2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct proc_dir_entry *ext4_proc_root;
59 static struct kset *ext4_kset;
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ext4_features *ext4_feat;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static void ext4_write_super(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
112 #define IS_EXT3_SB(sb) (0)
115 void *ext4_kvmalloc(size_t size, gfp_t flags)
119 ret = kmalloc(size, flags);
121 ret = __vmalloc(size, flags, PAGE_KERNEL);
125 void *ext4_kvzalloc(size_t size, gfp_t flags)
129 ret = kzalloc(size, flags);
131 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
135 void ext4_kvfree(void *ptr)
137 if (is_vmalloc_addr(ptr))
144 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
145 struct ext4_group_desc *bg)
147 return le32_to_cpu(bg->bg_block_bitmap_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
149 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
152 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
153 struct ext4_group_desc *bg)
155 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
156 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
157 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
160 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
161 struct ext4_group_desc *bg)
163 return le32_to_cpu(bg->bg_inode_table_lo) |
164 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
165 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
168 __u32 ext4_free_group_clusters(struct super_block *sb,
169 struct ext4_group_desc *bg)
171 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
172 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
173 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
176 __u32 ext4_free_inodes_count(struct super_block *sb,
177 struct ext4_group_desc *bg)
179 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
180 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
181 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
184 __u32 ext4_used_dirs_count(struct super_block *sb,
185 struct ext4_group_desc *bg)
187 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
189 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
192 __u32 ext4_itable_unused_count(struct super_block *sb,
193 struct ext4_group_desc *bg)
195 return le16_to_cpu(bg->bg_itable_unused_lo) |
196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
197 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
200 void ext4_block_bitmap_set(struct super_block *sb,
201 struct ext4_group_desc *bg, ext4_fsblk_t blk)
203 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
208 void ext4_inode_bitmap_set(struct super_block *sb,
209 struct ext4_group_desc *bg, ext4_fsblk_t blk)
211 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
212 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
213 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
216 void ext4_inode_table_set(struct super_block *sb,
217 struct ext4_group_desc *bg, ext4_fsblk_t blk)
219 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
220 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
221 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
224 void ext4_free_group_clusters_set(struct super_block *sb,
225 struct ext4_group_desc *bg, __u32 count)
227 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
228 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
229 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
232 void ext4_free_inodes_set(struct super_block *sb,
233 struct ext4_group_desc *bg, __u32 count)
235 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
236 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
237 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
240 void ext4_used_dirs_set(struct super_block *sb,
241 struct ext4_group_desc *bg, __u32 count)
243 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
248 void ext4_itable_unused_set(struct super_block *sb,
249 struct ext4_group_desc *bg, __u32 count)
251 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
257 /* Just increment the non-pointer handle value */
258 static handle_t *ext4_get_nojournal(void)
260 handle_t *handle = current->journal_info;
261 unsigned long ref_cnt = (unsigned long)handle;
263 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
266 handle = (handle_t *)ref_cnt;
268 current->journal_info = handle;
273 /* Decrement the non-pointer handle value */
274 static void ext4_put_nojournal(handle_t *handle)
276 unsigned long ref_cnt = (unsigned long)handle;
278 BUG_ON(ref_cnt == 0);
281 handle = (handle_t *)ref_cnt;
283 current->journal_info = handle;
287 * Wrappers for jbd2_journal_start/end.
289 * The only special thing we need to do here is to make sure that all
290 * journal_end calls result in the superblock being marked dirty, so
291 * that sync() will call the filesystem's write_super callback if
294 * To avoid j_barrier hold in userspace when a user calls freeze(),
295 * ext4 prevents a new handle from being started by s_frozen, which
296 * is in an upper layer.
298 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
303 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
304 if (sb->s_flags & MS_RDONLY)
305 return ERR_PTR(-EROFS);
307 journal = EXT4_SB(sb)->s_journal;
308 handle = ext4_journal_current_handle();
311 * If a handle has been started, it should be allowed to
312 * finish, otherwise deadlock could happen between freeze
313 * and others(e.g. truncate) due to the restart of the
314 * journal handle if the filesystem is forzen and active
315 * handles are not stopped.
318 vfs_check_frozen(sb, SB_FREEZE_TRANS);
321 return ext4_get_nojournal();
323 * Special case here: if the journal has aborted behind our
324 * backs (eg. EIO in the commit thread), then we still need to
325 * take the FS itself readonly cleanly.
327 if (is_journal_aborted(journal)) {
328 ext4_abort(sb, "Detected aborted journal");
329 return ERR_PTR(-EROFS);
331 return jbd2_journal_start(journal, nblocks);
335 * The only special thing we need to do here is to make sure that all
336 * jbd2_journal_stop calls result in the superblock being marked dirty, so
337 * that sync() will call the filesystem's write_super callback if
340 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
342 struct super_block *sb;
346 if (!ext4_handle_valid(handle)) {
347 ext4_put_nojournal(handle);
350 sb = handle->h_transaction->t_journal->j_private;
352 rc = jbd2_journal_stop(handle);
357 __ext4_std_error(sb, where, line, err);
361 void ext4_journal_abort_handle(const char *caller, unsigned int line,
362 const char *err_fn, struct buffer_head *bh,
363 handle_t *handle, int err)
366 const char *errstr = ext4_decode_error(NULL, err, nbuf);
368 BUG_ON(!ext4_handle_valid(handle));
371 BUFFER_TRACE(bh, "abort");
376 if (is_handle_aborted(handle))
379 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
380 caller, line, errstr, err_fn);
382 jbd2_journal_abort_handle(handle);
385 static void __save_error_info(struct super_block *sb, const char *func,
388 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
390 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
391 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
392 es->s_last_error_time = cpu_to_le32(get_seconds());
393 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
394 es->s_last_error_line = cpu_to_le32(line);
395 if (!es->s_first_error_time) {
396 es->s_first_error_time = es->s_last_error_time;
397 strncpy(es->s_first_error_func, func,
398 sizeof(es->s_first_error_func));
399 es->s_first_error_line = cpu_to_le32(line);
400 es->s_first_error_ino = es->s_last_error_ino;
401 es->s_first_error_block = es->s_last_error_block;
404 * Start the daily error reporting function if it hasn't been
407 if (!es->s_error_count)
408 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
409 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
412 static void save_error_info(struct super_block *sb, const char *func,
415 __save_error_info(sb, func, line);
416 ext4_commit_super(sb, 1);
420 * The del_gendisk() function uninitializes the disk-specific data
421 * structures, including the bdi structure, without telling anyone
422 * else. Once this happens, any attempt to call mark_buffer_dirty()
423 * (for example, by ext4_commit_super), will cause a kernel OOPS.
424 * This is a kludge to prevent these oops until we can put in a proper
425 * hook in del_gendisk() to inform the VFS and file system layers.
427 static int block_device_ejected(struct super_block *sb)
429 struct inode *bd_inode = sb->s_bdev->bd_inode;
430 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
432 return bdi->dev == NULL;
436 /* Deal with the reporting of failure conditions on a filesystem such as
437 * inconsistencies detected or read IO failures.
439 * On ext2, we can store the error state of the filesystem in the
440 * superblock. That is not possible on ext4, because we may have other
441 * write ordering constraints on the superblock which prevent us from
442 * writing it out straight away; and given that the journal is about to
443 * be aborted, we can't rely on the current, or future, transactions to
444 * write out the superblock safely.
446 * We'll just use the jbd2_journal_abort() error code to record an error in
447 * the journal instead. On recovery, the journal will complain about
448 * that error until we've noted it down and cleared it.
451 static void ext4_handle_error(struct super_block *sb)
453 if (sb->s_flags & MS_RDONLY)
456 if (!test_opt(sb, ERRORS_CONT)) {
457 journal_t *journal = EXT4_SB(sb)->s_journal;
459 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
461 jbd2_journal_abort(journal, -EIO);
463 if (test_opt(sb, ERRORS_RO)) {
464 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
465 sb->s_flags |= MS_RDONLY;
467 if (test_opt(sb, ERRORS_PANIC))
468 panic("EXT4-fs (device %s): panic forced after error\n",
472 void __ext4_error(struct super_block *sb, const char *function,
473 unsigned int line, const char *fmt, ...)
475 struct va_format vaf;
481 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
482 sb->s_id, function, line, current->comm, &vaf);
485 ext4_handle_error(sb);
488 void ext4_error_inode(struct inode *inode, const char *function,
489 unsigned int line, ext4_fsblk_t block,
490 const char *fmt, ...)
493 struct va_format vaf;
494 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
496 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
497 es->s_last_error_block = cpu_to_le64(block);
498 save_error_info(inode->i_sb, function, line);
502 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
503 inode->i_sb->s_id, function, line, inode->i_ino);
505 printk(KERN_CONT "block %llu: ", block);
506 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
509 ext4_handle_error(inode->i_sb);
512 void ext4_error_file(struct file *file, const char *function,
513 unsigned int line, ext4_fsblk_t block,
514 const char *fmt, ...)
517 struct va_format vaf;
518 struct ext4_super_block *es;
519 struct inode *inode = file->f_dentry->d_inode;
520 char pathname[80], *path;
522 es = EXT4_SB(inode->i_sb)->s_es;
523 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
524 save_error_info(inode->i_sb, function, line);
525 path = d_path(&(file->f_path), pathname, sizeof(pathname));
529 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
530 inode->i_sb->s_id, function, line, inode->i_ino);
532 printk(KERN_CONT "block %llu: ", block);
536 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
539 ext4_handle_error(inode->i_sb);
542 static const char *ext4_decode_error(struct super_block *sb, int errno,
549 errstr = "IO failure";
552 errstr = "Out of memory";
555 if (!sb || (EXT4_SB(sb)->s_journal &&
556 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
557 errstr = "Journal has aborted";
559 errstr = "Readonly filesystem";
562 /* If the caller passed in an extra buffer for unknown
563 * errors, textualise them now. Else we just return
566 /* Check for truncated error codes... */
567 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
576 /* __ext4_std_error decodes expected errors from journaling functions
577 * automatically and invokes the appropriate error response. */
579 void __ext4_std_error(struct super_block *sb, const char *function,
580 unsigned int line, int errno)
585 /* Special case: if the error is EROFS, and we're not already
586 * inside a transaction, then there's really no point in logging
588 if (errno == -EROFS && journal_current_handle() == NULL &&
589 (sb->s_flags & MS_RDONLY))
592 errstr = ext4_decode_error(sb, errno, nbuf);
593 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
594 sb->s_id, function, line, errstr);
595 save_error_info(sb, function, line);
597 ext4_handle_error(sb);
601 * ext4_abort is a much stronger failure handler than ext4_error. The
602 * abort function may be used to deal with unrecoverable failures such
603 * as journal IO errors or ENOMEM at a critical moment in log management.
605 * We unconditionally force the filesystem into an ABORT|READONLY state,
606 * unless the error response on the fs has been set to panic in which
607 * case we take the easy way out and panic immediately.
610 void __ext4_abort(struct super_block *sb, const char *function,
611 unsigned int line, const char *fmt, ...)
615 save_error_info(sb, function, line);
617 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
623 if ((sb->s_flags & MS_RDONLY) == 0) {
624 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
625 sb->s_flags |= MS_RDONLY;
626 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
627 if (EXT4_SB(sb)->s_journal)
628 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
629 save_error_info(sb, function, line);
631 if (test_opt(sb, ERRORS_PANIC))
632 panic("EXT4-fs panic from previous error\n");
635 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
637 struct va_format vaf;
643 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
647 void __ext4_warning(struct super_block *sb, const char *function,
648 unsigned int line, const char *fmt, ...)
650 struct va_format vaf;
656 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
657 sb->s_id, function, line, &vaf);
661 void __ext4_grp_locked_error(const char *function, unsigned int line,
662 struct super_block *sb, ext4_group_t grp,
663 unsigned long ino, ext4_fsblk_t block,
664 const char *fmt, ...)
