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 static struct ext4_lazy_init *ext4_li_info;
58 static struct mutex ext4_li_mtx;
59 static 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 inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86 static struct file_system_type ext2_fs_type = {
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #define IS_EXT2_SB(sb) (0)
99 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100 static struct file_system_type ext3_fs_type = {
101 .owner = THIS_MODULE,
104 .kill_sb = kill_block_super,
105 .fs_flags = FS_REQUIRES_DEV,
107 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109 #define IS_EXT3_SB(sb) (0)
112 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
113 struct ext4_group_desc *bg)
115 return le32_to_cpu(bg->bg_block_bitmap_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
120 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
121 struct ext4_group_desc *bg)
123 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
128 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
129 struct ext4_group_desc *bg)
131 return le32_to_cpu(bg->bg_inode_table_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
136 __u32 ext4_free_blks_count(struct super_block *sb,
137 struct ext4_group_desc *bg)
139 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
140 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
141 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
144 __u32 ext4_free_inodes_count(struct super_block *sb,
145 struct ext4_group_desc *bg)
147 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
149 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
152 __u32 ext4_used_dirs_count(struct super_block *sb,
153 struct ext4_group_desc *bg)
155 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
156 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
157 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
160 __u32 ext4_itable_unused_count(struct super_block *sb,
161 struct ext4_group_desc *bg)
163 return le16_to_cpu(bg->bg_itable_unused_lo) |
164 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
165 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
168 void ext4_block_bitmap_set(struct super_block *sb,
169 struct ext4_group_desc *bg, ext4_fsblk_t blk)
171 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
176 void ext4_inode_bitmap_set(struct super_block *sb,
177 struct ext4_group_desc *bg, ext4_fsblk_t blk)
179 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
184 void ext4_inode_table_set(struct super_block *sb,
185 struct ext4_group_desc *bg, ext4_fsblk_t blk)
187 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
192 void ext4_free_blks_set(struct super_block *sb,
193 struct ext4_group_desc *bg, __u32 count)
195 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
196 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
197 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
200 void ext4_free_inodes_set(struct super_block *sb,
201 struct ext4_group_desc *bg, __u32 count)
203 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
208 void ext4_used_dirs_set(struct super_block *sb,
209 struct ext4_group_desc *bg, __u32 count)
211 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
212 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
213 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
216 void ext4_itable_unused_set(struct super_block *sb,
217 struct ext4_group_desc *bg, __u32 count)
219 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
220 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
221 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
225 /* Just increment the non-pointer handle value */
226 static handle_t *ext4_get_nojournal(void)
228 handle_t *handle = current->journal_info;
229 unsigned long ref_cnt = (unsigned long)handle;
231 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
234 handle = (handle_t *)ref_cnt;
236 current->journal_info = handle;
241 /* Decrement the non-pointer handle value */
242 static void ext4_put_nojournal(handle_t *handle)
244 unsigned long ref_cnt = (unsigned long)handle;
246 BUG_ON(ref_cnt == 0);
249 handle = (handle_t *)ref_cnt;
251 current->journal_info = handle;
255 * Wrappers for jbd2_journal_start/end.
257 * The only special thing we need to do here is to make sure that all
258 * journal_end calls result in the superblock being marked dirty, so
259 * that sync() will call the filesystem's write_super callback if
262 * To avoid j_barrier hold in userspace when a user calls freeze(),
263 * ext4 prevents a new handle from being started by s_frozen, which
264 * is in an upper layer.
266 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
271 if (sb->s_flags & MS_RDONLY)
272 return ERR_PTR(-EROFS);
274 journal = EXT4_SB(sb)->s_journal;
275 handle = ext4_journal_current_handle();
278 * If a handle has been started, it should be allowed to
279 * finish, otherwise deadlock could happen between freeze
280 * and others(e.g. truncate) due to the restart of the
281 * journal handle if the filesystem is forzen and active
282 * handles are not stopped.
285 vfs_check_frozen(sb, SB_FREEZE_TRANS);
288 return ext4_get_nojournal();
290 * Special case here: if the journal has aborted behind our
291 * backs (eg. EIO in the commit thread), then we still need to
292 * take the FS itself readonly cleanly.
294 if (is_journal_aborted(journal)) {
295 ext4_abort(sb, "Detected aborted journal");
296 return ERR_PTR(-EROFS);
298 return jbd2_journal_start(journal, nblocks);
302 * The only special thing we need to do here is to make sure that all
303 * jbd2_journal_stop calls result in the superblock being marked dirty, so
304 * that sync() will call the filesystem's write_super callback if
307 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
309 struct super_block *sb;
313 if (!ext4_handle_valid(handle)) {
314 ext4_put_nojournal(handle);
317 sb = handle->h_transaction->t_journal->j_private;
319 rc = jbd2_journal_stop(handle);
324 __ext4_std_error(sb, where, line, err);
328 void ext4_journal_abort_handle(const char *caller, unsigned int line,
329 const char *err_fn, struct buffer_head *bh,
330 handle_t *handle, int err)
333 const char *errstr = ext4_decode_error(NULL, err, nbuf);
335 BUG_ON(!ext4_handle_valid(handle));
338 BUFFER_TRACE(bh, "abort");
343 if (is_handle_aborted(handle))
346 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
347 caller, line, errstr, err_fn);
349 jbd2_journal_abort_handle(handle);
352 static void __save_error_info(struct super_block *sb, const char *func,
355 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
357 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
358 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
359 es->s_last_error_time = cpu_to_le32(get_seconds());
360 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
361 es->s_last_error_line = cpu_to_le32(line);
362 if (!es->s_first_error_time) {
363 es->s_first_error_time = es->s_last_error_time;
364 strncpy(es->s_first_error_func, func,
365 sizeof(es->s_first_error_func));
366 es->s_first_error_line = cpu_to_le32(line);
367 es->s_first_error_ino = es->s_last_error_ino;
368 es->s_first_error_block = es->s_last_error_block;
371 * Start the daily error reporting function if it hasn't been
374 if (!es->s_error_count)
375 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
376 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
379 static void save_error_info(struct super_block *sb, const char *func,
382 __save_error_info(sb, func, line);
383 ext4_commit_super(sb, 1);
387 /* Deal with the reporting of failure conditions on a filesystem such as
388 * inconsistencies detected or read IO failures.
390 * On ext2, we can store the error state of the filesystem in the
391 * superblock. That is not possible on ext4, because we may have other
392 * write ordering constraints on the superblock which prevent us from
393 * writing it out straight away; and given that the journal is about to
394 * be aborted, we can't rely on the current, or future, transactions to
395 * write out the superblock safely.
397 * We'll just use the jbd2_journal_abort() error code to record an error in
398 * the journal instead. On recovery, the journal will complain about
399 * that error until we've noted it down and cleared it.
402 static void ext4_handle_error(struct super_block *sb)
404 if (sb->s_flags & MS_RDONLY)
407 if (!test_opt(sb, ERRORS_CONT)) {
408 journal_t *journal = EXT4_SB(sb)->s_journal;
410 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
412 jbd2_journal_abort(journal, -EIO);
414 if (test_opt(sb, ERRORS_RO)) {
415 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
416 sb->s_flags |= MS_RDONLY;
418 if (test_opt(sb, ERRORS_PANIC))
419 panic("EXT4-fs (device %s): panic forced after error\n",
423 void __ext4_error(struct super_block *sb, const char *function,
424 unsigned int line, const char *fmt, ...)
426 struct va_format vaf;
432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
433 sb->s_id, function, line, current->comm, &vaf);
436 ext4_handle_error(sb);
439 void ext4_error_inode(struct inode *inode, const char *function,
440 unsigned int line, ext4_fsblk_t block,
441 const char *fmt, ...)
444 struct va_format vaf;
445 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
447 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
448 es->s_last_error_block = cpu_to_le64(block);
449 save_error_info(inode->i_sb, function, line);
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
454 inode->i_sb->s_id, function, line, inode->i_ino);
456 printk(KERN_CONT "block %llu: ", block);
457 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
460 ext4_handle_error(inode->i_sb);
463 void ext4_error_file(struct file *file, const char *function,
464 unsigned int line, ext4_fsblk_t block,
465 const char *fmt, ...)
468 struct va_format vaf;
469 struct ext4_super_block *es;
470 struct inode *inode = file->f_dentry->d_inode;
471 char pathname[80], *path;
473 es = EXT4_SB(inode->i_sb)->s_es;
474 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
475 save_error_info(inode->i_sb, function, line);
476 path = d_path(&(file->f_path), pathname, sizeof(pathname));
480 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
481 inode->i_sb->s_id, function, line, inode->i_ino);
483 printk(KERN_CONT "block %llu: ", block);
487 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
490 ext4_handle_error(inode->i_sb);
493 static const char *ext4_decode_error(struct super_block *sb, int errno,
500 errstr = "IO failure";
503 errstr = "Out of memory";
506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
508 errstr = "Journal has aborted";
510 errstr = "Readonly filesystem";
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
527 /* __ext4_std_error decodes expected errors from journaling functions
528 * automatically and invokes the appropriate error response. */
530 void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
543 errstr = ext4_decode_error(sb, errno, nbuf);
544 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
545 sb->s_id, function, line, errstr);
546 save_error_info(sb, function, line);
548 ext4_handle_error(sb);
552 * ext4_abort is a much stronger failure handler than ext4_error. The
553 * abort function may be used to deal with unrecoverable failures such
554 * as journal IO errors or ENOMEM at a critical moment in log management.
556 * We unconditionally force the filesystem into an ABORT|READONLY state,
557 * unless the error response on the fs has been set to panic in which
558 * case we take the easy way out and panic immediately.
561 void __ext4_abort(struct super_block *sb, const char *function,
562 unsigned int line, const char *fmt, ...)
566 save_error_info(sb, function, line);
568 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
574 if ((sb->s_flags & MS_RDONLY) == 0) {
575 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
576 sb->s_flags |= MS_RDONLY;
577 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
578 if (EXT4_SB(sb)->s_journal)
579 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
580 save_error_info(sb, function, line);
582 if (test_opt(sb, ERRORS_PANIC))
583 panic("EXT4-fs panic from previous error\n");
586 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
588 struct va_format vaf;
594 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
598 void __ext4_warning(struct super_block *sb, const char *function,
599 unsigned int line, const char *fmt, ...)
601 struct va_format vaf;
607 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
612 void __ext4_grp_locked_error(const char *function, unsigned int line,
613 struct super_block *sb, ext4_group_t grp,
614 unsigned long ino, ext4_fsblk_t block,
615 const char *fmt, ...)
619 struct va_format vaf;
621 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
623 es->s_last_error_ino = cpu_to_le32(ino);
624 es->s_last_error_block = cpu_to_le64(block);
625 __save_error_info(sb, function, line);
631 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
632 sb->s_id, function, line, grp);
634 printk(KERN_CONT "inode %lu: ", ino);
636 printk(KERN_CONT "block %llu:", (unsigned long long) block);
637 printk(KERN_CONT "%pV\n", &vaf);
640 if (test_opt(sb, ERRORS_CONT)) {
641 ext4_commit_super(sb, 0);
645 ext4_unlock_group(sb, grp);
646 ext4_handle_error(sb);
648 * We only get here in the ERRORS_RO case; relocking the group
649 * may be dangerous, but nothing bad will happen since the
650 * filesystem will have already been marked read/only and the
651 * journal has been aborted. We return 1 as a hint to callers
652 * who might what to use the return value from
653 * ext4_grp_locked_error() to distinguish between the
654 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
655 * aggressively from the ext4 function in question, with a
656 * more appropriate error code.
658 ext4_lock_group(sb, grp);
662 void ext4_update_dynamic_rev(struct super_block *sb)
664 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
666 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
670 "updating to rev %d because of new feature flag, "
671 "running e2fsck is recommended",
674 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
675 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
676 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
677 /* leave es->s_feature_*compat flags alone */
678 /* es->s_uuid will be set by e2fsck if empty */
681 * The rest of the superblock fields should be zero, and if not it
682 * means they are likely already in use, so leave them alone. We
683 * can leave it up to e2fsck to clean up any inconsistencies there.
688 * Open the external journal device
690 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
692 struct block_device *bdev;
693 char b[BDEVNAME_SIZE];
695 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
701 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
702 __bdevname(dev, b), PTR_ERR(bdev));
707 * Release the journal device
709 static int ext4_blkdev_put(struct block_device *bdev)
711 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
714 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
716 struct block_device *bdev;
719 bdev = sbi->journal_bdev;
721 ret = ext4_blkdev_put(bdev);
722 sbi->journal_bdev = NULL;
727 static inline struct inode *orphan_list_entry(struct list_head *l)
729 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
732 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
736 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
737 le32_to_cpu(sbi->s_es->s_last_orphan));
739 printk(KERN_ERR "sb_info orphan list:\n");
740 list_for_each(l, &sbi->s_orphan) {
741 struct inode *inode = orphan_list_entry(l);
743 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
744 inode->i_sb->s_id, inode->i_ino, inode,
745 inode->i_mode, inode->i_nlink,
750 static void ext4_put_super(struct super_block *sb)
752 struct ext4_sb_info *sbi = EXT4_SB(sb);
753 struct ext4_super_block *es = sbi->s_es;
756 ext4_unregister_li_request(sb);
757 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
759 flush_workqueue(sbi->dio_unwritten_wq);
760 destroy_workqueue(sbi->dio_unwritten_wq);
764 ext4_commit_super(sb, 1);
766 if (sbi->s_journal) {
767 err = jbd2_journal_destroy(sbi->s_journal);
768 sbi->s_journal = NULL;
770 ext4_abort(sb, "Couldn't clean up the journal");
773 del_timer(&sbi->s_err_report);
774 ext4_release_system_zone(sb);
776 ext4_ext_release(sb);
777 ext4_xattr_put_super(sb);
779 if (!(sb->s_flags & MS_RDONLY)) {
780 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
781 es->s_state = cpu_to_le16(sbi->s_mount_state);
782 ext4_commit_super(sb, 1);
785 remove_proc_entry(sb->s_id, ext4_proc_root);
787 kobject_del(&sbi->s_kobj);
789 for (i = 0; i < sbi->s_gdb_count; i++)
790 brelse(sbi->s_group_desc[i]);
791 kfree(sbi->s_group_desc);
792 if (is_vmalloc_addr(sbi->s_flex_groups))
793 vfree(sbi->s_flex_groups);
795 kfree(sbi->s_flex_groups);
796 percpu_counter_destroy(&sbi->s_freeblocks_counter);
797 percpu_counter_destroy(&sbi->s_freeinodes_counter);
798 percpu_counter_destroy(&sbi->s_dirs_counter);
799 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
802 for (i = 0; i < MAXQUOTAS; i++)
803 kfree(sbi->s_qf_names[i]);
806 /* Debugging code just in case the in-memory inode orphan list
807 * isn't empty. The on-disk one can be non-empty if we've
808 * detected an error and taken the fs readonly, but the
809 * in-memory list had better be clean by this point. */
810 if (!list_empty(&sbi->s_orphan))
811 dump_orphan_list(sb, sbi);
812 J_ASSERT(list_empty(&sbi->s_orphan));
814 invalidate_bdev(sb->s_bdev);
815 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
817 * Invalidate the journal device's buffers. We don't want them
818 * floating about in memory - the physical journal device may
819 * hotswapped, and it breaks the `ro-after' testing code.
