2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_jbd2.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/ext4.h>
55 static struct proc_dir_entry *ext4_proc_root;
56 static struct kset *ext4_kset;
57 struct ext4_lazy_init *ext4_li_info;
58 struct mutex ext4_li_mtx;
59 struct ext4_features *ext4_feat;
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static const char *ext4_decode_error(struct super_block *sb, int errno,
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static void ext4_write_super(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
83 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
84 static struct file_system_type ext3_fs_type = {
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
91 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
93 #define IS_EXT3_SB(sb) (0)
96 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
97 struct ext4_group_desc *bg)
99 return le32_to_cpu(bg->bg_block_bitmap_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
104 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
105 struct ext4_group_desc *bg)
107 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
112 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
113 struct ext4_group_desc *bg)
115 return le32_to_cpu(bg->bg_inode_table_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
120 __u32 ext4_free_blks_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
123 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
128 __u32 ext4_free_inodes_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
131 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
136 __u32 ext4_used_dirs_count(struct super_block *sb,
137 struct ext4_group_desc *bg)
139 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
140 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
141 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
144 __u32 ext4_itable_unused_count(struct super_block *sb,
145 struct ext4_group_desc *bg)
147 return le16_to_cpu(bg->bg_itable_unused_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
149 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
152 void ext4_block_bitmap_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
155 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
160 void ext4_inode_bitmap_set(struct super_block *sb,
161 struct ext4_group_desc *bg, ext4_fsblk_t blk)
163 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
168 void ext4_inode_table_set(struct super_block *sb,
169 struct ext4_group_desc *bg, ext4_fsblk_t blk)
171 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
176 void ext4_free_blks_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
179 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
184 void ext4_free_inodes_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
187 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
192 void ext4_used_dirs_set(struct super_block *sb,
193 struct ext4_group_desc *bg, __u32 count)
195 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
196 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
197 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
200 void ext4_itable_unused_set(struct super_block *sb,
201 struct ext4_group_desc *bg, __u32 count)
203 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
209 /* Just increment the non-pointer handle value */
210 static handle_t *ext4_get_nojournal(void)
212 handle_t *handle = current->journal_info;
213 unsigned long ref_cnt = (unsigned long)handle;
215 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
218 handle = (handle_t *)ref_cnt;
220 current->journal_info = handle;
225 /* Decrement the non-pointer handle value */
226 static void ext4_put_nojournal(handle_t *handle)
228 unsigned long ref_cnt = (unsigned long)handle;
230 BUG_ON(ref_cnt == 0);
233 handle = (handle_t *)ref_cnt;
235 current->journal_info = handle;
239 * Wrappers for jbd2_journal_start/end.
241 * The only special thing we need to do here is to make sure that all
242 * journal_end calls result in the superblock being marked dirty, so
243 * that sync() will call the filesystem's write_super callback if
246 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
250 if (sb->s_flags & MS_RDONLY)
251 return ERR_PTR(-EROFS);
253 vfs_check_frozen(sb, SB_FREEZE_TRANS);
254 /* Special case here: if the journal has aborted behind our
255 * backs (eg. EIO in the commit thread), then we still need to
256 * take the FS itself readonly cleanly. */
257 journal = EXT4_SB(sb)->s_journal;
259 if (is_journal_aborted(journal)) {
260 ext4_abort(sb, "Detected aborted journal");
261 return ERR_PTR(-EROFS);
263 return jbd2_journal_start(journal, nblocks);
265 return ext4_get_nojournal();
269 * The only special thing we need to do here is to make sure that all
270 * jbd2_journal_stop calls result in the superblock being marked dirty, so
271 * that sync() will call the filesystem's write_super callback if
274 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
276 struct super_block *sb;
280 if (!ext4_handle_valid(handle)) {
281 ext4_put_nojournal(handle);
284 sb = handle->h_transaction->t_journal->j_private;
286 rc = jbd2_journal_stop(handle);
291 __ext4_std_error(sb, where, line, err);
295 void ext4_journal_abort_handle(const char *caller, unsigned int line,
296 const char *err_fn, struct buffer_head *bh,
297 handle_t *handle, int err)
300 const char *errstr = ext4_decode_error(NULL, err, nbuf);
302 BUG_ON(!ext4_handle_valid(handle));
305 BUFFER_TRACE(bh, "abort");
310 if (is_handle_aborted(handle))
313 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
314 caller, line, errstr, err_fn);
316 jbd2_journal_abort_handle(handle);
319 static void __save_error_info(struct super_block *sb, const char *func,
322 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
324 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
325 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
326 es->s_last_error_time = cpu_to_le32(get_seconds());
327 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
328 es->s_last_error_line = cpu_to_le32(line);
329 if (!es->s_first_error_time) {
330 es->s_first_error_time = es->s_last_error_time;
331 strncpy(es->s_first_error_func, func,
332 sizeof(es->s_first_error_func));
333 es->s_first_error_line = cpu_to_le32(line);
334 es->s_first_error_ino = es->s_last_error_ino;
335 es->s_first_error_block = es->s_last_error_block;
338 * Start the daily error reporting function if it hasn't been
341 if (!es->s_error_count)
342 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
343 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
346 static void save_error_info(struct super_block *sb, const char *func,
349 __save_error_info(sb, func, line);
350 ext4_commit_super(sb, 1);
354 /* Deal with the reporting of failure conditions on a filesystem such as
355 * inconsistencies detected or read IO failures.
357 * On ext2, we can store the error state of the filesystem in the
358 * superblock. That is not possible on ext4, because we may have other
359 * write ordering constraints on the superblock which prevent us from
360 * writing it out straight away; and given that the journal is about to
361 * be aborted, we can't rely on the current, or future, transactions to
362 * write out the superblock safely.
364 * We'll just use the jbd2_journal_abort() error code to record an error in
365 * the journal instead. On recovery, the journal will complain about
366 * that error until we've noted it down and cleared it.
369 static void ext4_handle_error(struct super_block *sb)
371 if (sb->s_flags & MS_RDONLY)
374 if (!test_opt(sb, ERRORS_CONT)) {
375 journal_t *journal = EXT4_SB(sb)->s_journal;
377 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
379 jbd2_journal_abort(journal, -EIO);
381 if (test_opt(sb, ERRORS_RO)) {
382 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
383 sb->s_flags |= MS_RDONLY;
385 if (test_opt(sb, ERRORS_PANIC))
386 panic("EXT4-fs (device %s): panic forced after error\n",
390 void __ext4_error(struct super_block *sb, const char *function,
391 unsigned int line, const char *fmt, ...)
393 struct va_format vaf;
399 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
400 sb->s_id, function, line, current->comm, &vaf);
403 ext4_handle_error(sb);
406 void ext4_error_inode(struct inode *inode, const char *function,
407 unsigned int line, ext4_fsblk_t block,
408 const char *fmt, ...)
411 struct va_format vaf;
412 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
414 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
415 es->s_last_error_block = cpu_to_le64(block);
416 save_error_info(inode->i_sb, function, line);
420 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
421 inode->i_sb->s_id, function, line, inode->i_ino);
423 printk(KERN_CONT "block %llu: ", block);
424 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
427 ext4_handle_error(inode->i_sb);
430 void ext4_error_file(struct file *file, const char *function,
431 unsigned int line, ext4_fsblk_t block,
432 const char *fmt, ...)
435 struct va_format vaf;
436 struct ext4_super_block *es;
437 struct inode *inode = file->f_dentry->d_inode;
438 char pathname[80], *path;
440 es = EXT4_SB(inode->i_sb)->s_es;
441 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
442 save_error_info(inode->i_sb, function, line);
443 path = d_path(&(file->f_path), pathname, sizeof(pathname));
447 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
448 inode->i_sb->s_id, function, line, inode->i_ino);
450 printk(KERN_CONT "block %llu: ", block);
454 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
457 ext4_handle_error(inode->i_sb);
460 static const char *ext4_decode_error(struct super_block *sb, int errno,
467 errstr = "IO failure";
470 errstr = "Out of memory";
473 if (!sb || (EXT4_SB(sb)->s_journal &&
474 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
475 errstr = "Journal has aborted";
477 errstr = "Readonly filesystem";
480 /* If the caller passed in an extra buffer for unknown
481 * errors, textualise them now. Else we just return
484 /* Check for truncated error codes... */
485 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
494 /* __ext4_std_error decodes expected errors from journaling functions
495 * automatically and invokes the appropriate error response. */
497 void __ext4_std_error(struct super_block *sb, const char *function,
498 unsigned int line, int errno)
503 /* Special case: if the error is EROFS, and we're not already
504 * inside a transaction, then there's really no point in logging
506 if (errno == -EROFS && journal_current_handle() == NULL &&
507 (sb->s_flags & MS_RDONLY))
510 errstr = ext4_decode_error(sb, errno, nbuf);
511 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
512 sb->s_id, function, line, errstr);
513 save_error_info(sb, function, line);
515 ext4_handle_error(sb);
519 * ext4_abort is a much stronger failure handler than ext4_error. The
520 * abort function may be used to deal with unrecoverable failures such
521 * as journal IO errors or ENOMEM at a critical moment in log management.
523 * We unconditionally force the filesystem into an ABORT|READONLY state,
524 * unless the error response on the fs has been set to panic in which
525 * case we take the easy way out and panic immediately.
528 void __ext4_abort(struct super_block *sb, const char *function,
529 unsigned int line, const char *fmt, ...)
533 save_error_info(sb, function, line);
535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
541 if ((sb->s_flags & MS_RDONLY) == 0) {
542 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
543 sb->s_flags |= MS_RDONLY;
544 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
545 if (EXT4_SB(sb)->s_journal)
546 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
547 save_error_info(sb, function, line);
549 if (test_opt(sb, ERRORS_PANIC))
550 panic("EXT4-fs panic from previous error\n");
553 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
555 struct va_format vaf;
561 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
565 void __ext4_warning(struct super_block *sb, const char *function,
566 unsigned int line, const char *fmt, ...)
568 struct va_format vaf;
574 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
575 sb->s_id, function, line, &vaf);
579 void __ext4_grp_locked_error(const char *function, unsigned int line,
580 struct super_block *sb, ext4_group_t grp,
581 unsigned long ino, ext4_fsblk_t block,
582 const char *fmt, ...)
586 struct va_format vaf;
588 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
590 es->s_last_error_ino = cpu_to_le32(ino);
591 es->s_last_error_block = cpu_to_le64(block);
592 __save_error_info(sb, function, line);
598 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
599 sb->s_id, function, line, grp);
601 printk(KERN_CONT "inode %lu: ", ino);
603 printk(KERN_CONT "block %llu:", (unsigned long long) block);
604 printk(KERN_CONT "%pV\n", &vaf);
607 if (test_opt(sb, ERRORS_CONT)) {
608 ext4_commit_super(sb, 0);
612 ext4_unlock_group(sb, grp);
613 ext4_handle_error(sb);
615 * We only get here in the ERRORS_RO case; relocking the group
616 * may be dangerous, but nothing bad will happen since the
617 * filesystem will have already been marked read/only and the
618 * journal has been aborted. We return 1 as a hint to callers
619 * who might what to use the return value from
620 * ext4_grp_locked_error() to distinguish beween the
621 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
622 * aggressively from the ext4 function in question, with a
623 * more appropriate error code.
625 ext4_lock_group(sb, grp);
629 void ext4_update_dynamic_rev(struct super_block *sb)
631 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
633 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
637 "updating to rev %d because of new feature flag, "
638 "running e2fsck is recommended",
641 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
642 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
643 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
644 /* leave es->s_feature_*compat flags alone */
645 /* es->s_uuid will be set by e2fsck if empty */
648 * The rest of the superblock fields should be zero, and if not it
649 * means they are likely already in use, so leave them alone. We
650 * can leave it up to e2fsck to clean up any inconsistencies there.
655 * Open the external journal device
657 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
659 struct block_device *bdev;
660 char b[BDEVNAME_SIZE];
662 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
668 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
669 __bdevname(dev, b), PTR_ERR(bdev));
674 * Release the journal device
676 static int ext4_blkdev_put(struct block_device *bdev)
678 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
681 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
683 struct block_device *bdev;
686 bdev = sbi->journal_bdev;
688 ret = ext4_blkdev_put(bdev);
689 sbi->journal_bdev = NULL;
694 static inline struct inode *orphan_list_entry(struct list_head *l)
696 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
699 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
703 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
704 le32_to_cpu(sbi->s_es->s_last_orphan));
706 printk(KERN_ERR "sb_info orphan list:\n");
707 list_for_each(l, &sbi->s_orphan) {
708 struct inode *inode = orphan_list_entry(l);
710 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
711 inode->i_sb->s_id, inode->i_ino, inode,
712 inode->i_mode, inode->i_nlink,
717 static void ext4_put_super(struct super_block *sb)
719 struct ext4_sb_info *sbi = EXT4_SB(sb);
720 struct ext4_super_block *es = sbi->s_es;
723 ext4_unregister_li_request(sb);
724 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
726 flush_workqueue(sbi->dio_unwritten_wq);
727 destroy_workqueue(sbi->dio_unwritten_wq);
731 ext4_commit_super(sb, 1);
733 if (sbi->s_journal) {
734 err = jbd2_journal_destroy(sbi->s_journal);
735 sbi->s_journal = NULL;
737 ext4_abort(sb, "Couldn't clean up the journal");
740 del_timer(&sbi->s_err_report);
741 ext4_release_system_zone(sb);
743 ext4_ext_release(sb);
744 ext4_xattr_put_super(sb);
746 if (!(sb->s_flags & MS_RDONLY)) {
747 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
748 es->s_state = cpu_to_le16(sbi->s_mount_state);
749 ext4_commit_super(sb, 1);
752 remove_proc_entry(sb->s_id, ext4_proc_root);
754 kobject_del(&sbi->s_kobj);
756 for (i = 0; i < sbi->s_gdb_count; i++)
757 brelse(sbi->s_group_desc[i]);
758 kfree(sbi->s_group_desc);
759 if (is_vmalloc_addr(sbi->s_flex_groups))
760 vfree(sbi->s_flex_groups);
762 kfree(sbi->s_flex_groups);
763 percpu_counter_destroy(&sbi->s_freeblocks_counter);
764 percpu_counter_destroy(&sbi->s_freeinodes_counter);
765 percpu_counter_destroy(&sbi->s_dirs_counter);
766 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
769 for (i = 0; i < MAXQUOTAS; i++)
770 kfree(sbi->s_qf_names[i]);
773 /* Debugging code just in case the in-memory inode orphan list
774 * isn't empty. The on-disk one can be non-empty if we've
775 * detected an error and taken the fs readonly, but the
776 * in-memory list had better be clean by this point. */
777 if (!list_empty(&sbi->s_orphan))
778 dump_orphan_list(sb, sbi);
779 J_ASSERT(list_empty(&sbi->s_orphan));
781 invalidate_bdev(sb->s_bdev);
782 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
784 * Invalidate the journal device's buffers. We don't want them
785 * floating about in memory - the physical journal device may
786 * hotswapped, and it breaks the `ro-after' testing code.