668 struct va_format vaf;
670 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
672 es->s_last_error_ino = cpu_to_le32(ino);
673 es->s_last_error_block = cpu_to_le64(block);
674 __save_error_info(sb, function, line);
680 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
681 sb->s_id, function, line, grp);
683 printk(KERN_CONT "inode %lu: ", ino);
685 printk(KERN_CONT "block %llu:", (unsigned long long) block);
686 printk(KERN_CONT "%pV\n", &vaf);
689 if (test_opt(sb, ERRORS_CONT)) {
690 ext4_commit_super(sb, 0);
694 ext4_unlock_group(sb, grp);
695 ext4_handle_error(sb);
697 * We only get here in the ERRORS_RO case; relocking the group
698 * may be dangerous, but nothing bad will happen since the
699 * filesystem will have already been marked read/only and the
700 * journal has been aborted. We return 1 as a hint to callers
701 * who might what to use the return value from
702 * ext4_grp_locked_error() to distinguish between the
703 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
704 * aggressively from the ext4 function in question, with a
705 * more appropriate error code.
707 ext4_lock_group(sb, grp);
711 void ext4_update_dynamic_rev(struct super_block *sb)
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
715 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
719 "updating to rev %d because of new feature flag, "
720 "running e2fsck is recommended",
723 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
724 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
725 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
726 /* leave es->s_feature_*compat flags alone */
727 /* es->s_uuid will be set by e2fsck if empty */
730 * The rest of the superblock fields should be zero, and if not it
731 * means they are likely already in use, so leave them alone. We
732 * can leave it up to e2fsck to clean up any inconsistencies there.
737 * Open the external journal device
739 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
741 struct block_device *bdev;
742 char b[BDEVNAME_SIZE];
744 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
750 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
751 __bdevname(dev, b), PTR_ERR(bdev));
756 * Release the journal device
758 static int ext4_blkdev_put(struct block_device *bdev)
760 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
763 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
765 struct block_device *bdev;
768 bdev = sbi->journal_bdev;
770 ret = ext4_blkdev_put(bdev);
771 sbi->journal_bdev = NULL;
776 static inline struct inode *orphan_list_entry(struct list_head *l)
778 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
781 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
785 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
786 le32_to_cpu(sbi->s_es->s_last_orphan));
788 printk(KERN_ERR "sb_info orphan list:\n");
789 list_for_each(l, &sbi->s_orphan) {
790 struct inode *inode = orphan_list_entry(l);
792 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
793 inode->i_sb->s_id, inode->i_ino, inode,
794 inode->i_mode, inode->i_nlink,
799 static void ext4_put_super(struct super_block *sb)
801 struct ext4_sb_info *sbi = EXT4_SB(sb);
802 struct ext4_super_block *es = sbi->s_es;
805 ext4_unregister_li_request(sb);
806 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
808 flush_workqueue(sbi->dio_unwritten_wq);
809 destroy_workqueue(sbi->dio_unwritten_wq);
813 ext4_commit_super(sb, 1);
815 if (sbi->s_journal) {
816 err = jbd2_journal_destroy(sbi->s_journal);
817 sbi->s_journal = NULL;
819 ext4_abort(sb, "Couldn't clean up the journal");
822 del_timer(&sbi->s_err_report);
823 ext4_release_system_zone(sb);
825 ext4_ext_release(sb);
826 ext4_xattr_put_super(sb);
828 if (!(sb->s_flags & MS_RDONLY)) {
829 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
830 es->s_state = cpu_to_le16(sbi->s_mount_state);
831 ext4_commit_super(sb, 1);
834 remove_proc_entry(sb->s_id, ext4_proc_root);
836 kobject_del(&sbi->s_kobj);
838 for (i = 0; i < sbi->s_gdb_count; i++)
839 brelse(sbi->s_group_desc[i]);
840 ext4_kvfree(sbi->s_group_desc);
841 ext4_kvfree(sbi->s_flex_groups);
842 percpu_counter_destroy(&sbi->s_freeclusters_counter);
843 percpu_counter_destroy(&sbi->s_freeinodes_counter);
844 percpu_counter_destroy(&sbi->s_dirs_counter);
845 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
848 for (i = 0; i < MAXQUOTAS; i++)
849 kfree(sbi->s_qf_names[i]);
852 /* Debugging code just in case the in-memory inode orphan list
853 * isn't empty. The on-disk one can be non-empty if we've
854 * detected an error and taken the fs readonly, but the
855 * in-memory list had better be clean by this point. */
856 if (!list_empty(&sbi->s_orphan))
857 dump_orphan_list(sb, sbi);
858 J_ASSERT(list_empty(&sbi->s_orphan));
860 invalidate_bdev(sb->s_bdev);
861 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
863 * Invalidate the journal device's buffers. We don't want them
864 * floating about in memory - the physical journal device may
865 * hotswapped, and it breaks the `ro-after' testing code.
867 sync_blockdev(sbi->journal_bdev);
868 invalidate_bdev(sbi->journal_bdev);
869 ext4_blkdev_remove(sbi);
872 kthread_stop(sbi->s_mmp_tsk);
873 sb->s_fs_info = NULL;
875 * Now that we are completely done shutting down the
876 * superblock, we need to actually destroy the kobject.
879 kobject_put(&sbi->s_kobj);
880 wait_for_completion(&sbi->s_kobj_unregister);
881 kfree(sbi->s_blockgroup_lock);
885 static struct kmem_cache *ext4_inode_cachep;
888 * Called inside transaction, so use GFP_NOFS
890 static struct inode *ext4_alloc_inode(struct super_block *sb)
892 struct ext4_inode_info *ei;
894 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
897 #ifdef CONFIG_EXT4_FS_RICHACL
898 ei->i_richacl = EXT4_RICHACL_NOT_CACHED;
900 ei->vfs_inode.i_version = 1;
901 ei->vfs_inode.i_data.writeback_index = 0;
902 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
903 INIT_LIST_HEAD(&ei->i_prealloc_list);
904 spin_lock_init(&ei->i_prealloc_lock);
905 ei->i_reserved_data_blocks = 0;
906 ei->i_reserved_meta_blocks = 0;
907 ei->i_allocated_meta_blocks = 0;
908 ei->i_da_metadata_calc_len = 0;
909 spin_lock_init(&(ei->i_block_reservation_lock));
911 ei->i_reserved_quota = 0;
914 INIT_LIST_HEAD(&ei->i_completed_io_list);
915 spin_lock_init(&ei->i_completed_io_lock);
916 ei->cur_aio_dio = NULL;
918 ei->i_datasync_tid = 0;
919 atomic_set(&ei->i_ioend_count, 0);
920 atomic_set(&ei->i_aiodio_unwritten, 0);
922 return &ei->vfs_inode;
925 static int ext4_drop_inode(struct inode *inode)
927 int drop = generic_drop_inode(inode);
929 trace_ext4_drop_inode(inode, drop);
933 static void ext4_i_callback(struct rcu_head *head)
935 struct inode *inode = container_of(head, struct inode, i_rcu);
936 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
939 static void ext4_destroy_inode(struct inode *inode)
941 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
942 ext4_msg(inode->i_sb, KERN_ERR,
943 "Inode %lu (%p): orphan list check failed!",
944 inode->i_ino, EXT4_I(inode));
945 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
946 EXT4_I(inode), sizeof(struct ext4_inode_info),
950 call_rcu(&inode->i_rcu, ext4_i_callback);
953 static void init_once(void *foo)
955 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
957 INIT_LIST_HEAD(&ei->i_orphan);
958 #ifdef CONFIG_EXT4_FS_XATTR
959 init_rwsem(&ei->xattr_sem);
961 init_rwsem(&ei->i_data_sem);
962 inode_init_once(&ei->vfs_inode);
965 static int init_inodecache(void)
967 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
968 sizeof(struct ext4_inode_info),
969 0, (SLAB_RECLAIM_ACCOUNT|
972 if (ext4_inode_cachep == NULL)
977 static void destroy_inodecache(void)
979 kmem_cache_destroy(ext4_inode_cachep);
982 void ext4_clear_inode(struct inode *inode)
984 invalidate_inode_buffers(inode);
985 end_writeback(inode);
987 #ifdef CONFIG_EXT4_FS_RICHACL
988 if (EXT4_I(inode)->i_richacl &&
989 EXT4_I(inode)->i_richacl != EXT4_RICHACL_NOT_CACHED) {
990 richacl_put(EXT4_I(inode)->i_richacl);
991 EXT4_I(inode)->i_richacl = EXT4_RICHACL_NOT_CACHED;
994 ext4_discard_preallocations(inode);
995 if (EXT4_I(inode)->jinode) {
996 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
997 EXT4_I(inode)->jinode);
998 jbd2_free_inode(EXT4_I(inode)->jinode);
999 EXT4_I(inode)->jinode = NULL;
1003 static inline void ext4_show_quota_options(struct seq_file *seq,
1004 struct super_block *sb)
1006 #if defined(CONFIG_QUOTA)
1007 struct ext4_sb_info *sbi = EXT4_SB(sb);
1009 if (sbi->s_jquota_fmt) {
1012 switch (sbi->s_jquota_fmt) {
1023 seq_printf(seq, ",jqfmt=%s", fmtname);
1026 if (sbi->s_qf_names[USRQUOTA])
1027 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1029 if (sbi->s_qf_names[GRPQUOTA])
1030 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1032 if (test_opt(sb, USRQUOTA))
1033 seq_puts(seq, ",usrquota");
1035 if (test_opt(sb, GRPQUOTA))
1036 seq_puts(seq, ",grpquota");
1042 * - it's set to a non-default value OR
1043 * - if the per-sb default is different from the global default
1045 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1048 unsigned long def_mount_opts;
1049 struct super_block *sb = root->d_sb;
1050 struct ext4_sb_info *sbi = EXT4_SB(sb);
1051 struct ext4_super_block *es = sbi->s_es;
1053 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1054 def_errors = le16_to_cpu(es->s_errors);
1056 if (sbi->s_sb_block != 1)
1057 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1058 if (test_opt(sb, MINIX_DF))
1059 seq_puts(seq, ",minixdf");
1060 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1061 seq_puts(seq, ",grpid");
1062 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1063 seq_puts(seq, ",nogrpid");
1064 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1065 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1066 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1068 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1069 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1070 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1072 if (test_opt(sb, ERRORS_RO)) {
1073 if (def_errors == EXT4_ERRORS_PANIC ||
1074 def_errors == EXT4_ERRORS_CONTINUE) {
1075 seq_puts(seq, ",errors=remount-ro");
1078 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1079 seq_puts(seq, ",errors=continue");
1080 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1081 seq_puts(seq, ",errors=panic");
1082 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1083 seq_puts(seq, ",nouid32");
1084 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1085 seq_puts(seq, ",debug");
1086 #ifdef CONFIG_EXT4_FS_XATTR
1087 if (test_opt(sb, XATTR_USER))
1088 seq_puts(seq, ",user_xattr");
1089 if (!test_opt(sb, XATTR_USER))
1090 seq_puts(seq, ",nouser_xattr");
1092 #if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4_FS_RICHACL)
1093 if (sb->s_flags & MS_POSIXACL) {
1094 if (!(def_mount_opts & EXT4_DEFM_ACL))
1095 seq_puts(seq, ",acl");
1096 } else if (sb->s_flags & MS_RICHACL)
1097 seq_puts(seq, ",richacl");
1098 else if (def_mount_opts & EXT4_DEFM_ACL)
1099 seq_puts(seq, ",noacl");
1101 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1102 seq_printf(seq, ",commit=%u",
1103 (unsigned) (sbi->s_commit_interval / HZ));
1105 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1106 seq_printf(seq, ",min_batch_time=%u",
1107 (unsigned) sbi->s_min_batch_time);
1109 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1110 seq_printf(seq, ",max_batch_time=%u",
1111 (unsigned) sbi->s_max_batch_time);
1115 * We're changing the default of barrier mount option, so
1116 * let's always display its mount state so it's clear what its
1119 seq_puts(seq, ",barrier=");
1120 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1121 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1122 seq_puts(seq, ",journal_async_commit");
1123 else if (test_opt(sb, JOURNAL_CHECKSUM))
1124 seq_puts(seq, ",journal_checksum");
1125 if (test_opt(sb, I_VERSION))
1126 seq_puts(seq, ",i_version");
1127 if (!test_opt(sb, DELALLOC) &&
1128 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1129 seq_puts(seq, ",nodelalloc");
1131 if (!test_opt(sb, MBLK_IO_SUBMIT))
1132 seq_puts(seq, ",nomblk_io_submit");
1134 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1136 * journal mode get enabled in different ways
1137 * So just print the value even if we didn't specify it
1139 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1140 seq_puts(seq, ",data=journal");
1141 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1142 seq_puts(seq, ",data=ordered");
1143 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1144 seq_puts(seq, ",data=writeback");
1146 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1147 seq_printf(seq, ",inode_readahead_blks=%u",
1148 sbi->s_inode_readahead_blks);
1150 if (test_opt(sb, DATA_ERR_ABORT))
1151 seq_puts(seq, ",data_err=abort");
1153 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1154 seq_puts(seq, ",noauto_da_alloc");
1156 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1157 seq_puts(seq, ",discard");
1159 if (test_opt(sb, NOLOAD))
1160 seq_puts(seq, ",norecovery");
1162 if (test_opt(sb, DIOREAD_NOLOCK))
1163 seq_puts(seq, ",dioread_nolock");
1165 if (test_opt(sb, BLOCK_VALIDITY) &&
1166 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1167 seq_puts(seq, ",block_validity");
1169 if (!test_opt(sb, INIT_INODE_TABLE))
1170 seq_puts(seq, ",noinit_itable");
1171 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1172 seq_printf(seq, ",init_itable=%u",
1173 (unsigned) sbi->s_li_wait_mult);
1175 ext4_show_quota_options(seq, sb);
1180 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1181 u64 ino, u32 generation)
1183 struct inode *inode;
1185 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1186 return ERR_PTR(-ESTALE);
1187 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1188 return ERR_PTR(-ESTALE);
1190 /* iget isn't really right if the inode is currently unallocated!!