821 sync_blockdev(sbi->journal_bdev);
822 invalidate_bdev(sbi->journal_bdev);
823 ext4_blkdev_remove(sbi);
825 sb->s_fs_info = NULL;
827 * Now that we are completely done shutting down the
828 * superblock, we need to actually destroy the kobject.
831 kobject_put(&sbi->s_kobj);
832 wait_for_completion(&sbi->s_kobj_unregister);
833 kfree(sbi->s_blockgroup_lock);
837 static struct kmem_cache *ext4_inode_cachep;
840 * Called inside transaction, so use GFP_NOFS
842 static struct inode *ext4_alloc_inode(struct super_block *sb)
844 struct ext4_inode_info *ei;
846 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
850 ei->vfs_inode.i_version = 1;
851 ei->vfs_inode.i_data.writeback_index = 0;
852 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
853 INIT_LIST_HEAD(&ei->i_prealloc_list);
854 spin_lock_init(&ei->i_prealloc_lock);
855 ei->i_reserved_data_blocks = 0;
856 ei->i_reserved_meta_blocks = 0;
857 ei->i_allocated_meta_blocks = 0;
858 ei->i_da_metadata_calc_len = 0;
859 spin_lock_init(&(ei->i_block_reservation_lock));
861 ei->i_reserved_quota = 0;
864 INIT_LIST_HEAD(&ei->i_completed_io_list);
865 spin_lock_init(&ei->i_completed_io_lock);
866 ei->cur_aio_dio = NULL;
868 ei->i_datasync_tid = 0;
869 atomic_set(&ei->i_ioend_count, 0);
870 atomic_set(&ei->i_aiodio_unwritten, 0);
872 return &ei->vfs_inode;
875 static int ext4_drop_inode(struct inode *inode)
877 int drop = generic_drop_inode(inode);
879 trace_ext4_drop_inode(inode, drop);
883 static void ext4_i_callback(struct rcu_head *head)
885 struct inode *inode = container_of(head, struct inode, i_rcu);
886 INIT_LIST_HEAD(&inode->i_dentry);
887 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
890 static void ext4_destroy_inode(struct inode *inode)
892 ext4_ioend_wait(inode);
893 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
894 ext4_msg(inode->i_sb, KERN_ERR,
895 "Inode %lu (%p): orphan list check failed!",
896 inode->i_ino, EXT4_I(inode));
897 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
898 EXT4_I(inode), sizeof(struct ext4_inode_info),
902 call_rcu(&inode->i_rcu, ext4_i_callback);
905 static void init_once(void *foo)
907 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
909 INIT_LIST_HEAD(&ei->i_orphan);
910 #ifdef CONFIG_EXT4_FS_XATTR
911 init_rwsem(&ei->xattr_sem);
913 init_rwsem(&ei->i_data_sem);
914 inode_init_once(&ei->vfs_inode);
917 static int init_inodecache(void)
919 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
920 sizeof(struct ext4_inode_info),
921 0, (SLAB_RECLAIM_ACCOUNT|
924 if (ext4_inode_cachep == NULL)
929 static void destroy_inodecache(void)
931 kmem_cache_destroy(ext4_inode_cachep);
934 void ext4_clear_inode(struct inode *inode)
936 invalidate_inode_buffers(inode);
937 end_writeback(inode);
939 ext4_discard_preallocations(inode);
940 if (EXT4_I(inode)->jinode) {
941 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
942 EXT4_I(inode)->jinode);
943 jbd2_free_inode(EXT4_I(inode)->jinode);
944 EXT4_I(inode)->jinode = NULL;
948 static inline void ext4_show_quota_options(struct seq_file *seq,
949 struct super_block *sb)
951 #if defined(CONFIG_QUOTA)
952 struct ext4_sb_info *sbi = EXT4_SB(sb);
954 if (sbi->s_jquota_fmt) {
957 switch (sbi->s_jquota_fmt) {
968 seq_printf(seq, ",jqfmt=%s", fmtname);
971 if (sbi->s_qf_names[USRQUOTA])
972 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
974 if (sbi->s_qf_names[GRPQUOTA])
975 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
977 if (test_opt(sb, USRQUOTA))
978 seq_puts(seq, ",usrquota");
980 if (test_opt(sb, GRPQUOTA))
981 seq_puts(seq, ",grpquota");
987 * - it's set to a non-default value OR
988 * - if the per-sb default is different from the global default
990 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
993 unsigned long def_mount_opts;
994 struct super_block *sb = vfs->mnt_sb;
995 struct ext4_sb_info *sbi = EXT4_SB(sb);
996 struct ext4_super_block *es = sbi->s_es;
998 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
999 def_errors = le16_to_cpu(es->s_errors);
1001 if (sbi->s_sb_block != 1)
1002 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1003 if (test_opt(sb, MINIX_DF))
1004 seq_puts(seq, ",minixdf");
1005 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1006 seq_puts(seq, ",grpid");
1007 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1008 seq_puts(seq, ",nogrpid");
1009 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1010 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1011 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1013 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1014 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1015 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1017 if (test_opt(sb, ERRORS_RO)) {
1018 if (def_errors == EXT4_ERRORS_PANIC ||
1019 def_errors == EXT4_ERRORS_CONTINUE) {
1020 seq_puts(seq, ",errors=remount-ro");
1023 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1024 seq_puts(seq, ",errors=continue");
1025 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1026 seq_puts(seq, ",errors=panic");
1027 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1028 seq_puts(seq, ",nouid32");
1029 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1030 seq_puts(seq, ",debug");
1031 if (test_opt(sb, OLDALLOC))
1032 seq_puts(seq, ",oldalloc");
1033 #ifdef CONFIG_EXT4_FS_XATTR
1034 if (test_opt(sb, XATTR_USER))
1035 seq_puts(seq, ",user_xattr");
1036 if (!test_opt(sb, XATTR_USER))
1037 seq_puts(seq, ",nouser_xattr");
1039 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1040 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1041 seq_puts(seq, ",acl");
1042 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1043 seq_puts(seq, ",noacl");
1045 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1046 seq_printf(seq, ",commit=%u",
1047 (unsigned) (sbi->s_commit_interval / HZ));
1049 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1050 seq_printf(seq, ",min_batch_time=%u",
1051 (unsigned) sbi->s_min_batch_time);
1053 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1054 seq_printf(seq, ",max_batch_time=%u",
1055 (unsigned) sbi->s_min_batch_time);
1059 * We're changing the default of barrier mount option, so
1060 * let's always display its mount state so it's clear what its
1063 seq_puts(seq, ",barrier=");
1064 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1065 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1066 seq_puts(seq, ",journal_async_commit");
1067 else if (test_opt(sb, JOURNAL_CHECKSUM))
1068 seq_puts(seq, ",journal_checksum");
1069 if (test_opt(sb, I_VERSION))
1070 seq_puts(seq, ",i_version");
1071 if (!test_opt(sb, DELALLOC) &&
1072 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1073 seq_puts(seq, ",nodelalloc");
1075 if (!test_opt(sb, MBLK_IO_SUBMIT))
1076 seq_puts(seq, ",nomblk_io_submit");
1078 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1080 * journal mode get enabled in different ways
1081 * So just print the value even if we didn't specify it
1083 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1084 seq_puts(seq, ",data=journal");
1085 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1086 seq_puts(seq, ",data=ordered");
1087 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1088 seq_puts(seq, ",data=writeback");
1090 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1091 seq_printf(seq, ",inode_readahead_blks=%u",
1092 sbi->s_inode_readahead_blks);
1094 if (test_opt(sb, DATA_ERR_ABORT))
1095 seq_puts(seq, ",data_err=abort");
1097 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1098 seq_puts(seq, ",noauto_da_alloc");
1100 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1101 seq_puts(seq, ",discard");
1103 if (test_opt(sb, NOLOAD))
1104 seq_puts(seq, ",norecovery");
1106 if (test_opt(sb, DIOREAD_NOLOCK))
1107 seq_puts(seq, ",dioread_nolock");
1109 if (test_opt(sb, BLOCK_VALIDITY) &&
1110 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1111 seq_puts(seq, ",block_validity");
1113 if (!test_opt(sb, INIT_INODE_TABLE))
1114 seq_puts(seq, ",noinit_inode_table");
1115 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1116 seq_printf(seq, ",init_inode_table=%u",
1117 (unsigned) sbi->s_li_wait_mult);
1119 ext4_show_quota_options(seq, sb);
1124 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1125 u64 ino, u32 generation)
1127 struct inode *inode;
1129 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1130 return ERR_PTR(-ESTALE);
1131 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1132 return ERR_PTR(-ESTALE);
1134 /* iget isn't really right if the inode is currently unallocated!!
1136 * ext4_read_inode will return a bad_inode if the inode had been
1137 * deleted, so we should be safe.
1139 * Currently we don't know the generation for parent directory, so
1140 * a generation of 0 means "accept any"
1142 inode = ext4_iget(sb, ino);
1144 return ERR_CAST(inode);
1145 if (generation && inode->i_generation != generation) {
1147 return ERR_PTR(-ESTALE);
1153 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1154 int fh_len, int fh_type)
1156 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1157 ext4_nfs_get_inode);
1160 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1161 int fh_len, int fh_type)
1163 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1164 ext4_nfs_get_inode);
1168 * Try to release metadata pages (indirect blocks, directories) which are
1169 * mapped via the block device. Since these pages could have journal heads
1170 * which would prevent try_to_free_buffers() from freeing them, we must use
1171 * jbd2 layer's try_to_free_buffers() function to release them.