788 sync_blockdev(sbi->journal_bdev);
789 invalidate_bdev(sbi->journal_bdev);
790 ext4_blkdev_remove(sbi);
792 sb->s_fs_info = NULL;
794 * Now that we are completely done shutting down the
795 * superblock, we need to actually destroy the kobject.
798 kobject_put(&sbi->s_kobj);
799 wait_for_completion(&sbi->s_kobj_unregister);
800 kfree(sbi->s_blockgroup_lock);
804 static struct kmem_cache *ext4_inode_cachep;
807 * Called inside transaction, so use GFP_NOFS
809 static struct inode *ext4_alloc_inode(struct super_block *sb)
811 struct ext4_inode_info *ei;
813 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
817 ei->vfs_inode.i_version = 1;
818 ei->vfs_inode.i_data.writeback_index = 0;
819 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
820 INIT_LIST_HEAD(&ei->i_prealloc_list);
821 spin_lock_init(&ei->i_prealloc_lock);
822 ei->i_reserved_data_blocks = 0;
823 ei->i_reserved_meta_blocks = 0;
824 ei->i_allocated_meta_blocks = 0;
825 ei->i_da_metadata_calc_len = 0;
826 spin_lock_init(&(ei->i_block_reservation_lock));
828 ei->i_reserved_quota = 0;
831 INIT_LIST_HEAD(&ei->i_completed_io_list);
832 spin_lock_init(&ei->i_completed_io_lock);
833 ei->cur_aio_dio = NULL;
835 ei->i_datasync_tid = 0;
836 atomic_set(&ei->i_ioend_count, 0);
837 atomic_set(&ei->i_aiodio_unwritten, 0);
839 return &ei->vfs_inode;
842 static int ext4_drop_inode(struct inode *inode)
844 int drop = generic_drop_inode(inode);
846 trace_ext4_drop_inode(inode, drop);
850 static void ext4_i_callback(struct rcu_head *head)
852 struct inode *inode = container_of(head, struct inode, i_rcu);
853 INIT_LIST_HEAD(&inode->i_dentry);
854 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
857 static void ext4_destroy_inode(struct inode *inode)
859 ext4_ioend_wait(inode);
860 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
861 ext4_msg(inode->i_sb, KERN_ERR,
862 "Inode %lu (%p): orphan list check failed!",
863 inode->i_ino, EXT4_I(inode));
864 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
865 EXT4_I(inode), sizeof(struct ext4_inode_info),
869 call_rcu(&inode->i_rcu, ext4_i_callback);
872 static void init_once(void *foo)
874 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
876 INIT_LIST_HEAD(&ei->i_orphan);
877 #ifdef CONFIG_EXT4_FS_XATTR
878 init_rwsem(&ei->xattr_sem);
880 init_rwsem(&ei->i_data_sem);
881 inode_init_once(&ei->vfs_inode);
884 static int init_inodecache(void)
886 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
887 sizeof(struct ext4_inode_info),
888 0, (SLAB_RECLAIM_ACCOUNT|
891 if (ext4_inode_cachep == NULL)
896 static void destroy_inodecache(void)
898 kmem_cache_destroy(ext4_inode_cachep);
901 void ext4_clear_inode(struct inode *inode)
903 invalidate_inode_buffers(inode);
904 end_writeback(inode);
906 ext4_discard_preallocations(inode);
907 if (EXT4_I(inode)->jinode) {
908 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
909 EXT4_I(inode)->jinode);
910 jbd2_free_inode(EXT4_I(inode)->jinode);
911 EXT4_I(inode)->jinode = NULL;
915 static inline void ext4_show_quota_options(struct seq_file *seq,
916 struct super_block *sb)
918 #if defined(CONFIG_QUOTA)
919 struct ext4_sb_info *sbi = EXT4_SB(sb);
921 if (sbi->s_jquota_fmt) {
924 switch (sbi->s_jquota_fmt) {
935 seq_printf(seq, ",jqfmt=%s", fmtname);
938 if (sbi->s_qf_names[USRQUOTA])
939 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
941 if (sbi->s_qf_names[GRPQUOTA])
942 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
944 if (test_opt(sb, USRQUOTA))
945 seq_puts(seq, ",usrquota");
947 if (test_opt(sb, GRPQUOTA))
948 seq_puts(seq, ",grpquota");
954 * - it's set to a non-default value OR
955 * - if the per-sb default is different from the global default
957 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
960 unsigned long def_mount_opts;
961 struct super_block *sb = vfs->mnt_sb;
962 struct ext4_sb_info *sbi = EXT4_SB(sb);
963 struct ext4_super_block *es = sbi->s_es;
965 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
966 def_errors = le16_to_cpu(es->s_errors);
968 if (sbi->s_sb_block != 1)
969 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
970 if (test_opt(sb, MINIX_DF))
971 seq_puts(seq, ",minixdf");
972 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
973 seq_puts(seq, ",grpid");
974 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
975 seq_puts(seq, ",nogrpid");
976 if (sbi->s_resuid != EXT4_DEF_RESUID ||
977 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
978 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
980 if (sbi->s_resgid != EXT4_DEF_RESGID ||
981 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
982 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
984 if (test_opt(sb, ERRORS_RO)) {
985 if (def_errors == EXT4_ERRORS_PANIC ||
986 def_errors == EXT4_ERRORS_CONTINUE) {
987 seq_puts(seq, ",errors=remount-ro");
990 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
991 seq_puts(seq, ",errors=continue");
992 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
993 seq_puts(seq, ",errors=panic");
994 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
995 seq_puts(seq, ",nouid32");
996 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
997 seq_puts(seq, ",debug");
998 if (test_opt(sb, OLDALLOC))
999 seq_puts(seq, ",oldalloc");
1000 #ifdef CONFIG_EXT4_FS_XATTR
1001 if (test_opt(sb, XATTR_USER) &&
1002 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
1003 seq_puts(seq, ",user_xattr");
1004 if (!test_opt(sb, XATTR_USER) &&
1005 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
1006 seq_puts(seq, ",nouser_xattr");
1009 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1010 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1011 seq_puts(seq, ",acl");
1012 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1013 seq_puts(seq, ",noacl");
1015 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1016 seq_printf(seq, ",commit=%u",
1017 (unsigned) (sbi->s_commit_interval / HZ));
1019 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1020 seq_printf(seq, ",min_batch_time=%u",
1021 (unsigned) sbi->s_min_batch_time);
1023 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1024 seq_printf(seq, ",max_batch_time=%u",
1025 (unsigned) sbi->s_min_batch_time);
1029 * We're changing the default of barrier mount option, so
1030 * let's always display its mount state so it's clear what its
1033 seq_puts(seq, ",barrier=");
1034 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1035 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1036 seq_puts(seq, ",journal_async_commit");
1037 else if (test_opt(sb, JOURNAL_CHECKSUM))
1038 seq_puts(seq, ",journal_checksum");
1039 if (test_opt(sb, I_VERSION))
1040 seq_puts(seq, ",i_version");
1041 if (!test_opt(sb, DELALLOC) &&
1042 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1043 seq_puts(seq, ",nodelalloc");
1045 if (test_opt(sb, MBLK_IO_SUBMIT))
1046 seq_puts(seq, ",mblk_io_submit");
1048 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1050 * journal mode get enabled in different ways
1051 * So just print the value even if we didn't specify it
1053 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1054 seq_puts(seq, ",data=journal");
1055 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1056 seq_puts(seq, ",data=ordered");
1057 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1058 seq_puts(seq, ",data=writeback");
1060 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1061 seq_printf(seq, ",inode_readahead_blks=%u",
1062 sbi->s_inode_readahead_blks);
1064 if (test_opt(sb, DATA_ERR_ABORT))
1065 seq_puts(seq, ",data_err=abort");
1067 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1068 seq_puts(seq, ",noauto_da_alloc");
1070 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1071 seq_puts(seq, ",discard");
1073 if (test_opt(sb, NOLOAD))
1074 seq_puts(seq, ",norecovery");
1076 if (test_opt(sb, DIOREAD_NOLOCK))
1077 seq_puts(seq, ",dioread_nolock");
1079 if (test_opt(sb, BLOCK_VALIDITY) &&
1080 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1081 seq_puts(seq, ",block_validity");
1083 if (!test_opt(sb, INIT_INODE_TABLE))
1084 seq_puts(seq, ",noinit_inode_table");
1085 else if (sbi->s_li_wait_mult)
1086 seq_printf(seq, ",init_inode_table=%u",
1087 (unsigned) sbi->s_li_wait_mult);
1089 ext4_show_quota_options(seq, sb);
1094 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1095 u64 ino, u32 generation)
1097 struct inode *inode;
1099 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1100 return ERR_PTR(-ESTALE);
1101 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1102 return ERR_PTR(-ESTALE);
1104 /* iget isn't really right if the inode is currently unallocated!!
1106 * ext4_read_inode will return a bad_inode if the inode had been
1107 * deleted, so we should be safe.
1109 * Currently we don't know the generation for parent directory, so
1110 * a generation of 0 means "accept any"
1112 inode = ext4_iget(sb, ino);
1114 return ERR_CAST(inode);
1115 if (generation && inode->i_generation != generation) {
1117 return ERR_PTR(-ESTALE);
1123 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1124 int fh_len, int fh_type)
1126 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1127 ext4_nfs_get_inode);
1130 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1131 int fh_len, int fh_type)
1133 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1134 ext4_nfs_get_inode);
1138 * Try to release metadata pages (indirect blocks, directories) which are
1139 * mapped via the block device. Since these pages could have journal heads
1140 * which would prevent try_to_free_buffers() from freeing them, we must use
1141 * jbd2 layer's try_to_free_buffers() function to release them.
