (test_opt(sb, OLDALLOC))
993 seq_puts(seq, ",oldalloc");
994 #ifdef CONFIG_EXT4_FS_XATTR
995 if (test_opt(sb, XATTR_USER) &&
996 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
997 seq_puts(seq, ",user_xattr");
998 if (!test_opt(sb, XATTR_USER) &&
999 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
1000 seq_puts(seq, ",nouser_xattr");
1003 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1004 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1005 seq_puts(seq, ",acl");
1006 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1007 seq_puts(seq, ",noacl");
1009 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1010 seq_printf(seq, ",commit=%u",
1011 (unsigned) (sbi->s_commit_interval / HZ));
1013 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1014 seq_printf(seq, ",min_batch_time=%u",
1015 (unsigned) sbi->s_min_batch_time);
1017 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1018 seq_printf(seq, ",max_batch_time=%u",
1019 (unsigned) sbi->s_min_batch_time);
1023 * We're changing the default of barrier mount option, so
1024 * let's always display its mount state so it's clear what its
1027 seq_puts(seq, ",barrier=");
1028 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1029 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1030 seq_puts(seq, ",journal_async_commit");
1031 else if (test_opt(sb, JOURNAL_CHECKSUM))
1032 seq_puts(seq, ",journal_checksum");
1033 if (test_opt(sb, I_VERSION))
1034 seq_puts(seq, ",i_version");
1035 if (!test_opt(sb, DELALLOC) &&
1036 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1037 seq_puts(seq, ",nodelalloc");
1039 if (test_opt(sb, MBLK_IO_SUBMIT))
1040 seq_puts(seq, ",mblk_io_submit");
1042 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1044 * journal mode get enabled in different ways
1045 * So just print the value even if we didn't specify it
1047 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1048 seq_puts(seq, ",data=journal");
1049 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1050 seq_puts(seq, ",data=ordered");
1051 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1052 seq_puts(seq, ",data=writeback");
1054 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1055 seq_printf(seq, ",inode_readahead_blks=%u",
1056 sbi->s_inode_readahead_blks);
1058 if (test_opt(sb, DATA_ERR_ABORT))
1059 seq_puts(seq, ",data_err=abort");
1061 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1062 seq_puts(seq, ",noauto_da_alloc");
1064 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1065 seq_puts(seq, ",discard");
1067 if (test_opt(sb, NOLOAD))
1068 seq_puts(seq, ",norecovery");
1070 if (test_opt(sb, DIOREAD_NOLOCK))
1071 seq_puts(seq, ",dioread_nolock");
1073 if (test_opt(sb, BLOCK_VALIDITY) &&
1074 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1075 seq_puts(seq, ",block_validity");
1077 if (!test_opt(sb, INIT_INODE_TABLE))
1078 seq_puts(seq, ",noinit_inode_table");
1079 else if (sbi->s_li_wait_mult)
1080 seq_printf(seq, ",init_inode_table=%u",
1081 (unsigned) sbi->s_li_wait_mult);
1083 ext4_show_quota_options(seq, sb);
1088 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1089 u64 ino, u32 generation)
1091 struct inode *inode;
1093 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1094 return ERR_PTR(-ESTALE);
1095 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1096 return ERR_PTR(-ESTALE);
1098 /* iget isn't really right if the inode is currently unallocated!!
1100 * ext4_read_inode will return a bad_inode if the inode had been
1101 * deleted, so we should be safe.
1103 * Currently we don't know the generation for parent directory, so
1104 * a generation of 0 means "accept any"
1106 inode = ext4_iget(sb, ino);
1108 return ERR_CAST(inode);
1109 if (generation && inode->i_generation != generation) {
1111 return ERR_PTR(-ESTALE);
1117 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1118 int fh_len, int fh_type)
1120 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1121 ext4_nfs_get_inode);
1124 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1125 int fh_len, int fh_type)
1127 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1128 ext4_nfs_get_inode);
1132 * Try to release metadata pages (indirect blocks, directories) which are
1133 * mapped via the block device. Since these pages could have journal heads
1134 * which would prevent try_to_free_buffers() from freeing them, we must use
1135 * jbd2 layer's try_to_free_buffers() function to release them.
1137 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1140 journal_t *journal = EXT4_SB(sb)->s_journal;
1142 WARN_ON(PageChecked(page));
1143 if (!page_has_buffers(page))
1146 return jbd2_journal_try_to_free_buffers(journal, page,
1147 wait & ~__GFP_WAIT);
1148 return try_to_free_buffers(page);
1152 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1153 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1155 static int ext4_write_dquot(struct dquot *dquot);
1156 static int ext4_acquire_dquot(struct dquot *dquot);
1157 static int ext4_release_dquot(struct dquot *dquot);
1158 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1159 static int ext4_write_info(struct super_block *sb, int type);
1160 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1162 static int ext4_quota_off(struct super_block *sb, int type);
1163 static int ext4_quota_on_mount(struct super_block *sb, int type);
1164 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1165 size_t len, loff_t off);
1166 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1167 const char *data, size_t len, loff_t off);
1169 static const struct dquot_operations ext4_quota_operations = {
1171 .get_reserved_space = ext4_get_reserved_space,
1173 .write_dquot = ext4_write_dquot,
1174 .acquire_dquot = ext4_acquire_dquot,
1175 .release_dquot = ext4_release_dquot,
1176 .mark_dirty = ext4_mark_dquot_dirty,
1177 .write_info = ext4_write_info,
1178 .alloc_dquot = dquot_alloc,
1179 .destroy_dquot = dquot_destroy,
1182 static const struct quotactl_ops ext4_qctl_operations = {
1183 .quota_on = ext4_quota_on,
1184 .quota_off = ext4_quota_off,
1185 .quota_sync = dquot_quota_sync,
1186 .get_info = dquot_get_dqinfo,
1187 .set_info = dquot_set_dqinfo,
1188 .get_dqblk = dquot_get_dqblk,
1189 .set_dqblk = dquot_set_dqblk
1193 static const struct super_operations ext4_sops = {
1194 .alloc_inode = ext4_alloc_inode,
1195 .destroy_inode = ext4_destroy_inode,
1196 .write_inode = ext4_write_inode,
1197 .dirty_inode = ext4_dirty_inode,
1198 .drop_inode = ext4_drop_inode,
1199 .evict_inode = ext4_evict_inode,
1200 .put_super = ext4_put_super,
1201 .sync_fs = ext4_sync_fs,
1202 .freeze_fs = ext4_freeze,
1203 .unfreeze_fs = ext4_unfreeze,
1204 .statfs = ext4_statfs,
1205 .remount_fs = ext4_remount,
1206 .show_options = ext4_show_options,
1208 .quota_read = ext4_quota_read,
1209 .quota_write = ext4_quota_write,
1211 .bdev_try_to_free_page = bdev_try_to_free_page,
1214 static const struct super_operations ext4_nojournal_sops = {
1215 .alloc_inode = ext4_alloc_inode,
1216 .destroy_inode = ext4_destroy_inode,
1217 .write_inode = ext4_write_inode,
1218 .dirty_inode = ext4_dirty_inode,
1219 .drop_inode = ext4_drop_inode,
1220 .evict_inode = ext4_evict_inode,
1221 .write_super = ext4_write_super,
1222 .put_super = ext4_put_super,
1223 .statfs = ext4_statfs,
1224 .remount_fs = ext4_remount,
1225 .show_options = ext4_show_options,
1227 .quota_read = ext4_quota_read,
1228 .quota_write = ext4_quota_write,
1230 .bdev_try_to_free_page = bdev_try_to_free_page,
1233 static const struct export_operations ext4_export_ops = {
1234 .fh_to_dentry = ext4_fh_to_dentry,
1235 .fh_to_parent = ext4_fh_to_parent,
1236 .get_parent = ext4_get_parent,
1240 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1241 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1242 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1243 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1244 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1245 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1246 Opt_journal_update, Opt_journal_dev,
1247 Opt_journal_checksum, Opt_journal_async_commit,
1248 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1249 Opt_data_err_abort, Opt_data_err_ignore,
1250 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1251 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1252 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1253 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1254 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1255 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1256 Opt_inode_readahead_blks, Opt_journal_ioprio,
1257 Opt_dioread_nolock, Opt_dioread_lock,
1258 Opt_discard, Opt_nodiscard,
1259 Opt_init_inode_table, Opt_noinit_inode_table,
1262 static const match_table_t tokens = {
1263 {Opt_bsd_df, "bsddf"},
1264 {Opt_minix_df, "minixdf"},
1265 {Opt_grpid, "grpid"},
1266 {Opt_grpid, "bsdgroups"},
1267 {Opt_nogrpid, "nogrpid"},
1268 {Opt_nogrpid, "sysvgroups"},
1269 {Opt_resgid, "resgid=%u"},
1270 {Opt_resuid, "resuid=%u"},
1272 {Opt_err_cont, "errors=continue"},
1273 {Opt_err_panic, "errors=panic"},
1274 {Opt_err_ro, "errors=remount-ro"},
1275 {Opt_nouid32, "nouid32"},
1276 {Opt_debug, "debug"},
1277 {Opt_oldalloc, "oldalloc"},
1278 {Opt_orlov, "orlov"},
1279 {Opt_user_xattr, "user_xattr"},
1280 {Opt_nouser_xattr, "nouser_xattr"},
1282 {Opt_noacl, "noacl"},
1283 {Opt_noload, "noload"},
1284 {Opt_noload, "norecovery"},
1287 {Opt_commit, "commit=%u"},
1288 {Opt_min_batch_time, "min_batch_time=%u"},
1289 {Opt_max_batch_time, "max_batch_time=%u"},
1290 {Opt_journal_update, "journal=update"},
1291 {Opt_journal_dev, "journal_dev=%u"},
1292 {Opt_journal_checksum, "journal_checksum"},
1293 {Opt_journal_async_commit, "journal_async_commit"},
1294 {Opt_abort, "abort"},
1295 {Opt_data_journal, "data=journal"},
1296 {Opt_data_ordered, "data=ordered"},
1297 {Opt_data_writeback, "data=writeback"},
1298 {Opt_data_err_abort, "data_err=abort"},
1299 {Opt_data_err_ignore, "data_err=ignore"},
1300 {Opt_offusrjquota, "usrjquota="},
1301 {Opt_usrjquota, "usrjquota=%s"},
1302 {Opt_offgrpjquota, "grpjquota="},
1303 {Opt_grpjquota, "grpjquota=%s"},
1304 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1305 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1306 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1307 {Opt_grpquota, "grpquota"},
1308 {Opt_noquota, "noquota"},
1309 {Opt_quota, "quota"},
1310 {Opt_usrquota, "usrquota"},
1311 {Opt_barrier, "barrier=%u"},
1312 {Opt_barrier, "barrier"},
1313 {Opt_nobarrier, "nobarrier"},
1314 {Opt_i_version, "i_version"},
1315 {Opt_stripe, "stripe=%u"},
1316 {Opt_resize, "resize"},
1317 {Opt_delalloc, "delalloc"},
1318 {Opt_nodelalloc, "nodelalloc"},
1319 {Opt_mblk_io_submit, "mblk_io_submit"},
1320 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1321 {Opt_block_validity, "block_validity"},
1322 {Opt_noblock_validity, "noblock_validity"},
1323 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1324 {Opt_journal_ioprio, "journal_ioprio=%u"},
1325 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1326 {Opt_auto_da_alloc, "auto_da_alloc"},
1327 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1328 {Opt_dioread_nolock, "dioread_nolock"},
1329 {Opt_dioread_lock, "dioread_lock"},
1330 {Opt_discard, "discard"},
1331 {Opt_nodiscard, "nodiscard"},
1332 {Opt_init_inode_table, "init_itable=%u"},
1333 {Opt_init_inode_table, "init_itable"},
1334 {Opt_noinit_inode_table, "noinit_itable"},
1338 static ext4_fsblk_t get_sb_block(void **data)
1340 ext4_fsblk_t sb_block;
1341 char *options = (char *) *data;
1343 if (!options || strncmp(options, "sb=", 3) != 0)
1344 return 1; /* Default location */
1347 /* TODO: use simple_strtoll with >32bit ext4 */
1348 sb_block = simple_strtoul(options, &options, 0);
1349 if (*options && *options != ',') {
1350 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1354 if (*options == ',')
1356 *data = (void *) options;
1361 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1362 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1363 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1366 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1368 struct ext4_sb_info *sbi = EXT4_SB(sb);
1371 if (sb_any_quota_loaded(sb) &&
1372 !sbi->s_qf_names[qtype]) {
1373 ext4_msg(sb, KERN_ERR,
1374 "Cannot change journaled "
1375 "quota options when quota turned on");
1378 qname = match_strdup(args);
1380 ext4_msg(sb, KERN_ERR,
1381 "Not enough memory for storing quotafile name");
1384 if (sbi->s_qf_names[qtype] &&
1385 strcmp(sbi->s_qf_names[qtype], qname)) {
1386 ext4_msg(sb, KERN_ERR,
1387 "%s quota file already specified", QTYPE2NAME(qtype));
1391 sbi->s_qf_names[qtype] = qname;
1392 if (strchr(sbi->s_qf_names[qtype], '/')) {
1393 ext4_msg(sb, KERN_ERR,
1394 "quotafile must be on filesystem root");
1395 kfree(sbi->s_qf_names[qtype]);
1396 sbi->s_qf_names[qtype] = NULL;
1403 static int clear_qf_name(struct super_block *sb, int qtype)
1406 struct ext4_sb_info *sbi = EXT4_SB(sb);
1408 if (sb_any_quota_loaded(sb) &&
1409 sbi->s_qf_names[qtype]) {
1410 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1411 " when quota turned on");
1415 * The space will be released later when all options are confirmed
1418 sbi->s_qf_names[qtype] = NULL;
1423 static int parse_options(char *options, struct super_block *sb,
1424 unsigned long *journal_devnum,
1425 unsigned int *journal_ioprio,
1426 ext4_fsblk_t *n_blocks_count, int is_remount)
1428 struct ext4_sb_info *sbi = EXT4_SB(sb);
1430 substring_t args[MAX_OPT_ARGS];
1440 while ((p = strsep(&options, ",")) != NULL) {
1446 * Initialize args struct so we know whether arg was
1447 * found; some options take optional arguments.