1192 * ext4_read_inode will return a bad_inode if the inode had been
1193 * deleted, so we should be safe.
1195 * Currently we don't know the generation for parent directory, so
1196 * a generation of 0 means "accept any"
1198 inode = ext4_iget(sb, ino);
1200 return ERR_CAST(inode);
1201 if (generation && inode->i_generation != generation) {
1203 return ERR_PTR(-ESTALE);
1209 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1210 int fh_len, int fh_type)
1212 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1213 ext4_nfs_get_inode);
1216 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1217 int fh_len, int fh_type)
1219 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1220 ext4_nfs_get_inode);
1224 * Try to release metadata pages (indirect blocks, directories) which are
1225 * mapped via the block device. Since these pages could have journal heads
1226 * which would prevent try_to_free_buffers() from freeing them, we must use
1227 * jbd2 layer's try_to_free_buffers() function to release them.
1229 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1232 journal_t *journal = EXT4_SB(sb)->s_journal;
1234 WARN_ON(PageChecked(page));
1235 if (!page_has_buffers(page))
1238 return jbd2_journal_try_to_free_buffers(journal, page,
1239 wait & ~__GFP_WAIT);
1240 return try_to_free_buffers(page);
1244 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1245 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1247 static int ext4_write_dquot(struct dquot *dquot);
1248 static int ext4_acquire_dquot(struct dquot *dquot);
1249 static int ext4_release_dquot(struct dquot *dquot);
1250 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1251 static int ext4_write_info(struct super_block *sb, int type);
1252 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1254 static int ext4_quota_off(struct super_block *sb, int type);
1255 static int ext4_quota_on_mount(struct super_block *sb, int type);
1256 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1257 size_t len, loff_t off);
1258 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1259 const char *data, size_t len, loff_t off);
1261 static const struct dquot_operations ext4_quota_operations = {
1262 .get_reserved_space = ext4_get_reserved_space,
1263 .write_dquot = ext4_write_dquot,
1264 .acquire_dquot = ext4_acquire_dquot,
1265 .release_dquot = ext4_release_dquot,
1266 .mark_dirty = ext4_mark_dquot_dirty,
1267 .write_info = ext4_write_info,
1268 .alloc_dquot = dquot_alloc,
1269 .destroy_dquot = dquot_destroy,
1272 static const struct quotactl_ops ext4_qctl_operations = {
1273 .quota_on = ext4_quota_on,
1274 .quota_off = ext4_quota_off,
1275 .quota_sync = dquot_quota_sync,
1276 .get_info = dquot_get_dqinfo,
1277 .set_info = dquot_set_dqinfo,
1278 .get_dqblk = dquot_get_dqblk,
1279 .set_dqblk = dquot_set_dqblk
1283 static const struct super_operations ext4_sops = {
1284 .alloc_inode = ext4_alloc_inode,
1285 .destroy_inode = ext4_destroy_inode,
1286 .write_inode = ext4_write_inode,
1287 .dirty_inode = ext4_dirty_inode,
1288 .drop_inode = ext4_drop_inode,
1289 .evict_inode = ext4_evict_inode,
1290 .put_super = ext4_put_super,
1291 .sync_fs = ext4_sync_fs,
1292 .freeze_fs = ext4_freeze,
1293 .unfreeze_fs = ext4_unfreeze,
1294 .statfs = ext4_statfs,
1295 .remount_fs = ext4_remount,
1296 .show_options = ext4_show_options,
1298 .quota_read = ext4_quota_read,
1299 .quota_write = ext4_quota_write,
1301 .bdev_try_to_free_page = bdev_try_to_free_page,
1304 static const struct super_operations ext4_nojournal_sops = {
1305 .alloc_inode = ext4_alloc_inode,
1306 .destroy_inode = ext4_destroy_inode,
1307 .write_inode = ext4_write_inode,
1308 .dirty_inode = ext4_dirty_inode,
1309 .drop_inode = ext4_drop_inode,
1310 .evict_inode = ext4_evict_inode,
1311 .write_super = ext4_write_super,
1312 .put_super = ext4_put_super,
1313 .statfs = ext4_statfs,
1314 .remount_fs = ext4_remount,
1315 .show_options = ext4_show_options,
1317 .quota_read = ext4_quota_read,
1318 .quota_write = ext4_quota_write,
1320 .bdev_try_to_free_page = bdev_try_to_free_page,
1323 static const struct export_operations ext4_export_ops = {
1324 .fh_to_dentry = ext4_fh_to_dentry,
1325 .fh_to_parent = ext4_fh_to_parent,
1326 .get_parent = ext4_get_parent,
1330 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1331 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1332 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1333 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_richacl, Opt_noacl,
1334 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1335 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1336 Opt_journal_update, Opt_journal_dev,
1337 Opt_journal_checksum, Opt_journal_async_commit,
1338 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1339 Opt_data_err_abort, Opt_data_err_ignore,
1340 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1341 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1342 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1343 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1344 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1345 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1346 Opt_inode_readahead_blks, Opt_journal_ioprio,
1347 Opt_dioread_nolock, Opt_dioread_lock,
1348 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1351 static const match_table_t tokens = {
1352 {Opt_bsd_df, "bsddf"},
1353 {Opt_minix_df, "minixdf"},
1354 {Opt_grpid, "grpid"},
1355 {Opt_grpid, "bsdgroups"},
1356 {Opt_nogrpid, "nogrpid"},
1357 {Opt_nogrpid, "sysvgroups"},
1358 {Opt_resgid, "resgid=%u"},
1359 {Opt_resuid, "resuid=%u"},
1361 {Opt_err_cont, "errors=continue"},
1362 {Opt_err_panic, "errors=panic"},
1363 {Opt_err_ro, "errors=remount-ro"},
1364 {Opt_nouid32, "nouid32"},
1365 {Opt_debug, "debug"},
1366 {Opt_oldalloc, "oldalloc"},
1367 {Opt_orlov, "orlov"},
1368 {Opt_user_xattr, "user_xattr"},
1369 {Opt_nouser_xattr, "nouser_xattr"},
1371 {Opt_richacl, "richacl"},
1372 {Opt_noacl, "noacl"},
1373 {Opt_noload, "noload"},
1374 {Opt_noload, "norecovery"},
1377 {Opt_commit, "commit=%u"},
1378 {Opt_min_batch_time, "min_batch_time=%u"},
1379 {Opt_max_batch_time, "max_batch_time=%u"},
1380 {Opt_journal_update, "journal=update"},
1381 {Opt_journal_dev, "journal_dev=%u"},
1382 {Opt_journal_checksum, "journal_checksum"},
1383 {Opt_journal_async_commit, "journal_async_commit"},
1384 {Opt_abort, "abort"},
1385 {Opt_data_journal, "data=journal"},
1386 {Opt_data_ordered, "data=ordered"},
1387 {Opt_data_writeback, "data=writeback"},
1388 {Opt_data_err_abort, "data_err=abort"},
1389 {Opt_data_err_ignore, "data_err=ignore"},
1390 {Opt_offusrjquota, "usrjquota="},
1391 {Opt_usrjquota, "usrjquota=%s"},
1392 {Opt_offgrpjquota, "grpjquota="},
1393 {Opt_grpjquota, "grpjquota=%s"},
1394 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1395 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1396 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1397 {Opt_grpquota, "grpquota"},
1398 {Opt_noquota, "noquota"},
1399 {Opt_quota, "quota"},
1400 {Opt_usrquota, "usrquota"},
1401 {Opt_barrier, "barrier=%u"},
1402 {Opt_barrier, "barrier"},
1403 {Opt_nobarrier, "nobarrier"},
1404 {Opt_i_version, "i_version"},
1405 {Opt_stripe, "stripe=%u"},
1406 {Opt_resize, "resize"},
1407 {Opt_delalloc, "delalloc"},
1408 {Opt_nodelalloc, "nodelalloc"},
1409 {Opt_mblk_io_submit, "mblk_io_submit"},
1410 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1411 {Opt_block_validity, "block_validity"},
1412 {Opt_noblock_validity, "noblock_validity"},
1413 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1414 {Opt_journal_ioprio, "journal_ioprio=%u"},
1415 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1416 {Opt_auto_da_alloc, "auto_da_alloc"},
1417 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1418 {Opt_dioread_nolock, "dioread_nolock"},
1419 {Opt_dioread_lock, "dioread_lock"},
1420 {Opt_discard, "discard"},
1421 {Opt_nodiscard, "nodiscard"},
1422 {Opt_init_itable, "init_itable=%u"},
1423 {Opt_init_itable, "init_itable"},
1424 {Opt_noinit_itable, "noinit_itable"},
1428 static ext4_fsblk_t get_sb_block(void **data)
1430 ext4_fsblk_t sb_block;
1431 char *options = (char *) *data;
1433 if (!options || strncmp(options, "sb=", 3) != 0)
1434 return 1; /* Default location */
1437 /* TODO: use simple_strtoll with >32bit ext4 */
1438 sb_block = simple_strtoul(options, &options, 0);
1439 if (*options && *options != ',') {
1440 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1444 if (*options == ',')
1446 *data = (void *) options;
1451 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1452 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1453 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1456 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1458 struct ext4_sb_info *sbi = EXT4_SB(sb);
1461 if (sb_any_quota_loaded(sb) &&
1462 !sbi->s_qf_names[qtype]) {
1463 ext4_msg(sb, KERN_ERR,
1464 "Cannot change journaled "
1465 "quota options when quota turned on");
1468 qname = match_strdup(args);
1470 ext4_msg(sb, KERN_ERR,
1471 "Not enough memory for storing quotafile name");
1474 if (sbi->s_qf_names[qtype] &&
1475 strcmp(sbi->s_qf_names[qtype], qname)) {
1476 ext4_msg(sb, KERN_ERR,
1477 "%s quota file already specified", QTYPE2NAME(qtype));
1481 sbi->s_qf_names[qtype] = qname;
1482 if (strchr(sbi->s_qf_names[qtype], '/')) {
1483 ext4_msg(sb, KERN_ERR,
1484 "quotafile must be on filesystem root");
1485 kfree(sbi->s_qf_names[qtype]);
1486 sbi->s_qf_names[qtype] = NULL;
1493 static int clear_qf_name(struct super_block *sb, int qtype)
1496 struct ext4_sb_info *sbi = EXT4_SB(sb);
1498 if (sb_any_quota_loaded(sb) &&
1499 sbi->s_qf_names[qtype]) {
1500 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1501 " when quota turned on");
1505 * The space will be released later when all options are confirmed
1508 sbi->s_qf_names[qtype] = NULL;
1513 static int parse_options(char *options, struct super_block *sb,
1514 unsigned long *journal_devnum,
1515 unsigned int *journal_ioprio,
1516 ext4_fsblk_t *n_blocks_count, int is_remount)
1518 struct ext4_sb_info *sbi = EXT4_SB(sb);
1520 substring_t args[MAX_OPT_ARGS];
1530 while ((p = strsep(&options, ",")) != NULL) {
1536 * Initialize args struct so we know whether arg was
1537 * found; some options take optional arguments.
