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/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 struct ext4_lazy_init *ext4_li_info;
59 struct mutex ext4_li_mtx;
60 struct ext4_features *ext4_feat;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static void ext4_write_super(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data, struct vfsmount *mnt);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
82 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83 static struct file_system_type ext3_fs_type = {
86 .get_sb = ext4_get_sb,
87 .kill_sb = kill_block_super,
88 .fs_flags = FS_REQUIRES_DEV,
90 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
92 #define IS_EXT3_SB(sb) (0)
95 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
98 return le32_to_cpu(bg->bg_block_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
103 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
104 struct ext4_group_desc *bg)
106 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
111 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
112 struct ext4_group_desc *bg)
114 return le32_to_cpu(bg->bg_inode_table_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
119 __u32 ext4_free_blks_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
122 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
127 __u32 ext4_free_inodes_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
130 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
135 __u32 ext4_used_dirs_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
138 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
143 __u32 ext4_itable_unused_count(struct super_block *sb,
144 struct ext4_group_desc *bg)
146 return le16_to_cpu(bg->bg_itable_unused_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
151 void ext4_block_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
159 void ext4_inode_bitmap_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
167 void ext4_inode_table_set(struct super_block *sb,
168 struct ext4_group_desc *bg, ext4_fsblk_t blk)
170 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
175 void ext4_free_blks_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
178 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
183 void ext4_free_inodes_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
186 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
191 void ext4_used_dirs_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
194 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
199 void ext4_itable_unused_set(struct super_block *sb,
200 struct ext4_group_desc *bg, __u32 count)
202 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
208 /* Just increment the non-pointer handle value */
209 static handle_t *ext4_get_nojournal(void)
211 handle_t *handle = current->journal_info;
212 unsigned long ref_cnt = (unsigned long)handle;
214 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
217 handle = (handle_t *)ref_cnt;
219 current->journal_info = handle;
224 /* Decrement the non-pointer handle value */
225 static void ext4_put_nojournal(handle_t *handle)
227 unsigned long ref_cnt = (unsigned long)handle;
229 BUG_ON(ref_cnt == 0);
232 handle = (handle_t *)ref_cnt;
234 current->journal_info = handle;
238 * Wrappers for jbd2_journal_start/end.
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
245 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
249 if (sb->s_flags & MS_RDONLY)
250 return ERR_PTR(-EROFS);
252 vfs_check_frozen(sb, SB_FREEZE_TRANS);
253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
256 journal = EXT4_SB(sb)->s_journal;
258 if (is_journal_aborted(journal)) {
259 ext4_abort(sb, "Detected aborted journal");
260 return ERR_PTR(-EROFS);
262 return jbd2_journal_start(journal, nblocks);
264 return ext4_get_nojournal();
268 * The only special thing we need to do here is to make sure that all
269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
270 * that sync() will call the filesystem's write_super callback if
273 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
275 struct super_block *sb;
279 if (!ext4_handle_valid(handle)) {
280 ext4_put_nojournal(handle);
283 sb = handle->h_transaction->t_journal->j_private;
285 rc = jbd2_journal_stop(handle);
290 __ext4_std_error(sb, where, line, err);
294 void ext4_journal_abort_handle(const char *caller, unsigned int line,
295 const char *err_fn, struct buffer_head *bh,
296 handle_t *handle, int err)
299 const char *errstr = ext4_decode_error(NULL, err, nbuf);
301 BUG_ON(!ext4_handle_valid(handle));
304 BUFFER_TRACE(bh, "abort");
309 if (is_handle_aborted(handle))
312 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
313 caller, line, errstr, err_fn);
315 jbd2_journal_abort_handle(handle);
318 static void __save_error_info(struct super_block *sb, const char *func,
321 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
337 * Start the daily error reporting function if it hasn't been
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
342 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
345 static void save_error_info(struct super_block *sb, const char *func,
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
353 /* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
356 * On ext2, we can store the error state of the filesystem in the
357 * superblock. That is not possible on ext4, because we may have other
358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
363 * We'll just use the jbd2_journal_abort() error code to record an error in
364 * the journal instead. On recovery, the journal will complain about
365 * that error until we've noted it down and cleared it.
368 static void ext4_handle_error(struct super_block *sb)
370 if (sb->s_flags & MS_RDONLY)
373 if (!test_opt(sb, ERRORS_CONT)) {
374 journal_t *journal = EXT4_SB(sb)->s_journal;
376 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
378 jbd2_journal_abort(journal, -EIO);
380 if (test_opt(sb, ERRORS_RO)) {
381 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
382 sb->s_flags |= MS_RDONLY;
384 if (test_opt(sb, ERRORS_PANIC))
385 panic("EXT4-fs (device %s): panic forced after error\n",
389 void __ext4_error(struct super_block *sb, const char *function,
390 unsigned int line, const char *fmt, ...)
395 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
396 sb->s_id, function, line, current->comm);
401 ext4_handle_error(sb);
404 void ext4_error_inode(struct inode *inode, const char *function,
405 unsigned int line, ext4_fsblk_t block,
406 const char *fmt, ...)
409 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
411 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
412 es->s_last_error_block = cpu_to_le64(block);
413 save_error_info(inode->i_sb, function, line);
415 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
416 inode->i_sb->s_id, function, line, inode->i_ino);
418 printk("block %llu: ", block);
419 printk("comm %s: ", current->comm);
424 ext4_handle_error(inode->i_sb);
427 void ext4_error_file(struct file *file, const char *function,
428 unsigned int line, const char *fmt, ...)
431 struct ext4_super_block *es;
432 struct inode *inode = file->f_dentry->d_inode;
433 char pathname[80], *path;
435 es = EXT4_SB(inode->i_sb)->s_es;
436 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
437 save_error_info(inode->i_sb, function, line);
439 path = d_path(&(file->f_path), pathname, sizeof(pathname));
443 "EXT4-fs error (device %s): %s:%d: inode #%lu "
444 "(comm %s path %s): ",
445 inode->i_sb->s_id, function, line, inode->i_ino,
446 current->comm, path);
451 ext4_handle_error(inode->i_sb);
454 static const char *ext4_decode_error(struct super_block *sb, int errno,
461 errstr = "IO failure";
464 errstr = "Out of memory";
467 if (!sb || (EXT4_SB(sb)->s_journal &&
468 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
469 errstr = "Journal has aborted";
471 errstr = "Readonly filesystem";
474 /* If the caller passed in an extra buffer for unknown
475 * errors, textualise them now. Else we just return
478 /* Check for truncated error codes... */
479 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
488 /* __ext4_std_error decodes expected errors from journaling functions
489 * automatically and invokes the appropriate error response. */
491 void __ext4_std_error(struct super_block *sb, const char *function,
492 unsigned int line, int errno)
497 /* Special case: if the error is EROFS, and we're not already
498 * inside a transaction, then there's really no point in logging
500 if (errno == -EROFS && journal_current_handle() == NULL &&
501 (sb->s_flags & MS_RDONLY))
504 errstr = ext4_decode_error(sb, errno, nbuf);
505 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
506 sb->s_id, function, line, errstr);
507 save_error_info(sb, function, line);
509 ext4_handle_error(sb);
513 * ext4_abort is a much stronger failure handler than ext4_error. The
514 * abort function may be used to deal with unrecoverable failures such
515 * as journal IO errors or ENOMEM at a critical moment in log management.
517 * We unconditionally force the filesystem into an ABORT|READONLY state,
518 * unless the error response on the fs has been set to panic in which
519 * case we take the easy way out and panic immediately.
522 void __ext4_abort(struct super_block *sb, const char *function,
523 unsigned int line, const char *fmt, ...)
527 save_error_info(sb, function, line);
529 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 if ((sb->s_flags & MS_RDONLY) == 0) {
536 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
537 sb->s_flags |= MS_RDONLY;
538 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
539 if (EXT4_SB(sb)->s_journal)
540 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
541 save_error_info(sb, function, line);
543 if (test_opt(sb, ERRORS_PANIC))
544 panic("EXT4-fs panic from previous error\n");
547 void ext4_msg (struct super_block * sb, const char *prefix,
548 const char *fmt, ...)
553 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
559 void __ext4_warning(struct super_block *sb, const char *function,
560 unsigned int line, const char *fmt, ...)
565 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
566 sb->s_id, function, line);
572 void __ext4_grp_locked_error(const char *function, unsigned int line,
573 struct super_block *sb, ext4_group_t grp,
574 unsigned long ino, ext4_fsblk_t block,
575 const char *fmt, ...)
580 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
582 es->s_last_error_ino = cpu_to_le32(ino);
583 es->s_last_error_block = cpu_to_le64(block);
584 __save_error_info(sb, function, line);
586 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
587 sb->s_id, function, line, grp);
589 printk("inode %lu: ", ino);
591 printk("block %llu:", (unsigned long long) block);
596 if (test_opt(sb, ERRORS_CONT)) {
597 ext4_commit_super(sb, 0);
601 ext4_unlock_group(sb, grp);
602 ext4_handle_error(sb);
604 * We only get here in the ERRORS_RO case; relocking the group
605 * may be dangerous, but nothing bad will happen since the
606 * filesystem will have already been marked read/only and the
607 * journal has been aborted. We return 1 as a hint to callers
608 * who might what to use the return value from
609 * ext4_grp_locked_error() to distinguish beween the
610 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
611 * aggressively from the ext4 function in question, with a
612 * more appropriate error code.
614 ext4_lock_group(sb, grp);
618 void ext4_update_dynamic_rev(struct super_block *sb)
620 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
622 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
626 "updating to rev %d because of new feature flag, "
627 "running e2fsck is recommended",
630 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
631 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
632 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
633 /* leave es->s_feature_*compat flags alone */
634 /* es->s_uuid will be set by e2fsck if empty */
637 * The rest of the superblock fields should be zero, and if not it
638 * means they are likely already in use, so leave them alone. We
639 * can leave it up to e2fsck to clean up any inconsistencies there.
644 * Open the external journal device
646 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
648 struct block_device *bdev;
649 char b[BDEVNAME_SIZE];
651 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
657 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
658 __bdevname(dev, b), PTR_ERR(bdev));
663 * Release the journal device
665 static int ext4_blkdev_put(struct block_device *bdev)
668 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
671 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
673 struct block_device *bdev;
676 bdev = sbi->journal_bdev;
678 ret = ext4_blkdev_put(bdev);
679 sbi->journal_bdev = NULL;
684 static inline struct inode *orphan_list_entry(struct list_head *l)
686 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
689 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
693 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
694 le32_to_cpu(sbi->s_es->s_last_orphan));
696 printk(KERN_ERR "sb_info orphan list:\n");
697 list_for_each(l, &sbi->s_orphan) {
698 struct inode *inode = orphan_list_entry(l);
700 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
701 inode->i_sb->s_id, inode->i_ino, inode,
702 inode->i_mode, inode->i_nlink,
707 static void ext4_put_super(struct super_block *sb)
709 struct ext4_sb_info *sbi = EXT4_SB(sb);
710 struct ext4_super_block *es = sbi->s_es;
713 ext4_unregister_li_request(sb);
714 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
716 flush_workqueue(sbi->dio_unwritten_wq);
717 destroy_workqueue(sbi->dio_unwritten_wq);
722 ext4_commit_super(sb, 1);
724 if (sbi->s_journal) {
725 err = jbd2_journal_destroy(sbi->s_journal);
726 sbi->s_journal = NULL;
728 ext4_abort(sb, "Couldn't clean up the journal");
731 del_timer(&sbi->s_err_report);
732 ext4_release_system_zone(sb);
734 ext4_ext_release(sb);
735 ext4_xattr_put_super(sb);
737 if (!(sb->s_flags & MS_RDONLY)) {
738 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
739 es->s_state = cpu_to_le16(sbi->s_mount_state);
740 ext4_commit_super(sb, 1);
743 remove_proc_entry(sb->s_id, ext4_proc_root);
745 kobject_del(&sbi->s_kobj);
747 for (i = 0; i < sbi->s_gdb_count; i++)
748 brelse(sbi->s_group_desc[i]);
749 kfree(sbi->s_group_desc);
750 if (is_vmalloc_addr(sbi->s_flex_groups))
751 vfree(sbi->s_flex_groups);
753 kfree(sbi->s_flex_groups);
754 percpu_counter_destroy(&sbi->s_freeblocks_counter);
755 percpu_counter_destroy(&sbi->s_freeinodes_counter);
756 percpu_counter_destroy(&sbi->s_dirs_counter);
757 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
760 for (i = 0; i < MAXQUOTAS; i++)
761 kfree(sbi->s_qf_names[i]);
764 /* Debugging code just in case the in-memory inode orphan list
765 * isn't empty. The on-disk one can be non-empty if we've
766 * detected an error and taken the fs readonly, but the
767 * in-memory list had better be clean by this point. */
768 if (!list_empty(&sbi->s_orphan))
769 dump_orphan_list(sb, sbi);
770 J_ASSERT(list_empty(&sbi->s_orphan));
772 invalidate_bdev(sb->s_bdev);
773 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
775 * Invalidate the journal device's buffers. We don't want them
776 * floating about in memory - the physical journal device may
777 * hotswapped, and it breaks the `ro-after' testing code.
779 sync_blockdev(sbi->journal_bdev);
780 invalidate_bdev(sbi->journal_bdev);
781 ext4_blkdev_remove(sbi);
783 sb->s_fs_info = NULL;
785 * Now that we are completely done shutting down the
786 * superblock, we need to actually destroy the kobject.