1143 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1146 journal_t *journal = EXT4_SB(sb)->s_journal;
1148 WARN_ON(PageChecked(page));
1149 if (!page_has_buffers(page))
1152 return jbd2_journal_try_to_free_buffers(journal, page,
1153 wait & ~__GFP_WAIT);
1154 return try_to_free_buffers(page);
1158 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1159 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1161 static int ext4_write_dquot(struct dquot *dquot);
1162 static int ext4_acquire_dquot(struct dquot *dquot);
1163 static int ext4_release_dquot(struct dquot *dquot);
1164 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1165 static int ext4_write_info(struct super_block *sb, int type);
1166 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1168 static int ext4_quota_off(struct super_block *sb, int type);
1169 static int ext4_quota_on_mount(struct super_block *sb, int type);
1170 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1171 size_t len, loff_t off);
1172 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1173 const char *data, size_t len, loff_t off);
1175 static const struct dquot_operations ext4_quota_operations = {
1177 .get_reserved_space = ext4_get_reserved_space,
1179 .write_dquot = ext4_write_dquot,
1180 .acquire_dquot = ext4_acquire_dquot,
1181 .release_dquot = ext4_release_dquot,
1182 .mark_dirty = ext4_mark_dquot_dirty,
1183 .write_info = ext4_write_info,
1184 .alloc_dquot = dquot_alloc,
1185 .destroy_dquot = dquot_destroy,
1188 static const struct quotactl_ops ext4_qctl_operations = {
1189 .quota_on = ext4_quota_on,
1190 .quota_off = ext4_quota_off,
1191 .quota_sync = dquot_quota_sync,
1192 .get_info = dquot_get_dqinfo,
1193 .set_info = dquot_set_dqinfo,
1194 .get_dqblk = dquot_get_dqblk,
1195 .set_dqblk = dquot_set_dqblk
1199 static const struct super_operations ext4_sops = {
1200 .alloc_inode = ext4_alloc_inode,
1201 .destroy_inode = ext4_destroy_inode,
1202 .write_inode = ext4_write_inode,
1203 .dirty_inode = ext4_dirty_inode,
1204 .drop_inode = ext4_drop_inode,
1205 .evict_inode = ext4_evict_inode,
1206 .put_super = ext4_put_super,
1207 .sync_fs = ext4_sync_fs,
1208 .freeze_fs = ext4_freeze,
1209 .unfreeze_fs = ext4_unfreeze,
1210 .statfs = ext4_statfs,
1211 .remount_fs = ext4_remount,
1212 .show_options = ext4_show_options,
1214 .quota_read = ext4_quota_read,
1215 .quota_write = ext4_quota_write,
1217 .bdev_try_to_free_page = bdev_try_to_free_page,
1220 static const struct super_operations ext4_nojournal_sops = {
1221 .alloc_inode = ext4_alloc_inode,
1222 .destroy_inode = ext4_destroy_inode,
1223 .write_inode = ext4_write_inode,
1224 .dirty_inode = ext4_dirty_inode,
1225 .drop_inode = ext4_drop_inode,
1226 .evict_inode = ext4_evict_inode,
1227 .write_super = ext4_write_super,
1228 .put_super = ext4_put_super,
1229 .statfs = ext4_statfs,
1230 .remount_fs = ext4_remount,
1231 .show_options = ext4_show_options,
1233 .quota_read = ext4_quota_read,
1234 .quota_write = ext4_quota_write,
1236 .bdev_try_to_free_page = bdev_try_to_free_page,
1239 static const struct export_operations ext4_export_ops = {
1240 .fh_to_dentry = ext4_fh_to_dentry,
1241 .fh_to_parent = ext4_fh_to_parent,
1242 .get_parent = ext4_get_parent,
1246 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1247 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1248 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1249 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1250 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1251 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1252 Opt_journal_update, Opt_journal_dev,
1253 Opt_journal_checksum, Opt_journal_async_commit,
1254 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1255 Opt_data_err_abort, Opt_data_err_ignore,
1256 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1257 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1258 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1259 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1260 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1261 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1262 Opt_inode_readahead_blks, Opt_journal_ioprio,
1263 Opt_dioread_nolock, Opt_dioread_lock,
1264 Opt_discard, Opt_nodiscard,
1265 Opt_init_inode_table, Opt_noinit_inode_table,
1268 static const match_table_t tokens = {
1269 {Opt_bsd_df, "bsddf"},
1270 {Opt_minix_df, "minixdf"},
1271 {Opt_grpid, "grpid"},
1272 {Opt_grpid, "bsdgroups"},
1273 {Opt_nogrpid, "nogrpid"},
1274 {Opt_nogrpid, "sysvgroups"},
1275 {Opt_resgid, "resgid=%u"},
1276 {Opt_resuid, "resuid=%u"},
1278 {Opt_err_cont, "errors=continue"},
1279 {Opt_err_panic, "errors=panic"},
1280 {Opt_err_ro, "errors=remount-ro"},
1281 {Opt_nouid32, "nouid32"},
1282 {Opt_debug, "debug"},
1283 {Opt_oldalloc, "oldalloc"},
1284 {Opt_orlov, "orlov"},
1285 {Opt_user_xattr, "user_xattr"},
1286 {Opt_nouser_xattr, "nouser_xattr"},
1288 {Opt_noacl, "noacl"},
1289 {Opt_noload, "noload"},
1290 {Opt_noload, "norecovery"},
1293 {Opt_commit, "commit=%u"},
1294 {Opt_min_batch_time, "min_batch_time=%u"},
1295 {Opt_max_batch_time, "max_batch_time=%u"},
1296 {Opt_journal_update, "journal=update"},
1297 {Opt_journal_dev, "journal_dev=%u"},
1298 {Opt_journal_checksum, "journal_checksum"},
1299 {Opt_journal_async_commit, "journal_async_commit"},
1300 {Opt_abort, "abort"},
1301 {Opt_data_journal, "data=journal"},
1302 {Opt_data_ordered, "data=ordered"},
1303 {Opt_data_writeback, "data=writeback"},
1304 {Opt_data_err_abort, "data_err=abort"},
1305 {Opt_data_err_ignore, "data_err=ignore"},
1306 {Opt_offusrjquota, "usrjquota="},
1307 {Opt_usrjquota, "usrjquota=%s"},
1308 {Opt_offgrpjquota, "grpjquota="},
1309 {Opt_grpjquota, "grpjquota=%s"},
1310 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1311 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1312 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1313 {Opt_grpquota, "grpquota"},
1314 {Opt_noquota, "noquota"},
1315 {Opt_quota, "quota"},
1316 {Opt_usrquota, "usrquota"},
1317 {Opt_barrier, "barrier=%u"},
1318 {Opt_barrier, "barrier"},
1319 {Opt_nobarrier, "nobarrier"},
1320 {Opt_i_version, "i_version"},
1321 {Opt_stripe, "stripe=%u"},
1322 {Opt_resize, "resize"},
1323 {Opt_delalloc, "delalloc"},
1324 {Opt_nodelalloc, "nodelalloc"},
1325 {Opt_mblk_io_submit, "mblk_io_submit"},
1326 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1327 {Opt_block_validity, "block_validity"},
1328 {Opt_noblock_validity, "noblock_validity"},
1329 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1330 {Opt_journal_ioprio, "journal_ioprio=%u"},
1331 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1332 {Opt_auto_da_alloc, "auto_da_alloc"},
1333 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1334 {Opt_dioread_nolock, "dioread_nolock"},
1335 {Opt_dioread_lock, "dioread_lock"},
1336 {Opt_discard, "discard"},
1337 {Opt_nodiscard, "nodiscard"},
1338 {Opt_init_inode_table, "init_itable=%u"},
1339 {Opt_init_inode_table, "init_itable"},
1340 {Opt_noinit_inode_table, "noinit_itable"},
1344 static ext4_fsblk_t get_sb_block(void **data)
1346 ext4_fsblk_t sb_block;
1347 char *options = (char *) *data;
1349 if (!options || strncmp(options, "sb=", 3) != 0)
1350 return 1; /* Default location */
1353 /* TODO: use simple_strtoll with >32bit ext4 */
1354 sb_block = simple_strtoul(options, &options, 0);
1355 if (*options && *options != ',') {
1356 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1360 if (*options == ',')
1362 *data = (void *) options;
1367 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1368 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1369 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1372 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1374 struct ext4_sb_info *sbi = EXT4_SB(sb);
1377 if (sb_any_quota_loaded(sb) &&
1378 !sbi->s_qf_names[qtype]) {
1379 ext4_msg(sb, KERN_ERR,
1380 "Cannot change journaled "
1381 "quota options when quota turned on");
1384 qname = match_strdup(args);
1386 ext4_msg(sb, KERN_ERR,
1387 "Not enough memory for storing quotafile name");
1390 if (sbi->s_qf_names[qtype] &&
1391 strcmp(sbi->s_qf_names[qtype], qname)) {
1392 ext4_msg(sb, KERN_ERR,
1393 "%s quota file already specified", QTYPE2NAME(qtype));
1397 sbi->s_qf_names[qtype] = qname;
1398 if (strchr(sbi->s_qf_names[qtype], '/')) {
1399 ext4_msg(sb, KERN_ERR,
1400 "quotafile must be on filesystem root");
1401 kfree(sbi->s_qf_names[qtype]);
1402 sbi->s_qf_names[qtype] = NULL;
1409 static int clear_qf_name(struct super_block *sb, int qtype)
1412 struct ext4_sb_info *sbi = EXT4_SB(sb);
1414 if (sb_any_quota_loaded(sb) &&
1415 sbi->s_qf_names[qtype]) {
1416 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1417 " when quota turned on");
1421 * The space will be released later when all options are confirmed
1424 sbi->s_qf_names[qtype] = NULL;
1429 static int parse_options(char *options, struct super_block *sb,
1430 unsigned long *journal_devnum,
1431 unsigned int *journal_ioprio,
1432 ext4_fsblk_t *n_blocks_count, int is_remount)
1434 struct ext4_sb_info *sbi = EXT4_SB(sb);
1436 substring_t args[MAX_OPT_ARGS];
1446 while ((p = strsep(&options, ",")) != NULL) {
1452 * Initialize args struct so we know whether arg was
1453 * found; some options take optional arguments.
1455 args[0].to = args[0].from = 0;
1456 token = match_token(p, tokens, args);
1459 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1460 clear_opt(sb, MINIX_DF);
1463 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1464 set_opt(sb, MINIX_DF);
1468 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1473 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1474 clear_opt(sb, GRPID);
1478 if (match_int(&args[0], &option))
1480 sbi->s_resuid = option;
1483 if (match_int(&args[0], &option))
1485 sbi->s_resgid = option;
1488 /* handled by get_sb_block() instead of here */
1489 /* *sb_block = match_int(&args[0]); */
1492 clear_opt(sb, ERRORS_CONT);
1493 clear_opt(sb, ERRORS_RO);
1494 set_opt(sb, ERRORS_PANIC);
1497 clear_opt(sb, ERRORS_CONT);
1498 clear_opt(sb, ERRORS_PANIC);
1499 set_opt(sb, ERRORS_RO);
1502 clear_opt(sb, ERRORS_RO);
1503 clear_opt(sb, ERRORS_PANIC);
1504 set_opt(sb, ERRORS_CONT);
1507 set_opt(sb, NO_UID32);
1513 set_opt(sb, OLDALLOC);
1516 clear_opt(sb, OLDALLOC);
1518 #ifdef CONFIG_EXT4_FS_XATTR
1519 case Opt_user_xattr:
1520 set_opt(sb, XATTR_USER);
1522 case Opt_nouser_xattr:
1523 clear_opt(sb, XATTR_USER);
1526 case Opt_user_xattr:
1527 case Opt_nouser_xattr:
1528 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1531 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1533 set_opt(sb, POSIX_ACL);
1536 clear_opt(sb, POSIX_ACL);
1541 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1544 case Opt_journal_update:
1546 /* Eventually we will want to be able to create
1547 a journal file here. For now, only allow the
1548 user to specify an existing inode to be the
1551 ext4_msg(sb, KERN_ERR,
1552 "Cannot specify journal on remount");
1555 set_opt(sb, UPDATE_JOURNAL);
1557 case Opt_journal_dev:
1559 ext4_msg(sb, KERN_ERR,
1560 "Cannot specify journal on remount");
1563 if (match_int(&args[0], &option))
1565 *journal_devnum = option;
1567 case Opt_journal_checksum:
1568 set_opt(sb, JOURNAL_CHECKSUM);
1570 case Opt_journal_async_commit:
1571 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1572 set_opt(sb, JOURNAL_CHECKSUM);
1575 set_opt(sb, NOLOAD);
1578 if (match_int(&args[0], &option))
1583 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1584 sbi->s_commit_interval = HZ * option;
1586 case Opt_max_batch_time:
1587 if (match_int(&args[0], &option))
1592 option = EXT4_DEF_MAX_BATCH_TIME;
1593 sbi->s_max_batch_time = option;
1595 case Opt_min_batch_time:
1596 if (match_int(&args[0], &option))
1600 sbi->s_min_batch_time = option;
1602 case Opt_data_journal:
1603 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1605 case Opt_data_ordered:
1606 data_opt = EXT4_MOUNT_ORDERED_DATA;
1608 case Opt_data_writeback:
1609 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1612 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1613 ext4_msg(sb, KERN_ERR,
1614 "Cannot change data mode on remount");
1618 clear_opt(sb, DATA_FLAGS);
1619 sbi->s_mount_opt |= data_opt;
1622 case Opt_data_err_abort:
1623 set_opt(sb, DATA_ERR_ABORT);
1625 case Opt_data_err_ignore:
1626 clear_opt(sb, DATA_ERR_ABORT);
1630 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1634 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1637 case Opt_offusrjquota:
1638 if (!clear_qf_name(sb, USRQUOTA))
1641 case Opt_offgrpjquota:
1642 if (!clear_qf_name(sb, GRPQUOTA))
1646 case Opt_jqfmt_vfsold:
1647 qfmt = QFMT_VFS_OLD;
1649 case Opt_jqfmt_vfsv0:
1652 case Opt_jqfmt_vfsv1:
1655 if (sb_any_quota_loaded(sb) &&
1656 sbi->s_jquota_fmt != qfmt) {
1657 ext4_msg(sb, KERN_ERR, "Cannot change "
1658 "journaled quota options when "
1662 sbi->s_jquota_fmt = qfmt;
1667 set_opt(sb, USRQUOTA);
1671 set_opt(sb, GRPQUOTA);
1674 if (sb_any_quota_loaded(sb)) {
1675 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1676 "options when quota turned on");
1679 clear_opt(sb, QUOTA);
1680 clear_opt(sb, USRQUOTA);
1681 clear_opt(sb, GRPQUOTA);
1687 ext4_msg(sb, KERN_ERR,
1688 "quota options not supported");
1692 case Opt_offusrjquota:
1693 case Opt_offgrpjquota:
1694 case Opt_jqfmt_vfsold:
1695 case Opt_jqfmt_vfsv0:
1696 case Opt_jqfmt_vfsv1:
1697 ext4_msg(sb, KERN_ERR,
1698 "journaled quota options not supported");
1704 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1707 clear_opt(sb, BARRIER);
1711 if (match_int(&args[0], &option))
1714 option = 1; /* No argument, default to 1 */
1716 set_opt(sb, BARRIER);
1718 clear_opt(sb, BARRIER);
1724 ext4_msg(sb, KERN_ERR,
1725 "resize option only available "
1729 if (match_int(&args[0], &option) != 0)
1731 *n_blocks_count = option;
1734 ext4_msg(sb, KERN_WARNING,
1735 "Ignoring deprecated nobh option");
1738 ext4_msg(sb, KERN_WARNING,
1739 "Ignoring deprecated bh option");
1742 set_opt(sb, I_VERSION);
1743 sb->s_flags |= MS_I_VERSION;
1745 case Opt_nodelalloc:
1746 clear_opt(sb, DELALLOC);
1748 case Opt_mblk_io_submit:
1749 set_opt(sb, MBLK_IO_SUBMIT);
1751 case Opt_nomblk_io_submit:
1752 clear_opt(sb, MBLK_IO_SUBMIT);
1755 if (match_int(&args[0], &option))
1759 sbi->s_stripe = option;
1762 set_opt(sb, DELALLOC);
1764 case Opt_block_validity:
1765 set_opt(sb, BLOCK_VALIDITY);
1767 case Opt_noblock_validity:
1768 clear_opt(sb, BLOCK_VALIDITY);
1770 case Opt_inode_readahead_blks:
1771 if (match_int(&args[0], &option))
1773 if (option < 0 || option > (1 << 30))
1775 if (!is_power_of_2(option)) {
1776 ext4_msg(sb, KERN_ERR,
1777 "EXT4-fs: inode_readahead_blks"
1778 " must be a power of 2");
1781 sbi->s_inode_readahead_blks = option;
1783 case Opt_journal_ioprio:
1784 if (match_int(&args[0], &option))
1786 if (option < 0 || option > 7)
1788 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1791 case Opt_noauto_da_alloc:
1792 set_opt(sb, NO_AUTO_DA_ALLOC);
1794 case Opt_auto_da_alloc:
1796 if (match_int(&args[0], &option))
1799 option = 1; /* No argument, default to 1 */
1801 clear_opt(sb, NO_AUTO_DA_ALLOC);
1803 set_opt(sb,NO_AUTO_DA_ALLOC);
1806 set_opt(sb, DISCARD);
1809 clear_opt(sb, DISCARD);