1539 args[0].to = args[0].from = NULL;
1540 token = match_token(p, tokens, args);
1543 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1544 clear_opt(sb, MINIX_DF);
1547 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1548 set_opt(sb, MINIX_DF);
1552 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1557 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1558 clear_opt(sb, GRPID);
1562 if (match_int(&args[0], &option))
1564 sbi->s_resuid = option;
1567 if (match_int(&args[0], &option))
1569 sbi->s_resgid = option;
1572 /* handled by get_sb_block() instead of here */
1573 /* *sb_block = match_int(&args[0]); */
1576 clear_opt(sb, ERRORS_CONT);
1577 clear_opt(sb, ERRORS_RO);
1578 set_opt(sb, ERRORS_PANIC);
1581 clear_opt(sb, ERRORS_CONT);
1582 clear_opt(sb, ERRORS_PANIC);
1583 set_opt(sb, ERRORS_RO);
1586 clear_opt(sb, ERRORS_RO);
1587 clear_opt(sb, ERRORS_PANIC);
1588 set_opt(sb, ERRORS_CONT);
1591 set_opt(sb, NO_UID32);
1597 ext4_msg(sb, KERN_WARNING,
1598 "Ignoring deprecated oldalloc option");
1601 ext4_msg(sb, KERN_WARNING,
1602 "Ignoring deprecated orlov option");
1604 #ifdef CONFIG_EXT4_FS_XATTR
1605 case Opt_user_xattr:
1606 set_opt(sb, XATTR_USER);
1608 case Opt_nouser_xattr:
1609 clear_opt(sb, XATTR_USER);
1612 case Opt_user_xattr:
1613 case Opt_nouser_xattr:
1614 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1617 #if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4_FS_RICHACL)
1618 # ifdef CONFIG_EXT4_FS_POSIX_ACL
1620 if (!(sb->s_flags & MS_RICHACL))
1621 sb->s_flags |= MS_POSIXACL;
1624 # ifdef CONFIG_EXT4_FS_RICHACL
1626 sb->s_flags &= ~MS_POSIXACL;
1627 sb->s_flags |= MS_RICHACL;
1631 sb->s_flags &= ~(MS_POSIXACL | MS_RICHACL);
1636 ext4_msg(sb, KERN_ERR, "(no)acl/richacl options not supported");
1639 case Opt_journal_update:
1641 /* Eventually we will want to be able to create
1642 a journal file here. For now, only allow the
1643 user to specify an existing inode to be the
1646 ext4_msg(sb, KERN_ERR,
1647 "Cannot specify journal on remount");
1650 set_opt(sb, UPDATE_JOURNAL);
1652 case Opt_journal_dev:
1654 ext4_msg(sb, KERN_ERR,
1655 "Cannot specify journal on remount");
1658 if (match_int(&args[0], &option))
1660 *journal_devnum = option;
1662 case Opt_journal_checksum:
1663 set_opt(sb, JOURNAL_CHECKSUM);
1665 case Opt_journal_async_commit:
1666 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1667 set_opt(sb, JOURNAL_CHECKSUM);
1670 set_opt(sb, NOLOAD);
1673 if (match_int(&args[0], &option))
1678 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1679 sbi->s_commit_interval = HZ * option;
1681 case Opt_max_batch_time:
1682 if (match_int(&args[0], &option))
1687 option = EXT4_DEF_MAX_BATCH_TIME;
1688 sbi->s_max_batch_time = option;
1690 case Opt_min_batch_time:
1691 if (match_int(&args[0], &option))
1695 sbi->s_min_batch_time = option;
1697 case Opt_data_journal:
1698 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1700 case Opt_data_ordered:
1701 data_opt = EXT4_MOUNT_ORDERED_DATA;
1703 case Opt_data_writeback:
1704 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1707 if (!sbi->s_journal)
1708 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1709 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1710 ext4_msg(sb, KERN_ERR,
1711 "Cannot change data mode on remount");
1715 clear_opt(sb, DATA_FLAGS);
1716 sbi->s_mount_opt |= data_opt;
1719 case Opt_data_err_abort:
1720 set_opt(sb, DATA_ERR_ABORT);
1722 case Opt_data_err_ignore:
1723 clear_opt(sb, DATA_ERR_ABORT);
1727 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1731 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1734 case Opt_offusrjquota:
1735 if (!clear_qf_name(sb, USRQUOTA))
1738 case Opt_offgrpjquota:
1739 if (!clear_qf_name(sb, GRPQUOTA))
1743 case Opt_jqfmt_vfsold:
1744 qfmt = QFMT_VFS_OLD;
1746 case Opt_jqfmt_vfsv0:
1749 case Opt_jqfmt_vfsv1:
1752 if (sb_any_quota_loaded(sb) &&
1753 sbi->s_jquota_fmt != qfmt) {
1754 ext4_msg(sb, KERN_ERR, "Cannot change "
1755 "journaled quota options when "
1759 sbi->s_jquota_fmt = qfmt;
1764 set_opt(sb, USRQUOTA);
1768 set_opt(sb, GRPQUOTA);
1771 if (sb_any_quota_loaded(sb)) {
1772 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1773 "options when quota turned on");
1776 clear_opt(sb, QUOTA);
1777 clear_opt(sb, USRQUOTA);
1778 clear_opt(sb, GRPQUOTA);
1784 ext4_msg(sb, KERN_ERR,
1785 "quota options not supported");
1789 case Opt_offusrjquota:
1790 case Opt_offgrpjquota:
1791 case Opt_jqfmt_vfsold:
1792 case Opt_jqfmt_vfsv0:
1793 case Opt_jqfmt_vfsv1:
1794 ext4_msg(sb, KERN_ERR,
1795 "journaled quota options not supported");
1801 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1804 clear_opt(sb, BARRIER);
1808 if (match_int(&args[0], &option))
1811 option = 1; /* No argument, default to 1 */
1813 set_opt(sb, BARRIER);
1815 clear_opt(sb, BARRIER);
1821 ext4_msg(sb, KERN_ERR,
1822 "resize option only available "
1826 if (match_int(&args[0], &option) != 0)
1828 *n_blocks_count = option;
1831 ext4_msg(sb, KERN_WARNING,
1832 "Ignoring deprecated nobh option");
1835 ext4_msg(sb, KERN_WARNING,
1836 "Ignoring deprecated bh option");
1839 set_opt(sb, I_VERSION);
1840 sb->s_flags |= MS_I_VERSION;
1842 case Opt_nodelalloc:
1843 clear_opt(sb, DELALLOC);
1844 clear_opt2(sb, EXPLICIT_DELALLOC);
1846 case Opt_mblk_io_submit:
1847 set_opt(sb, MBLK_IO_SUBMIT);
1849 case Opt_nomblk_io_submit:
1850 clear_opt(sb, MBLK_IO_SUBMIT);
1853 if (match_int(&args[0], &option))
1857 sbi->s_stripe = option;
1860 set_opt(sb, DELALLOC);
1861 set_opt2(sb, EXPLICIT_DELALLOC);
1863 case Opt_block_validity:
1864 set_opt(sb, BLOCK_VALIDITY);
1866 case Opt_noblock_validity:
1867 clear_opt(sb, BLOCK_VALIDITY);
1869 case Opt_inode_readahead_blks:
1870 if (match_int(&args[0], &option))
1872 if (option < 0 || option > (1 << 30))
1874 if (option && !is_power_of_2(option)) {
1875 ext4_msg(sb, KERN_ERR,
1876 "EXT4-fs: inode_readahead_blks"
1877 " must be a power of 2");
1880 sbi->s_inode_readahead_blks = option;
1882 case Opt_journal_ioprio:
1883 if (match_int(&args[0], &option))
1885 if (option < 0 || option > 7)
1887 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1890 case Opt_noauto_da_alloc:
1891 set_opt(sb, NO_AUTO_DA_ALLOC);
1893 case Opt_auto_da_alloc:
1895 if (match_int(&args[0], &option))
1898 option = 1; /* No argument, default to 1 */
1900 clear_opt(sb, NO_AUTO_DA_ALLOC);
1902 set_opt(sb,NO_AUTO_DA_ALLOC);
1905 set_opt(sb, DISCARD);
1908 clear_opt(sb, DISCARD);
1910 case Opt_dioread_nolock:
1911 set_opt(sb, DIOREAD_NOLOCK);
1913 case Opt_dioread_lock:
1914 clear_opt(sb, DIOREAD_NOLOCK);
1916 case Opt_init_itable:
1917 set_opt(sb, INIT_INODE_TABLE);
1919 if (match_int(&args[0], &option))
1922 option = EXT4_DEF_LI_WAIT_MULT;
1925 sbi->s_li_wait_mult = option;
1927 case Opt_noinit_itable:
1928 clear_opt(sb, INIT_INODE_TABLE);
1931 ext4_msg(sb, KERN_ERR,
1932 "Unrecognized mount option \"%s\" "
1933 "or missing value", p);
1938 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1939 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1940 clear_opt(sb, USRQUOTA);
1942 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1943 clear_opt(sb, GRPQUOTA);
1945 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1946 ext4_msg(sb, KERN_ERR, "old and new quota "
1951 if (!sbi->s_jquota_fmt) {
1952 ext4_msg(sb, KERN_ERR, "journaled quota format "
1957 if (sbi->s_jquota_fmt) {
1958 ext4_msg(sb, KERN_ERR, "journaled quota format "
1959 "specified with no journaling "
1968 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1971 struct ext4_sb_info *sbi = EXT4_SB(sb);
1974 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1975 ext4_msg(sb, KERN_ERR, "revision level too high, "
1976 "forcing read-only mode");
1981 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1982 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1983 "running e2fsck is recommended");
1984 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1985 ext4_msg(sb, KERN_WARNING,
1986 "warning: mounting fs with errors, "
1987 "running e2fsck is recommended");
1988 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1989 le16_to_cpu(es->s_mnt_count) >=
1990 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1991 ext4_msg(sb, KERN_WARNING,
1992 "warning: maximal mount count reached, "
1993 "running e2fsck is recommended");
1994 else if (le32_to_cpu(es->s_checkinterval) &&
1995 (le32_to_cpu(es->s_lastcheck) +
1996 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1997 ext4_msg(sb, KERN_WARNING,
1998 "warning: checktime reached, "
1999 "running e2fsck is recommended");
2000 if (!sbi->s_journal)
2001 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2002 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2003 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2004 le16_add_cpu(&es->s_mnt_count, 1);
2005 es->s_mtime = cpu_to_le32(get_seconds());
2006 ext4_update_dynamic_rev(sb);
2008 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2010 ext4_commit_super(sb, 1);
2012 if (test_opt(sb, DEBUG))
2013 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2014 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2016 sbi->s_groups_count,
2017 EXT4_BLOCKS_PER_GROUP(sb),
2018 EXT4_INODES_PER_GROUP(sb),
2019 sbi->s_mount_opt, sbi->s_mount_opt2);
2021 cleancache_init_fs(sb);
2025 static int ext4_fill_flex_info(struct super_block *sb)
2027 struct ext4_sb_info *sbi = EXT4_SB(sb);
2028 struct ext4_group_desc *gdp = NULL;
2029 ext4_group_t flex_group_count;
2030 ext4_group_t flex_group;
2031 unsigned int groups_per_flex = 0;
2035 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2036 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2037 sbi->s_log_groups_per_flex = 0;
2040 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2042 /* We allocate both existing and potentially added groups */
2043 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2044 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2045 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2046 size = flex_group_count * sizeof(struct flex_groups);
2047 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2048 if (sbi->s_flex_groups == NULL) {
2049 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2054 for (i = 0; i < sbi->s_groups_count; i++) {
2055 gdp = ext4_get_group_desc(sb, i, NULL);
2057 flex_group = ext4_flex_group(sbi, i);
2058 atomic_add(ext4_free_inodes_count(sb, gdp),
2059 &sbi->s_flex_groups[flex_group].free_inodes);
2060 atomic_add(ext4_free_group_clusters(sb, gdp),
2061 &sbi->s_flex_groups[flex_group].free_clusters);
2062 atomic_add(ext4_used_dirs_count(sb, gdp),
2063 &sbi->s_flex_groups[flex_group].used_dirs);
2071 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2072 struct ext4_group_desc *gdp)
2076 if (sbi->s_es->s_feature_ro_compat &
2077 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2078 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2079 __le32 le_group = cpu_to_le32(block_group);
2081 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2082 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2083 crc = crc16(crc, (__u8 *)gdp, offset);
2084 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2085 /* for checksum of struct ext4_group_desc do the rest...*/
2086 if ((sbi->s_es->s_feature_incompat &
2087 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2088 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2089 crc = crc16(crc, (__u8 *)gdp + offset,
2090 le16_to_cpu(sbi->s_es->s_desc_size) -
2094 return cpu_to_le16(crc);
2097 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2098 struct ext4_group_desc *gdp)
2100 if ((sbi->s_es->s_feature_ro_compat &
2101 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2102 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2108 /* Called at mount-time, super-block is locked */
2109 static int ext4_check_descriptors(struct super_block *sb,
2110 ext4_group_t *first_not_zeroed)
2112 struct ext4_sb_info *sbi = EXT4_SB(sb);
2113 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2114 ext4_fsblk_t last_block;
2115 ext4_fsblk_t block_bitmap;
2116 ext4_fsblk_t inode_bitmap;
2117 ext4_fsblk_t inode_table;
2118 int flexbg_flag = 0;
2119 ext4_group_t i, grp = sbi->s_groups_count;
2121 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2124 ext4_debug("Checking group descriptors");
2126 for (i = 0; i < sbi->s_groups_count; i++) {
2127 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2129 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2130 last_block = ext4_blocks_count(sbi->s_es) - 1;
2132 last_block = first_block +
2133 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2135 if ((grp == sbi->s_groups_count) &&
2136 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2139 block_bitmap = ext4_block_bitmap(sb, gdp);
2140 if (block_bitmap < first_block || block_bitmap > last_block) {
2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2142 "Block bitmap for group %u not in group "
2143 "(block %llu)!", i, block_bitmap);
2146 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2147 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2148 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2149 "Inode bitmap for group %u not in group "
2150 "(block %llu)!", i, inode_bitmap);
2153 inode_table = ext4_inode_table(sb, gdp);
2154 if (inode_table < first_block ||
2155 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2156 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2157 "Inode table for group %u not in group "
2158 "(block %llu)!", i, inode_table);
2161 ext4_lock_group(sb, i);
2162 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2163 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2164 "Checksum for group %u failed (%u!=%u)",
2165 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2166 gdp)), le16_to_cpu(gdp->bg_checksum));
2167 if (!(sb->s_flags & MS_RDONLY)) {
2168 ext4_unlock_group(sb, i);
2172 ext4_unlock_group(sb, i);
2174 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2176 if (NULL != first_not_zeroed)
2177 *first_not_zeroed = grp;
2179 ext4_free_blocks_count_set(sbi->s_es,
2180 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2181 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2185 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2186 * the superblock) which were deleted from all directories, but held open by
2187 * a process at the time of a crash. We walk the list and try to delete these
2188 * inodes at recovery time (only with a read-write filesystem).