1449 args[0].to = args[0].from = 0;
1450 token = match_token(p, tokens, args);
1453 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1454 clear_opt(sb, MINIX_DF);
1457 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1458 set_opt(sb, MINIX_DF);
1462 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1467 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1468 clear_opt(sb, GRPID);
1472 if (match_int(&args[0], &option))
1474 sbi->s_resuid = option;
1477 if (match_int(&args[0], &option))
1479 sbi->s_resgid = option;
1482 /* handled by get_sb_block() instead of here */
1483 /* *sb_block = match_int(&args[0]); */
1486 clear_opt(sb, ERRORS_CONT);
1487 clear_opt(sb, ERRORS_RO);
1488 set_opt(sb, ERRORS_PANIC);
1491 clear_opt(sb, ERRORS_CONT);
1492 clear_opt(sb, ERRORS_PANIC);
1493 set_opt(sb, ERRORS_RO);
1496 clear_opt(sb, ERRORS_RO);
1497 clear_opt(sb, ERRORS_PANIC);
1498 set_opt(sb, ERRORS_CONT);
1501 set_opt(sb, NO_UID32);
1507 set_opt(sb, OLDALLOC);
1510 clear_opt(sb, OLDALLOC);
1512 #ifdef CONFIG_EXT4_FS_XATTR
1513 case Opt_user_xattr:
1514 set_opt(sb, XATTR_USER);
1516 case Opt_nouser_xattr:
1517 clear_opt(sb, XATTR_USER);
1520 case Opt_user_xattr:
1521 case Opt_nouser_xattr:
1522 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1525 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1527 set_opt(sb, POSIX_ACL);
1530 clear_opt(sb, POSIX_ACL);
1535 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1538 case Opt_journal_update:
1540 /* Eventually we will want to be able to create
1541 a journal file here. For now, only allow the
1542 user to specify an existing inode to be the
1545 ext4_msg(sb, KERN_ERR,
1546 "Cannot specify journal on remount");
1549 set_opt(sb, UPDATE_JOURNAL);
1551 case Opt_journal_dev:
1553 ext4_msg(sb, KERN_ERR,
1554 "Cannot specify journal on remount");
1557 if (match_int(&args[0], &option))
1559 *journal_devnum = option;
1561 case Opt_journal_checksum:
1562 set_opt(sb, JOURNAL_CHECKSUM);
1564 case Opt_journal_async_commit:
1565 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1566 set_opt(sb, JOURNAL_CHECKSUM);
1569 set_opt(sb, NOLOAD);
1572 if (match_int(&args[0], &option))
1577 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1578 sbi->s_commit_interval = HZ * option;
1580 case Opt_max_batch_time:
1581 if (match_int(&args[0], &option))
1586 option = EXT4_DEF_MAX_BATCH_TIME;
1587 sbi->s_max_batch_time = option;
1589 case Opt_min_batch_time:
1590 if (match_int(&args[0], &option))
1594 sbi->s_min_batch_time = option;
1596 case Opt_data_journal:
1597 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1599 case Opt_data_ordered:
1600 data_opt = EXT4_MOUNT_ORDERED_DATA;
1602 case Opt_data_writeback:
1603 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1606 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot change data mode on remount");
1612 clear_opt(sb, DATA_FLAGS);
1613 sbi->s_mount_opt |= data_opt;
1616 case Opt_data_err_abort:
1617 set_opt(sb, DATA_ERR_ABORT);
1619 case Opt_data_err_ignore:
1620 clear_opt(sb, DATA_ERR_ABORT);
1624 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1628 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1631 case Opt_offusrjquota:
1632 if (!clear_qf_name(sb, USRQUOTA))
1635 case Opt_offgrpjquota:
1636 if (!clear_qf_name(sb, GRPQUOTA))
1640 case Opt_jqfmt_vfsold:
1641 qfmt = QFMT_VFS_OLD;
1643 case Opt_jqfmt_vfsv0:
1646 case Opt_jqfmt_vfsv1:
1649 if (sb_any_quota_loaded(sb) &&
1650 sbi->s_jquota_fmt != qfmt) {
1651 ext4_msg(sb, KERN_ERR, "Cannot change "
1652 "journaled quota options when "
1656 sbi->s_jquota_fmt = qfmt;
1661 set_opt(sb, USRQUOTA);
1665 set_opt(sb, GRPQUOTA);
1668 if (sb_any_quota_loaded(sb)) {
1669 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1670 "options when quota turned on");
1673 clear_opt(sb, QUOTA);
1674 clear_opt(sb, USRQUOTA);
1675 clear_opt(sb, GRPQUOTA);
1681 ext4_msg(sb, KERN_ERR,
1682 "quota options not supported");
1686 case Opt_offusrjquota:
1687 case Opt_offgrpjquota:
1688 case Opt_jqfmt_vfsold:
1689 case Opt_jqfmt_vfsv0:
1690 case Opt_jqfmt_vfsv1:
1691 ext4_msg(sb, KERN_ERR,
1692 "journaled quota options not supported");
1698 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1701 clear_opt(sb, BARRIER);
1705 if (match_int(&args[0], &option))
1708 option = 1; /* No argument, default to 1 */
1710 set_opt(sb, BARRIER);
1712 clear_opt(sb, BARRIER);
1718 ext4_msg(sb, KERN_ERR,
1719 "resize option only available "
1723 if (match_int(&args[0], &option) != 0)
1725 *n_blocks_count = option;
1728 ext4_msg(sb, KERN_WARNING,
1729 "Ignoring deprecated nobh option");
1732 ext4_msg(sb, KERN_WARNING,
1733 "Ignoring deprecated bh option");
1736 set_opt(sb, I_VERSION);
1737 sb->s_flags |= MS_I_VERSION;
1739 case Opt_nodelalloc:
1740 clear_opt(sb, DELALLOC);
1742 case Opt_mblk_io_submit:
1743 set_opt(sb, MBLK_IO_SUBMIT);
1745 case Opt_nomblk_io_submit:
1746 clear_opt(sb, MBLK_IO_SUBMIT);
1749 if (match_int(&args[0], &option))
1753 sbi->s_stripe = option;
1756 set_opt(sb, DELALLOC);
1758 case Opt_block_validity:
1759 set_opt(sb, BLOCK_VALIDITY);
1761 case Opt_noblock_validity:
1762 clear_opt(sb, BLOCK_VALIDITY);
1764 case Opt_inode_readahead_blks:
1765 if (match_int(&args[0], &option))
1767 if (option < 0 || option > (1 << 30))
1769 if (!is_power_of_2(option)) {
1770 ext4_msg(sb, KERN_ERR,
1771 "EXT4-fs: inode_readahead_blks"
1772 " must be a power of 2");
1775 sbi->s_inode_readahead_blks = option;
1777 case Opt_journal_ioprio:
1778 if (match_int(&args[0], &option))
1780 if (option < 0 || option > 7)
1782 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1785 case Opt_noauto_da_alloc:
1786 set_opt(sb, NO_AUTO_DA_ALLOC);
1788 case Opt_auto_da_alloc:
1790 if (match_int(&args[0], &option))
1793 option = 1; /* No argument, default to 1 */
1795 clear_opt(sb, NO_AUTO_DA_ALLOC);
1797 set_opt(sb,NO_AUTO_DA_ALLOC);
1800 set_opt(sb, DISCARD);
1803 clear_opt(sb, DISCARD);
1805 case Opt_dioread_nolock:
1806 set_opt(sb, DIOREAD_NOLOCK);
1808 case Opt_dioread_lock:
1809 clear_opt(sb, DIOREAD_NOLOCK);
1811 case Opt_init_inode_table:
1812 set_opt(sb, INIT_INODE_TABLE);
1814 if (match_int(&args[0], &option))
1817 option = EXT4_DEF_LI_WAIT_MULT;
1820 sbi->s_li_wait_mult = option;
1822 case Opt_noinit_inode_table:
1823 clear_opt(sb, INIT_INODE_TABLE);
1826 ext4_msg(sb, KERN_ERR,
1827 "Unrecognized mount option \"%s\" "
1828 "or missing value", p);
1833 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1834 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1835 clear_opt(sb, USRQUOTA);
1837 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1838 clear_opt(sb, GRPQUOTA);
1840 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1841 ext4_msg(sb, KERN_ERR, "old and new quota "
1846 if (!sbi->s_jquota_fmt) {
1847 ext4_msg(sb, KERN_ERR, "journaled quota format "
1852 if (sbi->s_jquota_fmt) {
1853 ext4_msg(sb, KERN_ERR, "journaled quota format "
1854 "specified with no journaling "
1863 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1866 struct ext4_sb_info *sbi = EXT4_SB(sb);
1869 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1870 ext4_msg(sb, KERN_ERR, "revision level too high, "
1871 "forcing read-only mode");
1876 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1877 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1878 "running e2fsck is recommended");
1879 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1880 ext4_msg(sb, KERN_WARNING,
1881 "warning: mounting fs with errors, "
1882 "running e2fsck is recommended");
1883 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1884 le16_to_cpu(es->s_mnt_count) >=
1885 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1886 ext4_msg(sb, KERN_WARNING,
1887 "warning: maximal mount count reached, "
1888 "running e2fsck is recommended");
1889 else if (le32_to_cpu(es->s_checkinterval) &&
1890 (le32_to_cpu(es->s_lastcheck) +
1891 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1892 ext4_msg(sb, KERN_WARNING,
1893 "warning: checktime reached, "
1894 "running e2fsck is recommended");
1895 if (!sbi->s_journal)
1896 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1897 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1898 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1899 le16_add_cpu(&es->s_mnt_count, 1);
1900 es->s_mtime = cpu_to_le32(get_seconds());
1901 ext4_update_dynamic_rev(sb);
1903 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1905 ext4_commit_super(sb, 1);
1906 if (test_opt(sb, DEBUG))
1907 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1908 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1910 sbi->s_groups_count,
1911 EXT4_BLOCKS_PER_GROUP(sb),
1912 EXT4_INODES_PER_GROUP(sb),
1913 sbi->s_mount_opt, sbi->s_mount_opt2);
1918 static int ext4_fill_flex_info(struct super_block *sb)
1920 struct ext4_sb_info *sbi = EXT4_SB(sb);
1921 struct ext4_group_desc *gdp = NULL;
1922 ext4_group_t flex_group_count;
1923 ext4_group_t flex_group;
1924 int groups_per_flex = 0;
1928 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1929 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1931 if (groups_per_flex < 2) {
1932 sbi->s_log_groups_per_flex = 0;
1936 /* We allocate both existing and potentially added groups */
1937 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1938 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1939 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1940 size = flex_group_count * sizeof(struct flex_groups);
1941 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1942 if (sbi->s_flex_groups == NULL) {
1943 sbi->s_flex_groups = vzalloc(size);
1944 if (sbi->s_flex_groups == NULL) {
1945 ext4_msg(sb, KERN_ERR,
1946 "not enough memory for %u flex groups",
1952 for (i = 0; i < sbi->s_groups_count; i++) {
1953 gdp = ext4_get_group_desc(sb, i, NULL);
1955 flex_group = ext4_flex_group(sbi, i);
1956 atomic_add(ext4_free_inodes_count(sb, gdp),
1957 &sbi->s_flex_groups[flex_group].free_inodes);
1958 atomic_add(ext4_free_blks_count(sb, gdp),
1959 &sbi->s_flex_groups[flex_group].free_blocks);
1960 atomic_add(ext4_used_dirs_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].used_dirs);
1969 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1970 struct ext4_group_desc *gdp)