1811 case Opt_dioread_nolock:
1812 set_opt(sb, DIOREAD_NOLOCK);
1814 case Opt_dioread_lock:
1815 clear_opt(sb, DIOREAD_NOLOCK);
1817 case Opt_init_inode_table:
1818 set_opt(sb, INIT_INODE_TABLE);
1820 if (match_int(&args[0], &option))
1823 option = EXT4_DEF_LI_WAIT_MULT;
1826 sbi->s_li_wait_mult = option;
1828 case Opt_noinit_inode_table:
1829 clear_opt(sb, INIT_INODE_TABLE);
1832 ext4_msg(sb, KERN_ERR,
1833 "Unrecognized mount option \"%s\" "
1834 "or missing value", p);
1839 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1840 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1841 clear_opt(sb, USRQUOTA);
1843 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1844 clear_opt(sb, GRPQUOTA);
1846 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1847 ext4_msg(sb, KERN_ERR, "old and new quota "
1852 if (!sbi->s_jquota_fmt) {
1853 ext4_msg(sb, KERN_ERR, "journaled quota format "
1858 if (sbi->s_jquota_fmt) {
1859 ext4_msg(sb, KERN_ERR, "journaled quota format "
1860 "specified with no journaling "
1869 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1872 struct ext4_sb_info *sbi = EXT4_SB(sb);
1875 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1876 ext4_msg(sb, KERN_ERR, "revision level too high, "
1877 "forcing read-only mode");
1882 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1883 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1884 "running e2fsck is recommended");
1885 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1886 ext4_msg(sb, KERN_WARNING,
1887 "warning: mounting fs with errors, "
1888 "running e2fsck is recommended");
1889 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1890 le16_to_cpu(es->s_mnt_count) >=
1891 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1892 ext4_msg(sb, KERN_WARNING,
1893 "warning: maximal mount count reached, "
1894 "running e2fsck is recommended");
1895 else if (le32_to_cpu(es->s_checkinterval) &&
1896 (le32_to_cpu(es->s_lastcheck) +
1897 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1898 ext4_msg(sb, KERN_WARNING,
1899 "warning: checktime reached, "
1900 "running e2fsck is recommended");
1901 if (!sbi->s_journal)
1902 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1903 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1904 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1905 le16_add_cpu(&es->s_mnt_count, 1);
1906 es->s_mtime = cpu_to_le32(get_seconds());
1907 ext4_update_dynamic_rev(sb);
1909 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1911 ext4_commit_super(sb, 1);
1912 if (test_opt(sb, DEBUG))
1913 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1916 sbi->s_groups_count,
1917 EXT4_BLOCKS_PER_GROUP(sb),
1918 EXT4_INODES_PER_GROUP(sb),
1919 sbi->s_mount_opt, sbi->s_mount_opt2);
1924 static int ext4_fill_flex_info(struct super_block *sb)
1926 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 struct ext4_group_desc *gdp = NULL;
1928 ext4_group_t flex_group_count;
1929 ext4_group_t flex_group;
1930 int groups_per_flex = 0;
1934 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1935 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1937 if (groups_per_flex < 2) {
1938 sbi->s_log_groups_per_flex = 0;
1942 /* We allocate both existing and potentially added groups */
1943 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1944 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1945 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1946 size = flex_group_count * sizeof(struct flex_groups);
1947 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1948 if (sbi->s_flex_groups == NULL) {
1949 sbi->s_flex_groups = vzalloc(size);
1950 if (sbi->s_flex_groups == NULL) {
1951 ext4_msg(sb, KERN_ERR,
1952 "not enough memory for %u flex groups",
1958 for (i = 0; i < sbi->s_groups_count; i++) {
1959 gdp = ext4_get_group_desc(sb, i, NULL);
1961 flex_group = ext4_flex_group(sbi, i);
1962 atomic_add(ext4_free_inodes_count(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].free_inodes);
1964 atomic_add(ext4_free_blks_count(sb, gdp),
1965 &sbi->s_flex_groups[flex_group].free_blocks);
1966 atomic_add(ext4_used_dirs_count(sb, gdp),
1967 &sbi->s_flex_groups[flex_group].used_dirs);
1975 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1976 struct ext4_group_desc *gdp)
1980 if (sbi->s_es->s_feature_ro_compat &
1981 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1982 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1983 __le32 le_group = cpu_to_le32(block_group);
1985 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1986 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1987 crc = crc16(crc, (__u8 *)gdp, offset);
1988 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1989 /* for checksum of struct ext4_group_desc do the rest...*/
1990 if ((sbi->s_es->s_feature_incompat &
1991 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1992 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1993 crc = crc16(crc, (__u8 *)gdp + offset,
1994 le16_to_cpu(sbi->s_es->s_desc_size) -
1998 return cpu_to_le16(crc);
2001 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2002 struct ext4_group_desc *gdp)
2004 if ((sbi->s_es->s_feature_ro_compat &
2005 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2006 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2012 /* Called at mount-time, super-block is locked */
2013 static int ext4_check_descriptors(struct super_block *sb,
2014 ext4_group_t *first_not_zeroed)
2016 struct ext4_sb_info *sbi = EXT4_SB(sb);
2017 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2018 ext4_fsblk_t last_block;
2019 ext4_fsblk_t block_bitmap;
2020 ext4_fsblk_t inode_bitmap;
2021 ext4_fsblk_t inode_table;
2022 int flexbg_flag = 0;
2023 ext4_group_t i, grp = sbi->s_groups_count;
2025 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2028 ext4_debug("Checking group descriptors");
2030 for (i = 0; i < sbi->s_groups_count; i++) {
2031 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2033 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2034 last_block = ext4_blocks_count(sbi->s_es) - 1;
2036 last_block = first_block +
2037 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2039 if ((grp == sbi->s_groups_count) &&
2040 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2043 block_bitmap = ext4_block_bitmap(sb, gdp);
2044 if (block_bitmap < first_block || block_bitmap > last_block) {
2045 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2046 "Block bitmap for group %u not in group "
2047 "(block %llu)!", i, block_bitmap);
2050 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2051 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2052 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2053 "Inode bitmap for group %u not in group "
2054 "(block %llu)!", i, inode_bitmap);
2057 inode_table = ext4_inode_table(sb, gdp);
2058 if (inode_table < first_block ||
2059 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2060 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2061 "Inode table for group %u not in group "
2062 "(block %llu)!", i, inode_table);
2065 ext4_lock_group(sb, i);
2066 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2067 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2068 "Checksum for group %u failed (%u!=%u)",
2069 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2070 gdp)), le16_to_cpu(gdp->bg_checksum));
2071 if (!(sb->s_flags & MS_RDONLY)) {
2072 ext4_unlock_group(sb, i);
2076 ext4_unlock_group(sb, i);
2078 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2080 if (NULL != first_not_zeroed)
2081 *first_not_zeroed = grp;
2083 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2084 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2088 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2089 * the superblock) which were deleted from all directories, but held open by
2090 * a process at the time of a crash. We walk the list and try to delete these
2091 * inodes at recovery time (only with a read-write filesystem).
2093 * In order to keep the orphan inode chain consistent during traversal (in
2094 * case of crash during recovery), we link each inode into the superblock
2095 * orphan list_head and handle it the same way as an inode deletion during
2096 * normal operation (which journals the operations for us).
2098 * We only do an iget() and an iput() on each inode, which is very safe if we
2099 * accidentally point at an in-use or already deleted inode. The worst that
2100 * can happen in this case is that we get a "bit already cleared" message from
2101 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2102 * e2fsck was run on this filesystem, and it must have already done the orphan
2103 * inode cleanup for us, so we can safely abort without any further action.
2105 static void ext4_orphan_cleanup(struct super_block *sb,
2106 struct ext4_super_block *es)
2108 unsigned int s_flags = sb->s_flags;
2109 int nr_orphans = 0, nr_truncates = 0;
2113 if (!es->s_last_orphan) {
2114 jbd_debug(4, "no orphan inodes to clean up\n");
2118 if (bdev_read_only(sb->s_bdev)) {
2119 ext4_msg(sb, KERN_ERR, "write access "
2120 "unavailable, skipping orphan cleanup");
2124 /* Check if feature set would not allow a r/w mount */
2125 if (!ext4_feature_set_ok(sb, 0)) {
2126 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2127 "unknown ROCOMPAT features");
2131 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2132 if (es->s_last_orphan)
2133 jbd_debug(1, "Errors on filesystem, "
2134 "clearing orphan list.\n");
2135 es->s_last_orphan = 0;
2136 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2140 if (s_flags & MS_RDONLY) {
2141 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2142 sb->s_flags &= ~MS_RDONLY;
2145 /* Needed for iput() to work correctly and not trash data */
2146 sb->s_flags |= MS_ACTIVE;
2147 /* Turn on quotas so that they are updated correctly */
2148 for (i = 0; i < MAXQUOTAS; i++) {
2149 if (EXT4_SB(sb)->s_qf_names[i]) {
2150 int ret = ext4_quota_on_mount(sb, i);
2152 ext4_msg(sb, KERN_ERR,
2153 "Cannot turn on journaled "
2154 "quota: error %d", ret);
2159 while (es->s_last_orphan) {
2160 struct inode *inode;
2162 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2163 if (IS_ERR(inode)) {
2164 es->s_last_orphan = 0;
2168 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2169 dquot_initialize(inode);
2170 if (inode->i_nlink) {
2171 ext4_msg(sb, KERN_DEBUG,
2172 "%s: truncating inode %lu to %lld bytes",
2173 __func__, inode->i_ino, inode->i_size);
2174 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2175 inode->i_ino, inode->i_size);
2176 ext4_truncate(inode);
2179 ext4_msg(sb, KERN_DEBUG,
2180 "%s: deleting unreferenced inode %lu",
2181 __func__, inode->i_ino);
2182 jbd_debug(2, "deleting unreferenced inode %lu\n",
2186 iput(inode); /* The delete magic happens here! */
2189 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2192 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2193 PLURAL(nr_orphans));
2195 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2196 PLURAL(nr_truncates));
2198 /* Turn quotas off */
2199 for (i = 0; i < MAXQUOTAS; i++) {
2200 if (sb_dqopt(sb)->files[i])
2201 dquot_quota_off(sb, i);
2204 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2208 * Maximal extent format file size.
2209 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2210 * extent format containers, within a sector_t, and within i_blocks
2211 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2212 * so that won't be a limiting factor.
2214 * However there is other limiting factor. We do store extents in the form
2215 * of starting block and length, hence the resulting length of the extent
2216 * covering maximum file size must fit into on-disk format containers as
2217 * well. Given that length is always by 1 unit bigger than max unit (because
2218 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2220 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2222 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2225 loff_t upper_limit = MAX_LFS_FILESIZE;
2227 /* small i_blocks in vfs inode? */
2228 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2230 * CONFIG_LBDAF is not enabled implies the inode
2231 * i_block represent total blocks in 512 bytes
2232 * 32 == size of vfs inode i_blocks * 8
2234 upper_limit = (1LL << 32) - 1;
2236 /* total blocks in file system block size */
2237 upper_limit >>= (blkbits - 9);
2238 upper_limit <<= blkbits;
2242 * 32-bit extent-start container, ee_block. We lower the maxbytes
2243 * by one fs block, so ee_len can cover the extent of maximum file
2246 res = (1LL << 32) - 1;
2249 /* Sanity check against vm- & vfs- imposed limits */
2250 if (res > upper_limit)
2257 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2258 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2259 * We need to be 1 filesystem block less than the 2^48 sector limit.
2261 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2263 loff_t res = EXT4_NDIR_BLOCKS;
2266 /* This is calculated to be the largest file size for a dense, block
2267 * mapped file such that the file's total number of 512-byte sectors,
2268 * including data and all indirect blocks, does not exceed (2^48 - 1).
2270 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2271 * number of 512-byte sectors of the file.
2274 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2276 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2277 * the inode i_block field represents total file blocks in
2278 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2280 upper_limit = (1LL << 32) - 1;
2282 /* total blocks in file system block size */
2283 upper_limit >>= (bits - 9);
2287 * We use 48 bit ext4_inode i_blocks
2288 * With EXT4_HUGE_FILE_FL set the i_blocks
2289 * represent total number of blocks in
2290 * file system block size
2292 upper_limit = (1LL << 48) - 1;
2296 /* indirect blocks */
2298 /* double indirect blocks */
2299 meta_blocks += 1 + (1LL << (bits-2));
2300 /* tripple indirect blocks */
2301 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2303 upper_limit -= meta_blocks;
2304 upper_limit <<= bits;
2306 res += 1LL << (bits-2);
2307 res += 1LL << (2*(bits-2));
2308 res += 1LL << (3*(bits-2));
2310 if (res > upper_limit)
2313 if (res > MAX_LFS_FILESIZE)
2314 res = MAX_LFS_FILESIZE;
2319 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2320 ext4_fsblk_t logical_sb_block, int nr)
2322 struct ext4_sb_info *sbi = EXT4_SB(sb);
2323 ext4_group_t bg, first_meta_bg;
2326 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2328 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2330 return logical_sb_block + nr + 1;
2331 bg = sbi->s_desc_per_block * nr;
2332 if (ext4_bg_has_super(sb, bg))
2335 return (has_super + ext4_group_first_block_no(sb, bg));
2339 * ext4_get_stripe_size: Get the stripe size.
2340 * @sbi: In memory super block info
2342 * If we have specified it via mount option, then
2343 * use the mount option value. If the value specified at mount time is
2344 * greater than the blocks per group use the super block value.
2345 * If the super block value is greater than blocks per group return 0.
2346 * Allocator needs it be less than blocks per group.