1173 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1176 journal_t *journal = EXT4_SB(sb)->s_journal;
1178 WARN_ON(PageChecked(page));
1179 if (!page_has_buffers(page))
1182 return jbd2_journal_try_to_free_buffers(journal, page,
1183 wait & ~__GFP_WAIT);
1184 return try_to_free_buffers(page);
1188 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1189 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1191 static int ext4_write_dquot(struct dquot *dquot);
1192 static int ext4_acquire_dquot(struct dquot *dquot);
1193 static int ext4_release_dquot(struct dquot *dquot);
1194 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1195 static int ext4_write_info(struct super_block *sb, int type);
1196 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1198 static int ext4_quota_off(struct super_block *sb, int type);
1199 static int ext4_quota_on_mount(struct super_block *sb, int type);
1200 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1201 size_t len, loff_t off);
1202 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1203 const char *data, size_t len, loff_t off);
1205 static const struct dquot_operations ext4_quota_operations = {
1206 .get_reserved_space = ext4_get_reserved_space,
1207 .write_dquot = ext4_write_dquot,
1208 .acquire_dquot = ext4_acquire_dquot,
1209 .release_dquot = ext4_release_dquot,
1210 .mark_dirty = ext4_mark_dquot_dirty,
1211 .write_info = ext4_write_info,
1212 .alloc_dquot = dquot_alloc,
1213 .destroy_dquot = dquot_destroy,
1216 static const struct quotactl_ops ext4_qctl_operations = {
1217 .quota_on = ext4_quota_on,
1218 .quota_off = ext4_quota_off,
1219 .quota_sync = dquot_quota_sync,
1220 .get_info = dquot_get_dqinfo,
1221 .set_info = dquot_set_dqinfo,
1222 .get_dqblk = dquot_get_dqblk,
1223 .set_dqblk = dquot_set_dqblk
1227 static const struct super_operations ext4_sops = {
1228 .alloc_inode = ext4_alloc_inode,
1229 .destroy_inode = ext4_destroy_inode,
1230 .write_inode = ext4_write_inode,
1231 .dirty_inode = ext4_dirty_inode,
1232 .drop_inode = ext4_drop_inode,
1233 .evict_inode = ext4_evict_inode,
1234 .put_super = ext4_put_super,
1235 .sync_fs = ext4_sync_fs,
1236 .freeze_fs = ext4_freeze,
1237 .unfreeze_fs = ext4_unfreeze,
1238 .statfs = ext4_statfs,
1239 .remount_fs = ext4_remount,
1240 .show_options = ext4_show_options,
1242 .quota_read = ext4_quota_read,
1243 .quota_write = ext4_quota_write,
1245 .bdev_try_to_free_page = bdev_try_to_free_page,
1248 static const struct super_operations ext4_nojournal_sops = {
1249 .alloc_inode = ext4_alloc_inode,
1250 .destroy_inode = ext4_destroy_inode,
1251 .write_inode = ext4_write_inode,
1252 .dirty_inode = ext4_dirty_inode,
1253 .drop_inode = ext4_drop_inode,
1254 .evict_inode = ext4_evict_inode,
1255 .write_super = ext4_write_super,
1256 .put_super = ext4_put_super,
1257 .statfs = ext4_statfs,
1258 .remount_fs = ext4_remount,
1259 .show_options = ext4_show_options,
1261 .quota_read = ext4_quota_read,
1262 .quota_write = ext4_quota_write,
1264 .bdev_try_to_free_page = bdev_try_to_free_page,
1267 static const struct export_operations ext4_export_ops = {
1268 .fh_to_dentry = ext4_fh_to_dentry,
1269 .fh_to_parent = ext4_fh_to_parent,
1270 .get_parent = ext4_get_parent,
1274 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1275 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1276 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1277 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1278 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1279 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1280 Opt_journal_update, Opt_journal_dev,
1281 Opt_journal_checksum, Opt_journal_async_commit,
1282 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1283 Opt_data_err_abort, Opt_data_err_ignore,
1284 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1285 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1286 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1287 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1288 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1289 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1290 Opt_inode_readahead_blks, Opt_journal_ioprio,
1291 Opt_dioread_nolock, Opt_dioread_lock,
1292 Opt_discard, Opt_nodiscard,
1293 Opt_init_inode_table, Opt_noinit_inode_table,
1296 static const match_table_t tokens = {
1297 {Opt_bsd_df, "bsddf"},
1298 {Opt_minix_df, "minixdf"},
1299 {Opt_grpid, "grpid"},
1300 {Opt_grpid, "bsdgroups"},
1301 {Opt_nogrpid, "nogrpid"},
1302 {Opt_nogrpid, "sysvgroups"},
1303 {Opt_resgid, "resgid=%u"},
1304 {Opt_resuid, "resuid=%u"},
1306 {Opt_err_cont, "errors=continue"},
1307 {Opt_err_panic, "errors=panic"},
1308 {Opt_err_ro, "errors=remount-ro"},
1309 {Opt_nouid32, "nouid32"},
1310 {Opt_debug, "debug"},
1311 {Opt_oldalloc, "oldalloc"},
1312 {Opt_orlov, "orlov"},
1313 {Opt_user_xattr, "user_xattr"},
1314 {Opt_nouser_xattr, "nouser_xattr"},
1316 {Opt_noacl, "noacl"},
1317 {Opt_noload, "noload"},
1318 {Opt_noload, "norecovery"},
1321 {Opt_commit, "commit=%u"},
1322 {Opt_min_batch_time, "min_batch_time=%u"},
1323 {Opt_max_batch_time, "max_batch_time=%u"},
1324 {Opt_journal_update, "journal=update"},
1325 {Opt_journal_dev, "journal_dev=%u"},
1326 {Opt_journal_checksum, "journal_checksum"},
1327 {Opt_journal_async_commit, "journal_async_commit"},
1328 {Opt_abort, "abort"},
1329 {Opt_data_journal, "data=journal"},
1330 {Opt_data_ordered, "data=ordered"},
1331 {Opt_data_writeback, "data=writeback"},
1332 {Opt_data_err_abort, "data_err=abort"},
1333 {Opt_data_err_ignore, "data_err=ignore"},
1334 {Opt_offusrjquota, "usrjquota="},
1335 {Opt_usrjquota, "usrjquota=%s"},
1336 {Opt_offgrpjquota, "grpjquota="},
1337 {Opt_grpjquota, "grpjquota=%s"},
1338 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1339 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1340 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1341 {Opt_grpquota, "grpquota"},
1342 {Opt_noquota, "noquota"},
1343 {Opt_quota, "quota"},
1344 {Opt_usrquota, "usrquota"},
1345 {Opt_barrier, "barrier=%u"},
1346 {Opt_barrier, "barrier"},
1347 {Opt_nobarrier, "nobarrier"},
1348 {Opt_i_version, "i_version"},
1349 {Opt_stripe, "stripe=%u"},
1350 {Opt_resize, "resize"},
1351 {Opt_delalloc, "delalloc"},
1352 {Opt_nodelalloc, "nodelalloc"},
1353 {Opt_mblk_io_submit, "mblk_io_submit"},
1354 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1355 {Opt_block_validity, "block_validity"},
1356 {Opt_noblock_validity, "noblock_validity"},
1357 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1358 {Opt_journal_ioprio, "journal_ioprio=%u"},
1359 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1360 {Opt_auto_da_alloc, "auto_da_alloc"},
1361 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1362 {Opt_dioread_nolock, "dioread_nolock"},
1363 {Opt_dioread_lock, "dioread_lock"},
1364 {Opt_discard, "discard"},
1365 {Opt_nodiscard, "nodiscard"},
1366 {Opt_init_inode_table, "init_itable=%u"},
1367 {Opt_init_inode_table, "init_itable"},
1368 {Opt_noinit_inode_table, "noinit_itable"},
1372 static ext4_fsblk_t get_sb_block(void **data)
1374 ext4_fsblk_t sb_block;
1375 char *options = (char *) *data;
1377 if (!options || strncmp(options, "sb=", 3) != 0)
1378 return 1; /* Default location */
1381 /* TODO: use simple_strtoll with >32bit ext4 */
1382 sb_block = simple_strtoul(options, &options, 0);
1383 if (*options && *options != ',') {
1384 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1388 if (*options == ',')
1390 *data = (void *) options;
1395 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1396 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1397 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1400 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1402 struct ext4_sb_info *sbi = EXT4_SB(sb);
1405 if (sb_any_quota_loaded(sb) &&
1406 !sbi->s_qf_names[qtype]) {
1407 ext4_msg(sb, KERN_ERR,
1408 "Cannot change journaled "
1409 "quota options when quota turned on");
1412 qname = match_strdup(args);
1414 ext4_msg(sb, KERN_ERR,
1415 "Not enough memory for storing quotafile name");
1418 if (sbi->s_qf_names[qtype] &&
1419 strcmp(sbi->s_qf_names[qtype], qname)) {
1420 ext4_msg(sb, KERN_ERR,
1421 "%s quota file already specified", QTYPE2NAME(qtype));
1425 sbi->s_qf_names[qtype] = qname;
1426 if (strchr(sbi->s_qf_names[qtype], '/')) {
1427 ext4_msg(sb, KERN_ERR,
1428 "quotafile must be on filesystem root");
1429 kfree(sbi->s_qf_names[qtype]);
1430 sbi->s_qf_names[qtype] = NULL;
1437 static int clear_qf_name(struct super_block *sb, int qtype)
1440 struct ext4_sb_info *sbi = EXT4_SB(sb);
1442 if (sb_any_quota_loaded(sb) &&
1443 sbi->s_qf_names[qtype]) {
1444 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1445 " when quota turned on");
1449 * The space will be released later when all options are confirmed
1452 sbi->s_qf_names[qtype] = NULL;
1457 static int parse_options(char *options, struct super_block *sb,
1458 unsigned long *journal_devnum,
1459 unsigned int *journal_ioprio,
1460 ext4_fsblk_t *n_blocks_count, int is_remount)
1462 struct ext4_sb_info *sbi = EXT4_SB(sb);
1464 substring_t args[MAX_OPT_ARGS];
1474 while ((p = strsep(&options, ",")) != NULL) {
1480 * Initialize args struct so we know whether arg was
1481 * found; some options take optional arguments.
1483 args[0].to = args[0].from = NULL;
1484 token = match_token(p, tokens, args);
1487 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1488 clear_opt(sb, MINIX_DF);
1491 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1492 set_opt(sb, MINIX_DF);
1496 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1501 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1502 clear_opt(sb, GRPID);
1506 if (match_int(&args[0], &option))
1508 sbi->s_resuid = option;
1511 if (match_int(&args[0], &option))
1513 sbi->s_resgid = option;
1516 /* handled by get_sb_block() instead of here */
1517 /* *sb_block = match_int(&args[0]); */
1520 clear_opt(sb, ERRORS_CONT);
1521 clear_opt(sb, ERRORS_RO);
1522 set_opt(sb, ERRORS_PANIC);
1525 clear_opt(sb, ERRORS_CONT);
1526 clear_opt(sb, ERRORS_PANIC);
1527 set_opt(sb, ERRORS_RO);
1530 clear_opt(sb, ERRORS_RO);
1531 clear_opt(sb, ERRORS_PANIC);
1532 set_opt(sb, ERRORS_CONT);
1535 set_opt(sb, NO_UID32);
1541 set_opt(sb, OLDALLOC);
1544 clear_opt(sb, OLDALLOC);
1546 #ifdef CONFIG_EXT4_FS_XATTR
1547 case Opt_user_xattr:
1548 set_opt(sb, XATTR_USER);
1550 case Opt_nouser_xattr:
1551 clear_opt(sb, XATTR_USER);
1554 case Opt_user_xattr:
1555 case Opt_nouser_xattr:
1556 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1559 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1561 set_opt(sb, POSIX_ACL);
1564 clear_opt(sb, POSIX_ACL);
1569 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1572 case Opt_journal_update:
1574 /* Eventually we will want to be able to create
1575 a journal file here. For now, only allow the
1576 user to specify an existing inode to be the
1579 ext4_msg(sb, KERN_ERR,
1580 "Cannot specify journal on remount");
1583 set_opt(sb, UPDATE_JOURNAL);
1585 case Opt_journal_dev:
1587 ext4_msg(sb, KERN_ERR,
1588 "Cannot specify journal on remount");
1591 if (match_int(&args[0], &option))
1593 *journal_devnum = option;
1595 case Opt_journal_checksum:
1596 set_opt(sb, JOURNAL_CHECKSUM);
1598 case Opt_journal_async_commit:
1599 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1600 set_opt(sb, JOURNAL_CHECKSUM);
1603 set_opt(sb, NOLOAD);
1606 if (match_int(&args[0], &option))
1611 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1612 sbi->s_commit_interval = HZ * option;
1614 case Opt_max_batch_time:
1615 if (match_int(&args[0], &option))
1620 option = EXT4_DEF_MAX_BATCH_TIME;
1621 sbi->s_max_batch_time = option;
1623 case Opt_min_batch_time:
1624 if (match_int(&args[0], &option))
1628 sbi->s_min_batch_time = option;
1630 case Opt_data_journal:
1631 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1633 case Opt_data_ordered:
1634 data_opt = EXT4_MOUNT_ORDERED_DATA;
1636 case Opt_data_writeback:
1637 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1640 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1641 ext4_msg(sb, KERN_ERR,
1642 "Cannot change data mode on remount");
1646 clear_opt(sb, DATA_FLAGS);
1647 sbi->s_mount_opt |= data_opt;
1650 case Opt_data_err_abort:
1651 set_opt(sb, DATA_ERR_ABORT);
1653 case Opt_data_err_ignore:
1654 clear_opt(sb, DATA_ERR_ABORT);
1658 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1662 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1665 case Opt_offusrjquota:
1666 if (!clear_qf_name(sb, USRQUOTA))
1669 case Opt_offgrpjquota:
1670 if (!clear_qf_name(sb, GRPQUOTA))
1674 case Opt_jqfmt_vfsold:
1675 qfmt = QFMT_VFS_OLD;
1677 case Opt_jqfmt_vfsv0:
1680 case Opt_jqfmt_vfsv1:
1683 if (sb_any_quota_loaded(sb) &&
1684 sbi->s_jquota_fmt != qfmt) {
1685 ext4_msg(sb, KERN_ERR, "Cannot change "
1686 "journaled quota options when "
1690 sbi->s_jquota_fmt = qfmt;
1695 set_opt(sb, USRQUOTA);
1699 set_opt(sb, GRPQUOTA);
1702 if (sb_any_quota_loaded(sb)) {
1703 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1704 "options when quota turned on");
1707 clear_opt(sb, QUOTA);
1708 clear_opt(sb, USRQUOTA);
1709 clear_opt(sb, GRPQUOTA);
1715 ext4_msg(sb, KERN_ERR,
1716 "quota options not supported");
1720 case Opt_offusrjquota:
1721 case Opt_offgrpjquota:
1722 case Opt_jqfmt_vfsold:
1723 case Opt_jqfmt_vfsv0:
1724 case Opt_jqfmt_vfsv1:
1725 ext4_msg(sb, KERN_ERR,
1726 "journaled quota options not supported");
1732 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1735 clear_opt(sb, BARRIER);
1739 if (match_int(&args[0], &option))
1742 option = 1; /* No argument, default to 1 */
1744 set_opt(sb, BARRIER);
1746 clear_opt(sb, BARRIER);
1752 ext4_msg(sb, KERN_ERR,
1753 "resize option only available "
1757 if (match_int(&args[0], &option) != 0)
1759 *n_blocks_count = option;
1762 ext4_msg(sb, KERN_WARNING,
1763 "Ignoring deprecated nobh option");
1766 ext4_msg(sb, KERN_WARNING,
1767 "Ignoring deprecated bh option");
1770 set_opt(sb, I_VERSION);
1771 sb->s_flags |= MS_I_VERSION;
1773 case Opt_nodelalloc:
1774 clear_opt(sb, DELALLOC);
1776 case Opt_mblk_io_submit:
1777 set_opt(sb, MBLK_IO_SUBMIT);
1779 case Opt_nomblk_io_submit:
1780 clear_opt(sb, MBLK_IO_SUBMIT);
1783 if (match_int(&args[0], &option))
1787 sbi->s_stripe = option;
1790 set_opt(sb, DELALLOC);
1792 case Opt_block_validity:
1793 set_opt(sb, BLOCK_VALIDITY);
1795 case Opt_noblock_validity:
1796 clear_opt(sb, BLOCK_VALIDITY);
1798 case Opt_inode_readahead_blks:
1799 if (match_int(&args[0], &option))
1801 if (option < 0 || option > (1 << 30))
1803 if (option && !is_power_of_2(option)) {
1804 ext4_msg(sb, KERN_ERR,
1805 "EXT4-fs: inode_readahead_blks"
1806 " must be a power of 2");
1809 sbi->s_inode_readahead_blks = option;
1811 case Opt_journal_ioprio:
1812 if (match_int(&args[0], &option))
1814 if (option < 0 || option > 7)
1816 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1819 case Opt_noauto_da_alloc:
1820 set_opt(sb, NO_AUTO_DA_ALLOC);
1822 case Opt_auto_da_alloc:
1824 if (match_int(&args[0], &option))
1827 option = 1; /* No argument, default to 1 */
1829 clear_opt(sb, NO_AUTO_DA_ALLOC);
1831 set_opt(sb,NO_AUTO_DA_ALLOC);
1834 set_opt(sb, DISCARD);
1837 clear_opt(sb, DISCARD);
1839 case Opt_dioread_nolock:
1840 set_opt(sb, DIOREAD_NOLOCK);
1842 case Opt_dioread_lock:
1843 clear_opt(sb, DIOREAD_NOLOCK);
1845 case Opt_init_inode_table:
1846 set_opt(sb, INIT_INODE_TABLE);
1848 if (match_int(&args[0], &option))
1851 option = EXT4_DEF_LI_WAIT_MULT;
1854 sbi->s_li_wait_mult = option;
1856 case Opt_noinit_inode_table:
1857 clear_opt(sb, INIT_INODE_TABLE);
1860 ext4_msg(sb, KERN_ERR,
1861 "Unrecognized mount option \"%s\" "
1862 "or missing value", p);
1867 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1868 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1869 clear_opt(sb, USRQUOTA);
1871 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1872 clear_opt(sb, GRPQUOTA);
1874 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1875 ext4_msg(sb, KERN_ERR, "old and new quota "
1880 if (!sbi->s_jquota_fmt) {
1881 ext4_msg(sb, KERN_ERR, "journaled quota format "
1886 if (sbi->s_jquota_fmt) {
1887 ext4_msg(sb, KERN_ERR, "journaled quota format "
1888 "specified with no journaling "
1897 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1900 struct ext4_sb_info *sbi = EXT4_SB(sb);
1903 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1904 ext4_msg(sb, KERN_ERR, "revision level too high, "
1905 "forcing read-only mode");
1910 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1911 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1912 "running e2fsck is recommended");
1913 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1914 ext4_msg(sb, KERN_WARNING,
1915 "warning: mounting fs with errors, "
1916 "running e2fsck is recommended");
1917 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1918 le16_to_cpu(es->s_mnt_count) >=
1919 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1920 ext4_msg(sb, KERN_WARNING,
1921 "warning: maximal mount count reached, "
1922 "running e2fsck is recommended");
1923 else if (le32_to_cpu(es->s_checkinterval) &&
1924 (le32_to_cpu(es->s_lastcheck) +
1925 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1926 ext4_msg(sb, KERN_WARNING,
1927 "warning: checktime reached, "
1928 "running e2fsck is recommended");
1929 if (!sbi->s_journal)
1930 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1931 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1932 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1933 le16_add_cpu(&es->s_mnt_count, 1);
1934 es->s_mtime = cpu_to_le32(get_seconds());
1935 ext4_update_dynamic_rev(sb);
1937 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1939 ext4_commit_super(sb, 1);
1940 if (test_opt(sb, DEBUG))
1941 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1942 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1944 sbi->s_groups_count,
1945 EXT4_BLOCKS_PER_GROUP(sb),
1946 EXT4_INODES_PER_GROUP(sb),
1947 sbi->s_mount_opt, sbi->s_mount_opt2);
1952 static int ext4_fill_flex_info(struct super_block *sb)
1954 struct ext4_sb_info *sbi = EXT4_SB(sb);
1955 struct ext4_group_desc *gdp = NULL;
1956 ext4_group_t flex_group_count;
1957 ext4_group_t flex_group;
1958 int groups_per_flex = 0;
1962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1963 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1965 if (groups_per_flex < 2) {
1966 sbi->s_log_groups_per_flex = 0;
1970 /* We allocate both existing and potentially added groups */
1971 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1972 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1973 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1974 size = flex_group_count * sizeof(struct flex_groups);
1975 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1976 if (sbi->s_flex_groups == NULL) {
1977 sbi->s_flex_groups = vzalloc(size);
1978 if (sbi->s_flex_groups == NULL) {
1979 ext4_msg(sb, KERN_ERR,
1980 "not enough memory for %u flex groups",
1986 for (i = 0; i < sbi->s_groups_count; i++) {
1987 gdp = ext4_get_group_desc(sb, i, NULL);
1989 flex_group = ext4_flex_group(sbi, i);
1990 atomic_add(ext4_free_inodes_count(sb, gdp),
1991 &sbi->s_flex_groups[flex_group].free_inodes);
1992 atomic_add(ext4_free_blks_count(sb, gdp),
1993 &sbi->s_flex_groups[flex_group].free_blocks);
1994 atomic_add(ext4_used_dirs_count(sb, gdp),
1995 &sbi->s_flex_groups[flex_group].used_dirs);
2003 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2004 struct ext4_group_desc *gdp)
2008 if (sbi->s_es->s_feature_ro_compat &
2009 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2010 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2011 __le32 le_group = cpu_to_le32(block_group);
2013 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2014 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2015 crc = crc16(crc, (__u8 *)gdp, offset);
2016 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2017 /* for checksum of struct ext4_group_desc do the rest...*/
2018 if ((sbi->s_es->s_feature_incompat &
2019 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2020 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2021 crc = crc16(crc, (__u8 *)gdp + offset,
2022 le16_to_cpu(sbi->s_es->s_desc_size) -
2026 return cpu_to_le16(crc);
2029 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2030 struct ext4_group_desc *gdp)
2032 if ((sbi->s_es->s_feature_ro_compat &
2033 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2034 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2040 /* Called at mount-time, super-block is locked */
2041 static int ext4_check_descriptors(struct super_block *sb,
2042 ext4_group_t *first_not_zeroed)
2044 struct ext4_sb_info *sbi = EXT4_SB(sb);
2045 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2046 ext4_fsblk_t last_block;
2047 ext4_fsblk_t block_bitmap;
2048 ext4_fsblk_t inode_bitmap;
2049 ext4_fsblk_t inode_table;
2050 int flexbg_flag = 0;
2051 ext4_group_t i, grp = sbi->s_groups_count;
2053 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2056 ext4_debug("Checking group descriptors");
2058 for (i = 0; i < sbi->s_groups_count; i++) {
2059 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2061 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2062 last_block = ext4_blocks_count(sbi->s_es) - 1;
2064 last_block = first_block +
2065 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2067 if ((grp == sbi->s_groups_count) &&
2068 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2071 block_bitmap = ext4_block_bitmap(sb, gdp);
2072 if (block_bitmap < first_block || block_bitmap > last_block) {
2073 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2074 "Block bitmap for group %u not in group "
2075 "(block %llu)!", i, block_bitmap);
2078 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2079 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2080 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2081 "Inode bitmap for group %u not in group "
2082 "(block %llu)!", i, inode_bitmap);
2085 inode_table = ext4_inode_table(sb, gdp);
2086 if (inode_table < first_block ||
2087 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2089 "Inode table for group %u not in group "
2090 "(block %llu)!", i, inode_table);
2093 ext4_lock_group(sb, i);
2094 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 "Checksum for group %u failed (%u!=%u)",
2097 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2098 gdp)), le16_to_cpu(gdp->bg_checksum));
2099 if (!(sb->s_flags & MS_RDONLY)) {
2100 ext4_unlock_group(sb, i);
2104 ext4_unlock_group(sb, i);
2106 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2108 if (NULL != first_not_zeroed)
2109 *first_not_zeroed = grp;
2111 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2112 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2116 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2117 * the superblock) which were deleted from all directories, but held open by
2118 * a process at the time of a crash. We walk the list and try to delete these
2119 * inodes at recovery time (only with a read-write filesystem).
2121 * In order to keep the orphan inode chain consistent during traversal (in
2122 * case of crash during recovery), we link each inode into the superblock
2123 * orphan list_head and handle it the same way as an inode deletion during
2124 * normal operation (which journals the operations for us).
2126 * We only do an iget() and an iput() on each inode, which is very safe if we
2127 * accidentally point at an in-use or already deleted inode. The worst that
2128 * can happen in this case is that we get a "bit already cleared" message from
2129 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2130 * e2fsck was run on this filesystem, and it must have already done the orphan
2131 * inode cleanup for us, so we can safely abort without any further action.
2133 static void ext4_orphan_cleanup(struct super_block *sb,
2134 struct ext4_super_block *es)
2136 unsigned int s_flags = sb->s_flags;
2137 int nr_orphans = 0, nr_truncates = 0;
2141 if (!es->s_last_orphan) {
2142 jbd_debug(4, "no orphan inodes to clean up\n");
2146 if (bdev_read_only(sb->s_bdev)) {
2147 ext4_msg(sb, KERN_ERR, "write access "
2148 "unavailable, skipping orphan cleanup");
2152 /* Check if feature set would not allow a r/w mount */
2153 if (!ext4_feature_set_ok(sb, 0)) {
2154 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2155 "unknown ROCOMPAT features");
2159 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2160 if (es->s_last_orphan)
2161 jbd_debug(1, "Errors on filesystem, "
2162 "clearing orphan list.\n");
2163 es->s_last_orphan = 0;
2164 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2168 if (s_flags & MS_RDONLY) {
2169 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2170 sb->s_flags &= ~MS_RDONLY;
2173 /* Needed for iput() to work correctly and not trash data */
2174 sb->s_flags |= MS_ACTIVE;
2175 /* Turn on quotas so that they are updated correctly */
2176 for (i = 0; i < MAXQUOTAS; i++) {
2177 if (EXT4_SB(sb)->s_qf_names[i]) {
2178 int ret = ext4_quota_on_mount(sb, i);
2180 ext4_msg(sb, KERN_ERR,
2181 "Cannot turn on journaled "
2182 "quota: error %d", ret);
2187 while (es->s_last_orphan) {
2188 struct inode *inode;
2190 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2191 if (IS_ERR(inode)) {
2192 es->s_last_orphan = 0;
2196 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2197 dquot_initialize(inode);
2198 if (inode->i_nlink) {
2199 ext4_msg(sb, KERN_DEBUG,
2200 "%s: truncating inode %lu to %lld bytes",
2201 __func__, inode->i_ino, inode->i_size);
2202 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2203 inode->i_ino, inode->i_size);
2204 ext4_truncate(inode);
2207 ext4_msg(sb, KERN_DEBUG,
2208 "%s: deleting unreferenced inode %lu",
2209 __func__, inode->i_ino);
2210 jbd_debug(2, "deleting unreferenced inode %lu\n",
2214 iput(inode); /* The delete magic happens here! */
2217 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2220 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2221 PLURAL(nr_orphans));
2223 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2224 PLURAL(nr_truncates));
2226 /* Turn quotas off */
2227 for (i = 0; i < MAXQUOTAS; i++) {
2228 if (sb_dqopt(sb)->files[i])
2229 dquot_quota_off(sb, i);
2232 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2236 * Maximal extent format file size.
2237 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2238 * extent format containers, within a sector_t, and within i_blocks
2239 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2240 * so that won't be a limiting factor.
2242 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2244 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2247 loff_t upper_limit = MAX_LFS_FILESIZE;
2249 /* small i_blocks in vfs inode? */
2250 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2252 * CONFIG_LBDAF is not enabled implies the inode
2253 * i_block represent total blocks in 512 bytes
2254 * 32 == size of vfs inode i_blocks * 8
2256 upper_limit = (1LL << 32) - 1;
2258 /* total blocks in file system block size */
2259 upper_limit >>= (blkbits - 9);
2260 upper_limit <<= blkbits;
2263 /* 32-bit extent-start container, ee_block */
2268 /* Sanity check against vm- & vfs- imposed limits */
2269 if (res > upper_limit)
2276 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2277 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2278 * We need to be 1 filesystem block less than the 2^48 sector limit.
2280 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2282 loff_t res = EXT4_NDIR_BLOCKS;
2285 /* This is calculated to be the largest file size for a dense, block
2286 * mapped file such that the file's total number of 512-byte sectors,
2287 * including data and all indirect blocks, does not exceed (2^48 - 1).
2289 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2290 * number of 512-byte sectors of the file.
2293 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2295 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2296 * the inode i_block field represents total file blocks in
2297 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2299 upper_limit = (1LL << 32) - 1;
2301 /* total blocks in file system block size */
2302 upper_limit >>= (bits - 9);
2306 * We use 48 bit ext4_inode i_blocks
2307 * With EXT4_HUGE_FILE_FL set the i_blocks
2308 * represent total number of blocks in
2309 * file system block size
2311 upper_limit = (1LL << 48) - 1;
2315 /* indirect blocks */
2317 /* double indirect blocks */
2318 meta_blocks += 1 + (1LL << (bits-2));
2319 /* tripple indirect blocks */
2320 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2322 upper_limit -= meta_blocks;
2323 upper_limit <<= bits;
2325 res += 1LL << (bits-2);
2326 res += 1LL << (2*(bits-2));
2327 res += 1LL << (3*(bits-2));
2329 if (res > upper_limit)
2332 if (res > MAX_LFS_FILESIZE)
2333 res = MAX_LFS_FILESIZE;
2338 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2339 ext4_fsblk_t logical_sb_block, int nr)
2341 struct ext4_sb_info *sbi = EXT4_SB(sb);
2342 ext4_group_t bg, first_meta_bg;
2345 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2347 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2349 return logical_sb_block + nr + 1;
2350 bg = sbi->s_desc_per_block * nr;
2351 if (ext4_bg_has_super(sb, bg))
2354 return (has_super + ext4_group_first_block_no(sb, bg));
2358 * ext4_get_stripe_size: Get the stripe size.
2359 * @sbi: In memory super block info
2361 * If we have specified it via mount option, then
2362 * use the mount option value. If the value specified at mount time is
2363 * greater than the blocks per group use the super block value.
2364 * If the super block value is greater than blocks per group return 0.
2365 * Allocator needs it be less than blocks per group.