790 kobject_put(&sbi->s_kobj);
791 wait_for_completion(&sbi->s_kobj_unregister);
792 kfree(sbi->s_blockgroup_lock);
796 static struct kmem_cache *ext4_inode_cachep;
799 * Called inside transaction, so use GFP_NOFS
801 static struct inode *ext4_alloc_inode(struct super_block *sb)
803 struct ext4_inode_info *ei;
805 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
809 ei->vfs_inode.i_version = 1;
810 ei->vfs_inode.i_data.writeback_index = 0;
811 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
812 INIT_LIST_HEAD(&ei->i_prealloc_list);
813 spin_lock_init(&ei->i_prealloc_lock);
815 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
816 * therefore it can be null here. Don't check it, just initialize
819 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
820 ei->i_reserved_data_blocks = 0;
821 ei->i_reserved_meta_blocks = 0;
822 ei->i_allocated_meta_blocks = 0;
823 ei->i_da_metadata_calc_len = 0;
824 ei->i_delalloc_reserved_flag = 0;
825 spin_lock_init(&(ei->i_block_reservation_lock));
827 ei->i_reserved_quota = 0;
829 INIT_LIST_HEAD(&ei->i_completed_io_list);
830 spin_lock_init(&ei->i_completed_io_lock);
831 ei->cur_aio_dio = NULL;
833 ei->i_datasync_tid = 0;
835 return &ei->vfs_inode;
838 static void ext4_destroy_inode(struct inode *inode)
840 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
841 ext4_msg(inode->i_sb, KERN_ERR,
842 "Inode %lu (%p): orphan list check failed!",
843 inode->i_ino, EXT4_I(inode));
844 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
845 EXT4_I(inode), sizeof(struct ext4_inode_info),
849 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
852 static void init_once(void *foo)
854 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
856 INIT_LIST_HEAD(&ei->i_orphan);
857 #ifdef CONFIG_EXT4_FS_XATTR
858 init_rwsem(&ei->xattr_sem);
860 init_rwsem(&ei->i_data_sem);
861 inode_init_once(&ei->vfs_inode);
864 static int init_inodecache(void)
866 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
867 sizeof(struct ext4_inode_info),
868 0, (SLAB_RECLAIM_ACCOUNT|
871 if (ext4_inode_cachep == NULL)
876 static void destroy_inodecache(void)
878 kmem_cache_destroy(ext4_inode_cachep);
881 void ext4_clear_inode(struct inode *inode)
883 invalidate_inode_buffers(inode);
884 end_writeback(inode);
886 ext4_discard_preallocations(inode);
887 if (EXT4_JOURNAL(inode))
888 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
889 &EXT4_I(inode)->jinode);
892 static inline void ext4_show_quota_options(struct seq_file *seq,
893 struct super_block *sb)
895 #if defined(CONFIG_QUOTA)
896 struct ext4_sb_info *sbi = EXT4_SB(sb);
898 if (sbi->s_jquota_fmt) {
901 switch (sbi->s_jquota_fmt) {
912 seq_printf(seq, ",jqfmt=%s", fmtname);
915 if (sbi->s_qf_names[USRQUOTA])
916 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
918 if (sbi->s_qf_names[GRPQUOTA])
919 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
921 if (test_opt(sb, USRQUOTA))
922 seq_puts(seq, ",usrquota");
924 if (test_opt(sb, GRPQUOTA))
925 seq_puts(seq, ",grpquota");
931 * - it's set to a non-default value OR
932 * - if the per-sb default is different from the global default
934 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
937 unsigned long def_mount_opts;
938 struct super_block *sb = vfs->mnt_sb;
939 struct ext4_sb_info *sbi = EXT4_SB(sb);
940 struct ext4_super_block *es = sbi->s_es;
942 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
943 def_errors = le16_to_cpu(es->s_errors);
945 if (sbi->s_sb_block != 1)
946 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
947 if (test_opt(sb, MINIX_DF))
948 seq_puts(seq, ",minixdf");
949 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
950 seq_puts(seq, ",grpid");
951 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
952 seq_puts(seq, ",nogrpid");
953 if (sbi->s_resuid != EXT4_DEF_RESUID ||
954 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
955 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
957 if (sbi->s_resgid != EXT4_DEF_RESGID ||
958 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
959 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
961 if (test_opt(sb, ERRORS_RO)) {
962 if (def_errors == EXT4_ERRORS_PANIC ||
963 def_errors == EXT4_ERRORS_CONTINUE) {
964 seq_puts(seq, ",errors=remount-ro");
967 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
968 seq_puts(seq, ",errors=continue");
969 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
970 seq_puts(seq, ",errors=panic");
971 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
972 seq_puts(seq, ",nouid32");
973 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
974 seq_puts(seq, ",debug");
975 if (test_opt(sb, OLDALLOC))
976 seq_puts(seq, ",oldalloc");
977 #ifdef CONFIG_EXT4_FS_XATTR
978 if (test_opt(sb, XATTR_USER) &&
979 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
980 seq_puts(seq, ",user_xattr");
981 if (!test_opt(sb, XATTR_USER) &&
982 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
983 seq_puts(seq, ",nouser_xattr");
986 #ifdef CONFIG_EXT4_FS_POSIX_ACL
987 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
988 seq_puts(seq, ",acl");
989 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
990 seq_puts(seq, ",noacl");
992 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
993 seq_printf(seq, ",commit=%u",
994 (unsigned) (sbi->s_commit_interval / HZ));
996 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
997 seq_printf(seq, ",min_batch_time=%u",
998 (unsigned) sbi->s_min_batch_time);
1000 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1001 seq_printf(seq, ",max_batch_time=%u",
1002 (unsigned) sbi->s_min_batch_time);
1006 * We're changing the default of barrier mount option, so
1007 * let's always display its mount state so it's clear what its
1010 seq_puts(seq, ",barrier=");
1011 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1012 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1013 seq_puts(seq, ",journal_async_commit");
1014 else if (test_opt(sb, JOURNAL_CHECKSUM))
1015 seq_puts(seq, ",journal_checksum");
1016 if (test_opt(sb, I_VERSION))
1017 seq_puts(seq, ",i_version");
1018 if (!test_opt(sb, DELALLOC) &&
1019 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1020 seq_puts(seq, ",nodelalloc");
1023 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1025 * journal mode get enabled in different ways
1026 * So just print the value even if we didn't specify it
1028 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1029 seq_puts(seq, ",data=journal");
1030 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1031 seq_puts(seq, ",data=ordered");
1032 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1033 seq_puts(seq, ",data=writeback");
1035 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1036 seq_printf(seq, ",inode_readahead_blks=%u",
1037 sbi->s_inode_readahead_blks);
1039 if (test_opt(sb, DATA_ERR_ABORT))
1040 seq_puts(seq, ",data_err=abort");
1042 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1043 seq_puts(seq, ",noauto_da_alloc");
1045 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1046 seq_puts(seq, ",discard");
1048 if (test_opt(sb, NOLOAD))
1049 seq_puts(seq, ",norecovery");
1051 if (test_opt(sb, DIOREAD_NOLOCK))
1052 seq_puts(seq, ",dioread_nolock");
1054 if (test_opt(sb, BLOCK_VALIDITY) &&
1055 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1056 seq_puts(seq, ",block_validity");
1058 if (!test_opt(sb, INIT_INODE_TABLE))
1059 seq_puts(seq, ",noinit_inode_table");
1060 else if (sbi->s_li_wait_mult)
1061 seq_printf(seq, ",init_inode_table=%u",
1062 (unsigned) sbi->s_li_wait_mult);
1064 ext4_show_quota_options(seq, sb);
1069 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1070 u64 ino, u32 generation)
1072 struct inode *inode;
1074 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1075 return ERR_PTR(-ESTALE);
1076 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1077 return ERR_PTR(-ESTALE);
1079 /* iget isn't really right if the inode is currently unallocated!!
1081 * ext4_read_inode will return a bad_inode if the inode had been
1082 * deleted, so we should be safe.
1084 * Currently we don't know the generation for parent directory, so
1085 * a generation of 0 means "accept any"
1087 inode = ext4_iget(sb, ino);
1089 return ERR_CAST(inode);
1090 if (generation && inode->i_generation != generation) {
1092 return ERR_PTR(-ESTALE);
1098 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1099 int fh_len, int fh_type)
1101 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1102 ext4_nfs_get_inode);
1105 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1106 int fh_len, int fh_type)
1108 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1109 ext4_nfs_get_inode);
1113 * Try to release metadata pages (indirect blocks, directories) which are
1114 * mapped via the block device. Since these pages could have journal heads
1115 * which would prevent try_to_free_buffers() from freeing them, we must use
1116 * jbd2 layer's try_to_free_buffers() function to release them.
1118 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1121 journal_t *journal = EXT4_SB(sb)->s_journal;
1123 WARN_ON(PageChecked(page));
1124 if (!page_has_buffers(page))
1127 return jbd2_journal_try_to_free_buffers(journal, page,
1128 wait & ~__GFP_WAIT);
1129 return try_to_free_buffers(page);
1133 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1134 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1136 static int ext4_write_dquot(struct dquot *dquot);
1137 static int ext4_acquire_dquot(struct dquot *dquot);
1138 static int ext4_release_dquot(struct dquot *dquot);
1139 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1140 static int ext4_write_info(struct super_block *sb, int type);
1141 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1143 static int ext4_quota_off(struct super_block *sb, int type);
1144 static int ext4_quota_on_mount(struct super_block *sb, int type);
1145 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1146 size_t len, loff_t off);
1147 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1148 const char *data, size_t len, loff_t off);
1150 static const struct dquot_operations ext4_quota_operations = {
1152 .get_reserved_space = ext4_get_reserved_space,
1154 .write_dquot = ext4_write_dquot,
1155 .acquire_dquot = ext4_acquire_dquot,
1156 .release_dquot = ext4_release_dquot,
1157 .mark_dirty = ext4_mark_dquot_dirty,
1158 .write_info = ext4_write_info,
1159 .alloc_dquot = dquot_alloc,
1160 .destroy_dquot = dquot_destroy,
1163 static const struct quotactl_ops ext4_qctl_operations = {
1164 .quota_on = ext4_quota_on,
1165 .quota_off = ext4_quota_off,
1166 .quota_sync = dquot_quota_sync,
1167 .get_info = dquot_get_dqinfo,
1168 .set_info = dquot_set_dqinfo,
1169 .get_dqblk = dquot_get_dqblk,
1170 .set_dqblk = dquot_set_dqblk
1174 static const struct super_operations ext4_sops = {
1175 .alloc_inode = ext4_alloc_inode,
1176 .destroy_inode = ext4_destroy_inode,
1177 .write_inode = ext4_write_inode,
1178 .dirty_inode = ext4_dirty_inode,
1179 .evict_inode = ext4_evict_inode,
1180 .put_super = ext4_put_super,
1181 .sync_fs = ext4_sync_fs,
1182 .freeze_fs = ext4_freeze,
1183 .unfreeze_fs = ext4_unfreeze,
1184 .statfs = ext4_statfs,
1185 .remount_fs = ext4_remount,
1186 .show_options = ext4_show_options,
1188 .quota_read = ext4_quota_read,
1189 .quota_write = ext4_quota_write,
1191 .bdev_try_to_free_page = bdev_try_to_free_page,
1192 .trim_fs = ext4_trim_fs
1195 static const struct super_operations ext4_nojournal_sops = {
1196 .alloc_inode = ext4_alloc_inode,
1197 .destroy_inode = ext4_destroy_inode,
1198 .write_inode = ext4_write_inode,
1199 .dirty_inode = ext4_dirty_inode,
1200 .evict_inode = ext4_evict_inode,
1201 .write_super = ext4_write_super,
1202 .put_super = ext4_put_super,
1203 .statfs = ext4_statfs,
1204 .remount_fs = ext4_remount,
1205 .show_options = ext4_show_options,
1207 .quota_read = ext4_quota_read,
1208 .quota_write = ext4_quota_write,
1210 .bdev_try_to_free_page = bdev_try_to_free_page,
1213 static const struct export_operations ext4_export_ops = {
1214 .fh_to_dentry = ext4_fh_to_dentry,
1215 .fh_to_parent = ext4_fh_to_parent,
1216 .get_parent = ext4_get_parent,
1220 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1221 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1222 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1223 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1224 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1225 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1226 Opt_journal_update, Opt_journal_dev,
1227 Opt_journal_checksum, Opt_journal_async_commit,
1228 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1229 Opt_data_err_abort, Opt_data_err_ignore,
1230 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1231 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1232 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1233 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1234 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1235 Opt_block_validity, Opt_noblock_validity,
1236 Opt_inode_readahead_blks, Opt_journal_ioprio,
1237 Opt_dioread_nolock, Opt_dioread_lock,
1238 Opt_discard, Opt_nodiscard,
1239 Opt_init_inode_table, Opt_noinit_inode_table,