2349 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2351 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2352 unsigned long stripe_width =
2353 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2355 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2356 return sbi->s_stripe;
2358 if (stripe_width <= sbi->s_blocks_per_group)
2359 return stripe_width;
2361 if (stride <= sbi->s_blocks_per_group)
2370 struct attribute attr;
2371 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2372 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2373 const char *, size_t);
2377 static int parse_strtoul(const char *buf,
2378 unsigned long max, unsigned long *value)
2382 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2383 endp = skip_spaces(endp);
2384 if (*endp || *value > max)
2390 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2391 struct ext4_sb_info *sbi,
2394 return snprintf(buf, PAGE_SIZE, "%llu\n",
2395 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2398 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2399 struct ext4_sb_info *sbi, char *buf)
2401 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2403 if (!sb->s_bdev->bd_part)
2404 return snprintf(buf, PAGE_SIZE, "0\n");
2405 return snprintf(buf, PAGE_SIZE, "%lu\n",
2406 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2407 sbi->s_sectors_written_start) >> 1);
2410 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2411 struct ext4_sb_info *sbi, char *buf)
2413 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2415 if (!sb->s_bdev->bd_part)
2416 return snprintf(buf, PAGE_SIZE, "0\n");
2417 return snprintf(buf, PAGE_SIZE, "%llu\n",
2418 (unsigned long long)(sbi->s_kbytes_written +
2419 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2420 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2423 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2424 struct ext4_sb_info *sbi,
2425 const char *buf, size_t count)
2429 if (parse_strtoul(buf, 0x40000000, &t))
2432 if (!is_power_of_2(t))
2435 sbi->s_inode_readahead_blks = t;
2439 static ssize_t sbi_ui_show(struct ext4_attr *a,
2440 struct ext4_sb_info *sbi, char *buf)
2442 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2444 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2447 static ssize_t sbi_ui_store(struct ext4_attr *a,
2448 struct ext4_sb_info *sbi,
2449 const char *buf, size_t count)
2451 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2454 if (parse_strtoul(buf, 0xffffffff, &t))
2460 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2461 static struct ext4_attr ext4_attr_##_name = { \
2462 .attr = {.name = __stringify(_name), .mode = _mode }, \
2465 .offset = offsetof(struct ext4_sb_info, _elname), \
2467 #define EXT4_ATTR(name, mode, show, store) \
2468 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2470 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2471 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2472 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2473 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2474 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2475 #define ATTR_LIST(name) &ext4_attr_##name.attr
2477 EXT4_RO_ATTR(delayed_allocation_blocks);
2478 EXT4_RO_ATTR(session_write_kbytes);
2479 EXT4_RO_ATTR(lifetime_write_kbytes);
2480 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2481 inode_readahead_blks_store, s_inode_readahead_blks);
2482 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2483 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2484 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2485 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2486 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2487 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2488 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2489 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2491 static struct attribute *ext4_attrs[] = {
2492 ATTR_LIST(delayed_allocation_blocks),
2493 ATTR_LIST(session_write_kbytes),
2494 ATTR_LIST(lifetime_write_kbytes),
2495 ATTR_LIST(inode_readahead_blks),
2496 ATTR_LIST(inode_goal),
2497 ATTR_LIST(mb_stats),
2498 ATTR_LIST(mb_max_to_scan),
2499 ATTR_LIST(mb_min_to_scan),
2500 ATTR_LIST(mb_order2_req),
2501 ATTR_LIST(mb_stream_req),
2502 ATTR_LIST(mb_group_prealloc),
2503 ATTR_LIST(max_writeback_mb_bump),
2507 /* Features this copy of ext4 supports */
2508 EXT4_INFO_ATTR(lazy_itable_init);
2509 EXT4_INFO_ATTR(batched_discard);
2511 static struct attribute *ext4_feat_attrs[] = {
2512 ATTR_LIST(lazy_itable_init),
2513 ATTR_LIST(batched_discard),
2517 static ssize_t ext4_attr_show(struct kobject *kobj,
2518 struct attribute *attr, char *buf)
2520 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2522 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2524 return a->show ? a->show(a, sbi, buf) : 0;
2527 static ssize_t ext4_attr_store(struct kobject *kobj,
2528 struct attribute *attr,
2529 const char *buf, size_t len)
2531 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2533 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2535 return a->store ? a->store(a, sbi, buf, len) : 0;
2538 static void ext4_sb_release(struct kobject *kobj)
2540 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2542 complete(&sbi->s_kobj_unregister);
2545 static const struct sysfs_ops ext4_attr_ops = {
2546 .show = ext4_attr_show,
2547 .store = ext4_attr_store,
2550 static struct kobj_type ext4_ktype = {
2551 .default_attrs = ext4_attrs,
2552 .sysfs_ops = &ext4_attr_ops,
2553 .release = ext4_sb_release,
2556 static void ext4_feat_release(struct kobject *kobj)
2558 complete(&ext4_feat->f_kobj_unregister);
2561 static struct kobj_type ext4_feat_ktype = {
2562 .default_attrs = ext4_feat_attrs,
2563 .sysfs_ops = &ext4_attr_ops,
2564 .release = ext4_feat_release,
2568 * Check whether this filesystem can be mounted based on
2569 * the features present and the RDONLY/RDWR mount requested.
2570 * Returns 1 if this filesystem can be mounted as requested,
2571 * 0 if it cannot be.
2573 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2575 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2576 ext4_msg(sb, KERN_ERR,
2577 "Couldn't mount because of "
2578 "unsupported optional features (%x)",
2579 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2580 ~EXT4_FEATURE_INCOMPAT_SUPP));
2587 /* Check that feature set is OK for a read-write mount */
2588 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2589 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2590 "unsupported optional features (%x)",
2591 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2592 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2596 * Large file size enabled file system can only be mounted
2597 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2599 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2600 if (sizeof(blkcnt_t) < sizeof(u64)) {
2601 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2602 "cannot be mounted RDWR without "
2611 * This function is called once a day if we have errors logged
2612 * on the file system
2614 static void print_daily_error_info(unsigned long arg)
2616 struct super_block *sb = (struct super_block *) arg;
2617 struct ext4_sb_info *sbi;
2618 struct ext4_super_block *es;
2623 if (es->s_error_count)
2624 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2625 le32_to_cpu(es->s_error_count));
2626 if (es->s_first_error_time) {
2627 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2628 sb->s_id, le32_to_cpu(es->s_first_error_time),
2629 (int) sizeof(es->s_first_error_func),
2630 es->s_first_error_func,
2631 le32_to_cpu(es->s_first_error_line));
2632 if (es->s_first_error_ino)
2633 printk(": inode %u",
2634 le32_to_cpu(es->s_first_error_ino));
2635 if (es->s_first_error_block)
2636 printk(": block %llu", (unsigned long long)
2637 le64_to_cpu(es->s_first_error_block));
2640 if (es->s_last_error_time) {
2641 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2642 sb->s_id, le32_to_cpu(es->s_last_error_time),
2643 (int) sizeof(es->s_last_error_func),
2644 es->s_last_error_func,
2645 le32_to_cpu(es->s_last_error_line));
2646 if (es->s_last_error_ino)
2647 printk(": inode %u",
2648 le32_to_cpu(es->s_last_error_ino));
2649 if (es->s_last_error_block)
2650 printk(": block %llu", (unsigned long long)
2651 le64_to_cpu(es->s_last_error_block));
2654 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2657 /* Find next suitable group and run ext4_init_inode_table */
2658 static int ext4_run_li_request(struct ext4_li_request *elr)
2660 struct ext4_group_desc *gdp = NULL;
2661 ext4_group_t group, ngroups;
2662 struct super_block *sb;
2663 unsigned long timeout = 0;
2667 ngroups = EXT4_SB(sb)->s_groups_count;
2669 for (group = elr->lr_next_group; group < ngroups; group++) {
2670 gdp = ext4_get_group_desc(sb, group, NULL);
2676 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2680 if (group == ngroups)
2685 ret = ext4_init_inode_table(sb, group,
2686 elr->lr_timeout ? 0 : 1);
2687 if (elr->lr_timeout == 0) {
2688 timeout = jiffies - timeout;
2689 if (elr->lr_sbi->s_li_wait_mult)
2690 timeout *= elr->lr_sbi->s_li_wait_mult;
2693 elr->lr_timeout = timeout;
2695 elr->lr_next_sched = jiffies + elr->lr_timeout;
2696 elr->lr_next_group = group + 1;
2703 * Remove lr_request from the list_request and free the
2704 * request structure. Should be called with li_list_mtx held
2706 static void ext4_remove_li_request(struct ext4_li_request *elr)
2708 struct ext4_sb_info *sbi;
2715 list_del(&elr->lr_request);
2716 sbi->s_li_request = NULL;
2720 static void ext4_unregister_li_request(struct super_block *sb)
2722 mutex_lock(&ext4_li_mtx);
2723 if (!ext4_li_info) {
2724 mutex_unlock(&ext4_li_mtx);
2728 mutex_lock(&ext4_li_info->li_list_mtx);
2729 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2730 mutex_unlock(&ext4_li_info->li_list_mtx);
2731 mutex_unlock(&ext4_li_mtx);
2734 static struct task_struct *ext4_lazyinit_task;
2737 * This is the function where ext4lazyinit thread lives. It walks
2738 * through the request list searching for next scheduled filesystem.
2739 * When such a fs is found, run the lazy initialization request
2740 * (ext4_rn_li_request) and keep track of the time spend in this
2741 * function. Based on that time we compute next schedule time of
2742 * the request. When walking through the list is complete, compute
2743 * next waking time and put itself into sleep.
2745 static int ext4_lazyinit_thread(void *arg)
2747 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2748 struct list_head *pos, *n;
2749 struct ext4_li_request *elr;
2750 unsigned long next_wakeup, cur;
2752 BUG_ON(NULL == eli);
2754 eli->li_task = current;
2755 wake_up(&eli->li_wait_task);
2759 next_wakeup = MAX_JIFFY_OFFSET;
2761 mutex_lock(&eli->li_list_mtx);
2762 if (list_empty(&eli->li_request_list)) {
2763 mutex_unlock(&eli->li_list_mtx);
2767 list_for_each_safe(pos, n, &eli->li_request_list) {
2768 elr = list_entry(pos, struct ext4_li_request,
2771 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2772 if (ext4_run_li_request(elr) != 0) {
2773 /* error, remove the lazy_init job */
2774 ext4_remove_li_request(elr);
2779 if (time_before(elr->lr_next_sched, next_wakeup))
2780 next_wakeup = elr->lr_next_sched;
2782 mutex_unlock(&eli->li_list_mtx);
2784 if (freezing(current))
2788 if ((time_after_eq(cur, next_wakeup)) ||
2789 (MAX_JIFFY_OFFSET == next_wakeup)) {
2794 schedule_timeout_interruptible(next_wakeup - cur);
2796 if (kthread_should_stop()) {
2797 ext4_clear_request_list();
2804 * It looks like the request list is empty, but we need
2805 * to check it under the li_list_mtx lock, to prevent any
2806 * additions into it, and of course we should lock ext4_li_mtx
2807 * to atomically free the list and ext4_li_info, because at
2808 * this point another ext4 filesystem could be registering
2811 mutex_lock(&ext4_li_mtx);
2812 mutex_lock(&eli->li_list_mtx);
2813 if (!list_empty(&eli->li_request_list)) {
2814 mutex_unlock(&eli->li_list_mtx);
2815 mutex_unlock(&ext4_li_mtx);
2818 mutex_unlock(&eli->li_list_mtx);
2819 eli->li_task = NULL;
2820 wake_up(&eli->li_wait_task);
2822 kfree(ext4_li_info);
2823 ext4_li_info = NULL;
2824 mutex_unlock(&ext4_li_mtx);
2829 static void ext4_clear_request_list(void)
2831 struct list_head *pos, *n;
2832 struct ext4_li_request *elr;
2834 mutex_lock(&ext4_li_info->li_list_mtx);
2835 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2836 elr = list_entry(pos, struct ext4_li_request,
2838 ext4_remove_li_request(elr);
2840 mutex_unlock(&ext4_li_info->li_list_mtx);
2843 static int ext4_run_lazyinit_thread(void)
2845 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2846 ext4_li_info, "ext4lazyinit");
2847 if (IS_ERR(ext4_lazyinit_task)) {
2848 int err = PTR_ERR(ext4_lazyinit_task);
2849 ext4_clear_request_list();
2850 kfree(ext4_li_info);
2851 ext4_li_info = NULL;
2852 printk(KERN_CRIT "EXT4: error %d creating inode table "
2853 "initialization thread\n",
2857 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2859 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2864 * Check whether it make sense to run itable init. thread or not.
2865 * If there is at least one uninitialized inode table, return
2866 * corresponding group number, else the loop goes through all
2867 * groups and return total number of groups.
2869 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2871 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2872 struct ext4_group_desc *gdp = NULL;
2874 for (group = 0; group < ngroups; group++) {
2875 gdp = ext4_get_group_desc(sb, group, NULL);
2879 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2886 static int ext4_li_info_new(void)
2888 struct ext4_lazy_init *eli = NULL;
2890 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2894 eli->li_task = NULL;
2895 INIT_LIST_HEAD(&eli->li_request_list);
2896 mutex_init(&eli->li_list_mtx);
2898 init_waitqueue_head(&eli->li_wait_task);
2899 eli->li_state |= EXT4_LAZYINIT_QUIT;
2906 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2909 struct ext4_sb_info *sbi = EXT4_SB(sb);
2910 struct ext4_li_request *elr;
2913 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2919 elr->lr_next_group = start;
2922 * Randomize first schedule time of the request to
2923 * spread the inode table initialization requests
2926 get_random_bytes(&rnd, sizeof(rnd));
2927 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2928 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2933 static int ext4_register_li_request(struct super_block *sb,
2934 ext4_group_t first_not_zeroed)
2936 struct ext4_sb_info *sbi = EXT4_SB(sb);
2937 struct ext4_li_request *elr;
2938 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2941 if (sbi->s_li_request != NULL)
2944 if (first_not_zeroed == ngroups ||
2945 (sb->s_flags & MS_RDONLY) ||
2946 !test_opt(sb, INIT_INODE_TABLE)) {
2947 sbi->s_li_request = NULL;
2951 if (first_not_zeroed == ngroups) {
2952 sbi->s_li_request = NULL;
2956 elr = ext4_li_request_new(sb, first_not_zeroed);
2960 mutex_lock(&ext4_li_mtx);
2962 if (NULL == ext4_li_info) {
2963 ret = ext4_li_info_new();
2968 mutex_lock(&ext4_li_info->li_list_mtx);
2969 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2970 mutex_unlock(&ext4_li_info->li_list_mtx);
2972 sbi->s_li_request = elr;
2974 * set elr to NULL here since it has been inserted to
2975 * the request_list and the removal and free of it is
2976 * handled by ext4_clear_request_list from now on.
2980 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2981 ret = ext4_run_lazyinit_thread();
2986 mutex_unlock(&ext4_li_mtx);
2993 * We do not need to lock anything since this is called on
2996 static void ext4_destroy_lazyinit_thread(void)
2999 * If thread exited earlier
3000 * there's nothing to be done.