2368 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2370 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2371 unsigned long stripe_width =
2372 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2374 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2375 return sbi->s_stripe;
2377 if (stripe_width <= sbi->s_blocks_per_group)
2378 return stripe_width;
2380 if (stride <= sbi->s_blocks_per_group)
2389 struct attribute attr;
2390 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2391 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2392 const char *, size_t);
2396 static int parse_strtoul(const char *buf,
2397 unsigned long max, unsigned long *value)
2401 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2402 endp = skip_spaces(endp);
2403 if (*endp || *value > max)
2409 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2410 struct ext4_sb_info *sbi,
2413 return snprintf(buf, PAGE_SIZE, "%llu\n",
2414 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2417 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2418 struct ext4_sb_info *sbi, char *buf)
2420 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2422 if (!sb->s_bdev->bd_part)
2423 return snprintf(buf, PAGE_SIZE, "0\n");
2424 return snprintf(buf, PAGE_SIZE, "%lu\n",
2425 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2426 sbi->s_sectors_written_start) >> 1);
2429 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi, char *buf)
2432 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2434 if (!sb->s_bdev->bd_part)
2435 return snprintf(buf, PAGE_SIZE, "0\n");
2436 return snprintf(buf, PAGE_SIZE, "%llu\n",
2437 (unsigned long long)(sbi->s_kbytes_written +
2438 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2439 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2442 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2443 struct ext4_sb_info *sbi,
2444 const char *buf, size_t count)
2448 if (parse_strtoul(buf, 0x40000000, &t))
2451 if (t && !is_power_of_2(t))
2454 sbi->s_inode_readahead_blks = t;
2458 static ssize_t sbi_ui_show(struct ext4_attr *a,
2459 struct ext4_sb_info *sbi, char *buf)
2461 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2463 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2466 static ssize_t sbi_ui_store(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi,
2468 const char *buf, size_t count)
2470 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2473 if (parse_strtoul(buf, 0xffffffff, &t))
2479 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2480 static struct ext4_attr ext4_attr_##_name = { \
2481 .attr = {.name = __stringify(_name), .mode = _mode }, \
2484 .offset = offsetof(struct ext4_sb_info, _elname), \
2486 #define EXT4_ATTR(name, mode, show, store) \
2487 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2489 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2490 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2491 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2492 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2493 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2494 #define ATTR_LIST(name) &ext4_attr_##name.attr
2496 EXT4_RO_ATTR(delayed_allocation_blocks);
2497 EXT4_RO_ATTR(session_write_kbytes);
2498 EXT4_RO_ATTR(lifetime_write_kbytes);
2499 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2500 inode_readahead_blks_store, s_inode_readahead_blks);
2501 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2502 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2503 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2504 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2505 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2506 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2507 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2508 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2510 static struct attribute *ext4_attrs[] = {
2511 ATTR_LIST(delayed_allocation_blocks),
2512 ATTR_LIST(session_write_kbytes),
2513 ATTR_LIST(lifetime_write_kbytes),
2514 ATTR_LIST(inode_readahead_blks),
2515 ATTR_LIST(inode_goal),
2516 ATTR_LIST(mb_stats),
2517 ATTR_LIST(mb_max_to_scan),
2518 ATTR_LIST(mb_min_to_scan),
2519 ATTR_LIST(mb_order2_req),
2520 ATTR_LIST(mb_stream_req),
2521 ATTR_LIST(mb_group_prealloc),
2522 ATTR_LIST(max_writeback_mb_bump),
2526 /* Features this copy of ext4 supports */
2527 EXT4_INFO_ATTR(lazy_itable_init);
2528 EXT4_INFO_ATTR(batched_discard);
2530 static struct attribute *ext4_feat_attrs[] = {
2531 ATTR_LIST(lazy_itable_init),
2532 ATTR_LIST(batched_discard),
2536 static ssize_t ext4_attr_show(struct kobject *kobj,
2537 struct attribute *attr, char *buf)
2539 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2541 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2543 return a->show ? a->show(a, sbi, buf) : 0;
2546 static ssize_t ext4_attr_store(struct kobject *kobj,
2547 struct attribute *attr,
2548 const char *buf, size_t len)
2550 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2552 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2554 return a->store ? a->store(a, sbi, buf, len) : 0;
2557 static void ext4_sb_release(struct kobject *kobj)
2559 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2561 complete(&sbi->s_kobj_unregister);
2564 static const struct sysfs_ops ext4_attr_ops = {
2565 .show = ext4_attr_show,
2566 .store = ext4_attr_store,
2569 static struct kobj_type ext4_ktype = {
2570 .default_attrs = ext4_attrs,
2571 .sysfs_ops = &ext4_attr_ops,
2572 .release = ext4_sb_release,
2575 static void ext4_feat_release(struct kobject *kobj)
2577 complete(&ext4_feat->f_kobj_unregister);
2580 static struct kobj_type ext4_feat_ktype = {
2581 .default_attrs = ext4_feat_attrs,
2582 .sysfs_ops = &ext4_attr_ops,
2583 .release = ext4_feat_release,
2587 * Check whether this filesystem can be mounted based on
2588 * the features present and the RDONLY/RDWR mount requested.
2589 * Returns 1 if this filesystem can be mounted as requested,
2590 * 0 if it cannot be.
2592 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2594 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2595 ext4_msg(sb, KERN_ERR,
2596 "Couldn't mount because of "
2597 "unsupported optional features (%x)",
2598 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2599 ~EXT4_FEATURE_INCOMPAT_SUPP));
2606 /* Check that feature set is OK for a read-write mount */
2607 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2608 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2609 "unsupported optional features (%x)",
2610 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2611 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2615 * Large file size enabled file system can only be mounted
2616 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2618 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2619 if (sizeof(blkcnt_t) < sizeof(u64)) {
2620 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2621 "cannot be mounted RDWR without "
2630 * This function is called once a day if we have errors logged
2631 * on the file system
2633 static void print_daily_error_info(unsigned long arg)
2635 struct super_block *sb = (struct super_block *) arg;
2636 struct ext4_sb_info *sbi;
2637 struct ext4_super_block *es;
2642 if (es->s_error_count)
2643 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2644 le32_to_cpu(es->s_error_count));
2645 if (es->s_first_error_time) {
2646 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2647 sb->s_id, le32_to_cpu(es->s_first_error_time),
2648 (int) sizeof(es->s_first_error_func),
2649 es->s_first_error_func,
2650 le32_to_cpu(es->s_first_error_line));
2651 if (es->s_first_error_ino)
2652 printk(": inode %u",
2653 le32_to_cpu(es->s_first_error_ino));
2654 if (es->s_first_error_block)
2655 printk(": block %llu", (unsigned long long)
2656 le64_to_cpu(es->s_first_error_block));
2659 if (es->s_last_error_time) {
2660 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2661 sb->s_id, le32_to_cpu(es->s_last_error_time),
2662 (int) sizeof(es->s_last_error_func),
2663 es->s_last_error_func,
2664 le32_to_cpu(es->s_last_error_line));
2665 if (es->s_last_error_ino)
2666 printk(": inode %u",
2667 le32_to_cpu(es->s_last_error_ino));
2668 if (es->s_last_error_block)
2669 printk(": block %llu", (unsigned long long)
2670 le64_to_cpu(es->s_last_error_block));
2673 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2676 /* Find next suitable group and run ext4_init_inode_table */
2677 static int ext4_run_li_request(struct ext4_li_request *elr)
2679 struct ext4_group_desc *gdp = NULL;
2680 ext4_group_t group, ngroups;
2681 struct super_block *sb;
2682 unsigned long timeout = 0;
2686 ngroups = EXT4_SB(sb)->s_groups_count;
2688 for (group = elr->lr_next_group; group < ngroups; group++) {
2689 gdp = ext4_get_group_desc(sb, group, NULL);
2695 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2699 if (group == ngroups)
2704 ret = ext4_init_inode_table(sb, group,
2705 elr->lr_timeout ? 0 : 1);
2706 if (elr->lr_timeout == 0) {
2707 timeout = (jiffies - timeout) *
2708 elr->lr_sbi->s_li_wait_mult;
2709 elr->lr_timeout = timeout;
2711 elr->lr_next_sched = jiffies + elr->lr_timeout;
2712 elr->lr_next_group = group + 1;
2719 * Remove lr_request from the list_request and free the
2720 * request structure. Should be called with li_list_mtx held
2722 static void ext4_remove_li_request(struct ext4_li_request *elr)
2724 struct ext4_sb_info *sbi;
2731 list_del(&elr->lr_request);
2732 sbi->s_li_request = NULL;
2736 static void ext4_unregister_li_request(struct super_block *sb)
2738 mutex_lock(&ext4_li_mtx);
2739 if (!ext4_li_info) {
2740 mutex_unlock(&ext4_li_mtx);
2744 mutex_lock(&ext4_li_info->li_list_mtx);
2745 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2746 mutex_unlock(&ext4_li_info->li_list_mtx);
2747 mutex_unlock(&ext4_li_mtx);
2750 static struct task_struct *ext4_lazyinit_task;
2753 * This is the function where ext4lazyinit thread lives. It walks
2754 * through the request list searching for next scheduled filesystem.
2755 * When such a fs is found, run the lazy initialization request
2756 * (ext4_rn_li_request) and keep track of the time spend in this
2757 * function. Based on that time we compute next schedule time of
2758 * the request. When walking through the list is complete, compute
2759 * next waking time and put itself into sleep.
2761 static int ext4_lazyinit_thread(void *arg)
2763 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2764 struct list_head *pos, *n;
2765 struct ext4_li_request *elr;
2766 unsigned long next_wakeup, cur;
2768 BUG_ON(NULL == eli);
2772 next_wakeup = MAX_JIFFY_OFFSET;
2774 mutex_lock(&eli->li_list_mtx);
2775 if (list_empty(&eli->li_request_list)) {
2776 mutex_unlock(&eli->li_list_mtx);
2780 list_for_each_safe(pos, n, &eli->li_request_list) {
2781 elr = list_entry(pos, struct ext4_li_request,
2784 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2785 if (ext4_run_li_request(elr) != 0) {
2786 /* error, remove the lazy_init job */
2787 ext4_remove_li_request(elr);
2792 if (time_before(elr->lr_next_sched, next_wakeup))
2793 next_wakeup = elr->lr_next_sched;
2795 mutex_unlock(&eli->li_list_mtx);
2797 if (freezing(current))
2801 if ((time_after_eq(cur, next_wakeup)) ||
2802 (MAX_JIFFY_OFFSET == next_wakeup)) {
2807 schedule_timeout_interruptible(next_wakeup - cur);
2809 if (kthread_should_stop()) {
2810 ext4_clear_request_list();
2817 * It looks like the request list is empty, but we need
2818 * to check it under the li_list_mtx lock, to prevent any
2819 * additions into it, and of course we should lock ext4_li_mtx
2820 * to atomically free the list and ext4_li_info, because at
2821 * this point another ext4 filesystem could be registering
2824 mutex_lock(&ext4_li_mtx);
2825 mutex_lock(&eli->li_list_mtx);
2826 if (!list_empty(&eli->li_request_list)) {
2827 mutex_unlock(&eli->li_list_mtx);
2828 mutex_unlock(&ext4_li_mtx);
2831 mutex_unlock(&eli->li_list_mtx);
2832 kfree(ext4_li_info);
2833 ext4_li_info = NULL;
2834 mutex_unlock(&ext4_li_mtx);
2839 static void ext4_clear_request_list(void)
2841 struct list_head *pos, *n;
2842 struct ext4_li_request *elr;
2844 mutex_lock(&ext4_li_info->li_list_mtx);
2845 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2846 elr = list_entry(pos, struct ext4_li_request,
2848 ext4_remove_li_request(elr);
2850 mutex_unlock(&ext4_li_info->li_list_mtx);
2853 static int ext4_run_lazyinit_thread(void)
2855 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2856 ext4_li_info, "ext4lazyinit");
2857 if (IS_ERR(ext4_lazyinit_task)) {
2858 int err = PTR_ERR(ext4_lazyinit_task);
2859 ext4_clear_request_list();
2860 kfree(ext4_li_info);
2861 ext4_li_info = NULL;
2862 printk(KERN_CRIT "EXT4: error %d creating inode table "
2863 "initialization thread\n",
2867 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2872 * Check whether it make sense to run itable init. thread or not.
2873 * If there is at least one uninitialized inode table, return
2874 * corresponding group number, else the loop goes through all
2875 * groups and return total number of groups.
2877 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2879 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2880 struct ext4_group_desc *gdp = NULL;
2882 for (group = 0; group < ngroups; group++) {
2883 gdp = ext4_get_group_desc(sb, group, NULL);
2887 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2894 static int ext4_li_info_new(void)
2896 struct ext4_lazy_init *eli = NULL;
2898 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2902 INIT_LIST_HEAD(&eli->li_request_list);
2903 mutex_init(&eli->li_list_mtx);
2905 eli->li_state |= EXT4_LAZYINIT_QUIT;
2912 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2915 struct ext4_sb_info *sbi = EXT4_SB(sb);
2916 struct ext4_li_request *elr;
2919 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2925 elr->lr_next_group = start;
2928 * Randomize first schedule time of the request to
2929 * spread the inode table initialization requests
2932 get_random_bytes(&rnd, sizeof(rnd));
2933 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2934 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2939 static int ext4_register_li_request(struct super_block *sb,
2940 ext4_group_t first_not_zeroed)
2942 struct ext4_sb_info *sbi = EXT4_SB(sb);
2943 struct ext4_li_request *elr;
2944 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2947 if (sbi->s_li_request != NULL) {
2949 * Reset timeout so it can be computed again, because
2950 * s_li_wait_mult might have changed.
2952 sbi->s_li_request->lr_timeout = 0;
2956 if (first_not_zeroed == ngroups ||
2957 (sb->s_flags & MS_RDONLY) ||
2958 !test_opt(sb, INIT_INODE_TABLE))
2961 elr = ext4_li_request_new(sb, first_not_zeroed);
2965 mutex_lock(&ext4_li_mtx);
2967 if (NULL == ext4_li_info) {
2968 ret = ext4_li_info_new();
2973 mutex_lock(&ext4_li_info->li_list_mtx);
2974 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2975 mutex_unlock(&ext4_li_info->li_list_mtx);
2977 sbi->s_li_request = elr;
2979 * set elr to NULL here since it has been inserted to
2980 * the request_list and the removal and free of it is
2981 * handled by ext4_clear_request_list from now on.
2985 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2986 ret = ext4_run_lazyinit_thread();
2991 mutex_unlock(&ext4_li_mtx);
2998 * We do not need to lock anything since this is called on
3001 static void ext4_destroy_lazyinit_thread(void)
3004 * If thread exited earlier
3005 * there's nothing to be done.
3007 if (!ext4_li_info || !ext4_lazyinit_task)
3010 kthread_stop(ext4_lazyinit_task);
3013 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3014 __releases(kernel_lock)
3015 __acquires(kernel_lock)
3017 char *orig_data = kstrdup(data, GFP_KERNEL);
3018 struct buffer_head *bh;
3019 struct ext4_super_block *es = NULL;
3020 struct ext4_sb_info *sbi;
3022 ext4_fsblk_t sb_block = get_sb_block(&data);
3023 ext4_fsblk_t logical_sb_block;
3024 unsigned long offset = 0;
3025 unsigned long journal_devnum = 0;
3026 unsigned long def_mount_opts;
3032 unsigned int db_count;
3034 int needs_recovery, has_huge_files;
3037 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3038 ext4_group_t first_not_zeroed;
3040 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3044 sbi->s_blockgroup_lock =
3045 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3046 if (!sbi->s_blockgroup_lock) {
3050 sb->s_fs_info = sbi;
3051 sbi->s_mount_opt = 0;
3052 sbi->s_resuid = EXT4_DEF_RESUID;
3053 sbi->s_resgid = EXT4_DEF_RESGID;
3054 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3055 sbi->s_sb_block = sb_block;
3056 if (sb->s_bdev->bd_part)
3057 sbi->s_sectors_written_start =
3058 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3060 /* Cleanup superblock name */
3061 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3065 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3067 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3072 * The ext4 superblock will not be buffer aligned for other than 1kB
3073 * block sizes. We need to calculate the offset from buffer start.