1974 if (sbi->s_es->s_feature_ro_compat &
1975 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1976 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1977 __le32 le_group = cpu_to_le32(block_group);
1979 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1980 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1981 crc = crc16(crc, (__u8 *)gdp, offset);
1982 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1983 /* for checksum of struct ext4_group_desc do the rest...*/
1984 if ((sbi->s_es->s_feature_incompat &
1985 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1986 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1987 crc = crc16(crc, (__u8 *)gdp + offset,
1988 le16_to_cpu(sbi->s_es->s_desc_size) -
1992 return cpu_to_le16(crc);
1995 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1996 struct ext4_group_desc *gdp)
1998 if ((sbi->s_es->s_feature_ro_compat &
1999 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2000 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2006 /* Called at mount-time, super-block is locked */
2007 static int ext4_check_descriptors(struct super_block *sb,
2008 ext4_group_t *first_not_zeroed)
2010 struct ext4_sb_info *sbi = EXT4_SB(sb);
2011 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2012 ext4_fsblk_t last_block;
2013 ext4_fsblk_t block_bitmap;
2014 ext4_fsblk_t inode_bitmap;
2015 ext4_fsblk_t inode_table;
2016 int flexbg_flag = 0;
2017 ext4_group_t i, grp = sbi->s_groups_count;
2019 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2022 ext4_debug("Checking group descriptors");
2024 for (i = 0; i < sbi->s_groups_count; i++) {
2025 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2027 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2028 last_block = ext4_blocks_count(sbi->s_es) - 1;
2030 last_block = first_block +
2031 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2033 if ((grp == sbi->s_groups_count) &&
2034 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2037 block_bitmap = ext4_block_bitmap(sb, gdp);
2038 if (block_bitmap < first_block || block_bitmap > last_block) {
2039 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2040 "Block bitmap for group %u not in group "
2041 "(block %llu)!", i, block_bitmap);
2044 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2045 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2046 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2047 "Inode bitmap for group %u not in group "
2048 "(block %llu)!", i, inode_bitmap);
2051 inode_table = ext4_inode_table(sb, gdp);
2052 if (inode_table < first_block ||
2053 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2054 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2055 "Inode table for group %u not in group "
2056 "(block %llu)!", i, inode_table);
2059 ext4_lock_group(sb, i);
2060 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2061 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2062 "Checksum for group %u failed (%u!=%u)",
2063 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2064 gdp)), le16_to_cpu(gdp->bg_checksum));
2065 if (!(sb->s_flags & MS_RDONLY)) {
2066 ext4_unlock_group(sb, i);
2070 ext4_unlock_group(sb, i);
2072 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2074 if (NULL != first_not_zeroed)
2075 *first_not_zeroed = grp;
2077 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2078 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2082 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2083 * the superblock) which were deleted from all directories, but held open by
2084 * a process at the time of a crash. We walk the list and try to delete these
2085 * inodes at recovery time (only with a read-write filesystem).
2087 * In order to keep the orphan inode chain consistent during traversal (in
2088 * case of crash during recovery), we link each inode into the superblock
2089 * orphan list_head and handle it the same way as an inode deletion during
2090 * normal operation (which journals the operations for us).
2092 * We only do an iget() and an iput() on each inode, which is very safe if we
2093 * accidentally point at an in-use or already deleted inode. The worst that
2094 * can happen in this case is that we get a "bit already cleared" message from
2095 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2096 * e2fsck was run on this filesystem, and it must have already done the orphan
2097 * inode cleanup for us, so we can safely abort without any further action.
2099 static void ext4_orphan_cleanup(struct super_block *sb,
2100 struct ext4_super_block *es)
2102 unsigned int s_flags = sb->s_flags;
2103 int nr_orphans = 0, nr_truncates = 0;
2107 if (!es->s_last_orphan) {
2108 jbd_debug(4, "no orphan inodes to clean up\n");
2112 if (bdev_read_only(sb->s_bdev)) {
2113 ext4_msg(sb, KERN_ERR, "write access "
2114 "unavailable, skipping orphan cleanup");
2118 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2119 if (es->s_last_orphan)
2120 jbd_debug(1, "Errors on filesystem, "
2121 "clearing orphan list.\n");
2122 es->s_last_orphan = 0;
2123 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2127 if (s_flags & MS_RDONLY) {
2128 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2129 sb->s_flags &= ~MS_RDONLY;
2132 /* Needed for iput() to work correctly and not trash data */
2133 sb->s_flags |= MS_ACTIVE;
2134 /* Turn on quotas so that they are updated correctly */
2135 for (i = 0; i < MAXQUOTAS; i++) {
2136 if (EXT4_SB(sb)->s_qf_names[i]) {
2137 int ret = ext4_quota_on_mount(sb, i);
2139 ext4_msg(sb, KERN_ERR,
2140 "Cannot turn on journaled "
2141 "quota: error %d", ret);
2146 while (es->s_last_orphan) {
2147 struct inode *inode;
2149 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2150 if (IS_ERR(inode)) {
2151 es->s_last_orphan = 0;
2155 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2156 dquot_initialize(inode);
2157 if (inode->i_nlink) {
2158 ext4_msg(sb, KERN_DEBUG,
2159 "%s: truncating inode %lu to %lld bytes",
2160 __func__, inode->i_ino, inode->i_size);
2161 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2162 inode->i_ino, inode->i_size);
2163 ext4_truncate(inode);
2166 ext4_msg(sb, KERN_DEBUG,
2167 "%s: deleting unreferenced inode %lu",
2168 __func__, inode->i_ino);
2169 jbd_debug(2, "deleting unreferenced inode %lu\n",
2173 iput(inode); /* The delete magic happens here! */
2176 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2179 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2180 PLURAL(nr_orphans));
2182 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2183 PLURAL(nr_truncates));
2185 /* Turn quotas off */
2186 for (i = 0; i < MAXQUOTAS; i++) {
2187 if (sb_dqopt(sb)->files[i])
2188 dquot_quota_off(sb, i);
2191 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2195 * Maximal extent format file size.
2196 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2197 * extent format containers, within a sector_t, and within i_blocks
2198 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2199 * so that won't be a limiting factor.
2201 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2203 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2206 loff_t upper_limit = MAX_LFS_FILESIZE;
2208 /* small i_blocks in vfs inode? */
2209 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2211 * CONFIG_LBDAF is not enabled implies the inode
2212 * i_block represent total blocks in 512 bytes
2213 * 32 == size of vfs inode i_blocks * 8
2215 upper_limit = (1LL << 32) - 1;
2217 /* total blocks in file system block size */
2218 upper_limit >>= (blkbits - 9);
2219 upper_limit <<= blkbits;
2222 /* 32-bit extent-start container, ee_block */
2227 /* Sanity check against vm- & vfs- imposed limits */
2228 if (res > upper_limit)
2235 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2236 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2237 * We need to be 1 filesystem block less than the 2^48 sector limit.
2239 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2241 loff_t res = EXT4_NDIR_BLOCKS;
2244 /* This is calculated to be the largest file size for a dense, block
2245 * mapped file such that the file's total number of 512-byte sectors,
2246 * including data and all indirect blocks, does not exceed (2^48 - 1).
2248 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2249 * number of 512-byte sectors of the file.
2252 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2254 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2255 * the inode i_block field represents total file blocks in
2256 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2258 upper_limit = (1LL << 32) - 1;
2260 /* total blocks in file system block size */
2261 upper_limit >>= (bits - 9);
2265 * We use 48 bit ext4_inode i_blocks
2266 * With EXT4_HUGE_FILE_FL set the i_blocks
2267 * represent total number of blocks in
2268 * file system block size
2270 upper_limit = (1LL << 48) - 1;
2274 /* indirect blocks */
2276 /* double indirect blocks */
2277 meta_blocks += 1 + (1LL << (bits-2));
2278 /* tripple indirect blocks */
2279 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2281 upper_limit -= meta_blocks;
2282 upper_limit <<= bits;
2284 res += 1LL << (bits-2);
2285 res += 1LL << (2*(bits-2));
2286 res += 1LL << (3*(bits-2));
2288 if (res > upper_limit)
2291 if (res > MAX_LFS_FILESIZE)
2292 res = MAX_LFS_FILESIZE;
2297 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2298 ext4_fsblk_t logical_sb_block, int nr)
2300 struct ext4_sb_info *sbi = EXT4_SB(sb);
2301 ext4_group_t bg, first_meta_bg;
2304 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2306 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2308 return logical_sb_block + nr + 1;
2309 bg = sbi->s_desc_per_block * nr;
2310 if (ext4_bg_has_super(sb, bg))
2313 return (has_super + ext4_group_first_block_no(sb, bg));
2317 * ext4_get_stripe_size: Get the stripe size.
2318 * @sbi: In memory super block info
2320 * If we have specified it via mount option, then
2321 * use the mount option value. If the value specified at mount time is
2322 * greater than the blocks per group use the super block value.
2323 * If the super block value is greater than blocks per group return 0.
2324 * Allocator needs it be less than blocks per group.