1242 static const match_table_t tokens = {
1243 {Opt_bsd_df, "bsddf"},
1244 {Opt_minix_df, "minixdf"},
1245 {Opt_grpid, "grpid"},
1246 {Opt_grpid, "bsdgroups"},
1247 {Opt_nogrpid, "nogrpid"},
1248 {Opt_nogrpid, "sysvgroups"},
1249 {Opt_resgid, "resgid=%u"},
1250 {Opt_resuid, "resuid=%u"},
1252 {Opt_err_cont, "errors=continue"},
1253 {Opt_err_panic, "errors=panic"},
1254 {Opt_err_ro, "errors=remount-ro"},
1255 {Opt_nouid32, "nouid32"},
1256 {Opt_debug, "debug"},
1257 {Opt_oldalloc, "oldalloc"},
1258 {Opt_orlov, "orlov"},
1259 {Opt_user_xattr, "user_xattr"},
1260 {Opt_nouser_xattr, "nouser_xattr"},
1262 {Opt_noacl, "noacl"},
1263 {Opt_noload, "noload"},
1264 {Opt_noload, "norecovery"},
1267 {Opt_commit, "commit=%u"},
1268 {Opt_min_batch_time, "min_batch_time=%u"},
1269 {Opt_max_batch_time, "max_batch_time=%u"},
1270 {Opt_journal_update, "journal=update"},
1271 {Opt_journal_dev, "journal_dev=%u"},
1272 {Opt_journal_checksum, "journal_checksum"},
1273 {Opt_journal_async_commit, "journal_async_commit"},
1274 {Opt_abort, "abort"},
1275 {Opt_data_journal, "data=journal"},
1276 {Opt_data_ordered, "data=ordered"},
1277 {Opt_data_writeback, "data=writeback"},
1278 {Opt_data_err_abort, "data_err=abort"},
1279 {Opt_data_err_ignore, "data_err=ignore"},
1280 {Opt_offusrjquota, "usrjquota="},
1281 {Opt_usrjquota, "usrjquota=%s"},
1282 {Opt_offgrpjquota, "grpjquota="},
1283 {Opt_grpjquota, "grpjquota=%s"},
1284 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1285 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1286 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1287 {Opt_grpquota, "grpquota"},
1288 {Opt_noquota, "noquota"},
1289 {Opt_quota, "quota"},
1290 {Opt_usrquota, "usrquota"},
1291 {Opt_barrier, "barrier=%u"},
1292 {Opt_barrier, "barrier"},
1293 {Opt_nobarrier, "nobarrier"},
1294 {Opt_i_version, "i_version"},
1295 {Opt_stripe, "stripe=%u"},
1296 {Opt_resize, "resize"},
1297 {Opt_delalloc, "delalloc"},
1298 {Opt_nodelalloc, "nodelalloc"},
1299 {Opt_block_validity, "block_validity"},
1300 {Opt_noblock_validity, "noblock_validity"},
1301 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1302 {Opt_journal_ioprio, "journal_ioprio=%u"},
1303 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1304 {Opt_auto_da_alloc, "auto_da_alloc"},
1305 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1306 {Opt_dioread_nolock, "dioread_nolock"},
1307 {Opt_dioread_lock, "dioread_lock"},
1308 {Opt_discard, "discard"},
1309 {Opt_nodiscard, "nodiscard"},
1310 {Opt_init_inode_table, "init_itable=%u"},
1311 {Opt_init_inode_table, "init_itable"},
1312 {Opt_noinit_inode_table, "noinit_itable"},
1316 static ext4_fsblk_t get_sb_block(void **data)
1318 ext4_fsblk_t sb_block;
1319 char *options = (char *) *data;
1321 if (!options || strncmp(options, "sb=", 3) != 0)
1322 return 1; /* Default location */
1325 /* TODO: use simple_strtoll with >32bit ext4 */
1326 sb_block = simple_strtoul(options, &options, 0);
1327 if (*options && *options != ',') {
1328 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1332 if (*options == ',')
1334 *data = (void *) options;
1339 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1340 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1341 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1344 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1346 struct ext4_sb_info *sbi = EXT4_SB(sb);
1349 if (sb_any_quota_loaded(sb) &&
1350 !sbi->s_qf_names[qtype]) {
1351 ext4_msg(sb, KERN_ERR,
1352 "Cannot change journaled "
1353 "quota options when quota turned on");
1356 qname = match_strdup(args);
1358 ext4_msg(sb, KERN_ERR,
1359 "Not enough memory for storing quotafile name");
1362 if (sbi->s_qf_names[qtype] &&
1363 strcmp(sbi->s_qf_names[qtype], qname)) {
1364 ext4_msg(sb, KERN_ERR,
1365 "%s quota file already specified", QTYPE2NAME(qtype));
1369 sbi->s_qf_names[qtype] = qname;
1370 if (strchr(sbi->s_qf_names[qtype], '/')) {
1371 ext4_msg(sb, KERN_ERR,
1372 "quotafile must be on filesystem root");
1373 kfree(sbi->s_qf_names[qtype]);
1374 sbi->s_qf_names[qtype] = NULL;
1377 set_opt(sbi->s_mount_opt, QUOTA);
1381 static int clear_qf_name(struct super_block *sb, int qtype)
1384 struct ext4_sb_info *sbi = EXT4_SB(sb);
1386 if (sb_any_quota_loaded(sb) &&
1387 sbi->s_qf_names[qtype]) {
1388 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1389 " when quota turned on");
1393 * The space will be released later when all options are confirmed
1396 sbi->s_qf_names[qtype] = NULL;
1401 static int parse_options(char *options, struct super_block *sb,
1402 unsigned long *journal_devnum,
1403 unsigned int *journal_ioprio,
1404 ext4_fsblk_t *n_blocks_count, int is_remount)
1406 struct ext4_sb_info *sbi = EXT4_SB(sb);
1408 substring_t args[MAX_OPT_ARGS];
1418 while ((p = strsep(&options, ",")) != NULL) {
1424 * Initialize args struct so we know whether arg was
1425 * found; some options take optional arguments.
1427 args[0].to = args[0].from = 0;
1428 token = match_token(p, tokens, args);
1431 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1432 clear_opt(sbi->s_mount_opt, MINIX_DF);
1435 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1436 set_opt(sbi->s_mount_opt, MINIX_DF);
1440 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1441 set_opt(sbi->s_mount_opt, GRPID);
1445 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1446 clear_opt(sbi->s_mount_opt, GRPID);
1450 if (match_int(&args[0], &option))
1452 sbi->s_resuid = option;
1455 if (match_int(&args[0], &option))
1457 sbi->s_resgid = option;
1460 /* handled by get_sb_block() instead of here */
1461 /* *sb_block = match_int(&args[0]); */
1464 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1465 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1466 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1469 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1470 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1471 set_opt(sbi->s_mount_opt, ERRORS_RO);
1474 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1475 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1476 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1479 set_opt(sbi->s_mount_opt, NO_UID32);
1482 set_opt(sbi->s_mount_opt, DEBUG);
1485 set_opt(sbi->s_mount_opt, OLDALLOC);
1488 clear_opt(sbi->s_mount_opt, OLDALLOC);
1490 #ifdef CONFIG_EXT4_FS_XATTR
1491 case Opt_user_xattr:
1492 set_opt(sbi->s_mount_opt, XATTR_USER);
1494 case Opt_nouser_xattr:
1495 clear_opt(sbi->s_mount_opt, XATTR_USER);
1498 case Opt_user_xattr:
1499 case Opt_nouser_xattr:
1500 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1503 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1505 set_opt(sbi->s_mount_opt, POSIX_ACL);
1508 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1513 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1516 case Opt_journal_update:
1518 /* Eventually we will want to be able to create
1519 a journal file here. For now, only allow the
1520 user to specify an existing inode to be the
1523 ext4_msg(sb, KERN_ERR,
1524 "Cannot specify journal on remount");
1527 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1529 case Opt_journal_dev:
1531 ext4_msg(sb, KERN_ERR,
1532 "Cannot specify journal on remount");
1535 if (match_int(&args[0], &option))
1537 *journal_devnum = option;
1539 case Opt_journal_checksum:
1540 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1542 case Opt_journal_async_commit:
1543 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1544 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1547 set_opt(sbi->s_mount_opt, NOLOAD);
1550 if (match_int(&args[0], &option))
1555 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1556 sbi->s_commit_interval = HZ * option;
1558 case Opt_max_batch_time:
1559 if (match_int(&args[0], &option))
1564 option = EXT4_DEF_MAX_BATCH_TIME;
1565 sbi->s_max_batch_time = option;
1567 case Opt_min_batch_time:
1568 if (match_int(&args[0], &option))
1572 sbi->s_min_batch_time = option;
1574 case Opt_data_journal:
1575 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1577 case Opt_data_ordered:
1578 data_opt = EXT4_MOUNT_ORDERED_DATA;
1580 case Opt_data_writeback:
1581 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1584 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1585 ext4_msg(sb, KERN_ERR,
1586 "Cannot change data mode on remount");
1590 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1591 sbi->s_mount_opt |= data_opt;
1594 case Opt_data_err_abort:
1595 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1597 case Opt_data_err_ignore:
1598 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1602 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1606 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1609 case Opt_offusrjquota:
1610 if (!clear_qf_name(sb, USRQUOTA))
1613 case Opt_offgrpjquota:
1614 if (!clear_qf_name(sb, GRPQUOTA))
1618 case Opt_jqfmt_vfsold:
1619 qfmt = QFMT_VFS_OLD;
1621 case Opt_jqfmt_vfsv0:
1624 case Opt_jqfmt_vfsv1:
1627 if (sb_any_quota_loaded(sb) &&
1628 sbi->s_jquota_fmt != qfmt) {
1629 ext4_msg(sb, KERN_ERR, "Cannot change "
1630 "journaled quota options when "
1634 sbi->s_jquota_fmt = qfmt;
1638 set_opt(sbi->s_mount_opt, QUOTA);
1639 set_opt(sbi->s_mount_opt, USRQUOTA);
1642 set_opt(sbi->s_mount_opt, QUOTA);
1643 set_opt(sbi->s_mount_opt, GRPQUOTA);
1646 if (sb_any_quota_loaded(sb)) {
1647 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1648 "options when quota turned on");
1651 clear_opt(sbi->s_mount_opt, QUOTA);
1652 clear_opt(sbi->s_mount_opt, USRQUOTA);
1653 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1659 ext4_msg(sb, KERN_ERR,
1660 "quota options not supported");
1664 case Opt_offusrjquota:
1665 case Opt_offgrpjquota:
1666 case Opt_jqfmt_vfsold:
1667 case Opt_jqfmt_vfsv0:
1668 case Opt_jqfmt_vfsv1:
1669 ext4_msg(sb, KERN_ERR,
1670 "journaled quota options not supported");
1676 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1679 clear_opt(sbi->s_mount_opt, BARRIER);
1683 if (match_int(&args[0], &option))
1686 option = 1; /* No argument, default to 1 */
1688 set_opt(sbi->s_mount_opt, BARRIER);
1690 clear_opt(sbi->s_mount_opt, BARRIER);
1696 ext4_msg(sb, KERN_ERR,
1697 "resize option only available "
1701 if (match_int(&args[0], &option) != 0)
1703 *n_blocks_count = option;
1706 ext4_msg(sb, KERN_WARNING,
1707 "Ignoring deprecated nobh option");
1710 ext4_msg(sb, KERN_WARNING,
1711 "Ignoring deprecated bh option");
1714 set_opt(sbi->s_mount_opt, I_VERSION);
1715 sb->s_flags |= MS_I_VERSION;
1717 case Opt_nodelalloc:
1718 clear_opt(sbi->s_mount_opt, DELALLOC);
1721 if (match_int(&args[0], &option))
1725 sbi->s_stripe = option;
1728 set_opt(sbi->s_mount_opt, DELALLOC);
1730 case Opt_block_validity:
1731 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1733 case Opt_noblock_validity:
1734 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1736 case Opt_inode_readahead_blks:
1737 if (match_int(&args[0], &option))
1739 if (option < 0 || option > (1 << 30))
1741 if (!is_power_of_2(option)) {
1742 ext4_msg(sb, KERN_ERR,
1743 "EXT4-fs: inode_readahead_blks"
1744 " must be a power of 2");
1747 sbi->s_inode_readahead_blks = option;
1749 case Opt_journal_ioprio:
1750 if (match_int(&args[0], &option))
1752 if (option < 0 || option > 7)
1754 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1757 case Opt_noauto_da_alloc:
1758 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1760 case Opt_auto_da_alloc:
1762 if (match_int(&args[0], &option))
1765 option = 1; /* No argument, default to 1 */
1767 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1769 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1772 set_opt(sbi->s_mount_opt, DISCARD);
1775 clear_opt(sbi->s_mount_opt, DISCARD);
1777 case Opt_dioread_nolock:
1778 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1780 case Opt_dioread_lock:
1781 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1783 case Opt_init_inode_table:
1784 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1786 if (match_int(&args[0], &option))
1789 option = EXT4_DEF_LI_WAIT_MULT;
1792 sbi->s_li_wait_mult = option;
1794 case Opt_noinit_inode_table:
1795 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1798 ext4_msg(sb, KERN_ERR,
1799 "Unrecognized mount option \"%s\" "
1800 "or missing value", p);
1805 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1806 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1807 clear_opt(sbi->s_mount_opt, USRQUOTA);
1809 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1810 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1812 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1813 ext4_msg(sb, KERN_ERR, "old and new quota "
1818 if (!sbi->s_jquota_fmt) {
1819 ext4_msg(sb, KERN_ERR, "journaled quota format "
1824 if (sbi->s_jquota_fmt) {
1825 ext4_msg(sb, KERN_ERR, "journaled quota format "
1826 "specified with no journaling "
1835 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1838 struct ext4_sb_info *sbi = EXT4_SB(sb);
1841 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1842 ext4_msg(sb, KERN_ERR, "revision level too high, "
1843 "forcing read-only mode");
1848 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1849 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1850 "running e2fsck is recommended");