3002 if (!ext4_li_info || !ext4_lazyinit_task)
3005 kthread_stop(ext4_lazyinit_task);
3008 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3009 __releases(kernel_lock)
3010 __acquires(kernel_lock)
3012 char *orig_data = kstrdup(data, GFP_KERNEL);
3013 struct buffer_head *bh;
3014 struct ext4_super_block *es = NULL;
3015 struct ext4_sb_info *sbi;
3017 ext4_fsblk_t sb_block = get_sb_block(&data);
3018 ext4_fsblk_t logical_sb_block;
3019 unsigned long offset = 0;
3020 unsigned long journal_devnum = 0;
3021 unsigned long def_mount_opts;
3027 unsigned int db_count;
3029 int needs_recovery, has_huge_files;
3032 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3033 ext4_group_t first_not_zeroed;
3035 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3039 sbi->s_blockgroup_lock =
3040 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3041 if (!sbi->s_blockgroup_lock) {
3045 sb->s_fs_info = sbi;
3046 sbi->s_mount_opt = 0;
3047 sbi->s_resuid = EXT4_DEF_RESUID;
3048 sbi->s_resgid = EXT4_DEF_RESGID;
3049 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3050 sbi->s_sb_block = sb_block;
3051 if (sb->s_bdev->bd_part)
3052 sbi->s_sectors_written_start =
3053 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3055 /* Cleanup superblock name */
3056 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3060 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3062 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3067 * The ext4 superblock will not be buffer aligned for other than 1kB
3068 * block sizes. We need to calculate the offset from buffer start.
3070 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3071 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3072 offset = do_div(logical_sb_block, blocksize);
3074 logical_sb_block = sb_block;
3077 if (!(bh = sb_bread(sb, logical_sb_block))) {
3078 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3082 * Note: s_es must be initialized as soon as possible because
3083 * some ext4 macro-instructions depend on its value
3085 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3087 sb->s_magic = le16_to_cpu(es->s_magic);
3088 if (sb->s_magic != EXT4_SUPER_MAGIC)
3090 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3092 /* Set defaults before we parse the mount options */
3093 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3094 set_opt(sb, INIT_INODE_TABLE);
3095 if (def_mount_opts & EXT4_DEFM_DEBUG)
3097 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3098 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3102 if (def_mount_opts & EXT4_DEFM_UID16)
3103 set_opt(sb, NO_UID32);
3104 #ifdef CONFIG_EXT4_FS_XATTR
3105 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3106 set_opt(sb, XATTR_USER);
3108 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3109 if (def_mount_opts & EXT4_DEFM_ACL)
3110 set_opt(sb, POSIX_ACL);
3112 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3113 set_opt(sb, JOURNAL_DATA);
3114 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3115 set_opt(sb, ORDERED_DATA);
3116 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3117 set_opt(sb, WRITEBACK_DATA);
3119 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3120 set_opt(sb, ERRORS_PANIC);
3121 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3122 set_opt(sb, ERRORS_CONT);
3124 set_opt(sb, ERRORS_RO);
3125 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3126 set_opt(sb, BLOCK_VALIDITY);
3127 if (def_mount_opts & EXT4_DEFM_DISCARD)
3128 set_opt(sb, DISCARD);
3130 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3131 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3132 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3133 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3134 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3136 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3137 set_opt(sb, BARRIER);
3140 * enable delayed allocation by default
3141 * Use -o nodelalloc to turn it off
3143 if (!IS_EXT3_SB(sb) &&
3144 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3145 set_opt(sb, DELALLOC);
3147 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3148 &journal_devnum, &journal_ioprio, NULL, 0)) {
3149 ext4_msg(sb, KERN_WARNING,
3150 "failed to parse options in superblock: %s",
3151 sbi->s_es->s_mount_opts);
3153 if (!parse_options((char *) data, sb, &journal_devnum,
3154 &journal_ioprio, NULL, 0))
3157 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3158 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3160 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3161 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3162 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3163 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3164 ext4_msg(sb, KERN_WARNING,
3165 "feature flags set on rev 0 fs, "
3166 "running e2fsck is recommended");
3169 * Check feature flags regardless of the revision level, since we
3170 * previously didn't change the revision level when setting the flags,
3171 * so there is a chance incompat flags are set on a rev 0 filesystem.
3173 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3176 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3178 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3179 blocksize > EXT4_MAX_BLOCK_SIZE) {
3180 ext4_msg(sb, KERN_ERR,
3181 "Unsupported filesystem blocksize %d", blocksize);
3185 if (sb->s_blocksize != blocksize) {
3186 /* Validate the filesystem blocksize */
3187 if (!sb_set_blocksize(sb, blocksize)) {
3188 ext4_msg(sb, KERN_ERR, "bad block size %d",
3194 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3195 offset = do_div(logical_sb_block, blocksize);
3196 bh = sb_bread(sb, logical_sb_block);
3198 ext4_msg(sb, KERN_ERR,
3199 "Can't read superblock on 2nd try");
3202 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3204 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3205 ext4_msg(sb, KERN_ERR,
3206 "Magic mismatch, very weird!");
3211 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3212 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3213 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3215 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3217 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3218 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3219 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3221 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3222 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3223 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3224 (!is_power_of_2(sbi->s_inode_size)) ||
3225 (sbi->s_inode_size > blocksize)) {
3226 ext4_msg(sb, KERN_ERR,
3227 "unsupported inode size: %d",
3231 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3232 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3235 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3236 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3237 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3238 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3239 !is_power_of_2(sbi->s_desc_size)) {
3240 ext4_msg(sb, KERN_ERR,
3241 "unsupported descriptor size %lu",
3246 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3248 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3249 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3250 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3253 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3254 if (sbi->s_inodes_per_block == 0)
3256 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3257 sbi->s_inodes_per_block;
3258 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3260 sbi->s_mount_state = le16_to_cpu(es->s_state);
3261 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3262 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3264 for (i = 0; i < 4; i++)
3265 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3266 sbi->s_def_hash_version = es->s_def_hash_version;
3267 i = le32_to_cpu(es->s_flags);
3268 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3269 sbi->s_hash_unsigned = 3;
3270 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3271 #ifdef __CHAR_UNSIGNED__
3272 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3273 sbi->s_hash_unsigned = 3;
3275 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3280 if (sbi->s_blocks_per_group > blocksize * 8) {
3281 ext4_msg(sb, KERN_ERR,
3282 "#blocks per group too big: %lu",
3283 sbi->s_blocks_per_group);
3286 if (sbi->s_inodes_per_group > blocksize * 8) {
3287 ext4_msg(sb, KERN_ERR,
3288 "#inodes per group too big: %lu",
3289 sbi->s_inodes_per_group);
3294 * Test whether we have more sectors than will fit in sector_t,
3295 * and whether the max offset is addressable by the page cache.
3297 err = generic_check_addressable(sb->s_blocksize_bits,
3298 ext4_blocks_count(es));
3300 ext4_msg(sb, KERN_ERR, "filesystem"
3301 " too large to mount safely on this system");
3302 if (sizeof(sector_t) < 8)
3303 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3308 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3311 /* check blocks count against device size */
3312 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3313 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3314 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3315 "exceeds size of device (%llu blocks)",
3316 ext4_blocks_count(es), blocks_count);
3321 * It makes no sense for the first data block to be beyond the end
3322 * of the filesystem.
3324 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3325 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3326 "block %u is beyond end of filesystem (%llu)",
3327 le32_to_cpu(es->s_first_data_block),
3328 ext4_blocks_count(es));
3331 blocks_count = (ext4_blocks_count(es) -
3332 le32_to_cpu(es->s_first_data_block) +
3333 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3334 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3335 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3336 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3337 "(block count %llu, first data block %u, "
3338 "blocks per group %lu)", sbi->s_groups_count,
3339 ext4_blocks_count(es),
3340 le32_to_cpu(es->s_first_data_block),
3341 EXT4_BLOCKS_PER_GROUP(sb));
3344 sbi->s_groups_count = blocks_count;
3345 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3346 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3347 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3348 EXT4_DESC_PER_BLOCK(sb);
3349 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3351 if (sbi->s_group_desc == NULL) {
3352 ext4_msg(sb, KERN_ERR, "not enough memory");
3356 #ifdef CONFIG_PROC_FS
3358 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3361 bgl_lock_init(sbi->s_blockgroup_lock);
3363 for (i = 0; i < db_count; i++) {
3364 block = descriptor_loc(sb, logical_sb_block, i);
3365 sbi->s_group_desc[i] = sb_bread(sb, block);
3366 if (!sbi->s_group_desc[i]) {
3367 ext4_msg(sb, KERN_ERR,
3368 "can't read group descriptor %d", i);
3373 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3374 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3377 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3378 if (!ext4_fill_flex_info(sb)) {
3379 ext4_msg(sb, KERN_ERR,
3380 "unable to initialize "
3381 "flex_bg meta info!");
3385 sbi->s_gdb_count = db_count;
3386 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3387 spin_lock_init(&sbi->s_next_gen_lock);
3389 init_timer(&sbi->s_err_report);
3390 sbi->s_err_report.function = print_daily_error_info;
3391 sbi->s_err_report.data = (unsigned long) sb;
3393 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3394 ext4_count_free_blocks(sb));
3396 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3397 ext4_count_free_inodes(sb));
3400 err = percpu_counter_init(&sbi->s_dirs_counter,
3401 ext4_count_dirs(sb));
3404 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3407 ext4_msg(sb, KERN_ERR, "insufficient memory");
3411 sbi->s_stripe = ext4_get_stripe_size(sbi);
3412 sbi->s_max_writeback_mb_bump = 128;
3415 * set up enough so that it can read an inode
3417 if (!test_opt(sb, NOLOAD) &&
3418 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3419 sb->s_op = &ext4_sops;
3421 sb->s_op = &ext4_nojournal_sops;
3422 sb->s_export_op = &ext4_export_ops;
3423 sb->s_xattr = ext4_xattr_handlers;
3425 sb->s_qcop = &ext4_qctl_operations;
3426 sb->dq_op = &ext4_quota_operations;
3428 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3429 mutex_init(&sbi->s_orphan_lock);
3430 mutex_init(&sbi->s_resize_lock);
3434 needs_recovery = (es->s_last_orphan != 0 ||
3435 EXT4_HAS_INCOMPAT_FEATURE(sb,
3436 EXT4_FEATURE_INCOMPAT_RECOVER));
3439 * The first inode we look at is the journal inode. Don't try
3440 * root first: it may be modified in the journal!
3442 if (!test_opt(sb, NOLOAD) &&
3443 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3444 if (ext4_load_journal(sb, es, journal_devnum))
3446 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3447 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3448 ext4_msg(sb, KERN_ERR, "required journal recovery "
3449 "suppressed and not mounted read-only");
3450 goto failed_mount_wq;
3452 clear_opt(sb, DATA_FLAGS);
3453 set_opt(sb, WRITEBACK_DATA);
3454 sbi->s_journal = NULL;
3459 if (ext4_blocks_count(es) > 0xffffffffULL &&
3460 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3461 JBD2_FEATURE_INCOMPAT_64BIT)) {
3462 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3463 goto failed_mount_wq;
3466 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3467 jbd2_journal_set_features(sbi->s_journal,
3468 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3469 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3470 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3471 jbd2_journal_set_features(sbi->s_journal,
3472 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3473 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3474 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3476 jbd2_journal_clear_features(sbi->s_journal,
3477 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3478 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3481 /* We have now updated the journal if required, so we can
3482 * validate the data journaling mode. */
3483 switch (test_opt(sb, DATA_FLAGS)) {
3485 /* No mode set, assume a default based on the journal
3486 * capabilities: ORDERED_DATA if the journal can
3487 * cope, else JOURNAL_DATA
3489 if (jbd2_journal_check_available_features
3490 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3491 set_opt(sb, ORDERED_DATA);
3493 set_opt(sb, JOURNAL_DATA);
3496 case EXT4_MOUNT_ORDERED_DATA:
3497 case EXT4_MOUNT_WRITEBACK_DATA:
3498 if (!jbd2_journal_check_available_features
3499 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3500 ext4_msg(sb, KERN_ERR, "Journal does not support "
3501 "requested data journaling mode");
3502 goto failed_mount_wq;
3507 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3510 * The journal may have updated the bg summary counts, so we
3511 * need to update the global counters.
3513 percpu_counter_set(&sbi->s_freeblocks_counter,
3514 ext4_count_free_blocks(sb));
3515 percpu_counter_set(&sbi->s_freeinodes_counter,
3516 ext4_count_free_inodes(sb));
3517 percpu_counter_set(&sbi->s_dirs_counter,
3518 ext4_count_dirs(sb));
3519 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3522 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3523 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3524 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3525 goto failed_mount_wq;
3529 * The jbd2_journal_load will have done any necessary log recovery,
3530 * so we can safely mount the rest of the filesystem now.