2190 * In order to keep the orphan inode chain consistent during traversal (in
2191 * case of crash during recovery), we link each inode into the superblock
2192 * orphan list_head and handle it the same way as an inode deletion during
2193 * normal operation (which journals the operations for us).
2195 * We only do an iget() and an iput() on each inode, which is very safe if we
2196 * accidentally point at an in-use or already deleted inode. The worst that
2197 * can happen in this case is that we get a "bit already cleared" message from
2198 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2199 * e2fsck was run on this filesystem, and it must have already done the orphan
2200 * inode cleanup for us, so we can safely abort without any further action.
2202 static void ext4_orphan_cleanup(struct super_block *sb,
2203 struct ext4_super_block *es)
2205 unsigned int s_flags = sb->s_flags;
2206 int nr_orphans = 0, nr_truncates = 0;
2210 if (!es->s_last_orphan) {
2211 jbd_debug(4, "no orphan inodes to clean up\n");
2215 if (bdev_read_only(sb->s_bdev)) {
2216 ext4_msg(sb, KERN_ERR, "write access "
2217 "unavailable, skipping orphan cleanup");
2221 /* Check if feature set would not allow a r/w mount */
2222 if (!ext4_feature_set_ok(sb, 0)) {
2223 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2224 "unknown ROCOMPAT features");
2228 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2229 if (es->s_last_orphan)
2230 jbd_debug(1, "Errors on filesystem, "
2231 "clearing orphan list.\n");
2232 es->s_last_orphan = 0;
2233 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2237 if (s_flags & MS_RDONLY) {
2238 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2239 sb->s_flags &= ~MS_RDONLY;
2242 /* Needed for iput() to work correctly and not trash data */
2243 sb->s_flags |= MS_ACTIVE;
2244 /* Turn on quotas so that they are updated correctly */
2245 for (i = 0; i < MAXQUOTAS; i++) {
2246 if (EXT4_SB(sb)->s_qf_names[i]) {
2247 int ret = ext4_quota_on_mount(sb, i);
2249 ext4_msg(sb, KERN_ERR,
2250 "Cannot turn on journaled "
2251 "quota: error %d", ret);
2256 while (es->s_last_orphan) {
2257 struct inode *inode;
2259 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2260 if (IS_ERR(inode)) {
2261 es->s_last_orphan = 0;
2265 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2266 dquot_initialize(inode);
2267 if (inode->i_nlink) {
2268 ext4_msg(sb, KERN_DEBUG,
2269 "%s: truncating inode %lu to %lld bytes",
2270 __func__, inode->i_ino, inode->i_size);
2271 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2272 inode->i_ino, inode->i_size);
2273 ext4_truncate(inode);
2276 ext4_msg(sb, KERN_DEBUG,
2277 "%s: deleting unreferenced inode %lu",
2278 __func__, inode->i_ino);
2279 jbd_debug(2, "deleting unreferenced inode %lu\n",
2283 iput(inode); /* The delete magic happens here! */
2286 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2289 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2290 PLURAL(nr_orphans));
2292 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2293 PLURAL(nr_truncates));
2295 /* Turn quotas off */
2296 for (i = 0; i < MAXQUOTAS; i++) {
2297 if (sb_dqopt(sb)->files[i])
2298 dquot_quota_off(sb, i);
2301 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2305 * Maximal extent format file size.
2306 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2307 * extent format containers, within a sector_t, and within i_blocks
2308 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2309 * so that won't be a limiting factor.
2311 * However there is other limiting factor. We do store extents in the form
2312 * of starting block and length, hence the resulting length of the extent
2313 * covering maximum file size must fit into on-disk format containers as
2314 * well. Given that length is always by 1 unit bigger than max unit (because
2315 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2317 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2319 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2322 loff_t upper_limit = MAX_LFS_FILESIZE;
2324 /* small i_blocks in vfs inode? */
2325 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2327 * CONFIG_LBDAF is not enabled implies the inode
2328 * i_block represent total blocks in 512 bytes
2329 * 32 == size of vfs inode i_blocks * 8
2331 upper_limit = (1LL << 32) - 1;
2333 /* total blocks in file system block size */
2334 upper_limit >>= (blkbits - 9);
2335 upper_limit <<= blkbits;
2339 * 32-bit extent-start container, ee_block. We lower the maxbytes
2340 * by one fs block, so ee_len can cover the extent of maximum file
2343 res = (1LL << 32) - 1;
2346 /* Sanity check against vm- & vfs- imposed limits */
2347 if (res > upper_limit)
2354 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2355 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2356 * We need to be 1 filesystem block less than the 2^48 sector limit.
2358 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2360 loff_t res = EXT4_NDIR_BLOCKS;
2363 /* This is calculated to be the largest file size for a dense, block
2364 * mapped file such that the file's total number of 512-byte sectors,
2365 * including data and all indirect blocks, does not exceed (2^48 - 1).
2367 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2368 * number of 512-byte sectors of the file.
2371 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2373 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2374 * the inode i_block field represents total file blocks in
2375 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2377 upper_limit = (1LL << 32) - 1;
2379 /* total blocks in file system block size */
2380 upper_limit >>= (bits - 9);
2384 * We use 48 bit ext4_inode i_blocks
2385 * With EXT4_HUGE_FILE_FL set the i_blocks
2386 * represent total number of blocks in
2387 * file system block size
2389 upper_limit = (1LL << 48) - 1;
2393 /* indirect blocks */
2395 /* double indirect blocks */
2396 meta_blocks += 1 + (1LL << (bits-2));
2397 /* tripple indirect blocks */
2398 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2400 upper_limit -= meta_blocks;
2401 upper_limit <<= bits;
2403 res += 1LL << (bits-2);
2404 res += 1LL << (2*(bits-2));
2405 res += 1LL << (3*(bits-2));
2407 if (res > upper_limit)
2410 if (res > MAX_LFS_FILESIZE)
2411 res = MAX_LFS_FILESIZE;
2416 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2417 ext4_fsblk_t logical_sb_block, int nr)
2419 struct ext4_sb_info *sbi = EXT4_SB(sb);
2420 ext4_group_t bg, first_meta_bg;
2423 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2425 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2427 return logical_sb_block + nr + 1;
2428 bg = sbi->s_desc_per_block * nr;
2429 if (ext4_bg_has_super(sb, bg))
2432 return (has_super + ext4_group_first_block_no(sb, bg));
2436 * ext4_get_stripe_size: Get the stripe size.
2437 * @sbi: In memory super block info
2439 * If we have specified it via mount option, then
2440 * use the mount option value. If the value specified at mount time is
2441 * greater than the blocks per group use the super block value.
2442 * If the super block value is greater than blocks per group return 0.
2443 * Allocator needs it be less than blocks per group.
2446 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2448 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2449 unsigned long stripe_width =
2450 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2453 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2454 ret = sbi->s_stripe;
2455 else if (stripe_width <= sbi->s_blocks_per_group)
2457 else if (stride <= sbi->s_blocks_per_group)
2463 * If the stripe width is 1, this makes no sense and
2464 * we set it to 0 to turn off stripe handling code.