2327 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2329 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2330 unsigned long stripe_width =
2331 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2333 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2334 return sbi->s_stripe;
2336 if (stripe_width <= sbi->s_blocks_per_group)
2337 return stripe_width;
2339 if (stride <= sbi->s_blocks_per_group)
2348 struct attribute attr;
2349 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2350 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2351 const char *, size_t);
2355 static int parse_strtoul(const char *buf,
2356 unsigned long max, unsigned long *value)
2360 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2361 endp = skip_spaces(endp);
2362 if (*endp || *value > max)
2368 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2369 struct ext4_sb_info *sbi,
2372 return snprintf(buf, PAGE_SIZE, "%llu\n",
2373 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2376 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2377 struct ext4_sb_info *sbi, char *buf)
2379 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2381 if (!sb->s_bdev->bd_part)
2382 return snprintf(buf, PAGE_SIZE, "0\n");
2383 return snprintf(buf, PAGE_SIZE, "%lu\n",
2384 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2385 sbi->s_sectors_written_start) >> 1);
2388 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2389 struct ext4_sb_info *sbi, char *buf)
2391 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2393 if (!sb->s_bdev->bd_part)
2394 return snprintf(buf, PAGE_SIZE, "0\n");
2395 return snprintf(buf, PAGE_SIZE, "%llu\n",
2396 (unsigned long long)(sbi->s_kbytes_written +
2397 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2398 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2401 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2402 struct ext4_sb_info *sbi,
2403 const char *buf, size_t count)
2407 if (parse_strtoul(buf, 0x40000000, &t))
2410 if (!is_power_of_2(t))
2413 sbi->s_inode_readahead_blks = t;
2417 static ssize_t sbi_ui_show(struct ext4_attr *a,
2418 struct ext4_sb_info *sbi, char *buf)
2420 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2422 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2425 static ssize_t sbi_ui_store(struct ext4_attr *a,
2426 struct ext4_sb_info *sbi,
2427 const char *buf, size_t count)
2429 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2432 if (parse_strtoul(buf, 0xffffffff, &t))
2438 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2439 static struct ext4_attr ext4_attr_##_name = { \
2440 .attr = {.name = __stringify(_name), .mode = _mode }, \
2443 .offset = offsetof(struct ext4_sb_info, _elname), \
2445 #define EXT4_ATTR(name, mode, show, store) \
2446 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2448 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2449 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2450 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2451 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2452 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2453 #define ATTR_LIST(name) &ext4_attr_##name.attr
2455 EXT4_RO_ATTR(delayed_allocation_blocks);
2456 EXT4_RO_ATTR(session_write_kbytes);
2457 EXT4_RO_ATTR(lifetime_write_kbytes);
2458 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2459 inode_readahead_blks_store, s_inode_readahead_blks);
2460 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2461 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2462 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2463 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2464 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2465 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2466 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2467 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2469 static struct attribute *ext4_attrs[] = {
2470 ATTR_LIST(delayed_allocation_blocks),
2471 ATTR_LIST(session_write_kbytes),
2472 ATTR_LIST(lifetime_write_kbytes),
2473 ATTR_LIST(inode_readahead_blks),
2474 ATTR_LIST(inode_goal),
2475 ATTR_LIST(mb_stats),
2476 ATTR_LIST(mb_max_to_scan),
2477 ATTR_LIST(mb_min_to_scan),
2478 ATTR_LIST(mb_order2_req),
2479 ATTR_LIST(mb_stream_req),
2480 ATTR_LIST(mb_group_prealloc),
2481 ATTR_LIST(max_writeback_mb_bump),
2485 /* Features this copy of ext4 supports */
2486 EXT4_INFO_ATTR(lazy_itable_init);
2487 EXT4_INFO_ATTR(batched_discard);
2489 static struct attribute *ext4_feat_attrs[] = {
2490 ATTR_LIST(lazy_itable_init),
2491 ATTR_LIST(batched_discard),
2495 static ssize_t ext4_attr_show(struct kobject *kobj,
2496 struct attribute *attr, char *buf)
2498 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2500 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2502 return a->show ? a->show(a, sbi, buf) : 0;
2505 static ssize_t ext4_attr_store(struct kobject *kobj,
2506 struct attribute *attr,
2507 const char *buf, size_t len)
2509 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2511 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2513 return a->store ? a->store(a, sbi, buf, len) : 0;
2516 static void ext4_sb_release(struct kobject *kobj)
2518 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2520 complete(&sbi->s_kobj_unregister);
2523 static const struct sysfs_ops ext4_attr_ops = {
2524 .show = ext4_attr_show,
2525 .store = ext4_attr_store,
2528 static struct kobj_type ext4_ktype = {
2529 .default_attrs = ext4_attrs,
2530 .sysfs_ops = &ext4_attr_ops,
2531 .release = ext4_sb_release,
2534 static void ext4_feat_release(struct kobject *kobj)
2536 complete(&ext4_feat->f_kobj_unregister);
2539 static struct kobj_type ext4_feat_ktype = {
2540 .default_attrs = ext4_feat_attrs,
2541 .sysfs_ops = &ext4_attr_ops,
2542 .release = ext4_feat_release,
2546 * Check whether this filesystem can be mounted based on
2547 * the features present and the RDONLY/RDWR mount requested.
2548 * Returns 1 if this filesystem can be mounted as requested,
2549 * 0 if it cannot be.
2551 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2553 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2554 ext4_msg(sb, KERN_ERR,
2555 "Couldn't mount because of "
2556 "unsupported optional features (%x)",
2557 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2558 ~EXT4_FEATURE_INCOMPAT_SUPP));
2565 /* Check that feature set is OK for a read-write mount */
2566 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2567 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2568 "unsupported optional features (%x)",
2569 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2570 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2574 * Large file size enabled file system can only be mounted
2575 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2577 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2578 if (sizeof(blkcnt_t) < sizeof(u64)) {
2579 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2580 "cannot be mounted RDWR without "
2589 * This function is called once a day if we have errors logged
2590 * on the file system
2592 static void print_daily_error_info(unsigned long arg)
2594 struct super_block *sb = (struct super_block *) arg;
2595 struct ext4_sb_info *sbi;
2596 struct ext4_super_block *es;
2601 if (es->s_error_count)
2602 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2603 le32_to_cpu(es->s_error_count));
2604 if (es->s_first_error_time) {
2605 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2606 sb->s_id, le32_to_cpu(es->s_first_error_time),
2607 (int) sizeof(es->s_first_error_func),
2608 es->s_first_error_func,
2609 le32_to_cpu(es->s_first_error_line));
2610 if (es->s_first_error_ino)
2611 printk(": inode %u",
2612 le32_to_cpu(es->s_first_error_ino));
2613 if (es->s_first_error_block)
2614 printk(": block %llu", (unsigned long long)
2615 le64_to_cpu(es->s_first_error_block));
2618 if (es->s_last_error_time) {
2619 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2620 sb->s_id, le32_to_cpu(es->s_last_error_time),
2621 (int) sizeof(es->s_last_error_func),
2622 es->s_last_error_func,
2623 le32_to_cpu(es->s_last_error_line));
2624 if (es->s_last_error_ino)
2625 printk(": inode %u",
2626 le32_to_cpu(es->s_last_error_ino));
2627 if (es->s_last_error_block)
2628 printk(": block %llu", (unsigned long long)
2629 le64_to_cpu(es->s_last_error_block));
2632 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2635 static void ext4_lazyinode_timeout(unsigned long data)
2637 struct task_struct *p = (struct task_struct *)data;
2641 /* Find next suitable group and run ext4_init_inode_table */
2642 static int ext4_run_li_request(struct ext4_li_request *elr)
2644 struct ext4_group_desc *gdp = NULL;
2645 ext4_group_t group, ngroups;
2646 struct super_block *sb;
2647 unsigned long timeout = 0;
2651 ngroups = EXT4_SB(sb)->s_groups_count;
2653 for (group = elr->lr_next_group; group < ngroups; group++) {
2654 gdp = ext4_get_group_desc(sb, group, NULL);
2660 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2664 if (group == ngroups)
2669 ret = ext4_init_inode_table(sb, group,
2670 elr->lr_timeout ? 0 : 1);
2671 if (elr->lr_timeout == 0) {
2672 timeout = jiffies - timeout;
2673 if (elr->lr_sbi->s_li_wait_mult)
2674 timeout *= elr->lr_sbi->s_li_wait_mult;
2677 elr->lr_timeout = timeout;
2679 elr->lr_next_sched = jiffies + elr->lr_timeout;
2680 elr->lr_next_group = group + 1;
2687 * Remove lr_request from the list_request and free the
2688 * request tructure. Should be called with li_list_mtx held
2690 static void ext4_remove_li_request(struct ext4_li_request *elr)
2692 struct ext4_sb_info *sbi;
2699 list_del(&elr->lr_request);
2700 sbi->s_li_request = NULL;
2704 static void ext4_unregister_li_request(struct super_block *sb)
2706 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2711 mutex_lock(&ext4_li_info->li_list_mtx);
2712 ext4_remove_li_request(elr);
2713 mutex_unlock(&ext4_li_info->li_list_mtx);
2717 * This is the function where ext4lazyinit thread lives. It walks
2718 * through the request list searching for next scheduled filesystem.
2719 * When such a fs is found, run the lazy initialization request
2720 * (ext4_rn_li_request) and keep track of the time spend in this
2721 * function. Based on that time we compute next schedule time of
2722 * the request. When walking through the list is complete, compute
2723 * next waking time and put itself into sleep.
2725 static int ext4_lazyinit_thread(void *arg)
2727 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2728 struct list_head *pos, *n;
2729 struct ext4_li_request *elr;
2730 unsigned long next_wakeup;
2733 BUG_ON(NULL == eli);
2735 eli->li_timer.data = (unsigned long)current;
2736 eli->li_timer.function = ext4_lazyinode_timeout;
2738 eli->li_task = current;
2739 wake_up(&eli->li_wait_task);
2743 next_wakeup = MAX_JIFFY_OFFSET;
2745 mutex_lock(&eli->li_list_mtx);
2746 if (list_empty(&eli->li_request_list)) {
2747 mutex_unlock(&eli->li_list_mtx);
2751 list_for_each_safe(pos, n, &eli->li_request_list) {
2752 elr = list_entry(pos, struct ext4_li_request,
2755 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2756 if (ext4_run_li_request(elr) != 0) {
2757 /* error, remove the lazy_init job */
2758 ext4_remove_li_request(elr);
2763 if (time_before(elr->lr_next_sched, next_wakeup))
2764 next_wakeup = elr->lr_next_sched;
2766 mutex_unlock(&eli->li_list_mtx);
2768 if (freezing(current))
2771 if ((time_after_eq(jiffies, next_wakeup)) ||
2772 (MAX_JIFFY_OFFSET == next_wakeup)) {
2777 eli->li_timer.expires = next_wakeup;
2778 add_timer(&eli->li_timer);
2779 prepare_to_wait(&eli->li_wait_daemon, &wait,
2780 TASK_INTERRUPTIBLE);
2781 if (time_before(jiffies, next_wakeup))
2783 finish_wait(&eli->li_wait_daemon, &wait);
2788 * It looks like the request list is empty, but we need
2789 * to check it under the li_list_mtx lock, to prevent any
2790 * additions into it, and of course we should lock ext4_li_mtx
2791 * to atomically free the list and ext4_li_info, because at
2792 * this point another ext4 filesystem could be registering
2795 mutex_lock(&ext4_li_mtx);
2796 mutex_lock(&eli->li_list_mtx);
2797 if (!list_empty(&eli->li_request_list)) {
2798 mutex_unlock(&eli->li_list_mtx);
2799 mutex_unlock(&ext4_li_mtx);
2802 mutex_unlock(&eli->li_list_mtx);
2803 del_timer_sync(&ext4_li_info->li_timer);
2804 eli->li_task = NULL;
2805 wake_up(&eli->li_wait_task);
2807 kfree(ext4_li_info);
2808 ext4_li_info = NULL;
2809 mutex_unlock(&ext4_li_mtx);
2814 static void ext4_clear_request_list(void)
2816 struct list_head *pos, *n;
2817 struct ext4_li_request *elr;
2819 mutex_lock(&ext4_li_info->li_list_mtx);
2820 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2821 elr = list_entry(pos, struct ext4_li_request,
2823 ext4_remove_li_request(elr);
2825 mutex_unlock(&ext4_li_info->li_list_mtx);
2828 static int ext4_run_lazyinit_thread(void)
2830 struct task_struct *t;
2832 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2834 int err = PTR_ERR(t);
2835 ext4_clear_request_list();
2836 del_timer_sync(&ext4_li_info->li_timer);
2837 kfree(ext4_li_info);
2838 ext4_li_info = NULL;
2839 printk(KERN_CRIT "EXT4: error %d creating inode table "
2840 "initialization thread\n",
2844 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2846 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2851 * Check whether it make sense to run itable init. thread or not.