3533 root = ext4_iget(sb, EXT4_ROOT_INO);
3535 ext4_msg(sb, KERN_ERR, "get root inode failed");
3536 ret = PTR_ERR(root);
3539 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3541 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3544 sb->s_root = d_alloc_root(root);
3546 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3552 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3554 /* determine the minimum size of new large inodes, if present */
3555 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3556 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3557 EXT4_GOOD_OLD_INODE_SIZE;
3558 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3559 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3560 if (sbi->s_want_extra_isize <
3561 le16_to_cpu(es->s_want_extra_isize))
3562 sbi->s_want_extra_isize =
3563 le16_to_cpu(es->s_want_extra_isize);
3564 if (sbi->s_want_extra_isize <
3565 le16_to_cpu(es->s_min_extra_isize))
3566 sbi->s_want_extra_isize =
3567 le16_to_cpu(es->s_min_extra_isize);
3570 /* Check if enough inode space is available */
3571 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3572 sbi->s_inode_size) {
3573 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3574 EXT4_GOOD_OLD_INODE_SIZE;
3575 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3579 if (test_opt(sb, DELALLOC) &&
3580 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3581 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3582 "requested data journaling mode");
3583 clear_opt(sb, DELALLOC);
3585 if (test_opt(sb, DIOREAD_NOLOCK)) {
3586 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3587 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3588 "option - requested data journaling mode");
3589 clear_opt(sb, DIOREAD_NOLOCK);
3591 if (sb->s_blocksize < PAGE_SIZE) {
3592 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3593 "option - block size is too small");
3594 clear_opt(sb, DIOREAD_NOLOCK);
3598 err = ext4_setup_system_zone(sb);
3600 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3606 err = ext4_mb_init(sb, needs_recovery);
3608 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3613 err = ext4_register_li_request(sb, first_not_zeroed);
3617 sbi->s_kobj.kset = ext4_kset;
3618 init_completion(&sbi->s_kobj_unregister);
3619 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3622 ext4_mb_release(sb);
3623 ext4_ext_release(sb);
3627 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3628 ext4_orphan_cleanup(sb, es);
3629 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3630 if (needs_recovery) {
3631 ext4_msg(sb, KERN_INFO, "recovery complete");
3632 ext4_mark_recovery_complete(sb, es);
3634 if (EXT4_SB(sb)->s_journal) {
3635 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3636 descr = " journalled data mode";
3637 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3638 descr = " ordered data mode";
3640 descr = " writeback data mode";
3642 descr = "out journal";
3644 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3645 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3646 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3648 if (es->s_error_count)
3649 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3656 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3660 ext4_msg(sb, KERN_ERR, "mount failed");
3661 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3663 ext4_release_system_zone(sb);
3664 if (sbi->s_journal) {
3665 jbd2_journal_destroy(sbi->s_journal);
3666 sbi->s_journal = NULL;
3669 del_timer(&sbi->s_err_report);
3670 if (sbi->s_flex_groups) {
3671 if (is_vmalloc_addr(sbi->s_flex_groups))
3672 vfree(sbi->s_flex_groups);
3674 kfree(sbi->s_flex_groups);
3676 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3677 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3678 percpu_counter_destroy(&sbi->s_dirs_counter);
3679 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3681 for (i = 0; i < db_count; i++)
3682 brelse(sbi->s_group_desc[i]);
3683 kfree(sbi->s_group_desc);
3686 remove_proc_entry(sb->s_id, ext4_proc_root);
3689 for (i = 0; i < MAXQUOTAS; i++)
3690 kfree(sbi->s_qf_names[i]);
3692 ext4_blkdev_remove(sbi);
3695 sb->s_fs_info = NULL;
3696 kfree(sbi->s_blockgroup_lock);
3704 * Setup any per-fs journal parameters now. We'll do this both on
3705 * initial mount, once the journal has been initialised but before we've
3706 * done any recovery; and again on any subsequent remount.
3708 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3710 struct ext4_sb_info *sbi = EXT4_SB(sb);
3712 journal->j_commit_interval = sbi->s_commit_interval;
3713 journal->j_min_batch_time = sbi->s_min_batch_time;
3714 journal->j_max_batch_time = sbi->s_max_batch_time;
3716 write_lock(&journal->j_state_lock);
3717 if (test_opt(sb, BARRIER))
3718 journal->j_flags |= JBD2_BARRIER;
3720 journal->j_flags &= ~JBD2_BARRIER;
3721 if (test_opt(sb, DATA_ERR_ABORT))
3722 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3724 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3725 write_unlock(&journal->j_state_lock);
3728 static journal_t *ext4_get_journal(struct super_block *sb,
3729 unsigned int journal_inum)
3731 struct inode *journal_inode;
3734 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3736 /* First, test for the existence of a valid inode on disk. Bad
3737 * things happen if we iget() an unused inode, as the subsequent
3738 * iput() will try to delete it. */
3740 journal_inode = ext4_iget(sb, journal_inum);
3741 if (IS_ERR(journal_inode)) {
3742 ext4_msg(sb, KERN_ERR, "no journal found");
3745 if (!journal_inode->i_nlink) {
3746 make_bad_inode(journal_inode);
3747 iput(journal_inode);
3748 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3752 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3753 journal_inode, journal_inode->i_size);
3754 if (!S_ISREG(journal_inode->i_mode)) {
3755 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3756 iput(journal_inode);
3760 journal = jbd2_journal_init_inode(journal_inode);
3762 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3763 iput(journal_inode);
3766 journal->j_private = sb;
3767 ext4_init_journal_params(sb, journal);
3771 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3774 struct buffer_head *bh;
3778 int hblock, blocksize;
3779 ext4_fsblk_t sb_block;
3780 unsigned long offset;
3781 struct ext4_super_block *es;
3782 struct block_device *bdev;
3784 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3786 bdev = ext4_blkdev_get(j_dev, sb);
3790 blocksize = sb->s_blocksize;
3791 hblock = bdev_logical_block_size(bdev);
3792 if (blocksize < hblock) {
3793 ext4_msg(sb, KERN_ERR,
3794 "blocksize too small for journal device");
3798 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3799 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3800 set_blocksize(bdev, blocksize);
3801 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3802 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3803 "external journal");
3807 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3808 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3809 !(le32_to_cpu(es->s_feature_incompat) &
3810 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3811 ext4_msg(sb, KERN_ERR, "external journal has "
3817 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3818 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3823 len = ext4_blocks_count(es);
3824 start = sb_block + 1;
3825 brelse(bh); /* we're done with the superblock */
3827 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3828 start, len, blocksize);
3830 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3833 journal->j_private = sb;
3834 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3835 wait_on_buffer(journal->j_sb_buffer);
3836 if (!buffer_uptodate(journal->j_sb_buffer)) {
3837 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3840 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3841 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3842 "user (unsupported) - %d",
3843 be32_to_cpu(journal->j_superblock->s_nr_users));
3846 EXT4_SB(sb)->journal_bdev = bdev;
3847 ext4_init_journal_params(sb, journal);
3851 jbd2_journal_destroy(journal);
3853 ext4_blkdev_put(bdev);
3857 static int ext4_load_journal(struct super_block *sb,
3858 struct ext4_super_block *es,
3859 unsigned long journal_devnum)
3862 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3865 int really_read_only;
3867 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3869 if (journal_devnum &&
3870 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3871 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3872 "numbers have changed");
3873 journal_dev = new_decode_dev(journal_devnum);
3875 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3877 really_read_only = bdev_read_only(sb->s_bdev);
3880 * Are we loading a blank journal or performing recovery after a
3881 * crash? For recovery, we need to check in advance whether we
3882 * can get read-write access to the device.
3884 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3885 if (sb->s_flags & MS_RDONLY) {
3886 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3887 "required on readonly filesystem");
3888 if (really_read_only) {
3889 ext4_msg(sb, KERN_ERR, "write access "
3890 "unavailable, cannot proceed");
3893 ext4_msg(sb, KERN_INFO, "write access will "
3894 "be enabled during recovery");
3898 if (journal_inum && journal_dev) {
3899 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3900 "and inode journals!");
3905 if (!(journal = ext4_get_journal(sb, journal_inum)))
3908 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3912 if (!(journal->j_flags & JBD2_BARRIER))
3913 ext4_msg(sb, KERN_INFO, "barriers disabled");
3915 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3916 err = jbd2_journal_update_format(journal);
3918 ext4_msg(sb, KERN_ERR, "error updating journal");
3919 jbd2_journal_destroy(journal);
3924 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3925 err = jbd2_journal_wipe(journal, !really_read_only);
3927 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3929 memcpy(save, ((char *) es) +
3930 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3931 err = jbd2_journal_load(journal);
3933 memcpy(((char *) es) + EXT4_S_ERR_START,
3934 save, EXT4_S_ERR_LEN);
3939 ext4_msg(sb, KERN_ERR, "error loading journal");
3940 jbd2_journal_destroy(journal);
3944 EXT4_SB(sb)->s_journal = journal;
3945 ext4_clear_journal_err(sb, es);
3947 if (!really_read_only && journal_devnum &&
3948 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3949 es->s_journal_dev = cpu_to_le32(journal_devnum);
3951 /* Make sure we flush the recovery flag to disk. */
3952 ext4_commit_super(sb, 1);
3958 static int ext4_commit_super(struct super_block *sb, int sync)
3960 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3961 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3966 if (buffer_write_io_error(sbh)) {
3968 * Oh, dear. A previous attempt to write the
3969 * superblock failed. This could happen because the
3970 * USB device was yanked out. Or it could happen to
3971 * be a transient write error and maybe the block will
3972 * be remapped. Nothing we can do but to retry the
3973 * write and hope for the best.
3975 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3976 "superblock detected");
3977 clear_buffer_write_io_error(sbh);
3978 set_buffer_uptodate(sbh);
3981 * If the file system is mounted read-only, don't update the
3982 * superblock write time. This avoids updating the superblock
3983 * write time when we are mounting the root file system
3984 * read/only but we need to replay the journal; at that point,
3985 * for people who are east of GMT and who make their clock
3986 * tick in localtime for Windows bug-for-bug compatibility,
3987 * the clock is set in the future, and this will cause e2fsck
3988 * to complain and force a full file system check.
3990 if (!(sb->s_flags & MS_RDONLY))
3991 es->s_wtime = cpu_to_le32(get_seconds());
3992 if (sb->s_bdev->bd_part)
3993 es->s_kbytes_written =
3994 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3995 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3996 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3998 es->s_kbytes_written =
3999 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4000 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4001 &EXT4_SB(sb)->s_freeblocks_counter));
4002 es->s_free_inodes_count =
4003 cpu_to_le32(percpu_counter_sum_positive(
4004 &EXT4_SB(sb)->s_freeinodes_counter));
4006 BUFFER_TRACE(sbh, "marking dirty");
4007 mark_buffer_dirty(sbh);
4009 error = sync_dirty_buffer(sbh);
4013 error = buffer_write_io_error(sbh);
4015 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4017 clear_buffer_write_io_error(sbh);
4018 set_buffer_uptodate(sbh);
4025 * Have we just finished recovery? If so, and if we are mounting (or
4026 * remounting) the filesystem readonly, then we will end up with a
4027 * consistent fs on disk. Record that fact.
4029 static void ext4_mark_recovery_complete(struct super_block *sb,
4030 struct ext4_super_block *es)
4032 journal_t *journal = EXT4_SB(sb)->s_journal;
4034 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4035 BUG_ON(journal != NULL);
4038 jbd2_journal_lock_updates(journal);
4039 if (jbd2_journal_flush(journal) < 0)
4042 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4043 sb->s_flags & MS_RDONLY) {
4044 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4045 ext4_commit_super(sb, 1);
4049 jbd2_journal_unlock_updates(journal);
4053 * If we are mounting (or read-write remounting) a filesystem whose journal
4054 * has recorded an error from a previous lifetime, move that error to the
4055 * main filesystem now.
4057 static void ext4_clear_journal_err(struct super_block *sb,
4058 struct ext4_super_block *es)
4064 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4066 journal = EXT4_SB(sb)->s_journal;
4069 * Now check for any error status which may have been recorded in the
4070 * journal by a prior ext4_error() or ext4_abort()
4073 j_errno = jbd2_journal_errno(journal);
4077 errstr = ext4_decode_error(sb, j_errno, nbuf);
4078 ext4_warning(sb, "Filesystem error recorded "
4079 "from previous mount: %s", errstr);
4080 ext4_warning(sb, "Marking fs in need of filesystem check.");
4082 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4083 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4084 ext4_commit_super(sb, 1);
4086 jbd2_journal_clear_err(journal);
4091 * Force the running and committing transactions to commit,
4092 * and wait on the commit.
4094 int ext4_force_commit(struct super_block *sb)
4099 if (sb->s_flags & MS_RDONLY)
4102 journal = EXT4_SB(sb)->s_journal;
4104 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4105 ret = ext4_journal_force_commit(journal);
4111 static void ext4_write_super(struct super_block *sb)
4114 ext4_commit_super(sb, 1);
4118 static int ext4_sync_fs(struct super_block *sb, int wait)
4122 struct ext4_sb_info *sbi = EXT4_SB(sb);
4124 trace_ext4_sync_fs(sb, wait);
4125 flush_workqueue(sbi->dio_unwritten_wq);
4126 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4128 jbd2_log_wait_commit(sbi->s_journal, target);
4134 * LVM calls this function before a (read-only) snapshot is created. This
4135 * gives us a chance to flush the journal completely and mark the fs clean.
4137 static int ext4_freeze(struct super_block *sb)
4142 if (sb->s_flags & MS_RDONLY)
4145 journal = EXT4_SB(sb)->s_journal;
4147 /* Now we set up the journal barrier. */
4148 jbd2_journal_lock_updates(journal);
4151 * Don't clear the needs_recovery flag if we failed to flush
4154 error = jbd2_journal_flush(journal);
4158 /* Journal blocked and flushed, clear needs_recovery flag. */
4159 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4160 error = ext4_commit_super(sb, 1);
4162 /* we rely on s_frozen to stop further updates */
4163 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4168 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4169 * flag here, even though the filesystem is not technically dirty yet.
4171 static int ext4_unfreeze(struct super_block *sb)
4173 if (sb->s_flags & MS_RDONLY)
4177 /* Reset the needs_recovery flag before the fs is unlocked. */
4178 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4179 ext4_commit_super(sb, 1);
4185 * Structure to save mount options for ext4_remount's benefit
4187 struct ext4_mount_options {
4188 unsigned long s_mount_opt;
4189 unsigned long s_mount_opt2;
4192 unsigned long s_commit_interval;
4193 u32 s_min_batch_time, s_max_batch_time;
4196 char *s_qf_names[MAXQUOTAS];
4200 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4202 struct ext4_super_block *es;
4203 struct ext4_sb_info *sbi = EXT4_SB(sb);
4204 ext4_fsblk_t n_blocks_count = 0;
4205 unsigned long old_sb_flags;
4206 struct ext4_mount_options old_opts;
4207 int enable_quota = 0;
4209 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4214 char *orig_data = kstrdup(data, GFP_KERNEL);
4216 /* Store the original options */
4218 old_sb_flags = sb->s_flags;
4219 old_opts.s_mount_opt = sbi->s_mount_opt;
4220 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4221 old_opts.s_resuid = sbi->s_resuid;
4222 old_opts.s_resgid = sbi->s_resgid;
4223 old_opts.s_commit_interval = sbi->s_commit_interval;
4224 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4225 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4227 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4228 for (i = 0; i < MAXQUOTAS; i++)
4229 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4231 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4232 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4235 * Allow the "check" option to be passed as a remount option.