3075 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3076 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3077 offset = do_div(logical_sb_block, blocksize);
3079 logical_sb_block = sb_block;
3082 if (!(bh = sb_bread(sb, logical_sb_block))) {
3083 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3087 * Note: s_es must be initialized as soon as possible because
3088 * some ext4 macro-instructions depend on its value
3090 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3092 sb->s_magic = le16_to_cpu(es->s_magic);
3093 if (sb->s_magic != EXT4_SUPER_MAGIC)
3095 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3097 /* Set defaults before we parse the mount options */
3098 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3099 set_opt(sb, INIT_INODE_TABLE);
3100 if (def_mount_opts & EXT4_DEFM_DEBUG)
3102 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3103 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3107 if (def_mount_opts & EXT4_DEFM_UID16)
3108 set_opt(sb, NO_UID32);
3109 /* xattr user namespace & acls are now defaulted on */
3110 #ifdef CONFIG_EXT4_FS_XATTR
3111 set_opt(sb, XATTR_USER);
3113 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3114 set_opt(sb, POSIX_ACL);
3116 set_opt(sb, MBLK_IO_SUBMIT);
3117 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3118 set_opt(sb, JOURNAL_DATA);
3119 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3120 set_opt(sb, ORDERED_DATA);
3121 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3122 set_opt(sb, WRITEBACK_DATA);
3124 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3125 set_opt(sb, ERRORS_PANIC);
3126 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3127 set_opt(sb, ERRORS_CONT);
3129 set_opt(sb, ERRORS_RO);
3130 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3131 set_opt(sb, BLOCK_VALIDITY);
3132 if (def_mount_opts & EXT4_DEFM_DISCARD)
3133 set_opt(sb, DISCARD);
3135 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3136 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3137 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3138 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3139 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3141 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3142 set_opt(sb, BARRIER);
3145 * enable delayed allocation by default
3146 * Use -o nodelalloc to turn it off
3148 if (!IS_EXT3_SB(sb) &&
3149 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3150 set_opt(sb, DELALLOC);
3153 * set default s_li_wait_mult for lazyinit, for the case there is
3154 * no mount option specified.
3156 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3158 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3159 &journal_devnum, &journal_ioprio, NULL, 0)) {
3160 ext4_msg(sb, KERN_WARNING,
3161 "failed to parse options in superblock: %s",
3162 sbi->s_es->s_mount_opts);
3164 if (!parse_options((char *) data, sb, &journal_devnum,
3165 &journal_ioprio, NULL, 0))
3168 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3169 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3171 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3172 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3173 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3174 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3175 ext4_msg(sb, KERN_WARNING,
3176 "feature flags set on rev 0 fs, "
3177 "running e2fsck is recommended");
3179 if (IS_EXT2_SB(sb)) {
3180 if (ext2_feature_set_ok(sb))
3181 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3182 "using the ext4 subsystem");
3184 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3185 "to feature incompatibilities");
3190 if (IS_EXT3_SB(sb)) {
3191 if (ext3_feature_set_ok(sb))
3192 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3193 "using the ext4 subsystem");
3195 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3196 "to feature incompatibilities");
3202 * Check feature flags regardless of the revision level, since we
3203 * previously didn't change the revision level when setting the flags,
3204 * so there is a chance incompat flags are set on a rev 0 filesystem.
3206 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3209 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3211 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3212 blocksize > EXT4_MAX_BLOCK_SIZE) {
3213 ext4_msg(sb, KERN_ERR,
3214 "Unsupported filesystem blocksize %d", blocksize);
3218 if (sb->s_blocksize != blocksize) {
3219 /* Validate the filesystem blocksize */
3220 if (!sb_set_blocksize(sb, blocksize)) {
3221 ext4_msg(sb, KERN_ERR, "bad block size %d",
3227 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3228 offset = do_div(logical_sb_block, blocksize);
3229 bh = sb_bread(sb, logical_sb_block);
3231 ext4_msg(sb, KERN_ERR,
3232 "Can't read superblock on 2nd try");
3235 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3237 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3238 ext4_msg(sb, KERN_ERR,
3239 "Magic mismatch, very weird!");
3244 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3245 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3246 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3248 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3250 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3251 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3252 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3254 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3255 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3256 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3257 (!is_power_of_2(sbi->s_inode_size)) ||
3258 (sbi->s_inode_size > blocksize)) {
3259 ext4_msg(sb, KERN_ERR,
3260 "unsupported inode size: %d",
3264 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3265 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3268 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3269 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3270 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3271 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3272 !is_power_of_2(sbi->s_desc_size)) {
3273 ext4_msg(sb, KERN_ERR,
3274 "unsupported descriptor size %lu",
3279 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3281 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3282 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3283 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3286 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3287 if (sbi->s_inodes_per_block == 0)
3289 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3290 sbi->s_inodes_per_block;
3291 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3293 sbi->s_mount_state = le16_to_cpu(es->s_state);
3294 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3295 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3297 for (i = 0; i < 4; i++)
3298 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3299 sbi->s_def_hash_version = es->s_def_hash_version;
3300 i = le32_to_cpu(es->s_flags);
3301 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3302 sbi->s_hash_unsigned = 3;
3303 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3304 #ifdef __CHAR_UNSIGNED__
3305 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3306 sbi->s_hash_unsigned = 3;
3308 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3313 if (sbi->s_blocks_per_group > blocksize * 8) {
3314 ext4_msg(sb, KERN_ERR,
3315 "#blocks per group too big: %lu",
3316 sbi->s_blocks_per_group);
3319 if (sbi->s_inodes_per_group > blocksize * 8) {
3320 ext4_msg(sb, KERN_ERR,
3321 "#inodes per group too big: %lu",
3322 sbi->s_inodes_per_group);
3327 * Test whether we have more sectors than will fit in sector_t,
3328 * and whether the max offset is addressable by the page cache.
3330 err = generic_check_addressable(sb->s_blocksize_bits,
3331 ext4_blocks_count(es));
3333 ext4_msg(sb, KERN_ERR, "filesystem"
3334 " too large to mount safely on this system");
3335 if (sizeof(sector_t) < 8)
3336 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3341 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3344 /* check blocks count against device size */
3345 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3346 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3347 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3348 "exceeds size of device (%llu blocks)",
3349 ext4_blocks_count(es), blocks_count);
3354 * It makes no sense for the first data block to be beyond the end
3355 * of the filesystem.
3357 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3358 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3359 "block %u is beyond end of filesystem (%llu)",
3360 le32_to_cpu(es->s_first_data_block),
3361 ext4_blocks_count(es));
3364 blocks_count = (ext4_blocks_count(es) -
3365 le32_to_cpu(es->s_first_data_block) +
3366 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3367 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3368 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3369 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3370 "(block count %llu, first data block %u, "
3371 "blocks per group %lu)", sbi->s_groups_count,
3372 ext4_blocks_count(es),
3373 le32_to_cpu(es->s_first_data_block),
3374 EXT4_BLOCKS_PER_GROUP(sb));
3377 sbi->s_groups_count = blocks_count;
3378 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3379 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3380 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3381 EXT4_DESC_PER_BLOCK(sb);
3382 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3384 if (sbi->s_group_desc == NULL) {
3385 ext4_msg(sb, KERN_ERR, "not enough memory");
3389 #ifdef CONFIG_PROC_FS
3391 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3394 bgl_lock_init(sbi->s_blockgroup_lock);
3396 for (i = 0; i < db_count; i++) {
3397 block = descriptor_loc(sb, logical_sb_block, i);
3398 sbi->s_group_desc[i] = sb_bread(sb, block);
3399 if (!sbi->s_group_desc[i]) {
3400 ext4_msg(sb, KERN_ERR,
3401 "can't read group descriptor %d", i);
3406 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3407 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3410 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3411 if (!ext4_fill_flex_info(sb)) {
3412 ext4_msg(sb, KERN_ERR,
3413 "unable to initialize "
3414 "flex_bg meta info!");
3418 sbi->s_gdb_count = db_count;
3419 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3420 spin_lock_init(&sbi->s_next_gen_lock);
3422 init_timer(&sbi->s_err_report);
3423 sbi->s_err_report.function = print_daily_error_info;
3424 sbi->s_err_report.data = (unsigned long) sb;
3426 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3427 ext4_count_free_blocks(sb));
3429 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3430 ext4_count_free_inodes(sb));
3433 err = percpu_counter_init(&sbi->s_dirs_counter,
3434 ext4_count_dirs(sb));
3437 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3440 ext4_msg(sb, KERN_ERR, "insufficient memory");
3444 sbi->s_stripe = ext4_get_stripe_size(sbi);
3445 sbi->s_max_writeback_mb_bump = 128;
3448 * set up enough so that it can read an inode
3450 if (!test_opt(sb, NOLOAD) &&
3451 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3452 sb->s_op = &ext4_sops;
3454 sb->s_op = &ext4_nojournal_sops;
3455 sb->s_export_op = &ext4_export_ops;
3456 sb->s_xattr = ext4_xattr_handlers;
3458 sb->s_qcop = &ext4_qctl_operations;
3459 sb->dq_op = &ext4_quota_operations;
3461 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3463 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3464 mutex_init(&sbi->s_orphan_lock);
3465 mutex_init(&sbi->s_resize_lock);
3469 needs_recovery = (es->s_last_orphan != 0 ||
3470 EXT4_HAS_INCOMPAT_FEATURE(sb,
3471 EXT4_FEATURE_INCOMPAT_RECOVER));
3474 * The first inode we look at is the journal inode. Don't try
3475 * root first: it may be modified in the journal!
3477 if (!test_opt(sb, NOLOAD) &&
3478 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3479 if (ext4_load_journal(sb, es, journal_devnum))
3481 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3482 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3483 ext4_msg(sb, KERN_ERR, "required journal recovery "
3484 "suppressed and not mounted read-only");
3485 goto failed_mount_wq;
3487 clear_opt(sb, DATA_FLAGS);
3488 sbi->s_journal = NULL;
3493 if (ext4_blocks_count(es) > 0xffffffffULL &&
3494 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3495 JBD2_FEATURE_INCOMPAT_64BIT)) {
3496 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3497 goto failed_mount_wq;
3500 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3501 jbd2_journal_set_features(sbi->s_journal,
3502 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3503 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3504 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3505 jbd2_journal_set_features(sbi->s_journal,
3506 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3507 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3508 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3510 jbd2_journal_clear_features(sbi->s_journal,
3511 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3512 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3515 /* We have now updated the journal if required, so we can
3516 * validate the data journaling mode. */
3517 switch (test_opt(sb, DATA_FLAGS)) {
3519 /* No mode set, assume a default based on the journal
3520 * capabilities: ORDERED_DATA if the journal can
3521 * cope, else JOURNAL_DATA
3523 if (jbd2_journal_check_available_features
3524 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3525 set_opt(sb, ORDERED_DATA);
3527 set_opt(sb, JOURNAL_DATA);
3530 case EXT4_MOUNT_ORDERED_DATA:
3531 case EXT4_MOUNT_WRITEBACK_DATA:
3532 if (!jbd2_journal_check_available_features
3533 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3534 ext4_msg(sb, KERN_ERR, "Journal does not support "
3535 "requested data journaling mode");
3536 goto failed_mount_wq;
3541 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3544 * The journal may have updated the bg summary counts, so we
3545 * need to update the global counters.
3547 percpu_counter_set(&sbi->s_freeblocks_counter,
3548 ext4_count_free_blocks(sb));
3549 percpu_counter_set(&sbi->s_freeinodes_counter,
3550 ext4_count_free_inodes(sb));
3551 percpu_counter_set(&sbi->s_dirs_counter,
3552 ext4_count_dirs(sb));
3553 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3557 * The maximum number of concurrent works can be high and
3558 * concurrency isn't really necessary. Limit it to 1.
3560 EXT4_SB(sb)->dio_unwritten_wq =
3561 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3562 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3563 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3564 goto failed_mount_wq;
3568 * The jbd2_journal_load will have done any necessary log recovery,
3569 * so we can safely mount the rest of the filesystem now.