2852 * If there is at least one uninitialized inode table, return
2853 * corresponding group number, else the loop goes through all
2854 * groups and return total number of groups.
2856 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2858 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2859 struct ext4_group_desc *gdp = NULL;
2861 for (group = 0; group < ngroups; group++) {
2862 gdp = ext4_get_group_desc(sb, group, NULL);
2866 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2873 static int ext4_li_info_new(void)
2875 struct ext4_lazy_init *eli = NULL;
2877 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2881 eli->li_task = NULL;
2882 INIT_LIST_HEAD(&eli->li_request_list);
2883 mutex_init(&eli->li_list_mtx);
2885 init_waitqueue_head(&eli->li_wait_daemon);
2886 init_waitqueue_head(&eli->li_wait_task);
2887 init_timer(&eli->li_timer);
2888 eli->li_state |= EXT4_LAZYINIT_QUIT;
2895 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2898 struct ext4_sb_info *sbi = EXT4_SB(sb);
2899 struct ext4_li_request *elr;
2902 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2908 elr->lr_next_group = start;
2911 * Randomize first schedule time of the request to
2912 * spread the inode table initialization requests
2915 get_random_bytes(&rnd, sizeof(rnd));
2916 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2917 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2922 static int ext4_register_li_request(struct super_block *sb,
2923 ext4_group_t first_not_zeroed)
2925 struct ext4_sb_info *sbi = EXT4_SB(sb);
2926 struct ext4_li_request *elr;
2927 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2930 if (sbi->s_li_request != NULL)
2933 if (first_not_zeroed == ngroups ||
2934 (sb->s_flags & MS_RDONLY) ||
2935 !test_opt(sb, INIT_INODE_TABLE)) {
2936 sbi->s_li_request = NULL;
2940 if (first_not_zeroed == ngroups) {
2941 sbi->s_li_request = NULL;
2945 elr = ext4_li_request_new(sb, first_not_zeroed);
2949 mutex_lock(&ext4_li_mtx);
2951 if (NULL == ext4_li_info) {
2952 ret = ext4_li_info_new();
2957 mutex_lock(&ext4_li_info->li_list_mtx);
2958 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2959 mutex_unlock(&ext4_li_info->li_list_mtx);
2961 sbi->s_li_request = elr;
2963 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2964 ret = ext4_run_lazyinit_thread();
2969 mutex_unlock(&ext4_li_mtx);
2976 * We do not need to lock anything since this is called on
2979 static void ext4_destroy_lazyinit_thread(void)
2982 * If thread exited earlier
2983 * there's nothing to be done.
2988 ext4_clear_request_list();
2990 while (ext4_li_info->li_task) {
2991 wake_up(&ext4_li_info->li_wait_daemon);
2992 wait_event(ext4_li_info->li_wait_task,
2993 ext4_li_info->li_task == NULL);
2997 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2998 __releases(kernel_lock)
2999 __acquires(kernel_lock)
3001 char *orig_data = kstrdup(data, GFP_KERNEL);
3002 struct buffer_head *bh;
3003 struct ext4_super_block *es = NULL;
3004 struct ext4_sb_info *sbi;
3006 ext4_fsblk_t sb_block = get_sb_block(&data);
3007 ext4_fsblk_t logical_sb_block;
3008 unsigned long offset = 0;
3009 unsigned long journal_devnum = 0;
3010 unsigned long def_mount_opts;
3016 unsigned int db_count;
3018 int needs_recovery, has_huge_files;
3021 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3022 ext4_group_t first_not_zeroed;
3024 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3028 sbi->s_blockgroup_lock =
3029 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3030 if (!sbi->s_blockgroup_lock) {
3034 sb->s_fs_info = sbi;
3035 sbi->s_mount_opt = 0;
3036 sbi->s_resuid = EXT4_DEF_RESUID;
3037 sbi->s_resgid = EXT4_DEF_RESGID;
3038 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3039 sbi->s_sb_block = sb_block;
3040 if (sb->s_bdev->bd_part)
3041 sbi->s_sectors_written_start =
3042 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3044 /* Cleanup superblock name */
3045 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3049 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3051 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3056 * The ext4 superblock will not be buffer aligned for other than 1kB
3057 * block sizes. We need to calculate the offset from buffer start.
3059 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3060 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3061 offset = do_div(logical_sb_block, blocksize);
3063 logical_sb_block = sb_block;
3066 if (!(bh = sb_bread(sb, logical_sb_block))) {
3067 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3071 * Note: s_es must be initialized as soon as possible because
3072 * some ext4 macro-instructions depend on its value
3074 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3076 sb->s_magic = le16_to_cpu(es->s_magic);
3077 if (sb->s_magic != EXT4_SUPER_MAGIC)
3079 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3081 /* Set defaults before we parse the mount options */
3082 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3083 set_opt(sb, INIT_INODE_TABLE);
3084 if (def_mount_opts & EXT4_DEFM_DEBUG)
3086 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3087 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3091 if (def_mount_opts & EXT4_DEFM_UID16)
3092 set_opt(sb, NO_UID32);
3093 #ifdef CONFIG_EXT4_FS_XATTR
3094 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3095 set_opt(sb, XATTR_USER);
3097 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3098 if (def_mount_opts & EXT4_DEFM_ACL)
3099 set_opt(sb, POSIX_ACL);
3101 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3102 set_opt(sb, JOURNAL_DATA);
3103 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3104 set_opt(sb, ORDERED_DATA);
3105 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3106 set_opt(sb, WRITEBACK_DATA);
3108 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3109 set_opt(sb, ERRORS_PANIC);
3110 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3111 set_opt(sb, ERRORS_CONT);
3113 set_opt(sb, ERRORS_RO);
3114 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3115 set_opt(sb, BLOCK_VALIDITY);
3116 if (def_mount_opts & EXT4_DEFM_DISCARD)
3117 set_opt(sb, DISCARD);
3119 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3120 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3121 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3122 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3123 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3125 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3126 set_opt(sb, BARRIER);
3129 * enable delayed allocation by default
3130 * Use -o nodelalloc to turn it off
3132 if (!IS_EXT3_SB(sb) &&
3133 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3134 set_opt(sb, DELALLOC);
3136 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3137 &journal_devnum, &journal_ioprio, NULL, 0)) {
3138 ext4_msg(sb, KERN_WARNING,
3139 "failed to parse options in superblock: %s",
3140 sbi->s_es->s_mount_opts);
3142 if (!parse_options((char *) data, sb, &journal_devnum,
3143 &journal_ioprio, NULL, 0))
3146 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3147 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3149 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3150 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3151 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3152 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3153 ext4_msg(sb, KERN_WARNING,
3154 "feature flags set on rev 0 fs, "
3155 "running e2fsck is recommended");
3158 * Check feature flags regardless of the revision level, since we
3159 * previously didn't change the revision level when setting the flags,
3160 * so there is a chance incompat flags are set on a rev 0 filesystem.
3162 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3165 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3167 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3168 blocksize > EXT4_MAX_BLOCK_SIZE) {
3169 ext4_msg(sb, KERN_ERR,
3170 "Unsupported filesystem blocksize %d", blocksize);
3174 if (sb->s_blocksize != blocksize) {
3175 /* Validate the filesystem blocksize */
3176 if (!sb_set_blocksize(sb, blocksize)) {
3177 ext4_msg(sb, KERN_ERR, "bad block size %d",
3183 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3184 offset = do_div(logical_sb_block, blocksize);
3185 bh = sb_bread(sb, logical_sb_block);
3187 ext4_msg(sb, KERN_ERR,
3188 "Can't read superblock on 2nd try");
3191 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3193 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3194 ext4_msg(sb, KERN_ERR,
3195 "Magic mismatch, very weird!");
3200 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3201 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3202 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3204 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3206 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3207 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3208 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3210 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3211 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3212 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3213 (!is_power_of_2(sbi->s_inode_size)) ||
3214 (sbi->s_inode_size > blocksize)) {
3215 ext4_msg(sb, KERN_ERR,
3216 "unsupported inode size: %d",
3220 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3221 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3224 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3225 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3226 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3227 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3228 !is_power_of_2(sbi->s_desc_size)) {
3229 ext4_msg(sb, KERN_ERR,
3230 "unsupported descriptor size %lu",
3235 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3237 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3238 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3239 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3242 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3243 if (sbi->s_inodes_per_block == 0)
3245 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3246 sbi->s_inodes_per_block;
3247 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3249 sbi->s_mount_state = le16_to_cpu(es->s_state);
3250 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3251 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3253 for (i = 0; i < 4; i++)
3254 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3255 sbi->s_def_hash_version = es->s_def_hash_version;
3256 i = le32_to_cpu(es->s_flags);
3257 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3258 sbi->s_hash_unsigned = 3;
3259 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3260 #ifdef __CHAR_UNSIGNED__
3261 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3262 sbi->s_hash_unsigned = 3;
3264 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3269 if (sbi->s_blocks_per_group > blocksize * 8) {
3270 ext4_msg(sb, KERN_ERR,
3271 "#blocks per group too big: %lu",
3272 sbi->s_blocks_per_group);
3275 if (sbi->s_inodes_per_group > blocksize * 8) {
3276 ext4_msg(sb, KERN_ERR,
3277 "#inodes per group too big: %lu",
3278 sbi->s_inodes_per_group);
3283 * Test whether we have more sectors than will fit in sector_t,
3284 * and whether the max offset is addressable by the page cache.
3286 err = generic_check_addressable(sb->s_blocksize_bits,
3287 ext4_blocks_count(es));
3289 ext4_msg(sb, KERN_ERR, "filesystem"
3290 " too large to mount safely on this system");
3291 if (sizeof(sector_t) < 8)
3292 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3297 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3300 /* check blocks count against device size */
3301 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3302 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3303 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3304 "exceeds size of device (%llu blocks)",
3305 ext4_blocks_count(es), blocks_count);
3310 * It makes no sense for the first data block to be beyond the end
3311 * of the filesystem.