2475 struct attribute attr;
2476 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2477 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2478 const char *, size_t);
2482 static int parse_strtoul(const char *buf,
2483 unsigned long max, unsigned long *value)
2487 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2488 endp = skip_spaces(endp);
2489 if (*endp || *value > max)
2495 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi,
2499 return snprintf(buf, PAGE_SIZE, "%llu\n",
2501 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2504 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2505 struct ext4_sb_info *sbi, char *buf)
2507 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2509 if (!sb->s_bdev->bd_part)
2510 return snprintf(buf, PAGE_SIZE, "0\n");
2511 return snprintf(buf, PAGE_SIZE, "%lu\n",
2512 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2513 sbi->s_sectors_written_start) >> 1);
2516 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi, char *buf)
2519 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2521 if (!sb->s_bdev->bd_part)
2522 return snprintf(buf, PAGE_SIZE, "0\n");
2523 return snprintf(buf, PAGE_SIZE, "%llu\n",
2524 (unsigned long long)(sbi->s_kbytes_written +
2525 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2526 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2529 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2530 struct ext4_sb_info *sbi, char *buf)
2532 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2535 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi, char *buf)
2538 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2541 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi,
2543 const char *buf, size_t count)
2547 if (parse_strtoul(buf, 0x40000000, &t))
2550 if (t && !is_power_of_2(t))
2553 sbi->s_inode_readahead_blks = t;
2557 static ssize_t sbi_ui_show(struct ext4_attr *a,
2558 struct ext4_sb_info *sbi, char *buf)
2560 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2562 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2565 static ssize_t sbi_ui_store(struct ext4_attr *a,
2566 struct ext4_sb_info *sbi,
2567 const char *buf, size_t count)
2569 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2572 if (parse_strtoul(buf, 0xffffffff, &t))
2578 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2579 static struct ext4_attr ext4_attr_##_name = { \
2580 .attr = {.name = __stringify(_name), .mode = _mode }, \
2583 .offset = offsetof(struct ext4_sb_info, _elname), \
2585 #define EXT4_ATTR(name, mode, show, store) \
2586 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2588 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2589 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2590 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2591 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2592 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2593 #define ATTR_LIST(name) &ext4_attr_##name.attr
2595 EXT4_RO_ATTR(delayed_allocation_blocks);
2596 EXT4_RO_ATTR(session_write_kbytes);
2597 EXT4_RO_ATTR(lifetime_write_kbytes);
2598 EXT4_RO_ATTR(extent_cache_hits);
2599 EXT4_RO_ATTR(extent_cache_misses);
2600 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2601 inode_readahead_blks_store, s_inode_readahead_blks);
2602 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2603 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2604 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2605 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2606 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2607 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2608 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2609 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2611 static struct attribute *ext4_attrs[] = {
2612 ATTR_LIST(delayed_allocation_blocks),
2613 ATTR_LIST(session_write_kbytes),
2614 ATTR_LIST(lifetime_write_kbytes),
2615 ATTR_LIST(extent_cache_hits),
2616 ATTR_LIST(extent_cache_misses),
2617 ATTR_LIST(inode_readahead_blks),
2618 ATTR_LIST(inode_goal),
2619 ATTR_LIST(mb_stats),
2620 ATTR_LIST(mb_max_to_scan),
2621 ATTR_LIST(mb_min_to_scan),
2622 ATTR_LIST(mb_order2_req),
2623 ATTR_LIST(mb_stream_req),
2624 ATTR_LIST(mb_group_prealloc),
2625 ATTR_LIST(max_writeback_mb_bump),
2629 /* Features this copy of ext4 supports */
2630 EXT4_INFO_ATTR(lazy_itable_init);
2631 EXT4_INFO_ATTR(batched_discard);
2633 static struct attribute *ext4_feat_attrs[] = {
2634 ATTR_LIST(lazy_itable_init),
2635 ATTR_LIST(batched_discard),
2639 static ssize_t ext4_attr_show(struct kobject *kobj,
2640 struct attribute *attr, char *buf)
2642 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2644 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2646 return a->show ? a->show(a, sbi, buf) : 0;
2649 static ssize_t ext4_attr_store(struct kobject *kobj,
2650 struct attribute *attr,
2651 const char *buf, size_t len)
2653 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2655 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2657 return a->store ? a->store(a, sbi, buf, len) : 0;
2660 static void ext4_sb_release(struct kobject *kobj)
2662 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2664 complete(&sbi->s_kobj_unregister);
2667 static const struct sysfs_ops ext4_attr_ops = {
2668 .show = ext4_attr_show,
2669 .store = ext4_attr_store,
2672 static struct kobj_type ext4_ktype = {
2673 .default_attrs = ext4_attrs,
2674 .sysfs_ops = &ext4_attr_ops,
2675 .release = ext4_sb_release,
2678 static void ext4_feat_release(struct kobject *kobj)
2680 complete(&ext4_feat->f_kobj_unregister);
2683 static struct kobj_type ext4_feat_ktype = {
2684 .default_attrs = ext4_feat_attrs,
2685 .sysfs_ops = &ext4_attr_ops,
2686 .release = ext4_feat_release,
2690 * Check whether this filesystem can be mounted based on
2691 * the features present and the RDONLY/RDWR mount requested.
2692 * Returns 1 if this filesystem can be mounted as requested,
2693 * 0 if it cannot be.
2695 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2697 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2698 ext4_msg(sb, KERN_ERR,
2699 "Couldn't mount because of "
2700 "unsupported optional features (%x)",
2701 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2702 ~EXT4_FEATURE_INCOMPAT_SUPP));
2709 /* Check that feature set is OK for a read-write mount */
2710 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2711 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2712 "unsupported optional features (%x)",
2713 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2714 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2718 * Large file size enabled file system can only be mounted
2719 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2721 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2722 if (sizeof(blkcnt_t) < sizeof(u64)) {
2723 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2724 "cannot be mounted RDWR without "
2729 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2730 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2731 ext4_msg(sb, KERN_ERR,
2732 "Can't support bigalloc feature without "
2733 "extents feature\n");
2740 * This function is called once a day if we have errors logged
2741 * on the file system
2743 static void print_daily_error_info(unsigned long arg)
2745 struct super_block *sb = (struct super_block *) arg;
2746 struct ext4_sb_info *sbi;
2747 struct ext4_super_block *es;
2752 if (es->s_error_count)
2753 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2754 le32_to_cpu(es->s_error_count));
2755 if (es->s_first_error_time) {
2756 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2757 sb->s_id, le32_to_cpu(es->s_first_error_time),
2758 (int) sizeof(es->s_first_error_func),
2759 es->s_first_error_func,
2760 le32_to_cpu(es->s_first_error_line));
2761 if (es->s_first_error_ino)
2762 printk(": inode %u",
2763 le32_to_cpu(es->s_first_error_ino));
2764 if (es->s_first_error_block)
2765 printk(": block %llu", (unsigned long long)
2766 le64_to_cpu(es->s_first_error_block));
2769 if (es->s_last_error_time) {
2770 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2771 sb->s_id, le32_to_cpu(es->s_last_error_time),
2772 (int) sizeof(es->s_last_error_func),
2773 es->s_last_error_func,
2774 le32_to_cpu(es->s_last_error_line));
2775 if (es->s_last_error_ino)
2776 printk(": inode %u",
2777 le32_to_cpu(es->s_last_error_ino));
2778 if (es->s_last_error_block)
2779 printk(": block %llu", (unsigned long long)
2780 le64_to_cpu(es->s_last_error_block));
2783 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2786 /* Find next suitable group and run ext4_init_inode_table */
2787 static int ext4_run_li_request(struct ext4_li_request *elr)
2789 struct ext4_group_desc *gdp = NULL;
2790 ext4_group_t group, ngroups;
2791 struct super_block *sb;
2792 unsigned long timeout = 0;
2796 ngroups = EXT4_SB(sb)->s_groups_count;
2798 for (group = elr->lr_next_group; group < ngroups; group++) {
2799 gdp = ext4_get_group_desc(sb, group, NULL);
2805 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2809 if (group == ngroups)
2814 ret = ext4_init_inode_table(sb, group,
2815 elr->lr_timeout ? 0 : 1);
2816 if (elr->lr_timeout == 0) {
2817 timeout = (jiffies - timeout) *
2818 elr->lr_sbi->s_li_wait_mult;
2819 elr->lr_timeout = timeout;
2821 elr->lr_next_sched = jiffies + elr->lr_timeout;
2822 elr->lr_next_group = group + 1;
2829 * Remove lr_request from the list_request and free the
2830 * request structure. Should be called with li_list_mtx held
2832 static void ext4_remove_li_request(struct ext4_li_request *elr)
2834 struct ext4_sb_info *sbi;
2841 list_del(&elr->lr_request);
2842 sbi->s_li_request = NULL;
2846 static void ext4_unregister_li_request(struct super_block *sb)
2848 mutex_lock(&ext4_li_mtx);
2849 if (!ext4_li_info) {
2850 mutex_unlock(&ext4_li_mtx);
2854 mutex_lock(&ext4_li_info->li_list_mtx);
2855 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2856 mutex_unlock(&ext4_li_info->li_list_mtx);
2857 mutex_unlock(&ext4_li_mtx);
2860 static struct task_struct *ext4_lazyinit_task;
2863 * This is the function where ext4lazyinit thread lives. It walks
2864 * through the request list searching for next scheduled filesystem.
2865 * When such a fs is found, run the lazy initialization request
2866 * (ext4_rn_li_request) and keep track of the time spend in this
2867 * function. Based on that time we compute next schedule time of
2868 * the request. When walking through the list is complete, compute
2869 * next waking time and put itself into sleep.
2871 static int ext4_lazyinit_thread(void *arg)
2873 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2874 struct list_head *pos, *n;
2875 struct ext4_li_request *elr;
2876 unsigned long next_wakeup, cur;
2878 BUG_ON(NULL == eli);
2882 next_wakeup = MAX_JIFFY_OFFSET;
2884 mutex_lock(&eli->li_list_mtx);
2885 if (list_empty(&eli->li_request_list)) {
2886 mutex_unlock(&eli->li_list_mtx);
2890 list_for_each_safe(pos, n, &eli->li_request_list) {
2891 elr = list_entry(pos, struct ext4_li_request,
2894 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2895 if (ext4_run_li_request(elr) != 0) {
2896 /* error, remove the lazy_init job */
2897 ext4_remove_li_request(elr);
2902 if (time_before(elr->lr_next_sched, next_wakeup))
2903 next_wakeup = elr->lr_next_sched;
2905 mutex_unlock(&eli->li_list_mtx);
2910 if ((time_after_eq(cur, next_wakeup)) ||
2911 (MAX_JIFFY_OFFSET == next_wakeup)) {
2916 schedule_timeout_interruptible(next_wakeup - cur);
2918 if (kthread_should_stop()) {
2919 ext4_clear_request_list();
2926 * It looks like the request list is empty, but we need
2927 * to check it under the li_list_mtx lock, to prevent any
2928 * additions into it, and of course we should lock ext4_li_mtx
2929 * to atomically free the list and ext4_li_info, because at
2930 * this point another ext4 filesystem could be registering
2933 mutex_lock(&ext4_li_mtx);
2934 mutex_lock(&eli->li_list_mtx);
2935 if (!list_empty(&eli->li_request_list)) {
2936 mutex_unlock(&eli->li_list_mtx);
2937 mutex_unlock(&ext4_li_mtx);
2940 mutex_unlock(&eli->li_list_mtx);
2941 kfree(ext4_li_info);
2942 ext4_li_info = NULL;
2943 mutex_unlock(&ext4_li_mtx);
2948 static void ext4_clear_request_list(void)
2950 struct list_head *pos, *n;
2951 struct ext4_li_request *elr;
2953 mutex_lock(&ext4_li_info->li_list_mtx);
2954 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2955 elr = list_entry(pos, struct ext4_li_request,
2957 ext4_remove_li_request(elr);
2959 mutex_unlock(&ext4_li_info->li_list_mtx);
2962 static int ext4_run_lazyinit_thread(void)
2964 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2965 ext4_li_info, "ext4lazyinit");
2966 if (IS_ERR(ext4_lazyinit_task)) {
2967 int err = PTR_ERR(ext4_lazyinit_task);
2968 ext4_clear_request_list();
2969 kfree(ext4_li_info);
2970 ext4_li_info = NULL;
2971 printk(KERN_CRIT "EXT4: error %d creating inode table "
2972 "initialization thread\n",
2976 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2981 * Check whether it make sense to run itable init. thread or not.
2982 * If there is at least one uninitialized inode table, return
2983 * corresponding group number, else the loop goes through all
2984 * groups and return total number of groups.
2986 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2988 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2989 struct ext4_group_desc *gdp = NULL;
2991 for (group = 0; group < ngroups; group++) {
2992 gdp = ext4_get_group_desc(sb, group, NULL);
2996 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3003 static int ext4_li_info_new(void)
3005 struct ext4_lazy_init *eli = NULL;
3007 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3011 INIT_LIST_HEAD(&eli->li_request_list);
3012 mutex_init(&eli->li_list_mtx);
3014 eli->li_state |= EXT4_LAZYINIT_QUIT;
3021 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3024 struct ext4_sb_info *sbi = EXT4_SB(sb);
3025 struct ext4_li_request *elr;
3028 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3034 elr->lr_next_group = start;
3037 * Randomize first schedule time of the request to
3038 * spread the inode table initialization requests
3041 get_random_bytes(&rnd, sizeof(rnd));
3042 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3043 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3048 static int ext4_register_li_request(struct super_block *sb,
3049 ext4_group_t first_not_zeroed)
3051 struct ext4_sb_info *sbi = EXT4_SB(sb);
3052 struct ext4_li_request *elr;
3053 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3056 if (sbi->s_li_request != NULL) {
3058 * Reset timeout so it can be computed again, because
3059 * s_li_wait_mult might have changed.
3061 sbi->s_li_request->lr_timeout = 0;
3065 if (first_not_zeroed == ngroups ||
3066 (sb->s_flags & MS_RDONLY) ||
3067 !test_opt(sb, INIT_INODE_TABLE))
3070 elr = ext4_li_request_new(sb, first_not_zeroed);
3074 mutex_lock(&ext4_li_mtx);
3076 if (NULL == ext4_li_info) {
3077 ret = ext4_li_info_new();
3082 mutex_lock(&ext4_li_info->li_list_mtx);
3083 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3084 mutex_unlock(&ext4_li_info->li_list_mtx);
3086 sbi->s_li_request = elr;
3088 * set elr to NULL here since it has been inserted to
3089 * the request_list and the removal and free of it is
3090 * handled by ext4_clear_request_list from now on.