1851 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1852 ext4_msg(sb, KERN_WARNING,
1853 "warning: mounting fs with errors, "
1854 "running e2fsck is recommended");
1855 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1856 le16_to_cpu(es->s_mnt_count) >=
1857 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1858 ext4_msg(sb, KERN_WARNING,
1859 "warning: maximal mount count reached, "
1860 "running e2fsck is recommended");
1861 else if (le32_to_cpu(es->s_checkinterval) &&
1862 (le32_to_cpu(es->s_lastcheck) +
1863 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1864 ext4_msg(sb, KERN_WARNING,
1865 "warning: checktime reached, "
1866 "running e2fsck is recommended");
1867 if (!sbi->s_journal)
1868 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1869 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1870 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1871 le16_add_cpu(&es->s_mnt_count, 1);
1872 es->s_mtime = cpu_to_le32(get_seconds());
1873 ext4_update_dynamic_rev(sb);
1875 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1877 ext4_commit_super(sb, 1);
1878 if (test_opt(sb, DEBUG))
1879 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1880 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1882 sbi->s_groups_count,
1883 EXT4_BLOCKS_PER_GROUP(sb),
1884 EXT4_INODES_PER_GROUP(sb),
1890 static int ext4_fill_flex_info(struct super_block *sb)
1892 struct ext4_sb_info *sbi = EXT4_SB(sb);
1893 struct ext4_group_desc *gdp = NULL;
1894 ext4_group_t flex_group_count;
1895 ext4_group_t flex_group;
1896 int groups_per_flex = 0;
1900 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1901 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1903 if (groups_per_flex < 2) {
1904 sbi->s_log_groups_per_flex = 0;
1908 /* We allocate both existing and potentially added groups */
1909 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1910 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1911 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1912 size = flex_group_count * sizeof(struct flex_groups);
1913 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1914 if (sbi->s_flex_groups == NULL) {
1915 sbi->s_flex_groups = vmalloc(size);
1916 if (sbi->s_flex_groups)
1917 memset(sbi->s_flex_groups, 0, size);
1919 if (sbi->s_flex_groups == NULL) {
1920 ext4_msg(sb, KERN_ERR, "not enough memory for "
1921 "%u flex groups", flex_group_count);
1925 for (i = 0; i < sbi->s_groups_count; i++) {
1926 gdp = ext4_get_group_desc(sb, i, NULL);
1928 flex_group = ext4_flex_group(sbi, i);
1929 atomic_add(ext4_free_inodes_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].free_inodes);
1931 atomic_add(ext4_free_blks_count(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].free_blocks);
1933 atomic_add(ext4_used_dirs_count(sb, gdp),
1934 &sbi->s_flex_groups[flex_group].used_dirs);
1942 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1943 struct ext4_group_desc *gdp)
1947 if (sbi->s_es->s_feature_ro_compat &
1948 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1949 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1950 __le32 le_group = cpu_to_le32(block_group);
1952 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1953 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1954 crc = crc16(crc, (__u8 *)gdp, offset);
1955 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1956 /* for checksum of struct ext4_group_desc do the rest...*/
1957 if ((sbi->s_es->s_feature_incompat &
1958 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1959 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1960 crc = crc16(crc, (__u8 *)gdp + offset,
1961 le16_to_cpu(sbi->s_es->s_desc_size) -
1965 return cpu_to_le16(crc);
1968 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1969 struct ext4_group_desc *gdp)
1971 if ((sbi->s_es->s_feature_ro_compat &
1972 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1973 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1979 /* Called at mount-time, super-block is locked */
1980 static int ext4_check_descriptors(struct super_block *sb,
1981 ext4_group_t *first_not_zeroed)
1983 struct ext4_sb_info *sbi = EXT4_SB(sb);
1984 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1985 ext4_fsblk_t last_block;
1986 ext4_fsblk_t block_bitmap;
1987 ext4_fsblk_t inode_bitmap;
1988 ext4_fsblk_t inode_table;
1989 int flexbg_flag = 0;
1990 ext4_group_t i, grp = sbi->s_groups_count;
1992 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1995 ext4_debug("Checking group descriptors");
1997 for (i = 0; i < sbi->s_groups_count; i++) {
1998 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2000 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2001 last_block = ext4_blocks_count(sbi->s_es) - 1;
2003 last_block = first_block +
2004 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2006 if ((grp == sbi->s_groups_count) &&
2007 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2010 block_bitmap = ext4_block_bitmap(sb, gdp);
2011 if (block_bitmap < first_block || block_bitmap > last_block) {
2012 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2013 "Block bitmap for group %u not in group "
2014 "(block %llu)!", i, block_bitmap);
2017 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2018 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2019 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2020 "Inode bitmap for group %u not in group "
2021 "(block %llu)!", i, inode_bitmap);
2024 inode_table = ext4_inode_table(sb, gdp);
2025 if (inode_table < first_block ||
2026 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2027 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2028 "Inode table for group %u not in group "
2029 "(block %llu)!", i, inode_table);
2032 ext4_lock_group(sb, i);
2033 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2034 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2035 "Checksum for group %u failed (%u!=%u)",
2036 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2037 gdp)), le16_to_cpu(gdp->bg_checksum));
2038 if (!(sb->s_flags & MS_RDONLY)) {
2039 ext4_unlock_group(sb, i);
2043 ext4_unlock_group(sb, i);
2045 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2047 if (NULL != first_not_zeroed)
2048 *first_not_zeroed = grp;
2050 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2051 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2055 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2056 * the superblock) which were deleted from all directories, but held open by
2057 * a process at the time of a crash. We walk the list and try to delete these
2058 * inodes at recovery time (only with a read-write filesystem).
2060 * In order to keep the orphan inode chain consistent during traversal (in
2061 * case of crash during recovery), we link each inode into the superblock
2062 * orphan list_head and handle it the same way as an inode deletion during
2063 * normal operation (which journals the operations for us).
2065 * We only do an iget() and an iput() on each inode, which is very safe if we
2066 * accidentally point at an in-use or already deleted inode. The worst that
2067 * can happen in this case is that we get a "bit already cleared" message from
2068 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2069 * e2fsck was run on this filesystem, and it must have already done the orphan
2070 * inode cleanup for us, so we can safely abort without any further action.
2072 static void ext4_orphan_cleanup(struct super_block *sb,
2073 struct ext4_super_block *es)
2075 unsigned int s_flags = sb->s_flags;
2076 int nr_orphans = 0, nr_truncates = 0;
2080 if (!es->s_last_orphan) {
2081 jbd_debug(4, "no orphan inodes to clean up\n");
2085 if (bdev_read_only(sb->s_bdev)) {
2086 ext4_msg(sb, KERN_ERR, "write access "
2087 "unavailable, skipping orphan cleanup");
2091 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2092 if (es->s_last_orphan)
2093 jbd_debug(1, "Errors on filesystem, "
2094 "clearing orphan list.\n");
2095 es->s_last_orphan = 0;
2096 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2100 if (s_flags & MS_RDONLY) {
2101 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2102 sb->s_flags &= ~MS_RDONLY;
2105 /* Needed for iput() to work correctly and not trash data */
2106 sb->s_flags |= MS_ACTIVE;
2107 /* Turn on quotas so that they are updated correctly */
2108 for (i = 0; i < MAXQUOTAS; i++) {
2109 if (EXT4_SB(sb)->s_qf_names[i]) {
2110 int ret = ext4_quota_on_mount(sb, i);
2112 ext4_msg(sb, KERN_ERR,
2113 "Cannot turn on journaled "
2114 "quota: error %d", ret);
2119 while (es->s_last_orphan) {
2120 struct inode *inode;
2122 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2123 if (IS_ERR(inode)) {
2124 es->s_last_orphan = 0;
2128 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2129 dquot_initialize(inode);
2130 if (inode->i_nlink) {
2131 ext4_msg(sb, KERN_DEBUG,
2132 "%s: truncating inode %lu to %lld bytes",
2133 __func__, inode->i_ino, inode->i_size);
2134 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2135 inode->i_ino, inode->i_size);
2136 ext4_truncate(inode);
2139 ext4_msg(sb, KERN_DEBUG,
2140 "%s: deleting unreferenced inode %lu",
2141 __func__, inode->i_ino);
2142 jbd_debug(2, "deleting unreferenced inode %lu\n",
2146 iput(inode); /* The delete magic happens here! */
2149 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2152 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2153 PLURAL(nr_orphans));
2155 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2156 PLURAL(nr_truncates));
2158 /* Turn quotas off */
2159 for (i = 0; i < MAXQUOTAS; i++) {
2160 if (sb_dqopt(sb)->files[i])
2161 dquot_quota_off(sb, i);
2164 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2168 * Maximal extent format file size.
2169 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2170 * extent format containers, within a sector_t, and within i_blocks
2171 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2172 * so that won't be a limiting factor.
2174 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2176 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2179 loff_t upper_limit = MAX_LFS_FILESIZE;
2181 /* small i_blocks in vfs inode? */
2182 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2184 * CONFIG_LBDAF is not enabled implies the inode
2185 * i_block represent total blocks in 512 bytes
2186 * 32 == size of vfs inode i_blocks * 8
2188 upper_limit = (1LL << 32) - 1;
2190 /* total blocks in file system block size */
2191 upper_limit >>= (blkbits - 9);
2192 upper_limit <<= blkbits;
2195 /* 32-bit extent-start container, ee_block */
2200 /* Sanity check against vm- & vfs- imposed limits */
2201 if (res > upper_limit)
2208 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2209 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2210 * We need to be 1 filesystem block less than the 2^48 sector limit.
2212 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2214 loff_t res = EXT4_NDIR_BLOCKS;
2217 /* This is calculated to be the largest file size for a dense, block
2218 * mapped file such that the file's total number of 512-byte sectors,
2219 * including data and all indirect blocks, does not exceed (2^48 - 1).
2221 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2222 * number of 512-byte sectors of the file.
2225 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2227 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2228 * the inode i_block field represents total file blocks in
2229 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2231 upper_limit = (1LL << 32) - 1;
2233 /* total blocks in file system block size */
2234 upper_limit >>= (bits - 9);
2238 * We use 48 bit ext4_inode i_blocks
2239 * With EXT4_HUGE_FILE_FL set the i_blocks
2240 * represent total number of blocks in
2241 * file system block size
2243 upper_limit = (1LL << 48) - 1;
2247 /* indirect blocks */
2249 /* double indirect blocks */
2250 meta_blocks += 1 + (1LL << (bits-2));
2251 /* tripple indirect blocks */
2252 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2254 upper_limit -= meta_blocks;
2255 upper_limit <<= bits;
2257 res += 1LL << (bits-2);
2258 res += 1LL << (2*(bits-2));
2259 res += 1LL << (3*(bits-2));
2261 if (res > upper_limit)
2264 if (res > MAX_LFS_FILESIZE)
2265 res = MAX_LFS_FILESIZE;
2270 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2271 ext4_fsblk_t logical_sb_block, int nr)
2273 struct ext4_sb_info *sbi = EXT4_SB(sb);
2274 ext4_group_t bg, first_meta_bg;
2277 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2279 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2281 return logical_sb_block + nr + 1;
2282 bg = sbi->s_desc_per_block * nr;
2283 if (ext4_bg_has_super(sb, bg))
2286 return (has_super + ext4_group_first_block_no(sb, bg));
2290 * ext4_get_stripe_size: Get the stripe size.
2291 * @sbi: In memory super block info
2293 * If we have specified it via mount option, then
2294 * use the mount option value. If the value specified at mount time is
2295 * greater than the blocks per group use the super block value.
2296 * If the super block value is greater than blocks per group return 0.
2297 * Allocator needs it be less than blocks per group.