3313 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3314 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3315 "block %u is beyond end of filesystem (%llu)",
3316 le32_to_cpu(es->s_first_data_block),
3317 ext4_blocks_count(es));
3320 blocks_count = (ext4_blocks_count(es) -
3321 le32_to_cpu(es->s_first_data_block) +
3322 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3323 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3324 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3325 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3326 "(block count %llu, first data block %u, "
3327 "blocks per group %lu)", sbi->s_groups_count,
3328 ext4_blocks_count(es),
3329 le32_to_cpu(es->s_first_data_block),
3330 EXT4_BLOCKS_PER_GROUP(sb));
3333 sbi->s_groups_count = blocks_count;
3334 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3335 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3336 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3337 EXT4_DESC_PER_BLOCK(sb);
3338 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3340 if (sbi->s_group_desc == NULL) {
3341 ext4_msg(sb, KERN_ERR, "not enough memory");
3345 #ifdef CONFIG_PROC_FS
3347 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3350 bgl_lock_init(sbi->s_blockgroup_lock);
3352 for (i = 0; i < db_count; i++) {
3353 block = descriptor_loc(sb, logical_sb_block, i);
3354 sbi->s_group_desc[i] = sb_bread(sb, block);
3355 if (!sbi->s_group_desc[i]) {
3356 ext4_msg(sb, KERN_ERR,
3357 "can't read group descriptor %d", i);
3362 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3363 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3366 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3367 if (!ext4_fill_flex_info(sb)) {
3368 ext4_msg(sb, KERN_ERR,
3369 "unable to initialize "
3370 "flex_bg meta info!");
3374 sbi->s_gdb_count = db_count;
3375 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3376 spin_lock_init(&sbi->s_next_gen_lock);
3378 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3379 ext4_count_free_blocks(sb));
3381 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3382 ext4_count_free_inodes(sb));
3385 err = percpu_counter_init(&sbi->s_dirs_counter,
3386 ext4_count_dirs(sb));
3389 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3392 ext4_msg(sb, KERN_ERR, "insufficient memory");
3396 sbi->s_stripe = ext4_get_stripe_size(sbi);
3397 sbi->s_max_writeback_mb_bump = 128;
3400 * set up enough so that it can read an inode
3402 if (!test_opt(sb, NOLOAD) &&
3403 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3404 sb->s_op = &ext4_sops;
3406 sb->s_op = &ext4_nojournal_sops;
3407 sb->s_export_op = &ext4_export_ops;
3408 sb->s_xattr = ext4_xattr_handlers;
3410 sb->s_qcop = &ext4_qctl_operations;
3411 sb->dq_op = &ext4_quota_operations;
3413 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3414 mutex_init(&sbi->s_orphan_lock);
3415 mutex_init(&sbi->s_resize_lock);
3419 needs_recovery = (es->s_last_orphan != 0 ||
3420 EXT4_HAS_INCOMPAT_FEATURE(sb,
3421 EXT4_FEATURE_INCOMPAT_RECOVER));
3424 * The first inode we look at is the journal inode. Don't try
3425 * root first: it may be modified in the journal!
3427 if (!test_opt(sb, NOLOAD) &&
3428 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3429 if (ext4_load_journal(sb, es, journal_devnum))
3431 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3432 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3433 ext4_msg(sb, KERN_ERR, "required journal recovery "
3434 "suppressed and not mounted read-only");
3435 goto failed_mount_wq;
3437 clear_opt(sb, DATA_FLAGS);
3438 set_opt(sb, WRITEBACK_DATA);
3439 sbi->s_journal = NULL;
3444 if (ext4_blocks_count(es) > 0xffffffffULL &&
3445 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3446 JBD2_FEATURE_INCOMPAT_64BIT)) {
3447 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3448 goto failed_mount_wq;
3451 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3452 jbd2_journal_set_features(sbi->s_journal,
3453 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3454 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3455 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3456 jbd2_journal_set_features(sbi->s_journal,
3457 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3458 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3461 jbd2_journal_clear_features(sbi->s_journal,
3462 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3463 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3466 /* We have now updated the journal if required, so we can
3467 * validate the data journaling mode. */
3468 switch (test_opt(sb, DATA_FLAGS)) {
3470 /* No mode set, assume a default based on the journal
3471 * capabilities: ORDERED_DATA if the journal can
3472 * cope, else JOURNAL_DATA
3474 if (jbd2_journal_check_available_features
3475 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3476 set_opt(sb, ORDERED_DATA);
3478 set_opt(sb, JOURNAL_DATA);
3481 case EXT4_MOUNT_ORDERED_DATA:
3482 case EXT4_MOUNT_WRITEBACK_DATA:
3483 if (!jbd2_journal_check_available_features
3484 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3485 ext4_msg(sb, KERN_ERR, "Journal does not support "
3486 "requested data journaling mode");
3487 goto failed_mount_wq;
3492 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3495 * The journal may have updated the bg summary counts, so we
3496 * need to update the global counters.
3498 percpu_counter_set(&sbi->s_freeblocks_counter,
3499 ext4_count_free_blocks(sb));
3500 percpu_counter_set(&sbi->s_freeinodes_counter,
3501 ext4_count_free_inodes(sb));
3502 percpu_counter_set(&sbi->s_dirs_counter,
3503 ext4_count_dirs(sb));
3504 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3507 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3508 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3509 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3510 goto failed_mount_wq;
3514 * The jbd2_journal_load will have done any necessary log recovery,
3515 * so we can safely mount the rest of the filesystem now.
3518 root = ext4_iget(sb, EXT4_ROOT_INO);
3520 ext4_msg(sb, KERN_ERR, "get root inode failed");
3521 ret = PTR_ERR(root);
3524 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3526 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3529 sb->s_root = d_alloc_root(root);
3531 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3537 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3539 /* determine the minimum size of new large inodes, if present */
3540 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3541 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3542 EXT4_GOOD_OLD_INODE_SIZE;
3543 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3544 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3545 if (sbi->s_want_extra_isize <
3546 le16_to_cpu(es->s_want_extra_isize))
3547 sbi->s_want_extra_isize =
3548 le16_to_cpu(es->s_want_extra_isize);
3549 if (sbi->s_want_extra_isize <
3550 le16_to_cpu(es->s_min_extra_isize))
3551 sbi->s_want_extra_isize =
3552 le16_to_cpu(es->s_min_extra_isize);
3555 /* Check if enough inode space is available */
3556 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3557 sbi->s_inode_size) {
3558 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3559 EXT4_GOOD_OLD_INODE_SIZE;
3560 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3564 if (test_opt(sb, DELALLOC) &&
3565 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3566 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3567 "requested data journaling mode");
3568 clear_opt(sb, DELALLOC);
3570 if (test_opt(sb, DIOREAD_NOLOCK)) {
3571 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3572 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3573 "option - requested data journaling mode");
3574 clear_opt(sb, DIOREAD_NOLOCK);
3576 if (sb->s_blocksize < PAGE_SIZE) {
3577 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3578 "option - block size is too small");
3579 clear_opt(sb, DIOREAD_NOLOCK);
3583 err = ext4_setup_system_zone(sb);
3585 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3591 err = ext4_mb_init(sb, needs_recovery);
3593 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3598 err = ext4_register_li_request(sb, first_not_zeroed);
3602 sbi->s_kobj.kset = ext4_kset;
3603 init_completion(&sbi->s_kobj_unregister);
3604 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3607 ext4_mb_release(sb);
3608 ext4_ext_release(sb);
3612 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3613 ext4_orphan_cleanup(sb, es);
3614 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3615 if (needs_recovery) {
3616 ext4_msg(sb, KERN_INFO, "recovery complete");
3617 ext4_mark_recovery_complete(sb, es);
3619 if (EXT4_SB(sb)->s_journal) {
3620 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3621 descr = " journalled data mode";
3622 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3623 descr = " ordered data mode";
3625 descr = " writeback data mode";
3627 descr = "out journal";
3629 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3630 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3631 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3633 init_timer(&sbi->s_err_report);
3634 sbi->s_err_report.function = print_daily_error_info;
3635 sbi->s_err_report.data = (unsigned long) sb;
3636 if (es->s_error_count)
3637 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3644 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3648 ext4_msg(sb, KERN_ERR, "mount failed");
3649 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3651 ext4_release_system_zone(sb);
3652 if (sbi->s_journal) {
3653 jbd2_journal_destroy(sbi->s_journal);
3654 sbi->s_journal = NULL;
3657 if (sbi->s_flex_groups) {
3658 if (is_vmalloc_addr(sbi->s_flex_groups))
3659 vfree(sbi->s_flex_groups);
3661 kfree(sbi->s_flex_groups);
3663 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3664 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3665 percpu_counter_destroy(&sbi->s_dirs_counter);
3666 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3668 for (i = 0; i < db_count; i++)
3669 brelse(sbi->s_group_desc[i]);
3670 kfree(sbi->s_group_desc);
3673 remove_proc_entry(sb->s_id, ext4_proc_root);
3676 for (i = 0; i < MAXQUOTAS; i++)
3677 kfree(sbi->s_qf_names[i]);
3679 ext4_blkdev_remove(sbi);
3682 sb->s_fs_info = NULL;
3683 kfree(sbi->s_blockgroup_lock);
3691 * Setup any per-fs journal parameters now. We'll do this both on
3692 * initial mount, once the journal has been initialised but before we've
3693 * done any recovery; and again on any subsequent remount.
3695 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3697 struct ext4_sb_info *sbi = EXT4_SB(sb);
3699 journal->j_commit_interval = sbi->s_commit_interval;
3700 journal->j_min_batch_time = sbi->s_min_batch_time;
3701 journal->j_max_batch_time = sbi->s_max_batch_time;
3703 write_lock(&journal->j_state_lock);
3704 if (test_opt(sb, BARRIER))
3705 journal->j_flags |= JBD2_BARRIER;
3707 journal->j_flags &= ~JBD2_BARRIER;
3708 if (test_opt(sb, DATA_ERR_ABORT))
3709 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3711 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3712 write_unlock(&journal->j_state_lock);
3715 static journal_t *ext4_get_journal(struct super_block *sb,
3716 unsigned int journal_inum)
3718 struct inode *journal_inode;
3721 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3723 /* First, test for the existence of a valid inode on disk. Bad
3724 * things happen if we iget() an unused inode, as the subsequent
3725 * iput() will try to delete it. */
3727 journal_inode = ext4_iget(sb, journal_inum);
3728 if (IS_ERR(journal_inode)) {
3729 ext4_msg(sb, KERN_ERR, "no journal found");
3732 if (!journal_inode->i_nlink) {
3733 make_bad_inode(journal_inode);
3734 iput(journal_inode);
3735 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3739 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3740 journal_inode, journal_inode->i_size);
3741 if (!S_ISREG(journal_inode->i_mode)) {
3742 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3743 iput(journal_inode);
3747 journal = jbd2_journal_init_inode(journal_inode);
3749 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3750 iput(journal_inode);
3753 journal->j_private = sb;
3754 ext4_init_journal_params(sb, journal);
3758 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3761 struct buffer_head *bh;
3765 int hblock, blocksize;
3766 ext4_fsblk_t sb_block;
3767 unsigned long offset;
3768 struct ext4_super_block *es;
3769 struct block_device *bdev;
3771 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3773 bdev = ext4_blkdev_get(j_dev, sb);
3777 if (bd_claim(bdev, sb)) {
3778 ext4_msg(sb, KERN_ERR,
3779 "failed to claim external journal device");
3780 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3784 blocksize = sb->s_blocksize;
3785 hblock = bdev_logical_block_size(bdev);
3786 if (blocksize < hblock) {
3787 ext4_msg(sb, KERN_ERR,
3788 "blocksize too small for journal device");
3792 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3793 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3794 set_blocksize(bdev, blocksize);
3795 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3796 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3797 "external journal");
3801 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3802 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3803 !(le32_to_cpu(es->s_feature_incompat) &
3804 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3805 ext4_msg(sb, KERN_ERR, "external journal has "
3811 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3812 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3817 len = ext4_blocks_count(es);
3818 start = sb_block + 1;
3819 brelse(bh); /* we're done with the superblock */
3821 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3822 start, len, blocksize);
3824 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3827 journal->j_private = sb;
3828 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3829 wait_on_buffer(journal->j_sb_buffer);
3830 if (!buffer_uptodate(journal->j_sb_buffer)) {
3831 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3834 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3835 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3836 "user (unsupported) - %d",
3837 be32_to_cpu(journal->j_superblock->s_nr_users));
3840 EXT4_SB(sb)->journal_bdev = bdev;
3841 ext4_init_journal_params(sb, journal);
3845 jbd2_journal_destroy(journal);
3847 ext4_blkdev_put(bdev);
3851 static int ext4_load_journal(struct super_block *sb,
3852 struct ext4_super_block *es,
3853 unsigned long journal_devnum)
3856 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3859 int really_read_only;
3861 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3863 if (journal_devnum &&
3864 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3865 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3866 "numbers have changed");
3867 journal_dev = new_decode_dev(journal_devnum);
3869 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3871 really_read_only = bdev_read_only(sb->s_bdev);
3874 * Are we loading a blank journal or performing recovery after a
3875 * crash? For recovery, we need to check in advance whether we
3876 * can get read-write access to the device.