3094 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3095 ret = ext4_run_lazyinit_thread();
3100 mutex_unlock(&ext4_li_mtx);
3107 * We do not need to lock anything since this is called on
3110 static void ext4_destroy_lazyinit_thread(void)
3113 * If thread exited earlier
3114 * there's nothing to be done.
3116 if (!ext4_li_info || !ext4_lazyinit_task)
3119 kthread_stop(ext4_lazyinit_task);
3122 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3124 char *orig_data = kstrdup(data, GFP_KERNEL);
3125 struct buffer_head *bh;
3126 struct ext4_super_block *es = NULL;
3127 struct ext4_sb_info *sbi;
3129 ext4_fsblk_t sb_block = get_sb_block(&data);
3130 ext4_fsblk_t logical_sb_block;
3131 unsigned long offset = 0;
3132 unsigned long journal_devnum = 0;
3133 unsigned long def_mount_opts;
3138 int blocksize, clustersize;
3139 unsigned int db_count;
3141 int needs_recovery, has_huge_files, has_bigalloc;
3144 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3145 ext4_group_t first_not_zeroed;
3147 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3151 sbi->s_blockgroup_lock =
3152 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3153 if (!sbi->s_blockgroup_lock) {
3157 sb->s_fs_info = sbi;
3158 sbi->s_mount_opt = 0;
3159 sbi->s_resuid = EXT4_DEF_RESUID;
3160 sbi->s_resgid = EXT4_DEF_RESGID;
3161 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3162 sbi->s_sb_block = sb_block;
3163 if (sb->s_bdev->bd_part)
3164 sbi->s_sectors_written_start =
3165 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3167 /* Cleanup superblock name */
3168 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3172 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3174 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3179 * The ext4 superblock will not be buffer aligned for other than 1kB
3180 * block sizes. We need to calculate the offset from buffer start.
3182 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3183 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3184 offset = do_div(logical_sb_block, blocksize);
3186 logical_sb_block = sb_block;
3189 if (!(bh = sb_bread(sb, logical_sb_block))) {
3190 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3194 * Note: s_es must be initialized as soon as possible because
3195 * some ext4 macro-instructions depend on its value
3197 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3199 sb->s_magic = le16_to_cpu(es->s_magic);
3200 if (sb->s_magic != EXT4_SUPER_MAGIC)
3202 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3204 /* Set defaults before we parse the mount options */
3205 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3206 set_opt(sb, INIT_INODE_TABLE);
3207 if (def_mount_opts & EXT4_DEFM_DEBUG)
3209 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3210 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3214 if (def_mount_opts & EXT4_DEFM_UID16)
3215 set_opt(sb, NO_UID32);
3216 /* xattr user namespace & acls are now defaulted on */
3217 #ifdef CONFIG_EXT4_FS_XATTR
3218 set_opt(sb, XATTR_USER);
3220 #if defined(CONFIG_EXT4_FS_POSIX_ACL)
3221 sb->s_flags |= MS_POSIXACL;
3223 set_opt(sb, MBLK_IO_SUBMIT);
3224 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3225 set_opt(sb, JOURNAL_DATA);
3226 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3227 set_opt(sb, ORDERED_DATA);
3228 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3229 set_opt(sb, WRITEBACK_DATA);
3231 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3232 set_opt(sb, ERRORS_PANIC);
3233 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3234 set_opt(sb, ERRORS_CONT);
3236 set_opt(sb, ERRORS_RO);
3237 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3238 set_opt(sb, BLOCK_VALIDITY);
3239 if (def_mount_opts & EXT4_DEFM_DISCARD)
3240 set_opt(sb, DISCARD);
3242 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3243 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3244 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3245 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3246 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3248 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3249 set_opt(sb, BARRIER);
3252 * enable delayed allocation by default
3253 * Use -o nodelalloc to turn it off
3255 if (!IS_EXT3_SB(sb) &&
3256 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3257 set_opt(sb, DELALLOC);
3260 * set default s_li_wait_mult for lazyinit, for the case there is
3261 * no mount option specified.
3263 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3265 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3266 &journal_devnum, &journal_ioprio, NULL, 0)) {
3267 ext4_msg(sb, KERN_WARNING,
3268 "failed to parse options in superblock: %s",
3269 sbi->s_es->s_mount_opts);
3271 if (!parse_options((char *) data, sb, &journal_devnum,
3272 &journal_ioprio, NULL, 0))
3275 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3276 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3277 "with data=journal disables delayed "
3278 "allocation and O_DIRECT support!\n");
3279 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3280 ext4_msg(sb, KERN_ERR, "can't mount with "
3281 "both data=journal and delalloc");
3284 if (test_opt(sb, DIOREAD_NOLOCK)) {
3285 ext4_msg(sb, KERN_ERR, "can't mount with "
3286 "both data=journal and delalloc");
3289 if (test_opt(sb, DELALLOC))
3290 clear_opt(sb, DELALLOC);
3293 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3294 if (test_opt(sb, DIOREAD_NOLOCK)) {
3295 if (blocksize < PAGE_SIZE) {
3296 ext4_msg(sb, KERN_ERR, "can't mount with "
3297 "dioread_nolock if block size != PAGE_SIZE");
3302 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3303 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3304 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3305 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3306 ext4_msg(sb, KERN_WARNING,
3307 "feature flags set on rev 0 fs, "
3308 "running e2fsck is recommended");
3310 if (IS_EXT2_SB(sb)) {
3311 if (ext2_feature_set_ok(sb))
3312 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3313 "using the ext4 subsystem");
3315 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3316 "to feature incompatibilities");
3321 if (IS_EXT3_SB(sb)) {
3322 if (ext3_feature_set_ok(sb))
3323 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3324 "using the ext4 subsystem");
3326 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3327 "to feature incompatibilities");
3333 * Check feature flags regardless of the revision level, since we
3334 * previously didn't change the revision level when setting the flags,
3335 * so there is a chance incompat flags are set on a rev 0 filesystem.
3337 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3340 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3341 blocksize > EXT4_MAX_BLOCK_SIZE) {
3342 ext4_msg(sb, KERN_ERR,
3343 "Unsupported filesystem blocksize %d", blocksize);
3347 if (sb->s_blocksize != blocksize) {
3348 /* Validate the filesystem blocksize */
3349 if (!sb_set_blocksize(sb, blocksize)) {
3350 ext4_msg(sb, KERN_ERR, "bad block size %d",
3356 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3357 offset = do_div(logical_sb_block, blocksize);
3358 bh = sb_bread(sb, logical_sb_block);
3360 ext4_msg(sb, KERN_ERR,
3361 "Can't read superblock on 2nd try");
3364 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3366 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3367 ext4_msg(sb, KERN_ERR,
3368 "Magic mismatch, very weird!");
3373 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3374 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3375 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3377 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3379 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3380 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3381 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3383 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3384 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3385 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3386 (!is_power_of_2(sbi->s_inode_size)) ||
3387 (sbi->s_inode_size > blocksize)) {
3388 ext4_msg(sb, KERN_ERR,
3389 "unsupported inode size: %d",
3393 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3394 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3397 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3398 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3399 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3400 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3401 !is_power_of_2(sbi->s_desc_size)) {
3402 ext4_msg(sb, KERN_ERR,
3403 "unsupported descriptor size %lu",
3408 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3410 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3411 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3412 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3415 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3416 if (sbi->s_inodes_per_block == 0)
3418 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3419 sbi->s_inodes_per_block;
3420 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3422 sbi->s_mount_state = le16_to_cpu(es->s_state);
3423 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3424 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3426 for (i = 0; i < 4; i++)
3427 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3428 sbi->s_def_hash_version = es->s_def_hash_version;
3429 i = le32_to_cpu(es->s_flags);
3430 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3431 sbi->s_hash_unsigned = 3;
3432 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3433 #ifdef __CHAR_UNSIGNED__
3434 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3435 sbi->s_hash_unsigned = 3;
3437 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3442 /* Handle clustersize */
3443 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3444 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3445 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3447 if (clustersize < blocksize) {
3448 ext4_msg(sb, KERN_ERR,
3449 "cluster size (%d) smaller than "
3450 "block size (%d)", clustersize, blocksize);
3453 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3454 le32_to_cpu(es->s_log_block_size);
3455 sbi->s_clusters_per_group =
3456 le32_to_cpu(es->s_clusters_per_group);
3457 if (sbi->s_clusters_per_group > blocksize * 8) {
3458 ext4_msg(sb, KERN_ERR,
3459 "#clusters per group too big: %lu",
3460 sbi->s_clusters_per_group);
3463 if (sbi->s_blocks_per_group !=
3464 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3465 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3466 "clusters per group (%lu) inconsistent",
3467 sbi->s_blocks_per_group,
3468 sbi->s_clusters_per_group);
3472 if (clustersize != blocksize) {
3473 ext4_warning(sb, "fragment/cluster size (%d) != "
3474 "block size (%d)", clustersize,
3476 clustersize = blocksize;
3478 if (sbi->s_blocks_per_group > blocksize * 8) {
3479 ext4_msg(sb, KERN_ERR,
3480 "#blocks per group too big: %lu",
3481 sbi->s_blocks_per_group);
3484 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3485 sbi->s_cluster_bits = 0;
3487 sbi->s_cluster_ratio = clustersize / blocksize;
3489 if (sbi->s_inodes_per_group > blocksize * 8) {
3490 ext4_msg(sb, KERN_ERR,
3491 "#inodes per group too big: %lu",
3492 sbi->s_inodes_per_group);
3497 * Test whether we have more sectors than will fit in sector_t,
3498 * and whether the max offset is addressable by the page cache.
3500 err = generic_check_addressable(sb->s_blocksize_bits,
3501 ext4_blocks_count(es));
3503 ext4_msg(sb, KERN_ERR, "filesystem"
3504 " too large to mount safely on this system");
3505 if (sizeof(sector_t) < 8)
3506 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3511 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3514 /* check blocks count against device size */
3515 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3516 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3517 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3518 "exceeds size of device (%llu blocks)",
3519 ext4_blocks_count(es), blocks_count);
3524 * It makes no sense for the first data block to be beyond the end
3525 * of the filesystem.
3527 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3528 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3529 "block %u is beyond end of filesystem (%llu)",
3530 le32_to_cpu(es->s_first_data_block),
3531 ext4_blocks_count(es));
3534 blocks_count = (ext4_blocks_count(es) -
3535 le32_to_cpu(es->s_first_data_block) +
3536 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3537 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3538 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3539 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3540 "(block count %llu, first data block %u, "
3541 "blocks per group %lu)", sbi->s_groups_count,
3542 ext4_blocks_count(es),
3543 le32_to_cpu(es->s_first_data_block),
3544 EXT4_BLOCKS_PER_GROUP(sb));
3547 sbi->s_groups_count = blocks_count;
3548 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3549 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3550 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3551 EXT4_DESC_PER_BLOCK(sb);
3552 sbi->s_group_desc = ext4_kvmalloc(db_count *
3553 sizeof(struct buffer_head *),
3555 if (sbi->s_group_desc == NULL) {
3556 ext4_msg(sb, KERN_ERR, "not enough memory");
3561 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3563 bgl_lock_init(sbi->s_blockgroup_lock);
3565 for (i = 0; i < db_count; i++) {
3566 block = descriptor_loc(sb, logical_sb_block, i);
3567 sbi->s_group_desc[i] = sb_bread(sb, block);
3568 if (!sbi->s_group_desc[i]) {
3569 ext4_msg(sb, KERN_ERR,
3570 "can't read group descriptor %d", i);
3575 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3576 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3579 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3580 if (!ext4_fill_flex_info(sb)) {
3581 ext4_msg(sb, KERN_ERR,
3582 "unable to initialize "
3583 "flex_bg meta info!");
3587 sbi->s_gdb_count = db_count;
3588 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3589 spin_lock_init(&sbi->s_next_gen_lock);
3591 init_timer(&sbi->s_err_report);
3592 sbi->s_err_report.function = print_daily_error_info;
3593 sbi->s_err_report.data = (unsigned long) sb;
3595 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3596 ext4_count_free_clusters(sb));
3598 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3599 ext4_count_free_inodes(sb));
3602 err = percpu_counter_init(&sbi->s_dirs_counter,
3603 ext4_count_dirs(sb));
3606 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3609 ext4_msg(sb, KERN_ERR, "insufficient memory");
3613 sbi->s_stripe = ext4_get_stripe_size(sbi);
3614 sbi->s_max_writeback_mb_bump = 128;
3617 * set up enough so that it can read an inode
3619 if (!test_opt(sb, NOLOAD) &&
3620 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3621 sb->s_op = &ext4_sops;
3623 sb->s_op = &ext4_nojournal_sops;
3624 sb->s_export_op = &ext4_export_ops;
3625 sb->s_xattr = ext4_xattr_handlers;
3627 sb->s_qcop = &ext4_qctl_operations;
3628 sb->dq_op = &ext4_quota_operations;
3630 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3632 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3633 mutex_init(&sbi->s_orphan_lock);
3634 sbi->s_resize_flags = 0;
3638 needs_recovery = (es->s_last_orphan != 0 ||
3639 EXT4_HAS_INCOMPAT_FEATURE(sb,
3640 EXT4_FEATURE_INCOMPAT_RECOVER));
3642 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3643 !(sb->s_flags & MS_RDONLY))
3644 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3648 * The first inode we look at is the journal inode. Don't try
3649 * root first: it may be modified in the journal!