2300 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2302 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2303 unsigned long stripe_width =
2304 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2306 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2307 return sbi->s_stripe;
2309 if (stripe_width <= sbi->s_blocks_per_group)
2310 return stripe_width;
2312 if (stride <= sbi->s_blocks_per_group)
2321 struct attribute attr;
2322 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2323 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2324 const char *, size_t);
2328 static int parse_strtoul(const char *buf,
2329 unsigned long max, unsigned long *value)
2333 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2334 endp = skip_spaces(endp);
2335 if (*endp || *value > max)
2341 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2342 struct ext4_sb_info *sbi,
2345 return snprintf(buf, PAGE_SIZE, "%llu\n",
2346 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2349 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2350 struct ext4_sb_info *sbi, char *buf)
2352 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2354 if (!sb->s_bdev->bd_part)
2355 return snprintf(buf, PAGE_SIZE, "0\n");
2356 return snprintf(buf, PAGE_SIZE, "%lu\n",
2357 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2358 sbi->s_sectors_written_start) >> 1);
2361 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2362 struct ext4_sb_info *sbi, char *buf)
2364 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2366 if (!sb->s_bdev->bd_part)
2367 return snprintf(buf, PAGE_SIZE, "0\n");
2368 return snprintf(buf, PAGE_SIZE, "%llu\n",
2369 (unsigned long long)(sbi->s_kbytes_written +
2370 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2371 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2374 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2375 struct ext4_sb_info *sbi,
2376 const char *buf, size_t count)
2380 if (parse_strtoul(buf, 0x40000000, &t))
2383 if (!is_power_of_2(t))
2386 sbi->s_inode_readahead_blks = t;
2390 static ssize_t sbi_ui_show(struct ext4_attr *a,
2391 struct ext4_sb_info *sbi, char *buf)
2393 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2395 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2398 static ssize_t sbi_ui_store(struct ext4_attr *a,
2399 struct ext4_sb_info *sbi,
2400 const char *buf, size_t count)
2402 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2405 if (parse_strtoul(buf, 0xffffffff, &t))
2411 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2412 static struct ext4_attr ext4_attr_##_name = { \
2413 .attr = {.name = __stringify(_name), .mode = _mode }, \
2416 .offset = offsetof(struct ext4_sb_info, _elname), \
2418 #define EXT4_ATTR(name, mode, show, store) \
2419 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2421 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2422 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2423 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2424 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2425 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2426 #define ATTR_LIST(name) &ext4_attr_##name.attr
2428 EXT4_RO_ATTR(delayed_allocation_blocks);
2429 EXT4_RO_ATTR(session_write_kbytes);
2430 EXT4_RO_ATTR(lifetime_write_kbytes);
2431 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2432 inode_readahead_blks_store, s_inode_readahead_blks);
2433 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2434 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2435 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2436 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2437 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2438 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2439 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2440 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2442 static struct attribute *ext4_attrs[] = {
2443 ATTR_LIST(delayed_allocation_blocks),
2444 ATTR_LIST(session_write_kbytes),
2445 ATTR_LIST(lifetime_write_kbytes),
2446 ATTR_LIST(inode_readahead_blks),
2447 ATTR_LIST(inode_goal),
2448 ATTR_LIST(mb_stats),
2449 ATTR_LIST(mb_max_to_scan),
2450 ATTR_LIST(mb_min_to_scan),
2451 ATTR_LIST(mb_order2_req),
2452 ATTR_LIST(mb_stream_req),
2453 ATTR_LIST(mb_group_prealloc),
2454 ATTR_LIST(max_writeback_mb_bump),
2458 /* Features this copy of ext4 supports */
2459 EXT4_INFO_ATTR(lazy_itable_init);
2460 EXT4_INFO_ATTR(batched_discard);
2462 static struct attribute *ext4_feat_attrs[] = {
2463 ATTR_LIST(lazy_itable_init),
2464 ATTR_LIST(batched_discard),
2468 static ssize_t ext4_attr_show(struct kobject *kobj,
2469 struct attribute *attr, char *buf)
2471 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2473 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2475 return a->show ? a->show(a, sbi, buf) : 0;
2478 static ssize_t ext4_attr_store(struct kobject *kobj,
2479 struct attribute *attr,
2480 const char *buf, size_t len)
2482 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2484 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2486 return a->store ? a->store(a, sbi, buf, len) : 0;
2489 static void ext4_sb_release(struct kobject *kobj)
2491 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2493 complete(&sbi->s_kobj_unregister);
2496 static const struct sysfs_ops ext4_attr_ops = {
2497 .show = ext4_attr_show,
2498 .store = ext4_attr_store,
2501 static struct kobj_type ext4_ktype = {
2502 .default_attrs = ext4_attrs,
2503 .sysfs_ops = &ext4_attr_ops,
2504 .release = ext4_sb_release,
2507 static void ext4_feat_release(struct kobject *kobj)
2509 complete(&ext4_feat->f_kobj_unregister);
2512 static struct kobj_type ext4_feat_ktype = {
2513 .default_attrs = ext4_feat_attrs,
2514 .sysfs_ops = &ext4_attr_ops,
2515 .release = ext4_feat_release,
2519 * Check whether this filesystem can be mounted based on
2520 * the features present and the RDONLY/RDWR mount requested.
2521 * Returns 1 if this filesystem can be mounted as requested,
2522 * 0 if it cannot be.
2524 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2526 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2527 ext4_msg(sb, KERN_ERR,
2528 "Couldn't mount because of "
2529 "unsupported optional features (%x)",
2530 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2531 ~EXT4_FEATURE_INCOMPAT_SUPP));
2538 /* Check that feature set is OK for a read-write mount */
2539 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2540 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2541 "unsupported optional features (%x)",
2542 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2543 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2547 * Large file size enabled file system can only be mounted
2548 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2550 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2551 if (sizeof(blkcnt_t) < sizeof(u64)) {
2552 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2553 "cannot be mounted RDWR without "
2562 * This function is called once a day if we have errors logged
2563 * on the file system
2565 static void print_daily_error_info(unsigned long arg)
2567 struct super_block *sb = (struct super_block *) arg;
2568 struct ext4_sb_info *sbi;
2569 struct ext4_super_block *es;
2574 if (es->s_error_count)
2575 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2576 le32_to_cpu(es->s_error_count));
2577 if (es->s_first_error_time) {
2578 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2579 sb->s_id, le32_to_cpu(es->s_first_error_time),
2580 (int) sizeof(es->s_first_error_func),
2581 es->s_first_error_func,
2582 le32_to_cpu(es->s_first_error_line));
2583 if (es->s_first_error_ino)
2584 printk(": inode %u",
2585 le32_to_cpu(es->s_first_error_ino));
2586 if (es->s_first_error_block)
2587 printk(": block %llu", (unsigned long long)
2588 le64_to_cpu(es->s_first_error_block));
2591 if (es->s_last_error_time) {
2592 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2593 sb->s_id, le32_to_cpu(es->s_last_error_time),
2594 (int) sizeof(es->s_last_error_func),
2595 es->s_last_error_func,
2596 le32_to_cpu(es->s_last_error_line));
2597 if (es->s_last_error_ino)
2598 printk(": inode %u",
2599 le32_to_cpu(es->s_last_error_ino));
2600 if (es->s_last_error_block)
2601 printk(": block %llu", (unsigned long long)
2602 le64_to_cpu(es->s_last_error_block));
2605 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2608 static void ext4_lazyinode_timeout(unsigned long data)
2610 struct task_struct *p = (struct task_struct *)data;
2614 /* Find next suitable group and run ext4_init_inode_table */
2615 static int ext4_run_li_request(struct ext4_li_request *elr)
2617 struct ext4_group_desc *gdp = NULL;
2618 ext4_group_t group, ngroups;
2619 struct super_block *sb;
2620 unsigned long timeout = 0;
2624 ngroups = EXT4_SB(sb)->s_groups_count;
2626 for (group = elr->lr_next_group; group < ngroups; group++) {
2627 gdp = ext4_get_group_desc(sb, group, NULL);
2633 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2637 if (group == ngroups)
2642 ret = ext4_init_inode_table(sb, group,
2643 elr->lr_timeout ? 0 : 1);
2644 if (elr->lr_timeout == 0) {
2645 timeout = jiffies - timeout;
2646 if (elr->lr_sbi->s_li_wait_mult)
2647 timeout *= elr->lr_sbi->s_li_wait_mult;
2650 elr->lr_timeout = timeout;
2652 elr->lr_next_sched = jiffies + elr->lr_timeout;
2653 elr->lr_next_group = group + 1;
2660 * Remove lr_request from the list_request and free the
2661 * request tructure. Should be called with li_list_mtx held
2663 static void ext4_remove_li_request(struct ext4_li_request *elr)
2665 struct ext4_sb_info *sbi;
2672 list_del(&elr->lr_request);
2673 sbi->s_li_request = NULL;
2677 static void ext4_unregister_li_request(struct super_block *sb)
2679 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2684 mutex_lock(&ext4_li_info->li_list_mtx);
2685 ext4_remove_li_request(elr);
2686 mutex_unlock(&ext4_li_info->li_list_mtx);
2690 * This is the function where ext4lazyinit thread lives. It walks
2691 * through the request list searching for next scheduled filesystem.
2692 * When such a fs is found, run the lazy initialization request
2693 * (ext4_rn_li_request) and keep track of the time spend in this
2694 * function. Based on that time we compute next schedule time of
2695 * the request. When walking through the list is complete, compute
2696 * next waking time and put itself into sleep.
2698 static int ext4_lazyinit_thread(void *arg)
2700 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2701 struct list_head *pos, *n;
2702 struct ext4_li_request *elr;
2703 unsigned long next_wakeup;
2707 BUG_ON(NULL == eli);
2709 eli->li_timer.data = (unsigned long)current;
2710 eli->li_timer.function = ext4_lazyinode_timeout;
2712 eli->li_task = current;
2713 wake_up(&eli->li_wait_task);
2717 next_wakeup = MAX_JIFFY_OFFSET;
2719 mutex_lock(&eli->li_list_mtx);
2720 if (list_empty(&eli->li_request_list)) {
2721 mutex_unlock(&eli->li_list_mtx);
2725 list_for_each_safe(pos, n, &eli->li_request_list) {
2726 elr = list_entry(pos, struct ext4_li_request,
2729 if (time_after_eq(jiffies, elr->lr_next_sched))
2730 ret = ext4_run_li_request(elr);
2734 ext4_remove_li_request(elr);
2738 if (time_before(elr->lr_next_sched, next_wakeup))
2739 next_wakeup = elr->lr_next_sched;
2741 mutex_unlock(&eli->li_list_mtx);
2743 if (freezing(current))
2746 if (time_after_eq(jiffies, next_wakeup)) {
2751 eli->li_timer.expires = next_wakeup;
2752 add_timer(&eli->li_timer);
2753 prepare_to_wait(&eli->li_wait_daemon, &wait,
2754 TASK_INTERRUPTIBLE);
2755 if (time_before(jiffies, next_wakeup))
2757 finish_wait(&eli->li_wait_daemon, &wait);
2762 * It looks like the request list is empty, but we need
2763 * to check it under the li_list_mtx lock, to prevent any
2764 * additions into it, and of course we should lock ext4_li_mtx
2765 * to atomically free the list and ext4_li_info, because at
2766 * this point another ext4 filesystem could be registering
2769 mutex_lock(&ext4_li_mtx);
2770 mutex_lock(&eli->li_list_mtx);
2771 if (!list_empty(&eli->li_request_list)) {
2772 mutex_unlock(&eli->li_list_mtx);
2773 mutex_unlock(&ext4_li_mtx);
2776 mutex_unlock(&eli->li_list_mtx);
2777 del_timer_sync(&ext4_li_info->li_timer);
2778 eli->li_task = NULL;
2779 wake_up(&eli->li_wait_task);
2781 kfree(ext4_li_info);
2782 ext4_li_info = NULL;
2783 mutex_unlock(&ext4_li_mtx);
2788 static void ext4_clear_request_list(void)
2790 struct list_head *pos, *n;
2791 struct ext4_li_request *elr;
2793 mutex_lock(&ext4_li_info->li_list_mtx);
2794 if (list_empty(&ext4_li_info->li_request_list))
2797 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2798 elr = list_entry(pos, struct ext4_li_request,
2800 ext4_remove_li_request(elr);
2802 mutex_unlock(&ext4_li_info->li_list_mtx);
2805 static int ext4_run_lazyinit_thread(void)
2807 struct task_struct *t;
2809 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2811 int err = PTR_ERR(t);
2812 ext4_clear_request_list();
2813 del_timer_sync(&ext4_li_info->li_timer);
2814 kfree(ext4_li_info);
2815 ext4_li_info = NULL;
2816 printk(KERN_CRIT "EXT4: error %d creating inode table "
2817 "initialization thread\n",
2821 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2823 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2828 * Check whether it make sense to run itable init. thread or not.
2829 * If there is at least one uninitialized inode table, return
2830 * corresponding group number, else the loop goes through all
2831 * groups and return total number of groups.
2833 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2835 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2836 struct ext4_group_desc *gdp = NULL;
2838 for (group = 0; group < ngroups; group++) {
2839 gdp = ext4_get_group_desc(sb, group, NULL);
2843 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2850 static int ext4_li_info_new(void)
2852 struct ext4_lazy_init *eli = NULL;
2854 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2858 eli->li_task = NULL;
2859 INIT_LIST_HEAD(&eli->li_request_list);
2860 mutex_init(&eli->li_list_mtx);
2862 init_waitqueue_head(&eli->li_wait_daemon);
2863 init_waitqueue_head(&eli->li_wait_task);
2864 init_timer(&eli->li_timer);
2865 eli->li_state |= EXT4_LAZYINIT_QUIT;
2872 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2875 struct ext4_sb_info *sbi = EXT4_SB(sb);
2876 struct ext4_li_request *elr;
2879 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2885 elr->lr_next_group = start;
2888 * Randomize first schedule time of the request to
2889 * spread the inode table initialization requests
2892 get_random_bytes(&rnd, sizeof(rnd));
2893 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2894 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2899 static int ext4_register_li_request(struct super_block *sb,
2900 ext4_group_t first_not_zeroed)
2902 struct ext4_sb_info *sbi = EXT4_SB(sb);
2903 struct ext4_li_request *elr;
2904 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2907 if (sbi->s_li_request != NULL)
2910 if (first_not_zeroed == ngroups ||
2911 (sb->s_flags & MS_RDONLY) ||
2912 !test_opt(sb, INIT_INODE_TABLE)) {
2913 sbi->s_li_request = NULL;
2917 if (first_not_zeroed == ngroups) {
2918 sbi->s_li_request = NULL;
2922 elr = ext4_li_request_new(sb, first_not_zeroed);
2926 mutex_lock(&ext4_li_mtx);
2928 if (NULL == ext4_li_info) {
2929 ret = ext4_li_info_new();
2934 mutex_lock(&ext4_li_info->li_list_mtx);
2935 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2936 mutex_unlock(&ext4_li_info->li_list_mtx);
2938 sbi->s_li_request = elr;
2940 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2941 ret = ext4_run_lazyinit_thread();
2946 mutex_unlock(&ext4_li_mtx);
2953 * We do not need to lock anything since this is called on
2956 static void ext4_destroy_lazyinit_thread(void)
2959 * If thread exited earlier
2960 * there's nothing to be done.