3572 root = ext4_iget(sb, EXT4_ROOT_INO);
3574 ext4_msg(sb, KERN_ERR, "get root inode failed");
3575 ret = PTR_ERR(root);
3579 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3580 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3583 sb->s_root = d_alloc_root(root);
3585 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3590 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3592 /* determine the minimum size of new large inodes, if present */
3593 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3594 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3595 EXT4_GOOD_OLD_INODE_SIZE;
3596 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3597 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3598 if (sbi->s_want_extra_isize <
3599 le16_to_cpu(es->s_want_extra_isize))
3600 sbi->s_want_extra_isize =
3601 le16_to_cpu(es->s_want_extra_isize);
3602 if (sbi->s_want_extra_isize <
3603 le16_to_cpu(es->s_min_extra_isize))
3604 sbi->s_want_extra_isize =
3605 le16_to_cpu(es->s_min_extra_isize);
3608 /* Check if enough inode space is available */
3609 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3610 sbi->s_inode_size) {
3611 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3612 EXT4_GOOD_OLD_INODE_SIZE;
3613 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3617 if (test_opt(sb, DELALLOC) &&
3618 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3619 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3620 "requested data journaling mode");
3621 clear_opt(sb, DELALLOC);
3623 if (test_opt(sb, DIOREAD_NOLOCK)) {
3624 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3625 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3626 "option - requested data journaling mode");
3627 clear_opt(sb, DIOREAD_NOLOCK);
3629 if (sb->s_blocksize < PAGE_SIZE) {
3630 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3631 "option - block size is too small");
3632 clear_opt(sb, DIOREAD_NOLOCK);
3636 err = ext4_setup_system_zone(sb);
3638 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3644 err = ext4_mb_init(sb, needs_recovery);
3646 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3651 err = ext4_register_li_request(sb, first_not_zeroed);
3655 sbi->s_kobj.kset = ext4_kset;
3656 init_completion(&sbi->s_kobj_unregister);
3657 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3660 ext4_mb_release(sb);
3661 ext4_ext_release(sb);
3665 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3666 ext4_orphan_cleanup(sb, es);
3667 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3668 if (needs_recovery) {
3669 ext4_msg(sb, KERN_INFO, "recovery complete");
3670 ext4_mark_recovery_complete(sb, es);
3672 if (EXT4_SB(sb)->s_journal) {
3673 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3674 descr = " journalled data mode";
3675 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3676 descr = " ordered data mode";
3678 descr = " writeback data mode";
3680 descr = "out journal";
3682 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3683 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3684 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3686 if (es->s_error_count)
3687 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3694 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3700 ext4_msg(sb, KERN_ERR, "mount failed");
3701 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3703 ext4_release_system_zone(sb);
3704 if (sbi->s_journal) {
3705 jbd2_journal_destroy(sbi->s_journal);
3706 sbi->s_journal = NULL;
3709 del_timer(&sbi->s_err_report);
3710 if (sbi->s_flex_groups) {
3711 if (is_vmalloc_addr(sbi->s_flex_groups))
3712 vfree(sbi->s_flex_groups);
3714 kfree(sbi->s_flex_groups);
3716 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3717 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3718 percpu_counter_destroy(&sbi->s_dirs_counter);
3719 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3721 for (i = 0; i < db_count; i++)
3722 brelse(sbi->s_group_desc[i]);
3723 kfree(sbi->s_group_desc);
3726 remove_proc_entry(sb->s_id, ext4_proc_root);
3729 for (i = 0; i < MAXQUOTAS; i++)
3730 kfree(sbi->s_qf_names[i]);
3732 ext4_blkdev_remove(sbi);
3735 sb->s_fs_info = NULL;
3736 kfree(sbi->s_blockgroup_lock);
3744 * Setup any per-fs journal parameters now. We'll do this both on
3745 * initial mount, once the journal has been initialised but before we've
3746 * done any recovery; and again on any subsequent remount.
3748 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3750 struct ext4_sb_info *sbi = EXT4_SB(sb);
3752 journal->j_commit_interval = sbi->s_commit_interval;
3753 journal->j_min_batch_time = sbi->s_min_batch_time;
3754 journal->j_max_batch_time = sbi->s_max_batch_time;
3756 write_lock(&journal->j_state_lock);
3757 if (test_opt(sb, BARRIER))
3758 journal->j_flags |= JBD2_BARRIER;
3760 journal->j_flags &= ~JBD2_BARRIER;
3761 if (test_opt(sb, DATA_ERR_ABORT))
3762 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3764 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3765 write_unlock(&journal->j_state_lock);
3768 static journal_t *ext4_get_journal(struct super_block *sb,
3769 unsigned int journal_inum)
3771 struct inode *journal_inode;
3774 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3776 /* First, test for the existence of a valid inode on disk. Bad
3777 * things happen if we iget() an unused inode, as the subsequent
3778 * iput() will try to delete it. */
3780 journal_inode = ext4_iget(sb, journal_inum);
3781 if (IS_ERR(journal_inode)) {
3782 ext4_msg(sb, KERN_ERR, "no journal found");
3785 if (!journal_inode->i_nlink) {
3786 make_bad_inode(journal_inode);
3787 iput(journal_inode);
3788 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3792 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3793 journal_inode, journal_inode->i_size);
3794 if (!S_ISREG(journal_inode->i_mode)) {
3795 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3796 iput(journal_inode);
3800 journal = jbd2_journal_init_inode(journal_inode);
3802 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3803 iput(journal_inode);
3806 journal->j_private = sb;
3807 ext4_init_journal_params(sb, journal);
3811 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3814 struct buffer_head *bh;
3818 int hblock, blocksize;
3819 ext4_fsblk_t sb_block;
3820 unsigned long offset;
3821 struct ext4_super_block *es;
3822 struct block_device *bdev;
3824 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3826 bdev = ext4_blkdev_get(j_dev, sb);
3830 blocksize = sb->s_blocksize;
3831 hblock = bdev_logical_block_size(bdev);
3832 if (blocksize < hblock) {
3833 ext4_msg(sb, KERN_ERR,
3834 "blocksize too small for journal device");
3838 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3839 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3840 set_blocksize(bdev, blocksize);
3841 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3842 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3843 "external journal");
3847 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3848 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3849 !(le32_to_cpu(es->s_feature_incompat) &
3850 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3851 ext4_msg(sb, KERN_ERR, "external journal has "
3857 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3858 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3863 len = ext4_blocks_count(es);
3864 start = sb_block + 1;
3865 brelse(bh); /* we're done with the superblock */
3867 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3868 start, len, blocksize);
3870 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3873 journal->j_private = sb;
3874 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3875 wait_on_buffer(journal->j_sb_buffer);
3876 if (!buffer_uptodate(journal->j_sb_buffer)) {
3877 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3880 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3881 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3882 "user (unsupported) - %d",
3883 be32_to_cpu(journal->j_superblock->s_nr_users));
3886 EXT4_SB(sb)->journal_bdev = bdev;
3887 ext4_init_journal_params(sb, journal);
3891 jbd2_journal_destroy(journal);
3893 ext4_blkdev_put(bdev);
3897 static int ext4_load_journal(struct super_block *sb,
3898 struct ext4_super_block *es,
3899 unsigned long journal_devnum)
3902 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3905 int really_read_only;
3907 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3909 if (journal_devnum &&
3910 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3911 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3912 "numbers have changed");
3913 journal_dev = new_decode_dev(journal_devnum);
3915 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3917 really_read_only = bdev_read_only(sb->s_bdev);
3920 * Are we loading a blank journal or performing recovery after a
3921 * crash? For recovery, we need to check in advance whether we
3922 * can get read-write access to the device.
3924 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3925 if (sb->s_flags & MS_RDONLY) {
3926 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3927 "required on readonly filesystem");
3928 if (really_read_only) {
3929 ext4_msg(sb, KERN_ERR, "write access "
3930 "unavailable, cannot proceed");
3933 ext4_msg(sb, KERN_INFO, "write access will "
3934 "be enabled during recovery");
3938 if (journal_inum && journal_dev) {
3939 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3940 "and inode journals!");
3945 if (!(journal = ext4_get_journal(sb, journal_inum)))
3948 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3952 if (!(journal->j_flags & JBD2_BARRIER))
3953 ext4_msg(sb, KERN_INFO, "barriers disabled");
3955 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3956 err = jbd2_journal_update_format(journal);
3958 ext4_msg(sb, KERN_ERR, "error updating journal");
3959 jbd2_journal_destroy(journal);
3964 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3965 err = jbd2_journal_wipe(journal, !really_read_only);
3967 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3969 memcpy(save, ((char *) es) +
3970 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3971 err = jbd2_journal_load(journal);
3973 memcpy(((char *) es) + EXT4_S_ERR_START,
3974 save, EXT4_S_ERR_LEN);
3979 ext4_msg(sb, KERN_ERR, "error loading journal");
3980 jbd2_journal_destroy(journal);
3984 EXT4_SB(sb)->s_journal = journal;
3985 ext4_clear_journal_err(sb, es);
3987 if (!really_read_only && journal_devnum &&
3988 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3989 es->s_journal_dev = cpu_to_le32(journal_devnum);
3991 /* Make sure we flush the recovery flag to disk. */
3992 ext4_commit_super(sb, 1);
3998 static int ext4_commit_super(struct super_block *sb, int sync)
4000 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4001 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4006 if (buffer_write_io_error(sbh)) {
4008 * Oh, dear. A previous attempt to write the
4009 * superblock failed. This could happen because the
4010 * USB device was yanked out. Or it could happen to
4011 * be a transient write error and maybe the block will
4012 * be remapped. Nothing we can do but to retry the
4013 * write and hope for the best.
4015 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4016 "superblock detected");
4017 clear_buffer_write_io_error(sbh);
4018 set_buffer_uptodate(sbh);
4021 * If the file system is mounted read-only, don't update the
4022 * superblock write time. This avoids updating the superblock
4023 * write time when we are mounting the root file system
4024 * read/only but we need to replay the journal; at that point,
4025 * for people who are east of GMT and who make their clock
4026 * tick in localtime for Windows bug-for-bug compatibility,
4027 * the clock is set in the future, and this will cause e2fsck
4028 * to complain and force a full file system check.
4030 if (!(sb->s_flags & MS_RDONLY))
4031 es->s_wtime = cpu_to_le32(get_seconds());
4032 if (sb->s_bdev->bd_part)
4033 es->s_kbytes_written =
4034 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4035 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4036 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4038 es->s_kbytes_written =
4039 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4040 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4041 &EXT4_SB(sb)->s_freeblocks_counter));
4042 es->s_free_inodes_count =
4043 cpu_to_le32(percpu_counter_sum_positive(
4044 &EXT4_SB(sb)->s_freeinodes_counter));
4046 BUFFER_TRACE(sbh, "marking dirty");
4047 mark_buffer_dirty(sbh);
4049 error = sync_dirty_buffer(sbh);
4053 error = buffer_write_io_error(sbh);
4055 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4057 clear_buffer_write_io_error(sbh);
4058 set_buffer_uptodate(sbh);
4065 * Have we just finished recovery? If so, and if we are mounting (or
4066 * remounting) the filesystem readonly, then we will end up with a
4067 * consistent fs on disk. Record that fact.
4069 static void ext4_mark_recovery_complete(struct super_block *sb,
4070 struct ext4_super_block *es)
4072 journal_t *journal = EXT4_SB(sb)->s_journal;
4074 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4075 BUG_ON(journal != NULL);
4078 jbd2_journal_lock_updates(journal);
4079 if (jbd2_journal_flush(journal) < 0)
4082 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4083 sb->s_flags & MS_RDONLY) {
4084 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4085 ext4_commit_super(sb, 1);
4089 jbd2_journal_unlock_updates(journal);
4093 * If we are mounting (or read-write remounting) a filesystem whose journal
4094 * has recorded an error from a previous lifetime, move that error to the
4095 * main filesystem now.
4097 static void ext4_clear_journal_err(struct super_block *sb,
4098 struct ext4_super_block *es)
4104 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4106 journal = EXT4_SB(sb)->s_journal;
4109 * Now check for any error status which may have been recorded in the
4110 * journal by a prior ext4_error() or ext4_abort()
4113 j_errno = jbd2_journal_errno(journal);
4117 errstr = ext4_decode_error(sb, j_errno, nbuf);
4118 ext4_warning(sb, "Filesystem error recorded "
4119 "from previous mount: %s", errstr);
4120 ext4_warning(sb, "Marking fs in need of filesystem check.");
4122 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4123 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4124 ext4_commit_super(sb, 1);
4126 jbd2_journal_clear_err(journal);
4131 * Force the running and committing transactions to commit,
4132 * and wait on the commit.
4134 int ext4_force_commit(struct super_block *sb)
4139 if (sb->s_flags & MS_RDONLY)
4142 journal = EXT4_SB(sb)->s_journal;
4144 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4145 ret = ext4_journal_force_commit(journal);
4151 static void ext4_write_super(struct super_block *sb)
4154 ext4_commit_super(sb, 1);
4158 static int ext4_sync_fs(struct super_block *sb, int wait)
4162 struct ext4_sb_info *sbi = EXT4_SB(sb);
4164 trace_ext4_sync_fs(sb, wait);
4165 flush_workqueue(sbi->dio_unwritten_wq);
4166 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4168 jbd2_log_wait_commit(sbi->s_journal, target);
4174 * LVM calls this function before a (read-only) snapshot is created. This
4175 * gives us a chance to flush the journal completely and mark the fs clean.
4177 * Note that only this function cannot bring a filesystem to be in a clean
4178 * state independently, because ext4 prevents a new handle from being started
4179 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4182 static int ext4_freeze(struct super_block *sb)
4187 if (sb->s_flags & MS_RDONLY)
4190 journal = EXT4_SB(sb)->s_journal;
4192 /* Now we set up the journal barrier. */
4193 jbd2_journal_lock_updates(journal);
4196 * Don't clear the needs_recovery flag if we failed to flush
4199 error = jbd2_journal_flush(journal);
4203 /* Journal blocked and flushed, clear needs_recovery flag. */
4204 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4205 error = ext4_commit_super(sb, 1);
4207 /* we rely on s_frozen to stop further updates */
4208 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4213 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4214 * flag here, even though the filesystem is not technically dirty yet.
4216 static int ext4_unfreeze(struct super_block *sb)
4218 if (sb->s_flags & MS_RDONLY)
4222 /* Reset the needs_recovery flag before the fs is unlocked. */
4223 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4224 ext4_commit_super(sb, 1);
4230 * Structure to save mount options for ext4_remount's benefit
4232 struct ext4_mount_options {
4233 unsigned long s_mount_opt;
4234 unsigned long s_mount_opt2;
4237 unsigned long s_commit_interval;
4238 u32 s_min_batch_time, s_max_batch_time;
4241 char *s_qf_names[MAXQUOTAS];
4245 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4247 struct ext4_super_block *es;
4248 struct ext4_sb_info *sbi = EXT4_SB(sb);
4249 ext4_fsblk_t n_blocks_count = 0;
4250 unsigned long old_sb_flags;
4251 struct ext4_mount_options old_opts;
4252 int enable_quota = 0;
4254 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4259 char *orig_data = kstrdup(data, GFP_KERNEL);
4261 /* Store the original options */
4263 old_sb_flags = sb->s_flags;
4264 old_opts.s_mount_opt = sbi->s_mount_opt;
4265 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4266 old_opts.s_resuid = sbi->s_resuid;
4267 old_opts.s_resgid = sbi->s_resgid;
4268 old_opts.s_commit_interval = sbi->s_commit_interval;
4269 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4270 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4272 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4273 for (i = 0; i < MAXQUOTAS; i++)
4274 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4276 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4277 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4280 * Allow the "check" option to be passed as a remount option.
4282 if (!parse_options(data, sb, NULL, &journal_ioprio,
4283 &n_blocks_count, 1)) {
4288 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4289 ext4_abort(sb, "Abort forced by user");
4291 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4292 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4296 if (sbi->s_journal) {
4297 ext4_init_journal_params(sb, sbi->s_journal);
4298 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4301 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4302 n_blocks_count > ext4_blocks_count(es)) {
4303 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4308 if (*flags & MS_RDONLY) {
4309 err = dquot_suspend(sb, -1);
4314 * First of all, the unconditional stuff we have to do
4315 * to disable replay of the journal when we next remount
4317 sb->s_flags |= MS_RDONLY;
4320 * OK, test if we are remounting a valid rw partition
4321 * readonly, and if so set the rdonly flag and then
4322 * mark the partition as valid again.