3878 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3879 if (sb->s_flags & MS_RDONLY) {
3880 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3881 "required on readonly filesystem");
3882 if (really_read_only) {
3883 ext4_msg(sb, KERN_ERR, "write access "
3884 "unavailable, cannot proceed");
3887 ext4_msg(sb, KERN_INFO, "write access will "
3888 "be enabled during recovery");
3892 if (journal_inum && journal_dev) {
3893 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3894 "and inode journals!");
3899 if (!(journal = ext4_get_journal(sb, journal_inum)))
3902 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3906 if (!(journal->j_flags & JBD2_BARRIER))
3907 ext4_msg(sb, KERN_INFO, "barriers disabled");
3909 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3910 err = jbd2_journal_update_format(journal);
3912 ext4_msg(sb, KERN_ERR, "error updating journal");
3913 jbd2_journal_destroy(journal);
3918 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3919 err = jbd2_journal_wipe(journal, !really_read_only);
3921 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3923 memcpy(save, ((char *) es) +
3924 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3925 err = jbd2_journal_load(journal);
3927 memcpy(((char *) es) + EXT4_S_ERR_START,
3928 save, EXT4_S_ERR_LEN);
3933 ext4_msg(sb, KERN_ERR, "error loading journal");
3934 jbd2_journal_destroy(journal);
3938 EXT4_SB(sb)->s_journal = journal;
3939 ext4_clear_journal_err(sb, es);
3941 if (!really_read_only && journal_devnum &&
3942 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3943 es->s_journal_dev = cpu_to_le32(journal_devnum);
3945 /* Make sure we flush the recovery flag to disk. */
3946 ext4_commit_super(sb, 1);
3952 static int ext4_commit_super(struct super_block *sb, int sync)
3954 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3955 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3960 if (buffer_write_io_error(sbh)) {
3962 * Oh, dear. A previous attempt to write the
3963 * superblock failed. This could happen because the
3964 * USB device was yanked out. Or it could happen to
3965 * be a transient write error and maybe the block will
3966 * be remapped. Nothing we can do but to retry the
3967 * write and hope for the best.
3969 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3970 "superblock detected");
3971 clear_buffer_write_io_error(sbh);
3972 set_buffer_uptodate(sbh);
3975 * If the file system is mounted read-only, don't update the
3976 * superblock write time. This avoids updating the superblock
3977 * write time when we are mounting the root file system
3978 * read/only but we need to replay the journal; at that point,
3979 * for people who are east of GMT and who make their clock
3980 * tick in localtime for Windows bug-for-bug compatibility,
3981 * the clock is set in the future, and this will cause e2fsck
3982 * to complain and force a full file system check.
3984 if (!(sb->s_flags & MS_RDONLY))
3985 es->s_wtime = cpu_to_le32(get_seconds());
3986 if (sb->s_bdev->bd_part)
3987 es->s_kbytes_written =
3988 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3989 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3990 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3992 es->s_kbytes_written =
3993 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3994 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3995 &EXT4_SB(sb)->s_freeblocks_counter));
3996 es->s_free_inodes_count =
3997 cpu_to_le32(percpu_counter_sum_positive(
3998 &EXT4_SB(sb)->s_freeinodes_counter));
4000 BUFFER_TRACE(sbh, "marking dirty");
4001 mark_buffer_dirty(sbh);
4003 error = sync_dirty_buffer(sbh);
4007 error = buffer_write_io_error(sbh);
4009 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4011 clear_buffer_write_io_error(sbh);
4012 set_buffer_uptodate(sbh);
4019 * Have we just finished recovery? If so, and if we are mounting (or
4020 * remounting) the filesystem readonly, then we will end up with a
4021 * consistent fs on disk. Record that fact.
4023 static void ext4_mark_recovery_complete(struct super_block *sb,
4024 struct ext4_super_block *es)
4026 journal_t *journal = EXT4_SB(sb)->s_journal;
4028 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4029 BUG_ON(journal != NULL);
4032 jbd2_journal_lock_updates(journal);
4033 if (jbd2_journal_flush(journal) < 0)
4036 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4037 sb->s_flags & MS_RDONLY) {
4038 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4039 ext4_commit_super(sb, 1);
4043 jbd2_journal_unlock_updates(journal);
4047 * If we are mounting (or read-write remounting) a filesystem whose journal
4048 * has recorded an error from a previous lifetime, move that error to the
4049 * main filesystem now.
4051 static void ext4_clear_journal_err(struct super_block *sb,
4052 struct ext4_super_block *es)
4058 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4060 journal = EXT4_SB(sb)->s_journal;
4063 * Now check for any error status which may have been recorded in the
4064 * journal by a prior ext4_error() or ext4_abort()
4067 j_errno = jbd2_journal_errno(journal);
4071 errstr = ext4_decode_error(sb, j_errno, nbuf);
4072 ext4_warning(sb, "Filesystem error recorded "
4073 "from previous mount: %s", errstr);
4074 ext4_warning(sb, "Marking fs in need of filesystem check.");
4076 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4077 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4078 ext4_commit_super(sb, 1);
4080 jbd2_journal_clear_err(journal);
4085 * Force the running and committing transactions to commit,
4086 * and wait on the commit.
4088 int ext4_force_commit(struct super_block *sb)
4093 if (sb->s_flags & MS_RDONLY)
4096 journal = EXT4_SB(sb)->s_journal;
4098 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4099 ret = ext4_journal_force_commit(journal);
4105 static void ext4_write_super(struct super_block *sb)
4108 ext4_commit_super(sb, 1);
4112 static int ext4_sync_fs(struct super_block *sb, int wait)
4116 struct ext4_sb_info *sbi = EXT4_SB(sb);
4118 trace_ext4_sync_fs(sb, wait);
4119 flush_workqueue(sbi->dio_unwritten_wq);
4120 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4122 jbd2_log_wait_commit(sbi->s_journal, target);
4128 * LVM calls this function before a (read-only) snapshot is created. This
4129 * gives us a chance to flush the journal completely and mark the fs clean.
4131 static int ext4_freeze(struct super_block *sb)
4136 if (sb->s_flags & MS_RDONLY)
4139 journal = EXT4_SB(sb)->s_journal;
4141 /* Now we set up the journal barrier. */
4142 jbd2_journal_lock_updates(journal);
4145 * Don't clear the needs_recovery flag if we failed to flush
4148 error = jbd2_journal_flush(journal);
4152 /* Journal blocked and flushed, clear needs_recovery flag. */
4153 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4154 error = ext4_commit_super(sb, 1);
4156 /* we rely on s_frozen to stop further updates */
4157 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4162 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4163 * flag here, even though the filesystem is not technically dirty yet.
4165 static int ext4_unfreeze(struct super_block *sb)
4167 if (sb->s_flags & MS_RDONLY)
4171 /* Reset the needs_recovery flag before the fs is unlocked. */
4172 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4173 ext4_commit_super(sb, 1);
4179 * Structure to save mount options for ext4_remount's benefit
4181 struct ext4_mount_options {
4182 unsigned long s_mount_opt;
4183 unsigned long s_mount_opt2;
4186 unsigned long s_commit_interval;
4187 u32 s_min_batch_time, s_max_batch_time;
4190 char *s_qf_names[MAXQUOTAS];
4194 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4196 struct ext4_super_block *es;
4197 struct ext4_sb_info *sbi = EXT4_SB(sb);
4198 ext4_fsblk_t n_blocks_count = 0;
4199 unsigned long old_sb_flags;
4200 struct ext4_mount_options old_opts;
4201 int enable_quota = 0;
4203 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4208 char *orig_data = kstrdup(data, GFP_KERNEL);
4210 /* Store the original options */
4212 old_sb_flags = sb->s_flags;
4213 old_opts.s_mount_opt = sbi->s_mount_opt;
4214 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4215 old_opts.s_resuid = sbi->s_resuid;
4216 old_opts.s_resgid = sbi->s_resgid;
4217 old_opts.s_commit_interval = sbi->s_commit_interval;
4218 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4219 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4221 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4222 for (i = 0; i < MAXQUOTAS; i++)
4223 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4225 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4226 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4229 * Allow the "check" option to be passed as a remount option.
4231 if (!parse_options(data, sb, NULL, &journal_ioprio,
4232 &n_blocks_count, 1)) {
4237 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4238 ext4_abort(sb, "Abort forced by user");
4240 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4241 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4245 if (sbi->s_journal) {
4246 ext4_init_journal_params(sb, sbi->s_journal);
4247 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4250 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4251 n_blocks_count > ext4_blocks_count(es)) {
4252 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4257 if (*flags & MS_RDONLY) {
4258 err = dquot_suspend(sb, -1);
4263 * First of all, the unconditional stuff we have to do
4264 * to disable replay of the journal when we next remount
4266 sb->s_flags |= MS_RDONLY;
4269 * OK, test if we are remounting a valid rw partition
4270 * readonly, and if so set the rdonly flag and then
4271 * mark the partition as valid again.
4273 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4274 (sbi->s_mount_state & EXT4_VALID_FS))
4275 es->s_state = cpu_to_le16(sbi->s_mount_state);
4278 ext4_mark_recovery_complete(sb, es);
4280 /* Make sure we can mount this feature set readwrite */
4281 if (!ext4_feature_set_ok(sb, 0)) {
4286 * Make sure the group descriptor checksums
4287 * are sane. If they aren't, refuse to remount r/w.
4289 for (g = 0; g < sbi->s_groups_count; g++) {
4290 struct ext4_group_desc *gdp =
4291 ext4_get_group_desc(sb, g, NULL);
4293 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4294 ext4_msg(sb, KERN_ERR,
4295 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4296 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4297 le16_to_cpu(gdp->bg_checksum));
4304 * If we have an unprocessed orphan list hanging
4305 * around from a previously readonly bdev mount,
4306 * require a full umount/remount for now.