2965 ext4_clear_request_list();
2967 while (ext4_li_info->li_task) {
2968 wake_up(&ext4_li_info->li_wait_daemon);
2969 wait_event(ext4_li_info->li_wait_task,
2970 ext4_li_info->li_task == NULL);
2974 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2975 __releases(kernel_lock)
2976 __acquires(kernel_lock)
2978 char *orig_data = kstrdup(data, GFP_KERNEL);
2979 struct buffer_head *bh;
2980 struct ext4_super_block *es = NULL;
2981 struct ext4_sb_info *sbi;
2983 ext4_fsblk_t sb_block = get_sb_block(&data);
2984 ext4_fsblk_t logical_sb_block;
2985 unsigned long offset = 0;
2986 unsigned long journal_devnum = 0;
2987 unsigned long def_mount_opts;
2993 unsigned int db_count;
2995 int needs_recovery, has_huge_files;
2998 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2999 ext4_group_t first_not_zeroed;
3001 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3005 sbi->s_blockgroup_lock =
3006 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3007 if (!sbi->s_blockgroup_lock) {
3011 sb->s_fs_info = sbi;
3012 sbi->s_mount_opt = 0;
3013 sbi->s_resuid = EXT4_DEF_RESUID;
3014 sbi->s_resgid = EXT4_DEF_RESGID;
3015 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3016 sbi->s_sb_block = sb_block;
3017 if (sb->s_bdev->bd_part)
3018 sbi->s_sectors_written_start =
3019 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3023 /* Cleanup superblock name */
3024 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3028 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3030 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3035 * The ext4 superblock will not be buffer aligned for other than 1kB
3036 * block sizes. We need to calculate the offset from buffer start.
3038 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3039 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3040 offset = do_div(logical_sb_block, blocksize);
3042 logical_sb_block = sb_block;
3045 if (!(bh = sb_bread(sb, logical_sb_block))) {
3046 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3050 * Note: s_es must be initialized as soon as possible because
3051 * some ext4 macro-instructions depend on its value
3053 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3055 sb->s_magic = le16_to_cpu(es->s_magic);
3056 if (sb->s_magic != EXT4_SUPER_MAGIC)
3058 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3060 /* Set defaults before we parse the mount options */
3061 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3062 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
3063 if (def_mount_opts & EXT4_DEFM_DEBUG)
3064 set_opt(sbi->s_mount_opt, DEBUG);
3065 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3066 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3068 set_opt(sbi->s_mount_opt, GRPID);
3070 if (def_mount_opts & EXT4_DEFM_UID16)
3071 set_opt(sbi->s_mount_opt, NO_UID32);
3072 #ifdef CONFIG_EXT4_FS_XATTR
3073 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
3074 set_opt(sbi->s_mount_opt, XATTR_USER);
3076 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3077 if (def_mount_opts & EXT4_DEFM_ACL)
3078 set_opt(sbi->s_mount_opt, POSIX_ACL);
3080 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3081 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3082 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3083 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3084 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3085 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3087 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3088 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
3089 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3090 set_opt(sbi->s_mount_opt, ERRORS_CONT);
3092 set_opt(sbi->s_mount_opt, ERRORS_RO);
3093 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3094 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3095 if (def_mount_opts & EXT4_DEFM_DISCARD)
3096 set_opt(sbi->s_mount_opt, DISCARD);
3098 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3099 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3100 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3101 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3102 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3104 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3105 set_opt(sbi->s_mount_opt, BARRIER);
3108 * enable delayed allocation by default
3109 * Use -o nodelalloc to turn it off
3111 if (!IS_EXT3_SB(sb) &&
3112 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3113 set_opt(sbi->s_mount_opt, DELALLOC);
3115 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3116 &journal_devnum, &journal_ioprio, NULL, 0)) {
3117 ext4_msg(sb, KERN_WARNING,
3118 "failed to parse options in superblock: %s",
3119 sbi->s_es->s_mount_opts);
3121 if (!parse_options((char *) data, sb, &journal_devnum,
3122 &journal_ioprio, NULL, 0))
3125 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3126 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3128 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3129 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3130 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3131 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3132 ext4_msg(sb, KERN_WARNING,
3133 "feature flags set on rev 0 fs, "
3134 "running e2fsck is recommended");
3137 * Check feature flags regardless of the revision level, since we
3138 * previously didn't change the revision level when setting the flags,
3139 * so there is a chance incompat flags are set on a rev 0 filesystem.
3141 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3144 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3146 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3147 blocksize > EXT4_MAX_BLOCK_SIZE) {
3148 ext4_msg(sb, KERN_ERR,
3149 "Unsupported filesystem blocksize %d", blocksize);
3153 if (sb->s_blocksize != blocksize) {
3154 /* Validate the filesystem blocksize */
3155 if (!sb_set_blocksize(sb, blocksize)) {
3156 ext4_msg(sb, KERN_ERR, "bad block size %d",
3162 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3163 offset = do_div(logical_sb_block, blocksize);
3164 bh = sb_bread(sb, logical_sb_block);
3166 ext4_msg(sb, KERN_ERR,
3167 "Can't read superblock on 2nd try");
3170 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3172 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3173 ext4_msg(sb, KERN_ERR,
3174 "Magic mismatch, very weird!");
3179 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3180 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3181 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3183 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3185 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3186 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3187 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3189 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3190 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3191 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3192 (!is_power_of_2(sbi->s_inode_size)) ||
3193 (sbi->s_inode_size > blocksize)) {
3194 ext4_msg(sb, KERN_ERR,
3195 "unsupported inode size: %d",
3199 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3200 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3203 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3204 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3205 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3206 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3207 !is_power_of_2(sbi->s_desc_size)) {
3208 ext4_msg(sb, KERN_ERR,
3209 "unsupported descriptor size %lu",
3214 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3216 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3217 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3218 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3221 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3222 if (sbi->s_inodes_per_block == 0)
3224 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3225 sbi->s_inodes_per_block;
3226 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3228 sbi->s_mount_state = le16_to_cpu(es->s_state);
3229 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3230 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3232 for (i = 0; i < 4; i++)
3233 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3234 sbi->s_def_hash_version = es->s_def_hash_version;
3235 i = le32_to_cpu(es->s_flags);
3236 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3237 sbi->s_hash_unsigned = 3;
3238 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3239 #ifdef __CHAR_UNSIGNED__
3240 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3241 sbi->s_hash_unsigned = 3;
3243 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3248 if (sbi->s_blocks_per_group > blocksize * 8) {
3249 ext4_msg(sb, KERN_ERR,
3250 "#blocks per group too big: %lu",
3251 sbi->s_blocks_per_group);
3254 if (sbi->s_inodes_per_group > blocksize * 8) {
3255 ext4_msg(sb, KERN_ERR,
3256 "#inodes per group too big: %lu",
3257 sbi->s_inodes_per_group);
3262 * Test whether we have more sectors than will fit in sector_t,
3263 * and whether the max offset is addressable by the page cache.
3265 if ((ext4_blocks_count(es) >
3266 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
3267 (ext4_blocks_count(es) >
3268 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
3269 ext4_msg(sb, KERN_ERR, "filesystem"
3270 " too large to mount safely on this system");
3271 if (sizeof(sector_t) < 8)
3272 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3277 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3280 /* check blocks count against device size */
3281 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3282 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3283 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3284 "exceeds size of device (%llu blocks)",
3285 ext4_blocks_count(es), blocks_count);
3290 * It makes no sense for the first data block to be beyond the end
3291 * of the filesystem.
3293 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3294 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3295 "block %u is beyond end of filesystem (%llu)",
3296 le32_to_cpu(es->s_first_data_block),
3297 ext4_blocks_count(es));
3300 blocks_count = (ext4_blocks_count(es) -
3301 le32_to_cpu(es->s_first_data_block) +
3302 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3303 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3304 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3305 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3306 "(block count %llu, first data block %u, "
3307 "blocks per group %lu)", sbi->s_groups_count,
3308 ext4_blocks_count(es),
3309 le32_to_cpu(es->s_first_data_block),
3310 EXT4_BLOCKS_PER_GROUP(sb));
3313 sbi->s_groups_count = blocks_count;
3314 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3315 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3316 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3317 EXT4_DESC_PER_BLOCK(sb);
3318 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
3320 if (sbi->s_group_desc == NULL) {
3321 ext4_msg(sb, KERN_ERR, "not enough memory");
3325 #ifdef CONFIG_PROC_FS
3327 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3330 bgl_lock_init(sbi->s_blockgroup_lock);
3332 for (i = 0; i < db_count; i++) {
3333 block = descriptor_loc(sb, logical_sb_block, i);
3334 sbi->s_group_desc[i] = sb_bread(sb, block);
3335 if (!sbi->s_group_desc[i]) {
3336 ext4_msg(sb, KERN_ERR,
3337 "can't read group descriptor %d", i);
3342 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3343 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3346 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3347 if (!ext4_fill_flex_info(sb)) {
3348 ext4_msg(sb, KERN_ERR,
3349 "unable to initialize "
3350 "flex_bg meta info!");
3354 sbi->s_gdb_count = db_count;
3355 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3356 spin_lock_init(&sbi->s_next_gen_lock);
3358 sbi->s_stripe = ext4_get_stripe_size(sbi);
3359 sbi->s_max_writeback_mb_bump = 128;
3362 * set up enough so that it can read an inode
3364 if (!test_opt(sb, NOLOAD) &&
3365 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3366 sb->s_op = &ext4_sops;
3368 sb->s_op = &ext4_nojournal_sops;
3369 sb->s_export_op = &ext4_export_ops;
3370 sb->s_xattr = ext4_xattr_handlers;
3372 sb->s_qcop = &ext4_qctl_operations;
3373 sb->dq_op = &ext4_quota_operations;
3375 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3376 mutex_init(&sbi->s_orphan_lock);
3377 mutex_init(&sbi->s_resize_lock);
3381 needs_recovery = (es->s_last_orphan != 0 ||
3382 EXT4_HAS_INCOMPAT_FEATURE(sb,
3383 EXT4_FEATURE_INCOMPAT_RECOVER));
3386 * The first inode we look at is the journal inode. Don't try
3387 * root first: it may be modified in the journal!
3389 if (!test_opt(sb, NOLOAD) &&
3390 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3391 if (ext4_load_journal(sb, es, journal_devnum))
3393 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3394 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3395 ext4_msg(sb, KERN_ERR, "required journal recovery "
3396 "suppressed and not mounted read-only");
3397 goto failed_mount_wq;
3399 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
3400 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3401 sbi->s_journal = NULL;
3406 if (ext4_blocks_count(es) > 0xffffffffULL &&
3407 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3408 JBD2_FEATURE_INCOMPAT_64BIT)) {
3409 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3410 goto failed_mount_wq;
3413 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3414 jbd2_journal_set_features(sbi->s_journal,
3415 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3416 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3417 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3418 jbd2_journal_set_features(sbi->s_journal,
3419 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3420 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3421 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3423 jbd2_journal_clear_features(sbi->s_journal,
3424 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3425 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3428 /* We have now updated the journal if required, so we can
3429 * validate the data journaling mode. */
3430 switch (test_opt(sb, DATA_FLAGS)) {
3432 /* No mode set, assume a default based on the journal
3433 * capabilities: ORDERED_DATA if the journal can
3434 * cope, else JOURNAL_DATA
3436 if (jbd2_journal_check_available_features
3437 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3438 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3440 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3443 case EXT4_MOUNT_ORDERED_DATA:
3444 case EXT4_MOUNT_WRITEBACK_DATA:
3445 if (!jbd2_journal_check_available_features
3446 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3447 ext4_msg(sb, KERN_ERR, "Journal does not support "
3448 "requested data journaling mode");
3449 goto failed_mount_wq;
3454 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3457 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3458 ext4_count_free_blocks(sb));
3460 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3461 ext4_count_free_inodes(sb));
3463 err = percpu_counter_init(&sbi->s_dirs_counter,
3464 ext4_count_dirs(sb));
3466 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3468 ext4_msg(sb, KERN_ERR, "insufficient memory");
3469 goto failed_mount_wq;
3472 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3473 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3474 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3475 goto failed_mount_wq;
3479 * The jbd2_journal_load will have done any necessary log recovery,
3480 * so we can safely mount the rest of the filesystem now.
3483 root = ext4_iget(sb, EXT4_ROOT_INO);
3485 ext4_msg(sb, KERN_ERR, "get root inode failed");
3486 ret = PTR_ERR(root);
3489 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3491 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3494 sb->s_root = d_alloc_root(root);
3496 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3502 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3504 /* determine the minimum size of new large inodes, if present */
3505 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3506 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3507 EXT4_GOOD_OLD_INODE_SIZE;
3508 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3509 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3510 if (sbi->s_want_extra_isize <
3511 le16_to_cpu(es->s_want_extra_isize))
3512 sbi->s_want_extra_isize =
3513 le16_to_cpu(es->s_want_extra_isize);
3514 if (sbi->s_want_extra_isize <
3515 le16_to_cpu(es->s_min_extra_isize))
3516 sbi->s_want_extra_isize =
3517 le16_to_cpu(es->s_min_extra_isize);
3520 /* Check if enough inode space is available */
3521 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3522 sbi->s_inode_size) {
3523 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3524 EXT4_GOOD_OLD_INODE_SIZE;
3525 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3529 if (test_opt(sb, DELALLOC) &&
3530 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3531 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3532 "requested data journaling mode");
3533 clear_opt(sbi->s_mount_opt, DELALLOC);
3535 if (test_opt(sb, DIOREAD_NOLOCK)) {
3536 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3537 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3538 "option - requested data journaling mode");
3539 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3541 if (sb->s_blocksize < PAGE_SIZE) {
3542 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3543 "option - block size is too small");
3544 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3548 err = ext4_setup_system_zone(sb);
3550 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3556 err = ext4_mb_init(sb, needs_recovery);
3558 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3563 err = ext4_register_li_request(sb, first_not_zeroed);
3567 sbi->s_kobj.kset = ext4_kset;
3568 init_completion(&sbi->s_kobj_unregister);
3569 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3572 ext4_mb_release(sb);
3573 ext4_ext_release(sb);
3577 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3578 ext4_orphan_cleanup(sb, es);
3579 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3580 if (needs_recovery) {
3581 ext4_msg(sb, KERN_INFO, "recovery complete");
3582 ext4_mark_recovery_complete(sb, es);
3584 if (EXT4_SB(sb)->s_journal) {
3585 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3586 descr = " journalled data mode";
3587 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3588 descr = " ordered data mode";
3590 descr = " writeback data mode";
3592 descr = "out journal";
3594 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3595 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3596 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3598 init_timer(&sbi->s_err_report);
3599 sbi->s_err_report.function = print_daily_error_info;
3600 sbi->s_err_report.data = (unsigned long) sb;
3601 if (es->s_error_count)
3602 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3610 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3614 ext4_msg(sb, KERN_ERR, "mount failed");
3615 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3617 ext4_release_system_zone(sb);
3618 if (sbi->s_journal) {
3619 jbd2_journal_destroy(sbi->s_journal);
3620 sbi->s_journal = NULL;
3622 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3623 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3624 percpu_counter_destroy(&sbi->s_dirs_counter);
3625 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3627 if (sbi->s_flex_groups) {
3628 if (is_vmalloc_addr(sbi->s_flex_groups))
3629 vfree(sbi->s_flex_groups);
3631 kfree(sbi->s_flex_groups);
3634 for (i = 0; i < db_count; i++)
3635 brelse(sbi->s_group_desc[i]);
3636 kfree(sbi->s_group_desc);
3639 remove_proc_entry(sb->s_id, ext4_proc_root);
3642 for (i = 0; i < MAXQUOTAS; i++)
3643 kfree(sbi->s_qf_names[i]);
3645 ext4_blkdev_remove(sbi);
3648 sb->s_fs_info = NULL;
3649 kfree(sbi->s_blockgroup_lock);
3658 * Setup any per-fs journal parameters now. We'll do this both on
3659 * initial mount, once the journal has been initialised but before we've
3660 * done any recovery; and again on any subsequent remount.