4324 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4325 (sbi->s_mount_state & EXT4_VALID_FS))
4326 es->s_state = cpu_to_le16(sbi->s_mount_state);
4329 ext4_mark_recovery_complete(sb, es);
4331 /* Make sure we can mount this feature set readwrite */
4332 if (!ext4_feature_set_ok(sb, 0)) {
4337 * Make sure the group descriptor checksums
4338 * are sane. If they aren't, refuse to remount r/w.
4340 for (g = 0; g < sbi->s_groups_count; g++) {
4341 struct ext4_group_desc *gdp =
4342 ext4_get_group_desc(sb, g, NULL);
4344 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4345 ext4_msg(sb, KERN_ERR,
4346 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4347 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4348 le16_to_cpu(gdp->bg_checksum));
4355 * If we have an unprocessed orphan list hanging
4356 * around from a previously readonly bdev mount,
4357 * require a full umount/remount for now.
4359 if (es->s_last_orphan) {
4360 ext4_msg(sb, KERN_WARNING, "Couldn't "
4361 "remount RDWR because of unprocessed "
4362 "orphan inode list. Please "
4363 "umount/remount instead");
4369 * Mounting a RDONLY partition read-write, so reread
4370 * and store the current valid flag. (It may have
4371 * been changed by e2fsck since we originally mounted
4375 ext4_clear_journal_err(sb, es);
4376 sbi->s_mount_state = le16_to_cpu(es->s_state);
4377 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4379 if (!ext4_setup_super(sb, es, 0))
4380 sb->s_flags &= ~MS_RDONLY;
4386 * Reinitialize lazy itable initialization thread based on
4389 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4390 ext4_unregister_li_request(sb);
4392 ext4_group_t first_not_zeroed;
4393 first_not_zeroed = ext4_has_uninit_itable(sb);
4394 ext4_register_li_request(sb, first_not_zeroed);
4397 ext4_setup_system_zone(sb);
4398 if (sbi->s_journal == NULL)
4399 ext4_commit_super(sb, 1);
4402 /* Release old quota file names */
4403 for (i = 0; i < MAXQUOTAS; i++)
4404 if (old_opts.s_qf_names[i] &&
4405 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4406 kfree(old_opts.s_qf_names[i]);
4410 dquot_resume(sb, -1);
4412 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4417 sb->s_flags = old_sb_flags;
4418 sbi->s_mount_opt = old_opts.s_mount_opt;
4419 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4420 sbi->s_resuid = old_opts.s_resuid;
4421 sbi->s_resgid = old_opts.s_resgid;
4422 sbi->s_commit_interval = old_opts.s_commit_interval;
4423 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4424 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4426 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4427 for (i = 0; i < MAXQUOTAS; i++) {
4428 if (sbi->s_qf_names[i] &&
4429 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4430 kfree(sbi->s_qf_names[i]);
4431 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4439 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4441 struct super_block *sb = dentry->d_sb;
4442 struct ext4_sb_info *sbi = EXT4_SB(sb);
4443 struct ext4_super_block *es = sbi->s_es;
4446 if (test_opt(sb, MINIX_DF)) {
4447 sbi->s_overhead_last = 0;
4448 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4449 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4450 ext4_fsblk_t overhead = 0;
4453 * Compute the overhead (FS structures). This is constant
4454 * for a given filesystem unless the number of block groups
4455 * changes so we cache the previous value until it does.
4459 * All of the blocks before first_data_block are
4462 overhead = le32_to_cpu(es->s_first_data_block);
4465 * Add the overhead attributed to the superblock and
4466 * block group descriptors. If the sparse superblocks
4467 * feature is turned on, then not all groups have this.
4469 for (i = 0; i < ngroups; i++) {
4470 overhead += ext4_bg_has_super(sb, i) +
4471 ext4_bg_num_gdb(sb, i);
4476 * Every block group has an inode bitmap, a block
4477 * bitmap, and an inode table.
4479 overhead += ngroups * (2 + sbi->s_itb_per_group);
4480 sbi->s_overhead_last = overhead;
4482 sbi->s_blocks_last = ext4_blocks_count(es);
4485 buf->f_type = EXT4_SUPER_MAGIC;
4486 buf->f_bsize = sb->s_blocksize;
4487 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4488 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4489 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4490 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4491 if (buf->f_bfree < ext4_r_blocks_count(es))
4493 buf->f_files = le32_to_cpu(es->s_inodes_count);
4494 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4495 buf->f_namelen = EXT4_NAME_LEN;
4496 fsid = le64_to_cpup((void *)es->s_uuid) ^
4497 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4498 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4499 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4504 /* Helper function for writing quotas on sync - we need to start transaction
4505 * before quota file is locked for write. Otherwise the are possible deadlocks:
4506 * Process 1 Process 2
4507 * ext4_create() quota_sync()
4508 * jbd2_journal_start() write_dquot()
4509 * dquot_initialize() down(dqio_mutex)
4510 * down(dqio_mutex) jbd2_journal_start()
4516 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4518 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4521 static int ext4_write_dquot(struct dquot *dquot)
4525 struct inode *inode;
4527 inode = dquot_to_inode(dquot);
4528 handle = ext4_journal_start(inode,
4529 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4531 return PTR_ERR(handle);
4532 ret = dquot_commit(dquot);
4533 err = ext4_journal_stop(handle);
4539 static int ext4_acquire_dquot(struct dquot *dquot)
4544 handle = ext4_journal_start(dquot_to_inode(dquot),
4545 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4547 return PTR_ERR(handle);
4548 ret = dquot_acquire(dquot);
4549 err = ext4_journal_stop(handle);
4555 static int ext4_release_dquot(struct dquot *dquot)
4560 handle = ext4_journal_start(dquot_to_inode(dquot),
4561 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4562 if (IS_ERR(handle)) {
4563 /* Release dquot anyway to avoid endless cycle in dqput() */
4564 dquot_release(dquot);
4565 return PTR_ERR(handle);
4567 ret = dquot_release(dquot);
4568 err = ext4_journal_stop(handle);
4574 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4576 /* Are we journaling quotas? */
4577 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4578 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4579 dquot_mark_dquot_dirty(dquot);
4580 return ext4_write_dquot(dquot);
4582 return dquot_mark_dquot_dirty(dquot);
4586 static int ext4_write_info(struct super_block *sb, int type)
4591 /* Data block + inode block */
4592 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4594 return PTR_ERR(handle);
4595 ret = dquot_commit_info(sb, type);
4596 err = ext4_journal_stop(handle);
4603 * Turn on quotas during mount time - we need to find
4604 * the quota file and such...
4606 static int ext4_quota_on_mount(struct super_block *sb, int type)
4608 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4609 EXT4_SB(sb)->s_jquota_fmt, type);
4613 * Standard function to be called on quota_on
4615 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4620 if (!test_opt(sb, QUOTA))
4623 /* Quotafile not on the same filesystem? */
4624 if (path->mnt->mnt_sb != sb)
4626 /* Journaling quota? */
4627 if (EXT4_SB(sb)->s_qf_names[type]) {
4628 /* Quotafile not in fs root? */
4629 if (path->dentry->d_parent != sb->s_root)
4630 ext4_msg(sb, KERN_WARNING,
4631 "Quota file not on filesystem root. "
4632 "Journaled quota will not work");
4636 * When we journal data on quota file, we have to flush journal to see
4637 * all updates to the file when we bypass pagecache...
4639 if (EXT4_SB(sb)->s_journal &&
4640 ext4_should_journal_data(path->dentry->d_inode)) {
4642 * We don't need to lock updates but journal_flush() could
4643 * otherwise be livelocked...
4645 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4646 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4647 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4652 return dquot_quota_on(sb, type, format_id, path);
4655 static int ext4_quota_off(struct super_block *sb, int type)
4657 struct inode *inode = sb_dqopt(sb)->files[type];
4660 /* Force all delayed allocation blocks to be allocated.
4661 * Caller already holds s_umount sem */
4662 if (test_opt(sb, DELALLOC))
4663 sync_filesystem(sb);
4668 /* Update modification times of quota files when userspace can
4669 * start looking at them */
4670 handle = ext4_journal_start(inode, 1);
4673 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4674 ext4_mark_inode_dirty(handle, inode);
4675 ext4_journal_stop(handle);
4678 return dquot_quota_off(sb, type);
4681 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4682 * acquiring the locks... As quota files are never truncated and quota code
4683 * itself serializes the operations (and no one else should touch the files)
4684 * we don't have to be afraid of races */
4685 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4686 size_t len, loff_t off)
4688 struct inode *inode = sb_dqopt(sb)->files[type];
4689 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4691 int offset = off & (sb->s_blocksize - 1);
4694 struct buffer_head *bh;
4695 loff_t i_size = i_size_read(inode);
4699 if (off+len > i_size)
4702 while (toread > 0) {
4703 tocopy = sb->s_blocksize - offset < toread ?
4704 sb->s_blocksize - offset : toread;
4705 bh = ext4_bread(NULL, inode, blk, 0, &err);
4708 if (!bh) /* A hole? */
4709 memset(data, 0, tocopy);
4711 memcpy(data, bh->b_data+offset, tocopy);
4721 /* Write to quotafile (we know the transaction is already started and has
4722 * enough credits) */
4723 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4724 const char *data, size_t len, loff_t off)
4726 struct inode *inode = sb_dqopt(sb)->files[type];
4727 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4729 int offset = off & (sb->s_blocksize - 1);
4730 struct buffer_head *bh;
4731 handle_t *handle = journal_current_handle();
4733 if (EXT4_SB(sb)->s_journal && !handle) {
4734 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4735 " cancelled because transaction is not started",
4736 (unsigned long long)off, (unsigned long long)len);
4740 * Since we account only one data block in transaction credits,
4741 * then it is impossible to cross a block boundary.
4743 if (sb->s_blocksize - offset < len) {
4744 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4745 " cancelled because not block aligned",
4746 (unsigned long long)off, (unsigned long long)len);
4750 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4751 bh = ext4_bread(handle, inode, blk, 1, &err);
4754 err = ext4_journal_get_write_access(handle, bh);
4760 memcpy(bh->b_data+offset, data, len);
4761 flush_dcache_page(bh->b_page);
4763 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4767 mutex_unlock(&inode->i_mutex);
4770 if (inode->i_size < off + len) {
4771 i_size_write(inode, off + len);
4772 EXT4_I(inode)->i_disksize = inode->i_size;
4773 ext4_mark_inode_dirty(handle, inode);
4775 mutex_unlock(&inode->i_mutex);
4781 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4782 const char *dev_name, void *data)
4784 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4787 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4788 static inline void register_as_ext2(void)
4790 int err = register_filesystem(&ext2_fs_type);
4793 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4796 static inline void unregister_as_ext2(void)
4798 unregister_filesystem(&ext2_fs_type);
4801 static inline int ext2_feature_set_ok(struct super_block *sb)
4803 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4805 if (sb->s_flags & MS_RDONLY)
4807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4811 MODULE_ALIAS("ext2");
4813 static inline void register_as_ext2(void) { }
4814 static inline void unregister_as_ext2(void) { }
4815 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4818 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4819 static inline void register_as_ext3(void)
4821 int err = register_filesystem(&ext3_fs_type);
4824 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4827 static inline void unregister_as_ext3(void)
4829 unregister_filesystem(&ext3_fs_type);
4832 static inline int ext3_feature_set_ok(struct super_block *sb)
4834 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4836 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4838 if (sb->s_flags & MS_RDONLY)
4840 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4844 MODULE_ALIAS("ext3");
4846 static inline void register_as_ext3(void) { }
4847 static inline void unregister_as_ext3(void) { }
4848 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4851 static struct file_system_type ext4_fs_type = {
4852 .owner = THIS_MODULE,
4854 .mount = ext4_mount,
4855 .kill_sb = kill_block_super,
4856 .fs_flags = FS_REQUIRES_DEV,
4859 static int __init ext4_init_feat_adverts(void)
4861 struct ext4_features *ef;
4864 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4868 ef->f_kobj.kset = ext4_kset;
4869 init_completion(&ef->f_kobj_unregister);
4870 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4883 static void ext4_exit_feat_adverts(void)
4885 kobject_put(&ext4_feat->f_kobj);
4886 wait_for_completion(&ext4_feat->f_kobj_unregister);
4890 /* Shared across all ext4 file systems */
4891 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4892 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4894 static int __init ext4_init_fs(void)
4898 ext4_check_flag_values();
4900 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4901 mutex_init(&ext4__aio_mutex[i]);
4902 init_waitqueue_head(&ext4__ioend_wq[i]);
4905 err = ext4_init_pageio();
4908 err = ext4_init_system_zone();
4911 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4914 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4915 if (!ext4_proc_root)
4918 err = ext4_init_feat_adverts();
4922 err = ext4_init_mballoc();
4926 err = ext4_init_xattr();
4929 err = init_inodecache();
4934 err = register_filesystem(&ext4_fs_type);
4938 ext4_li_info = NULL;
4939 mutex_init(&ext4_li_mtx);
4942 unregister_as_ext2();
4943 unregister_as_ext3();
4944 destroy_inodecache();
4948 ext4_exit_mballoc();
4950 ext4_exit_feat_adverts();
4952 remove_proc_entry("fs/ext4", NULL);
4954 kset_unregister(ext4_kset);
4956 ext4_exit_system_zone();
4962 static void __exit ext4_exit_fs(void)
4964 ext4_destroy_lazyinit_thread();
4965 unregister_as_ext2();
4966 unregister_as_ext3();
4967 unregister_filesystem(&ext4_fs_type);
4968 destroy_inodecache();
4970 ext4_exit_mballoc();
4971 ext4_exit_feat_adverts();
4972 remove_proc_entry("fs/ext4", NULL);
4973 kset_unregister(ext4_kset);
4974 ext4_exit_system_zone();
4978 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4979 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4980 MODULE_LICENSE("GPL");
4981 module_init(ext4_init_fs)
4982 module_exit(ext4_exit_fs)