4308 if (es->s_last_orphan) {
4309 ext4_msg(sb, KERN_WARNING, "Couldn't "
4310 "remount RDWR because of unprocessed "
4311 "orphan inode list. Please "
4312 "umount/remount instead");
4318 * Mounting a RDONLY partition read-write, so reread
4319 * and store the current valid flag. (It may have
4320 * been changed by e2fsck since we originally mounted
4324 ext4_clear_journal_err(sb, es);
4325 sbi->s_mount_state = le16_to_cpu(es->s_state);
4326 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4328 if (!ext4_setup_super(sb, es, 0))
4329 sb->s_flags &= ~MS_RDONLY;
4335 * Reinitialize lazy itable initialization thread based on
4338 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4339 ext4_unregister_li_request(sb);
4341 ext4_group_t first_not_zeroed;
4342 first_not_zeroed = ext4_has_uninit_itable(sb);
4343 ext4_register_li_request(sb, first_not_zeroed);
4346 ext4_setup_system_zone(sb);
4347 if (sbi->s_journal == NULL)
4348 ext4_commit_super(sb, 1);
4351 /* Release old quota file names */
4352 for (i = 0; i < MAXQUOTAS; i++)
4353 if (old_opts.s_qf_names[i] &&
4354 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4355 kfree(old_opts.s_qf_names[i]);
4359 dquot_resume(sb, -1);
4361 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4366 sb->s_flags = old_sb_flags;
4367 sbi->s_mount_opt = old_opts.s_mount_opt;
4368 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4369 sbi->s_resuid = old_opts.s_resuid;
4370 sbi->s_resgid = old_opts.s_resgid;
4371 sbi->s_commit_interval = old_opts.s_commit_interval;
4372 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4373 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4375 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4376 for (i = 0; i < MAXQUOTAS; i++) {
4377 if (sbi->s_qf_names[i] &&
4378 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4379 kfree(sbi->s_qf_names[i]);
4380 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4388 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4390 struct super_block *sb = dentry->d_sb;
4391 struct ext4_sb_info *sbi = EXT4_SB(sb);
4392 struct ext4_super_block *es = sbi->s_es;
4395 if (test_opt(sb, MINIX_DF)) {
4396 sbi->s_overhead_last = 0;
4397 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4398 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4399 ext4_fsblk_t overhead = 0;
4402 * Compute the overhead (FS structures). This is constant
4403 * for a given filesystem unless the number of block groups
4404 * changes so we cache the previous value until it does.
4408 * All of the blocks before first_data_block are
4411 overhead = le32_to_cpu(es->s_first_data_block);
4414 * Add the overhead attributed to the superblock and
4415 * block group descriptors. If the sparse superblocks
4416 * feature is turned on, then not all groups have this.
4418 for (i = 0; i < ngroups; i++) {
4419 overhead += ext4_bg_has_super(sb, i) +
4420 ext4_bg_num_gdb(sb, i);
4425 * Every block group has an inode bitmap, a block
4426 * bitmap, and an inode table.
4428 overhead += ngroups * (2 + sbi->s_itb_per_group);
4429 sbi->s_overhead_last = overhead;
4431 sbi->s_blocks_last = ext4_blocks_count(es);
4434 buf->f_type = EXT4_SUPER_MAGIC;
4435 buf->f_bsize = sb->s_blocksize;
4436 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4437 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4438 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4439 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4440 if (buf->f_bfree < ext4_r_blocks_count(es))
4442 buf->f_files = le32_to_cpu(es->s_inodes_count);
4443 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4444 buf->f_namelen = EXT4_NAME_LEN;
4445 fsid = le64_to_cpup((void *)es->s_uuid) ^
4446 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4447 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4448 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4453 /* Helper function for writing quotas on sync - we need to start transaction
4454 * before quota file is locked for write. Otherwise the are possible deadlocks:
4455 * Process 1 Process 2
4456 * ext4_create() quota_sync()
4457 * jbd2_journal_start() write_dquot()
4458 * dquot_initialize() down(dqio_mutex)
4459 * down(dqio_mutex) jbd2_journal_start()
4465 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4467 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4470 static int ext4_write_dquot(struct dquot *dquot)
4474 struct inode *inode;
4476 inode = dquot_to_inode(dquot);
4477 handle = ext4_journal_start(inode,
4478 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4480 return PTR_ERR(handle);
4481 ret = dquot_commit(dquot);
4482 err = ext4_journal_stop(handle);
4488 static int ext4_acquire_dquot(struct dquot *dquot)
4493 handle = ext4_journal_start(dquot_to_inode(dquot),
4494 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4496 return PTR_ERR(handle);
4497 ret = dquot_acquire(dquot);
4498 err = ext4_journal_stop(handle);
4504 static int ext4_release_dquot(struct dquot *dquot)
4509 handle = ext4_journal_start(dquot_to_inode(dquot),
4510 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4511 if (IS_ERR(handle)) {
4512 /* Release dquot anyway to avoid endless cycle in dqput() */
4513 dquot_release(dquot);
4514 return PTR_ERR(handle);
4516 ret = dquot_release(dquot);
4517 err = ext4_journal_stop(handle);
4523 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4525 /* Are we journaling quotas? */
4526 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4527 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4528 dquot_mark_dquot_dirty(dquot);
4529 return ext4_write_dquot(dquot);
4531 return dquot_mark_dquot_dirty(dquot);
4535 static int ext4_write_info(struct super_block *sb, int type)
4540 /* Data block + inode block */
4541 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4543 return PTR_ERR(handle);
4544 ret = dquot_commit_info(sb, type);
4545 err = ext4_journal_stop(handle);
4552 * Turn on quotas during mount time - we need to find
4553 * the quota file and such...
4555 static int ext4_quota_on_mount(struct super_block *sb, int type)
4557 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4558 EXT4_SB(sb)->s_jquota_fmt, type);
4562 * Standard function to be called on quota_on
4564 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4570 if (!test_opt(sb, QUOTA))
4573 err = kern_path(name, LOOKUP_FOLLOW, &path);
4577 /* Quotafile not on the same filesystem? */
4578 if (path.mnt->mnt_sb != sb) {
4582 /* Journaling quota? */
4583 if (EXT4_SB(sb)->s_qf_names[type]) {
4584 /* Quotafile not in fs root? */
4585 if (path.dentry->d_parent != sb->s_root)
4586 ext4_msg(sb, KERN_WARNING,
4587 "Quota file not on filesystem root. "
4588 "Journaled quota will not work");
4592 * When we journal data on quota file, we have to flush journal to see
4593 * all updates to the file when we bypass pagecache...
4595 if (EXT4_SB(sb)->s_journal &&
4596 ext4_should_journal_data(path.dentry->d_inode)) {
4598 * We don't need to lock updates but journal_flush() could
4599 * otherwise be livelocked...
4601 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4602 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4603 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4610 err = dquot_quota_on_path(sb, type, format_id, &path);
4615 static int ext4_quota_off(struct super_block *sb, int type)
4617 /* Force all delayed allocation blocks to be allocated.
4618 * Caller already holds s_umount sem */
4619 if (test_opt(sb, DELALLOC))
4620 sync_filesystem(sb);
4622 return dquot_quota_off(sb, type);
4625 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4626 * acquiring the locks... As quota files are never truncated and quota code
4627 * itself serializes the operations (and noone else should touch the files)
4628 * we don't have to be afraid of races */
4629 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4630 size_t len, loff_t off)
4632 struct inode *inode = sb_dqopt(sb)->files[type];
4633 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4635 int offset = off & (sb->s_blocksize - 1);
4638 struct buffer_head *bh;
4639 loff_t i_size = i_size_read(inode);
4643 if (off+len > i_size)
4646 while (toread > 0) {
4647 tocopy = sb->s_blocksize - offset < toread ?
4648 sb->s_blocksize - offset : toread;
4649 bh = ext4_bread(NULL, inode, blk, 0, &err);
4652 if (!bh) /* A hole? */
4653 memset(data, 0, tocopy);
4655 memcpy(data, bh->b_data+offset, tocopy);
4665 /* Write to quotafile (we know the transaction is already started and has
4666 * enough credits) */
4667 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4668 const char *data, size_t len, loff_t off)
4670 struct inode *inode = sb_dqopt(sb)->files[type];
4671 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4673 int offset = off & (sb->s_blocksize - 1);
4674 struct buffer_head *bh;
4675 handle_t *handle = journal_current_handle();
4677 if (EXT4_SB(sb)->s_journal && !handle) {
4678 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4679 " cancelled because transaction is not started",
4680 (unsigned long long)off, (unsigned long long)len);
4684 * Since we account only one data block in transaction credits,
4685 * then it is impossible to cross a block boundary.
4687 if (sb->s_blocksize - offset < len) {
4688 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4689 " cancelled because not block aligned",
4690 (unsigned long long)off, (unsigned long long)len);
4694 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4695 bh = ext4_bread(handle, inode, blk, 1, &err);
4698 err = ext4_journal_get_write_access(handle, bh);
4704 memcpy(bh->b_data+offset, data, len);
4705 flush_dcache_page(bh->b_page);
4707 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4711 mutex_unlock(&inode->i_mutex);
4714 if (inode->i_size < off + len) {
4715 i_size_write(inode, off + len);
4716 EXT4_I(inode)->i_disksize = inode->i_size;
4718 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4719 ext4_mark_inode_dirty(handle, inode);
4720 mutex_unlock(&inode->i_mutex);
4726 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4727 const char *dev_name, void *data)
4729 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4732 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4733 static struct file_system_type ext2_fs_type = {
4734 .owner = THIS_MODULE,
4736 .mount = ext4_mount,
4737 .kill_sb = kill_block_super,
4738 .fs_flags = FS_REQUIRES_DEV,
4741 static inline void register_as_ext2(void)
4743 int err = register_filesystem(&ext2_fs_type);
4746 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4749 static inline void unregister_as_ext2(void)
4751 unregister_filesystem(&ext2_fs_type);
4753 MODULE_ALIAS("ext2");
4755 static inline void register_as_ext2(void) { }
4756 static inline void unregister_as_ext2(void) { }
4759 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4760 static inline void register_as_ext3(void)
4762 int err = register_filesystem(&ext3_fs_type);
4765 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4768 static inline void unregister_as_ext3(void)
4770 unregister_filesystem(&ext3_fs_type);
4772 MODULE_ALIAS("ext3");
4774 static inline void register_as_ext3(void) { }
4775 static inline void unregister_as_ext3(void) { }
4778 static struct file_system_type ext4_fs_type = {
4779 .owner = THIS_MODULE,
4781 .mount = ext4_mount,
4782 .kill_sb = kill_block_super,
4783 .fs_flags = FS_REQUIRES_DEV,
4786 int __init ext4_init_feat_adverts(void)
4788 struct ext4_features *ef;
4791 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4795 ef->f_kobj.kset = ext4_kset;
4796 init_completion(&ef->f_kobj_unregister);
4797 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4810 static int __init ext4_init_fs(void)
4814 ext4_check_flag_values();
4815 err = ext4_init_pageio();
4818 err = ext4_init_system_zone();
4821 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4824 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4826 err = ext4_init_feat_adverts();
4828 err = ext4_init_mballoc();
4832 err = ext4_init_xattr();
4835 err = init_inodecache();
4840 err = register_filesystem(&ext4_fs_type);
4844 ext4_li_info = NULL;
4845 mutex_init(&ext4_li_mtx);
4848 unregister_as_ext2();
4849 unregister_as_ext3();
4850 destroy_inodecache();
4854 ext4_exit_mballoc();
4857 remove_proc_entry("fs/ext4", NULL);
4858 kset_unregister(ext4_kset);
4860 ext4_exit_system_zone();
4866 static void __exit ext4_exit_fs(void)
4868 ext4_destroy_lazyinit_thread();
4869 unregister_as_ext2();
4870 unregister_as_ext3();
4871 unregister_filesystem(&ext4_fs_type);
4872 destroy_inodecache();
4874 ext4_exit_mballoc();
4875 remove_proc_entry("fs/ext4", NULL);
4876 kset_unregister(ext4_kset);
4877 ext4_exit_system_zone();
4881 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4882 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4883 MODULE_LICENSE("GPL");
4884 module_init(ext4_init_fs)
4885 module_exit(ext4_exit_fs)