3662 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3664 struct ext4_sb_info *sbi = EXT4_SB(sb);
3666 journal->j_commit_interval = sbi->s_commit_interval;
3667 journal->j_min_batch_time = sbi->s_min_batch_time;
3668 journal->j_max_batch_time = sbi->s_max_batch_time;
3670 write_lock(&journal->j_state_lock);
3671 if (test_opt(sb, BARRIER))
3672 journal->j_flags |= JBD2_BARRIER;
3674 journal->j_flags &= ~JBD2_BARRIER;
3675 if (test_opt(sb, DATA_ERR_ABORT))
3676 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3678 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3679 write_unlock(&journal->j_state_lock);
3682 static journal_t *ext4_get_journal(struct super_block *sb,
3683 unsigned int journal_inum)
3685 struct inode *journal_inode;
3688 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3690 /* First, test for the existence of a valid inode on disk. Bad
3691 * things happen if we iget() an unused inode, as the subsequent
3692 * iput() will try to delete it. */
3694 journal_inode = ext4_iget(sb, journal_inum);
3695 if (IS_ERR(journal_inode)) {
3696 ext4_msg(sb, KERN_ERR, "no journal found");
3699 if (!journal_inode->i_nlink) {
3700 make_bad_inode(journal_inode);
3701 iput(journal_inode);
3702 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3706 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3707 journal_inode, journal_inode->i_size);
3708 if (!S_ISREG(journal_inode->i_mode)) {
3709 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3710 iput(journal_inode);
3714 journal = jbd2_journal_init_inode(journal_inode);
3716 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3717 iput(journal_inode);
3720 journal->j_private = sb;
3721 ext4_init_journal_params(sb, journal);
3725 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3728 struct buffer_head *bh;
3732 int hblock, blocksize;
3733 ext4_fsblk_t sb_block;
3734 unsigned long offset;
3735 struct ext4_super_block *es;
3736 struct block_device *bdev;
3738 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3740 bdev = ext4_blkdev_get(j_dev, sb);
3744 if (bd_claim(bdev, sb)) {
3745 ext4_msg(sb, KERN_ERR,
3746 "failed to claim external journal device");
3747 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3751 blocksize = sb->s_blocksize;
3752 hblock = bdev_logical_block_size(bdev);
3753 if (blocksize < hblock) {
3754 ext4_msg(sb, KERN_ERR,
3755 "blocksize too small for journal device");
3759 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3760 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3761 set_blocksize(bdev, blocksize);
3762 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3763 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3764 "external journal");
3768 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3769 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3770 !(le32_to_cpu(es->s_feature_incompat) &
3771 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3772 ext4_msg(sb, KERN_ERR, "external journal has "
3778 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3779 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3784 len = ext4_blocks_count(es);
3785 start = sb_block + 1;
3786 brelse(bh); /* we're done with the superblock */
3788 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3789 start, len, blocksize);
3791 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3794 journal->j_private = sb;
3795 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3796 wait_on_buffer(journal->j_sb_buffer);
3797 if (!buffer_uptodate(journal->j_sb_buffer)) {
3798 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3801 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3802 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3803 "user (unsupported) - %d",
3804 be32_to_cpu(journal->j_superblock->s_nr_users));
3807 EXT4_SB(sb)->journal_bdev = bdev;
3808 ext4_init_journal_params(sb, journal);
3812 jbd2_journal_destroy(journal);
3814 ext4_blkdev_put(bdev);
3818 static int ext4_load_journal(struct super_block *sb,
3819 struct ext4_super_block *es,
3820 unsigned long journal_devnum)
3823 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3826 int really_read_only;
3828 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3830 if (journal_devnum &&
3831 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3832 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3833 "numbers have changed");
3834 journal_dev = new_decode_dev(journal_devnum);
3836 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3838 really_read_only = bdev_read_only(sb->s_bdev);
3841 * Are we loading a blank journal or performing recovery after a
3842 * crash? For recovery, we need to check in advance whether we
3843 * can get read-write access to the device.
3845 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3846 if (sb->s_flags & MS_RDONLY) {
3847 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3848 "required on readonly filesystem");
3849 if (really_read_only) {
3850 ext4_msg(sb, KERN_ERR, "write access "
3851 "unavailable, cannot proceed");
3854 ext4_msg(sb, KERN_INFO, "write access will "
3855 "be enabled during recovery");
3859 if (journal_inum && journal_dev) {
3860 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3861 "and inode journals!");
3866 if (!(journal = ext4_get_journal(sb, journal_inum)))
3869 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3873 if (!(journal->j_flags & JBD2_BARRIER))
3874 ext4_msg(sb, KERN_INFO, "barriers disabled");
3876 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3877 err = jbd2_journal_update_format(journal);
3879 ext4_msg(sb, KERN_ERR, "error updating journal");
3880 jbd2_journal_destroy(journal);
3885 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3886 err = jbd2_journal_wipe(journal, !really_read_only);
3888 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3890 memcpy(save, ((char *) es) +
3891 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3892 err = jbd2_journal_load(journal);
3894 memcpy(((char *) es) + EXT4_S_ERR_START,
3895 save, EXT4_S_ERR_LEN);
3900 ext4_msg(sb, KERN_ERR, "error loading journal");
3901 jbd2_journal_destroy(journal);
3905 EXT4_SB(sb)->s_journal = journal;
3906 ext4_clear_journal_err(sb, es);
3908 if (!really_read_only && journal_devnum &&
3909 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3910 es->s_journal_dev = cpu_to_le32(journal_devnum);
3912 /* Make sure we flush the recovery flag to disk. */
3913 ext4_commit_super(sb, 1);
3919 static int ext4_commit_super(struct super_block *sb, int sync)
3921 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3922 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3927 if (buffer_write_io_error(sbh)) {
3929 * Oh, dear. A previous attempt to write the
3930 * superblock failed. This could happen because the
3931 * USB device was yanked out. Or it could happen to
3932 * be a transient write error and maybe the block will
3933 * be remapped. Nothing we can do but to retry the
3934 * write and hope for the best.
3936 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3937 "superblock detected");
3938 clear_buffer_write_io_error(sbh);
3939 set_buffer_uptodate(sbh);
3942 * If the file system is mounted read-only, don't update the
3943 * superblock write time. This avoids updating the superblock
3944 * write time when we are mounting the root file system
3945 * read/only but we need to replay the journal; at that point,
3946 * for people who are east of GMT and who make their clock
3947 * tick in localtime for Windows bug-for-bug compatibility,
3948 * the clock is set in the future, and this will cause e2fsck
3949 * to complain and force a full file system check.
3951 if (!(sb->s_flags & MS_RDONLY))
3952 es->s_wtime = cpu_to_le32(get_seconds());
3953 if (sb->s_bdev->bd_part)
3954 es->s_kbytes_written =
3955 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3956 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3957 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3959 es->s_kbytes_written =
3960 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3961 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeblocks_counter))
3962 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3963 &EXT4_SB(sb)->s_freeblocks_counter));
3964 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
3965 es->s_free_inodes_count =
3966 cpu_to_le32(percpu_counter_sum_positive(
3967 &EXT4_SB(sb)->s_freeinodes_counter));
3969 BUFFER_TRACE(sbh, "marking dirty");
3970 mark_buffer_dirty(sbh);
3972 error = sync_dirty_buffer(sbh);
3976 error = buffer_write_io_error(sbh);
3978 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3980 clear_buffer_write_io_error(sbh);
3981 set_buffer_uptodate(sbh);
3988 * Have we just finished recovery? If so, and if we are mounting (or
3989 * remounting) the filesystem readonly, then we will end up with a
3990 * consistent fs on disk. Record that fact.
3992 static void ext4_mark_recovery_complete(struct super_block *sb,
3993 struct ext4_super_block *es)
3995 journal_t *journal = EXT4_SB(sb)->s_journal;
3997 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3998 BUG_ON(journal != NULL);
4001 jbd2_journal_lock_updates(journal);
4002 if (jbd2_journal_flush(journal) < 0)
4005 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4006 sb->s_flags & MS_RDONLY) {
4007 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4008 ext4_commit_super(sb, 1);
4012 jbd2_journal_unlock_updates(journal);
4016 * If we are mounting (or read-write remounting) a filesystem whose journal
4017 * has recorded an error from a previous lifetime, move that error to the
4018 * main filesystem now.
4020 static void ext4_clear_journal_err(struct super_block *sb,
4021 struct ext4_super_block *es)
4027 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4029 journal = EXT4_SB(sb)->s_journal;
4032 * Now check for any error status which may have been recorded in the
4033 * journal by a prior ext4_error() or ext4_abort()
4036 j_errno = jbd2_journal_errno(journal);
4040 errstr = ext4_decode_error(sb, j_errno, nbuf);
4041 ext4_warning(sb, "Filesystem error recorded "
4042 "from previous mount: %s", errstr);
4043 ext4_warning(sb, "Marking fs in need of filesystem check.");
4045 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4046 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4047 ext4_commit_super(sb, 1);
4049 jbd2_journal_clear_err(journal);
4054 * Force the running and committing transactions to commit,
4055 * and wait on the commit.
4057 int ext4_force_commit(struct super_block *sb)
4062 if (sb->s_flags & MS_RDONLY)
4065 journal = EXT4_SB(sb)->s_journal;
4067 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4068 ret = ext4_journal_force_commit(journal);
4074 static void ext4_write_super(struct super_block *sb)
4077 ext4_commit_super(sb, 1);
4081 static int ext4_sync_fs(struct super_block *sb, int wait)
4085 struct ext4_sb_info *sbi = EXT4_SB(sb);
4087 trace_ext4_sync_fs(sb, wait);
4088 flush_workqueue(sbi->dio_unwritten_wq);
4089 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4091 jbd2_log_wait_commit(sbi->s_journal, target);
4097 * LVM calls this function before a (read-only) snapshot is created. This
4098 * gives us a chance to flush the journal completely and mark the fs clean.
4100 static int ext4_freeze(struct super_block *sb)
4105 if (sb->s_flags & MS_RDONLY)
4108 journal = EXT4_SB(sb)->s_journal;
4110 /* Now we set up the journal barrier. */
4111 jbd2_journal_lock_updates(journal);
4114 * Don't clear the needs_recovery flag if we failed to flush
4117 error = jbd2_journal_flush(journal);
4121 /* Journal blocked and flushed, clear needs_recovery flag. */
4122 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4123 error = ext4_commit_super(sb, 1);
4125 /* we rely on s_frozen to stop further updates */
4126 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4131 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4132 * flag here, even though the filesystem is not technically dirty yet.
4134 static int ext4_unfreeze(struct super_block *sb)
4136 if (sb->s_flags & MS_RDONLY)
4140 /* Reset the needs_recovery flag before the fs is unlocked. */
4141 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4142 ext4_commit_super(sb, 1);
4147 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4149 struct ext4_super_block *es;
4150 struct ext4_sb_info *sbi = EXT4_SB(sb);
4151 ext4_fsblk_t n_blocks_count = 0;
4152 unsigned long old_sb_flags;
4153 struct ext4_mount_options old_opts;
4154 int enable_quota = 0;
4156 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4161 char *orig_data = kstrdup(data, GFP_KERNEL);
4165 /* Store the original options */
4167 old_sb_flags = sb->s_flags;
4168 old_opts.s_mount_opt = sbi->s_mount_opt;
4169 old_opts.s_resuid = sbi->s_resuid;
4170 old_opts.s_resgid = sbi->s_resgid;
4171 old_opts.s_commit_interval = sbi->s_commit_interval;
4172 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4173 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4175 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4176 for (i = 0; i < MAXQUOTAS; i++)
4177 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4179 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4180 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4183 * Allow the "check" option to be passed as a remount option.