4237 if (!parse_options(data, sb, NULL, &journal_ioprio,
4238 &n_blocks_count, 1)) {
4243 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4244 ext4_abort(sb, "Abort forced by user");
4246 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4247 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4251 if (sbi->s_journal) {
4252 ext4_init_journal_params(sb, sbi->s_journal);
4253 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4256 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4257 n_blocks_count > ext4_blocks_count(es)) {
4258 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4263 if (*flags & MS_RDONLY) {
4264 err = dquot_suspend(sb, -1);
4269 * First of all, the unconditional stuff we have to do
4270 * to disable replay of the journal when we next remount
4272 sb->s_flags |= MS_RDONLY;
4275 * OK, test if we are remounting a valid rw partition
4276 * readonly, and if so set the rdonly flag and then
4277 * mark the partition as valid again.
4279 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4280 (sbi->s_mount_state & EXT4_VALID_FS))
4281 es->s_state = cpu_to_le16(sbi->s_mount_state);
4284 ext4_mark_recovery_complete(sb, es);
4286 /* Make sure we can mount this feature set readwrite */
4287 if (!ext4_feature_set_ok(sb, 0)) {
4292 * Make sure the group descriptor checksums
4293 * are sane. If they aren't, refuse to remount r/w.
4295 for (g = 0; g < sbi->s_groups_count; g++) {
4296 struct ext4_group_desc *gdp =
4297 ext4_get_group_desc(sb, g, NULL);
4299 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4300 ext4_msg(sb, KERN_ERR,
4301 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4302 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4303 le16_to_cpu(gdp->bg_checksum));
4310 * If we have an unprocessed orphan list hanging
4311 * around from a previously readonly bdev mount,
4312 * require a full umount/remount for now.
4314 if (es->s_last_orphan) {
4315 ext4_msg(sb, KERN_WARNING, "Couldn't "
4316 "remount RDWR because of unprocessed "
4317 "orphan inode list. Please "
4318 "umount/remount instead");
4324 * Mounting a RDONLY partition read-write, so reread
4325 * and store the current valid flag. (It may have
4326 * been changed by e2fsck since we originally mounted
4330 ext4_clear_journal_err(sb, es);
4331 sbi->s_mount_state = le16_to_cpu(es->s_state);
4332 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4334 if (!ext4_setup_super(sb, es, 0))
4335 sb->s_flags &= ~MS_RDONLY;
4341 * Reinitialize lazy itable initialization thread based on
4344 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4345 ext4_unregister_li_request(sb);
4347 ext4_group_t first_not_zeroed;
4348 first_not_zeroed = ext4_has_uninit_itable(sb);
4349 ext4_register_li_request(sb, first_not_zeroed);
4352 ext4_setup_system_zone(sb);
4353 if (sbi->s_journal == NULL)
4354 ext4_commit_super(sb, 1);
4357 /* Release old quota file names */
4358 for (i = 0; i < MAXQUOTAS; i++)
4359 if (old_opts.s_qf_names[i] &&
4360 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4361 kfree(old_opts.s_qf_names[i]);
4365 dquot_resume(sb, -1);
4367 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4372 sb->s_flags = old_sb_flags;
4373 sbi->s_mount_opt = old_opts.s_mount_opt;
4374 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4375 sbi->s_resuid = old_opts.s_resuid;
4376 sbi->s_resgid = old_opts.s_resgid;
4377 sbi->s_commit_interval = old_opts.s_commit_interval;
4378 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4379 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4381 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4382 for (i = 0; i < MAXQUOTAS; i++) {
4383 if (sbi->s_qf_names[i] &&
4384 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4385 kfree(sbi->s_qf_names[i]);
4386 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4394 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4396 struct super_block *sb = dentry->d_sb;
4397 struct ext4_sb_info *sbi = EXT4_SB(sb);
4398 struct ext4_super_block *es = sbi->s_es;
4401 if (test_opt(sb, MINIX_DF)) {
4402 sbi->s_overhead_last = 0;
4403 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4404 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4405 ext4_fsblk_t overhead = 0;
4408 * Compute the overhead (FS structures). This is constant
4409 * for a given filesystem unless the number of block groups
4410 * changes so we cache the previous value until it does.
4414 * All of the blocks before first_data_block are
4417 overhead = le32_to_cpu(es->s_first_data_block);
4420 * Add the overhead attributed to the superblock and
4421 * block group descriptors. If the sparse superblocks
4422 * feature is turned on, then not all groups have this.
4424 for (i = 0; i < ngroups; i++) {
4425 overhead += ext4_bg_has_super(sb, i) +
4426 ext4_bg_num_gdb(sb, i);
4431 * Every block group has an inode bitmap, a block
4432 * bitmap, and an inode table.
4434 overhead += ngroups * (2 + sbi->s_itb_per_group);
4435 sbi->s_overhead_last = overhead;
4437 sbi->s_blocks_last = ext4_blocks_count(es);
4440 buf->f_type = EXT4_SUPER_MAGIC;
4441 buf->f_bsize = sb->s_blocksize;
4442 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4443 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4444 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4445 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4446 if (buf->f_bfree < ext4_r_blocks_count(es))
4448 buf->f_files = le32_to_cpu(es->s_inodes_count);
4449 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4450 buf->f_namelen = EXT4_NAME_LEN;
4451 fsid = le64_to_cpup((void *)es->s_uuid) ^
4452 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4453 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4454 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4459 /* Helper function for writing quotas on sync - we need to start transaction
4460 * before quota file is locked for write. Otherwise the are possible deadlocks:
4461 * Process 1 Process 2
4462 * ext4_create() quota_sync()
4463 * jbd2_journal_start() write_dquot()
4464 * dquot_initialize() down(dqio_mutex)
4465 * down(dqio_mutex) jbd2_journal_start()
4471 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4473 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4476 static int ext4_write_dquot(struct dquot *dquot)
4480 struct inode *inode;
4482 inode = dquot_to_inode(dquot);
4483 handle = ext4_journal_start(inode,
4484 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4486 return PTR_ERR(handle);
4487 ret = dquot_commit(dquot);
4488 err = ext4_journal_stop(handle);
4494 static int ext4_acquire_dquot(struct dquot *dquot)
4499 handle = ext4_journal_start(dquot_to_inode(dquot),
4500 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4502 return PTR_ERR(handle);
4503 ret = dquot_acquire(dquot);
4504 err = ext4_journal_stop(handle);
4510 static int ext4_release_dquot(struct dquot *dquot)
4515 handle = ext4_journal_start(dquot_to_inode(dquot),
4516 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4517 if (IS_ERR(handle)) {
4518 /* Release dquot anyway to avoid endless cycle in dqput() */
4519 dquot_release(dquot);
4520 return PTR_ERR(handle);
4522 ret = dquot_release(dquot);
4523 err = ext4_journal_stop(handle);
4529 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4531 /* Are we journaling quotas? */
4532 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4533 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4534 dquot_mark_dquot_dirty(dquot);
4535 return ext4_write_dquot(dquot);
4537 return dquot_mark_dquot_dirty(dquot);
4541 static int ext4_write_info(struct super_block *sb, int type)
4546 /* Data block + inode block */
4547 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4549 return PTR_ERR(handle);
4550 ret = dquot_commit_info(sb, type);
4551 err = ext4_journal_stop(handle);
4558 * Turn on quotas during mount time - we need to find
4559 * the quota file and such...
4561 static int ext4_quota_on_mount(struct super_block *sb, int type)
4563 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4564 EXT4_SB(sb)->s_jquota_fmt, type);
4568 * Standard function to be called on quota_on
4570 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4575 if (!test_opt(sb, QUOTA))
4578 /* Quotafile not on the same filesystem? */
4579 if (path->mnt->mnt_sb != sb)
4581 /* Journaling quota? */
4582 if (EXT4_SB(sb)->s_qf_names[type]) {
4583 /* Quotafile not in fs root? */
4584 if (path->dentry->d_parent != sb->s_root)
4585 ext4_msg(sb, KERN_WARNING,
4586 "Quota file not on filesystem root. "
4587 "Journaled quota will not work");
4591 * When we journal data on quota file, we have to flush journal to see
4592 * all updates to the file when we bypass pagecache...
4594 if (EXT4_SB(sb)->s_journal &&
4595 ext4_should_journal_data(path->dentry->d_inode)) {
4597 * We don't need to lock updates but journal_flush() could
4598 * otherwise be livelocked...
4600 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4601 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4602 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4607 return dquot_quota_on(sb, type, format_id, path);
4610 static int ext4_quota_off(struct super_block *sb, int type)
4612 /* Force all delayed allocation blocks to be allocated.
4613 * Caller already holds s_umount sem */
4614 if (test_opt(sb, DELALLOC))
4615 sync_filesystem(sb);
4617 return dquot_quota_off(sb, type);
4620 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4621 * acquiring the locks... As quota files are never truncated and quota code
4622 * itself serializes the operations (and noone else should touch the files)
4623 * we don't have to be afraid of races */
4624 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4625 size_t len, loff_t off)
4627 struct inode *inode = sb_dqopt(sb)->files[type];
4628 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4630 int offset = off & (sb->s_blocksize - 1);
4633 struct buffer_head *bh;
4634 loff_t i_size = i_size_read(inode);
4638 if (off+len > i_size)
4641 while (toread > 0) {
4642 tocopy = sb->s_blocksize - offset < toread ?
4643 sb->s_blocksize - offset : toread;
4644 bh = ext4_bread(NULL, inode, blk, 0, &err);
4647 if (!bh) /* A hole? */
4648 memset(data, 0, tocopy);
4650 memcpy(data, bh->b_data+offset, tocopy);
4660 /* Write to quotafile (we know the transaction is already started and has
4661 * enough credits) */
4662 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4663 const char *data, size_t len, loff_t off)
4665 struct inode *inode = sb_dqopt(sb)->files[type];
4666 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4668 int offset = off & (sb->s_blocksize - 1);
4669 struct buffer_head *bh;
4670 handle_t *handle = journal_current_handle();
4672 if (EXT4_SB(sb)->s_journal && !handle) {
4673 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4674 " cancelled because transaction is not started",
4675 (unsigned long long)off, (unsigned long long)len);
4679 * Since we account only one data block in transaction credits,
4680 * then it is impossible to cross a block boundary.
4682 if (sb->s_blocksize - offset < len) {
4683 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4684 " cancelled because not block aligned",
4685 (unsigned long long)off, (unsigned long long)len);
4689 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4690 bh = ext4_bread(handle, inode, blk, 1, &err);
4693 err = ext4_journal_get_write_access(handle, bh);
4699 memcpy(bh->b_data+offset, data, len);
4700 flush_dcache_page(bh->b_page);
4702 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4706 mutex_unlock(&inode->i_mutex);
4709 if (inode->i_size < off + len) {
4710 i_size_write(inode, off + len);
4711 EXT4_I(inode)->i_disksize = inode->i_size;
4713 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4714 ext4_mark_inode_dirty(handle, inode);
4715 mutex_unlock(&inode->i_mutex);
4721 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4722 const char *dev_name, void *data)
4724 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4727 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4728 static struct file_system_type ext2_fs_type = {
4729 .owner = THIS_MODULE,
4731 .mount = ext4_mount,
4732 .kill_sb = kill_block_super,
4733 .fs_flags = FS_REQUIRES_DEV,
4736 static inline void register_as_ext2(void)
4738 int err = register_filesystem(&ext2_fs_type);
4741 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4744 static inline void unregister_as_ext2(void)
4746 unregister_filesystem(&ext2_fs_type);
4748 MODULE_ALIAS("ext2");
4750 static inline void register_as_ext2(void) { }
4751 static inline void unregister_as_ext2(void) { }
4754 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4755 static inline void register_as_ext3(void)
4757 int err = register_filesystem(&ext3_fs_type);
4760 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4763 static inline void unregister_as_ext3(void)
4765 unregister_filesystem(&ext3_fs_type);
4767 MODULE_ALIAS("ext3");
4769 static inline void register_as_ext3(void) { }
4770 static inline void unregister_as_ext3(void) { }
4773 static struct file_system_type ext4_fs_type = {
4774 .owner = THIS_MODULE,
4776 .mount = ext4_mount,
4777 .kill_sb = kill_block_super,
4778 .fs_flags = FS_REQUIRES_DEV,
4781 static int __init ext4_init_feat_adverts(void)
4783 struct ext4_features *ef;
4786 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4790 ef->f_kobj.kset = ext4_kset;
4791 init_completion(&ef->f_kobj_unregister);
4792 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4805 static void ext4_exit_feat_adverts(void)
4807 kobject_put(&ext4_feat->f_kobj);
4808 wait_for_completion(&ext4_feat->f_kobj_unregister);
4812 /* Shared across all ext4 file systems */
4813 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4814 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4816 static int __init ext4_init_fs(void)
4820 ext4_check_flag_values();
4822 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4823 mutex_init(&ext4__aio_mutex[i]);
4824 init_waitqueue_head(&ext4__ioend_wq[i]);
4827 err = ext4_init_pageio();
4830 err = ext4_init_system_zone();
4833 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4836 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4837 if (!ext4_proc_root)
4840 err = ext4_init_feat_adverts();
4844 err = ext4_init_mballoc();
4848 err = ext4_init_xattr();
4851 err = init_inodecache();
4856 err = register_filesystem(&ext4_fs_type);
4860 ext4_li_info = NULL;
4861 mutex_init(&ext4_li_mtx);
4864 unregister_as_ext2();
4865 unregister_as_ext3();
4866 destroy_inodecache();
4870 ext4_exit_mballoc();
4872 ext4_exit_feat_adverts();
4874 remove_proc_entry("fs/ext4", NULL);
4876 kset_unregister(ext4_kset);
4878 ext4_exit_system_zone();
4884 static void __exit ext4_exit_fs(void)
4886 ext4_destroy_lazyinit_thread();
4887 unregister_as_ext2();
4888 unregister_as_ext3();
4889 unregister_filesystem(&ext4_fs_type);
4890 destroy_inodecache();
4892 ext4_exit_mballoc();
4893 ext4_exit_feat_adverts();
4894 remove_proc_entry("fs/ext4", NULL);
4895 kset_unregister(ext4_kset);
4896 ext4_exit_system_zone();
4900 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4901 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4902 MODULE_LICENSE("GPL");
4903 module_init(ext4_init_fs)
4904 module_exit(ext4_exit_fs)