4185 if (!parse_options(data, sb, NULL, &journal_ioprio,
4186 &n_blocks_count, 1)) {
4191 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4192 ext4_abort(sb, "Abort forced by user");
4194 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4195 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4199 if (sbi->s_journal) {
4200 ext4_init_journal_params(sb, sbi->s_journal);
4201 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4204 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4205 n_blocks_count > ext4_blocks_count(es)) {
4206 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4211 if (*flags & MS_RDONLY) {
4212 err = dquot_suspend(sb, -1);
4217 * First of all, the unconditional stuff we have to do
4218 * to disable replay of the journal when we next remount
4220 sb->s_flags |= MS_RDONLY;
4223 * OK, test if we are remounting a valid rw partition
4224 * readonly, and if so set the rdonly flag and then
4225 * mark the partition as valid again.
4227 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4228 (sbi->s_mount_state & EXT4_VALID_FS))
4229 es->s_state = cpu_to_le16(sbi->s_mount_state);
4232 ext4_mark_recovery_complete(sb, es);
4234 /* Make sure we can mount this feature set readwrite */
4235 if (!ext4_feature_set_ok(sb, 0)) {
4240 * Make sure the group descriptor checksums
4241 * are sane. If they aren't, refuse to remount r/w.
4243 for (g = 0; g < sbi->s_groups_count; g++) {
4244 struct ext4_group_desc *gdp =
4245 ext4_get_group_desc(sb, g, NULL);
4247 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4248 ext4_msg(sb, KERN_ERR,
4249 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4250 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4251 le16_to_cpu(gdp->bg_checksum));
4258 * If we have an unprocessed orphan list hanging
4259 * around from a previously readonly bdev mount,
4260 * require a full umount/remount for now.
4262 if (es->s_last_orphan) {
4263 ext4_msg(sb, KERN_WARNING, "Couldn't "
4264 "remount RDWR because of unprocessed "
4265 "orphan inode list. Please "
4266 "umount/remount instead");
4272 * Mounting a RDONLY partition read-write, so reread
4273 * and store the current valid flag. (It may have
4274 * been changed by e2fsck since we originally mounted
4278 ext4_clear_journal_err(sb, es);
4279 sbi->s_mount_state = le16_to_cpu(es->s_state);
4280 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4282 if (!ext4_setup_super(sb, es, 0))
4283 sb->s_flags &= ~MS_RDONLY;
4289 * Reinitialize lazy itable initialization thread based on
4292 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4293 ext4_unregister_li_request(sb);
4295 ext4_group_t first_not_zeroed;
4296 first_not_zeroed = ext4_has_uninit_itable(sb);
4297 ext4_register_li_request(sb, first_not_zeroed);
4300 ext4_setup_system_zone(sb);
4301 if (sbi->s_journal == NULL)
4302 ext4_commit_super(sb, 1);
4305 /* Release old quota file names */
4306 for (i = 0; i < MAXQUOTAS; i++)
4307 if (old_opts.s_qf_names[i] &&
4308 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4309 kfree(old_opts.s_qf_names[i]);
4314 dquot_resume(sb, -1);
4316 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4321 sb->s_flags = old_sb_flags;
4322 sbi->s_mount_opt = old_opts.s_mount_opt;
4323 sbi->s_resuid = old_opts.s_resuid;
4324 sbi->s_resgid = old_opts.s_resgid;
4325 sbi->s_commit_interval = old_opts.s_commit_interval;
4326 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4327 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4329 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4330 for (i = 0; i < MAXQUOTAS; i++) {
4331 if (sbi->s_qf_names[i] &&
4332 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4333 kfree(sbi->s_qf_names[i]);
4334 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4343 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4345 struct super_block *sb = dentry->d_sb;
4346 struct ext4_sb_info *sbi = EXT4_SB(sb);
4347 struct ext4_super_block *es = sbi->s_es;
4350 if (test_opt(sb, MINIX_DF)) {
4351 sbi->s_overhead_last = 0;
4352 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4353 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4354 ext4_fsblk_t overhead = 0;
4357 * Compute the overhead (FS structures). This is constant
4358 * for a given filesystem unless the number of block groups
4359 * changes so we cache the previous value until it does.
4363 * All of the blocks before first_data_block are
4366 overhead = le32_to_cpu(es->s_first_data_block);
4369 * Add the overhead attributed to the superblock and
4370 * block group descriptors. If the sparse superblocks
4371 * feature is turned on, then not all groups have this.
4373 for (i = 0; i < ngroups; i++) {
4374 overhead += ext4_bg_has_super(sb, i) +
4375 ext4_bg_num_gdb(sb, i);
4380 * Every block group has an inode bitmap, a block
4381 * bitmap, and an inode table.
4383 overhead += ngroups * (2 + sbi->s_itb_per_group);
4384 sbi->s_overhead_last = overhead;
4386 sbi->s_blocks_last = ext4_blocks_count(es);
4389 buf->f_type = EXT4_SUPER_MAGIC;
4390 buf->f_bsize = sb->s_blocksize;
4391 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4392 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4393 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4394 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4395 if (buf->f_bfree < ext4_r_blocks_count(es))
4397 buf->f_files = le32_to_cpu(es->s_inodes_count);
4398 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4399 buf->f_namelen = EXT4_NAME_LEN;
4400 fsid = le64_to_cpup((void *)es->s_uuid) ^
4401 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4402 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4403 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4408 /* Helper function for writing quotas on sync - we need to start transaction
4409 * before quota file is locked for write. Otherwise the are possible deadlocks:
4410 * Process 1 Process 2
4411 * ext4_create() quota_sync()
4412 * jbd2_journal_start() write_dquot()
4413 * dquot_initialize() down(dqio_mutex)
4414 * down(dqio_mutex) jbd2_journal_start()
4420 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4422 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4425 static int ext4_write_dquot(struct dquot *dquot)
4429 struct inode *inode;
4431 inode = dquot_to_inode(dquot);
4432 handle = ext4_journal_start(inode,
4433 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4435 return PTR_ERR(handle);
4436 ret = dquot_commit(dquot);
4437 err = ext4_journal_stop(handle);
4443 static int ext4_acquire_dquot(struct dquot *dquot)
4448 handle = ext4_journal_start(dquot_to_inode(dquot),
4449 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4451 return PTR_ERR(handle);
4452 ret = dquot_acquire(dquot);
4453 err = ext4_journal_stop(handle);
4459 static int ext4_release_dquot(struct dquot *dquot)
4464 handle = ext4_journal_start(dquot_to_inode(dquot),
4465 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4466 if (IS_ERR(handle)) {
4467 /* Release dquot anyway to avoid endless cycle in dqput() */
4468 dquot_release(dquot);
4469 return PTR_ERR(handle);
4471 ret = dquot_release(dquot);
4472 err = ext4_journal_stop(handle);
4478 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4480 /* Are we journaling quotas? */
4481 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4482 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4483 dquot_mark_dquot_dirty(dquot);
4484 return ext4_write_dquot(dquot);
4486 return dquot_mark_dquot_dirty(dquot);
4490 static int ext4_write_info(struct super_block *sb, int type)
4495 /* Data block + inode block */
4496 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4498 return PTR_ERR(handle);
4499 ret = dquot_commit_info(sb, type);
4500 err = ext4_journal_stop(handle);
4507 * Turn on quotas during mount time - we need to find
4508 * the quota file and such...
4510 static int ext4_quota_on_mount(struct super_block *sb, int type)
4512 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4513 EXT4_SB(sb)->s_jquota_fmt, type);
4517 * Standard function to be called on quota_on
4519 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4525 if (!test_opt(sb, QUOTA))
4528 err = kern_path(name, LOOKUP_FOLLOW, &path);
4532 /* Quotafile not on the same filesystem? */
4533 if (path.mnt->mnt_sb != sb) {
4537 /* Journaling quota? */
4538 if (EXT4_SB(sb)->s_qf_names[type]) {
4539 /* Quotafile not in fs root? */
4540 if (path.dentry->d_parent != sb->s_root)
4541 ext4_msg(sb, KERN_WARNING,
4542 "Quota file not on filesystem root. "
4543 "Journaled quota will not work");
4547 * When we journal data on quota file, we have to flush journal to see
4548 * all updates to the file when we bypass pagecache...
4550 if (EXT4_SB(sb)->s_journal &&
4551 ext4_should_journal_data(path.dentry->d_inode)) {
4553 * We don't need to lock updates but journal_flush() could
4554 * otherwise be livelocked...
4556 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4557 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4558 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4565 err = dquot_quota_on_path(sb, type, format_id, &path);
4570 static int ext4_quota_off(struct super_block *sb, int type)
4572 /* Force all delayed allocation blocks to be allocated */
4573 if (test_opt(sb, DELALLOC)) {
4574 down_read(&sb->s_umount);
4575 sync_filesystem(sb);
4576 up_read(&sb->s_umount);
4579 return dquot_quota_off(sb, type);
4582 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4583 * acquiring the locks... As quota files are never truncated and quota code
4584 * itself serializes the operations (and noone else should touch the files)
4585 * we don't have to be afraid of races */
4586 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4587 size_t len, loff_t off)
4589 struct inode *inode = sb_dqopt(sb)->files[type];
4590 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4592 int offset = off & (sb->s_blocksize - 1);
4595 struct buffer_head *bh;
4596 loff_t i_size = i_size_read(inode);
4600 if (off+len > i_size)
4603 while (toread > 0) {
4604 tocopy = sb->s_blocksize - offset < toread ?
4605 sb->s_blocksize - offset : toread;
4606 bh = ext4_bread(NULL, inode, blk, 0, &err);
4609 if (!bh) /* A hole? */
4610 memset(data, 0, tocopy);
4612 memcpy(data, bh->b_data+offset, tocopy);
4622 /* Write to quotafile (we know the transaction is already started and has
4623 * enough credits) */
4624 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4625 const char *data, 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);
4631 struct buffer_head *bh;
4632 handle_t *handle = journal_current_handle();
4634 if (EXT4_SB(sb)->s_journal && !handle) {
4635 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4636 " cancelled because transaction is not started",
4637 (unsigned long long)off, (unsigned long long)len);
4641 * Since we account only one data block in transaction credits,
4642 * then it is impossible to cross a block boundary.
4644 if (sb->s_blocksize - offset < len) {
4645 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4646 " cancelled because not block aligned",
4647 (unsigned long long)off, (unsigned long long)len);
4651 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4652 bh = ext4_bread(handle, inode, blk, 1, &err);
4655 err = ext4_journal_get_write_access(handle, bh);
4661 memcpy(bh->b_data+offset, data, len);
4662 flush_dcache_page(bh->b_page);
4664 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4668 mutex_unlock(&inode->i_mutex);
4671 if (inode->i_size < off + len) {
4672 i_size_write(inode, off + len);
4673 EXT4_I(inode)->i_disksize = inode->i_size;
4675 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4676 ext4_mark_inode_dirty(handle, inode);
4677 mutex_unlock(&inode->i_mutex);
4683 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4684 const char *dev_name, void *data, struct vfsmount *mnt)
4686 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4689 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4690 static struct file_system_type ext2_fs_type = {
4691 .owner = THIS_MODULE,
4693 .get_sb = ext4_get_sb,
4694 .kill_sb = kill_block_super,
4695 .fs_flags = FS_REQUIRES_DEV,
4698 static inline void register_as_ext2(void)
4700 int err = register_filesystem(&ext2_fs_type);
4703 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4706 static inline void unregister_as_ext2(void)
4708 unregister_filesystem(&ext2_fs_type);
4710 MODULE_ALIAS("ext2");
4712 static inline void register_as_ext2(void) { }
4713 static inline void unregister_as_ext2(void) { }
4716 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4717 static inline void register_as_ext3(void)
4719 int err = register_filesystem(&ext3_fs_type);
4722 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4725 static inline void unregister_as_ext3(void)
4727 unregister_filesystem(&ext3_fs_type);
4729 MODULE_ALIAS("ext3");
4731 static inline void register_as_ext3(void) { }
4732 static inline void unregister_as_ext3(void) { }
4735 static struct file_system_type ext4_fs_type = {
4736 .owner = THIS_MODULE,
4738 .get_sb = ext4_get_sb,
4739 .kill_sb = kill_block_super,
4740 .fs_flags = FS_REQUIRES_DEV,
4743 int __init ext4_init_feat_adverts(void)
4745 struct ext4_features *ef;
4748 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4752 ef->f_kobj.kset = ext4_kset;
4753 init_completion(&ef->f_kobj_unregister);
4754 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4767 static int __init ext4_init_fs(void)
4771 ext4_check_flag_values();
4772 err = ext4_init_pageio();
4775 err = ext4_init_system_zone();
4778 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4781 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4783 err = ext4_init_feat_adverts();
4785 err = ext4_init_mballoc();
4789 err = ext4_init_xattr();
4792 err = init_inodecache();
4797 err = register_filesystem(&ext4_fs_type);
4801 ext4_li_info = NULL;
4802 mutex_init(&ext4_li_mtx);
4805 unregister_as_ext2();
4806 unregister_as_ext3();
4807 destroy_inodecache();
4811 ext4_exit_mballoc();
4814 remove_proc_entry("fs/ext4", NULL);
4815 kset_unregister(ext4_kset);
4817 ext4_exit_system_zone();
4823 static void __exit ext4_exit_fs(void)
4825 ext4_destroy_lazyinit_thread();
4826 unregister_as_ext2();
4827 unregister_as_ext3();
4828 unregister_filesystem(&ext4_fs_type);
4829 destroy_inodecache();
4831 ext4_exit_mballoc();
4832 remove_proc_entry("fs/ext4", NULL);
4833 kset_unregister(ext4_kset);
4834 ext4_exit_system_zone();
4838 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4839 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4840 MODULE_LICENSE("GPL");
4841 module_init(ext4_init_fs)
4842 module_exit(ext4_exit_fs)