3651 if (!test_opt(sb, NOLOAD) &&
3652 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3653 if (ext4_load_journal(sb, es, journal_devnum))
3655 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3656 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3657 ext4_msg(sb, KERN_ERR, "required journal recovery "
3658 "suppressed and not mounted read-only");
3659 goto failed_mount_wq;
3661 clear_opt(sb, DATA_FLAGS);
3662 sbi->s_journal = NULL;
3667 if (ext4_blocks_count(es) > 0xffffffffULL &&
3668 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3669 JBD2_FEATURE_INCOMPAT_64BIT)) {
3670 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3671 goto failed_mount_wq;
3674 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3675 jbd2_journal_set_features(sbi->s_journal,
3676 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3677 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3678 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3679 jbd2_journal_set_features(sbi->s_journal,
3680 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3681 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3682 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3684 jbd2_journal_clear_features(sbi->s_journal,
3685 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3686 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3689 /* We have now updated the journal if required, so we can
3690 * validate the data journaling mode. */
3691 switch (test_opt(sb, DATA_FLAGS)) {
3693 /* No mode set, assume a default based on the journal
3694 * capabilities: ORDERED_DATA if the journal can
3695 * cope, else JOURNAL_DATA
3697 if (jbd2_journal_check_available_features
3698 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3699 set_opt(sb, ORDERED_DATA);
3701 set_opt(sb, JOURNAL_DATA);
3704 case EXT4_MOUNT_ORDERED_DATA:
3705 case EXT4_MOUNT_WRITEBACK_DATA:
3706 if (!jbd2_journal_check_available_features
3707 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3708 ext4_msg(sb, KERN_ERR, "Journal does not support "
3709 "requested data journaling mode");
3710 goto failed_mount_wq;
3715 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3718 * The journal may have updated the bg summary counts, so we
3719 * need to update the global counters.
3721 percpu_counter_set(&sbi->s_freeclusters_counter,
3722 ext4_count_free_clusters(sb));
3723 percpu_counter_set(&sbi->s_freeinodes_counter,
3724 ext4_count_free_inodes(sb));
3725 percpu_counter_set(&sbi->s_dirs_counter,
3726 ext4_count_dirs(sb));
3727 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3731 * The maximum number of concurrent works can be high and
3732 * concurrency isn't really necessary. Limit it to 1.
3734 EXT4_SB(sb)->dio_unwritten_wq =
3735 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3736 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3737 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3738 goto failed_mount_wq;
3742 * The jbd2_journal_load will have done any necessary log recovery,
3743 * so we can safely mount the rest of the filesystem now.
3746 root = ext4_iget(sb, EXT4_ROOT_INO);
3748 ext4_msg(sb, KERN_ERR, "get root inode failed");
3749 ret = PTR_ERR(root);
3753 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3754 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3758 sb->s_root = d_alloc_root(root);
3761 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3766 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3768 /* determine the minimum size of new large inodes, if present */
3769 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3770 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3771 EXT4_GOOD_OLD_INODE_SIZE;
3772 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3773 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3774 if (sbi->s_want_extra_isize <
3775 le16_to_cpu(es->s_want_extra_isize))
3776 sbi->s_want_extra_isize =
3777 le16_to_cpu(es->s_want_extra_isize);
3778 if (sbi->s_want_extra_isize <
3779 le16_to_cpu(es->s_min_extra_isize))
3780 sbi->s_want_extra_isize =
3781 le16_to_cpu(es->s_min_extra_isize);
3784 /* Check if enough inode space is available */
3785 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3786 sbi->s_inode_size) {
3787 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3788 EXT4_GOOD_OLD_INODE_SIZE;
3789 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3793 err = ext4_setup_system_zone(sb);
3795 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3797 goto failed_mount4a;
3801 err = ext4_mb_init(sb, needs_recovery);
3803 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3808 err = ext4_register_li_request(sb, first_not_zeroed);
3812 sbi->s_kobj.kset = ext4_kset;
3813 init_completion(&sbi->s_kobj_unregister);
3814 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3819 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3820 ext4_orphan_cleanup(sb, es);
3821 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3822 if (needs_recovery) {
3823 ext4_msg(sb, KERN_INFO, "recovery complete");
3824 ext4_mark_recovery_complete(sb, es);
3826 if (EXT4_SB(sb)->s_journal) {
3827 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3828 descr = " journalled data mode";
3829 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3830 descr = " ordered data mode";
3832 descr = " writeback data mode";
3834 descr = "out journal";
3836 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3837 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3838 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3840 if (es->s_error_count)
3841 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3848 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3852 ext4_unregister_li_request(sb);
3854 ext4_mb_release(sb);
3856 ext4_ext_release(sb);
3857 ext4_release_system_zone(sb);
3862 ext4_msg(sb, KERN_ERR, "mount failed");
3863 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3865 if (sbi->s_journal) {
3866 jbd2_journal_destroy(sbi->s_journal);
3867 sbi->s_journal = NULL;
3870 del_timer(&sbi->s_err_report);
3871 if (sbi->s_flex_groups)
3872 ext4_kvfree(sbi->s_flex_groups);
3873 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3874 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3875 percpu_counter_destroy(&sbi->s_dirs_counter);
3876 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3878 kthread_stop(sbi->s_mmp_tsk);
3880 for (i = 0; i < db_count; i++)
3881 brelse(sbi->s_group_desc[i]);
3882 ext4_kvfree(sbi->s_group_desc);
3885 remove_proc_entry(sb->s_id, ext4_proc_root);
3888 for (i = 0; i < MAXQUOTAS; i++)
3889 kfree(sbi->s_qf_names[i]);
3891 ext4_blkdev_remove(sbi);
3894 sb->s_fs_info = NULL;
3895 kfree(sbi->s_blockgroup_lock);
3903 * Setup any per-fs journal parameters now. We'll do this both on
3904 * initial mount, once the journal has been initialised but before we've
3905 * done any recovery; and again on any subsequent remount.
3907 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3909 struct ext4_sb_info *sbi = EXT4_SB(sb);
3911 journal->j_commit_interval = sbi->s_commit_interval;
3912 journal->j_min_batch_time = sbi->s_min_batch_time;
3913 journal->j_max_batch_time = sbi->s_max_batch_time;
3915 write_lock(&journal->j_state_lock);
3916 if (test_opt(sb, BARRIER))
3917 journal->j_flags |= JBD2_BARRIER;
3919 journal->j_flags &= ~JBD2_BARRIER;
3920 if (test_opt(sb, DATA_ERR_ABORT))
3921 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3923 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3924 write_unlock(&journal->j_state_lock);
3927 static journal_t *ext4_get_journal(struct super_block *sb,
3928 unsigned int journal_inum)
3930 struct inode *journal_inode;
3933 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3935 /* First, test for the existence of a valid inode on disk. Bad
3936 * things happen if we iget() an unused inode, as the subsequent
3937 * iput() will try to delete it. */
3939 journal_inode = ext4_iget(sb, journal_inum);
3940 if (IS_ERR(journal_inode)) {
3941 ext4_msg(sb, KERN_ERR, "no journal found");
3944 if (!journal_inode->i_nlink) {
3945 make_bad_inode(journal_inode);
3946 iput(journal_inode);
3947 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3951 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3952 journal_inode, journal_inode->i_size);
3953 if (!S_ISREG(journal_inode->i_mode)) {
3954 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3955 iput(journal_inode);
3959 journal = jbd2_journal_init_inode(journal_inode);
3961 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3962 iput(journal_inode);
3965 journal->j_private = sb;
3966 ext4_init_journal_params(sb, journal);
3970 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3973 struct buffer_head *bh;
3977 int hblock, blocksize;
3978 ext4_fsblk_t sb_block;
3979 unsigned long offset;
3980 struct ext4_super_block *es;
3981 struct block_device *bdev;
3983 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3985 bdev = ext4_blkdev_get(j_dev, sb);
3989 blocksize = sb->s_blocksize;
3990 hblock = bdev_logical_block_size(bdev);
3991 if (blocksize < hblock) {
3992 ext4_msg(sb, KERN_ERR,
3993 "blocksize too small for journal device");
3997 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3998 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3999 set_blocksize(bdev, blocksize);
4000 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4001 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4002 "external journal");
4006 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4007 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4008 !(le32_to_cpu(es->s_feature_incompat) &
4009 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4010 ext4_msg(sb, KERN_ERR, "external journal has "
4016 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4017 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4022 len = ext4_blocks_count(es);
4023 start = sb_block + 1;
4024 brelse(bh); /* we're done with the superblock */
4026 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4027 start, len, blocksize);
4029 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4032 journal->j_private = sb;
4033 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4034 wait_on_buffer(journal->j_sb_buffer);
4035 if (!buffer_uptodate(journal->j_sb_buffer)) {
4036 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4039 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4040 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4041 "user (unsupported) - %d",
4042 be32_to_cpu(journal->j_superblock->s_nr_users));
4045 EXT4_SB(sb)->journal_bdev = bdev;
4046 ext4_init_journal_params(sb, journal);
4050 jbd2_journal_destroy(journal);
4052 ext4_blkdev_put(bdev);
4056 static int ext4_load_journal(struct super_block *sb,
4057 struct ext4_super_block *es,
4058 unsigned long journal_devnum)
4061 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4064 int really_read_only;
4066 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4068 if (journal_devnum &&
4069 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4070 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4071 "numbers have changed");
4072 journal_dev = new_decode_dev(journal_devnum);
4074 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4076 really_read_only = bdev_read_only(sb->s_bdev);
4079 * Are we loading a blank journal or performing recovery after a
4080 * crash? For recovery, we need to check in advance whether we
4081 * can get read-write access to the device.
4083 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4084 if (sb->s_flags & MS_RDONLY) {
4085 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4086 "required on readonly filesystem");
4087 if (really_read_only) {
4088 ext4_msg(sb, KERN_ERR, "write access "
4089 "unavailable, cannot proceed");
4092 ext4_msg(sb, KERN_INFO, "write access will "
4093 "be enabled during recovery");
4097 if (journal_inum && journal_dev) {
4098 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4099 "and inode journals!");
4104 if (!(journal = ext4_get_journal(sb, journal_inum)))
4107 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4111 if (!(journal->j_flags & JBD2_BARRIER))
4112 ext4_msg(sb, KERN_INFO, "barriers disabled");
4114 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4115 err = jbd2_journal_update_format(journal);
4117 ext4_msg(sb, KERN_ERR, "error updating journal");
4118 jbd2_journal_destroy(journal);
4123 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4124 err = jbd2_journal_wipe(journal, !really_read_only);
4126 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4128 memcpy(save, ((char *) es) +
4129 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4130 err = jbd2_journal_load(journal);
4132 memcpy(((char *) es) + EXT4_S_ERR_START,
4133 save, EXT4_S_ERR_LEN);
4138 ext4_msg(sb, KERN_ERR, "error loading journal");
4139 jbd2_journal_destroy(journal);
4143 EXT4_SB(sb)->s_journal = journal;
4144 ext4_clear_journal_err(sb, es);
4146 if (!really_read_only && journal_devnum &&
4147 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4148 es->s_journal_dev = cpu_to_le32(journal_devnum);
4150 /* Make sure we flush the recovery flag to disk. */
4151 ext4_commit_super(sb, 1);
4157 static int ext4_commit_super(struct super_block *sb, int sync)
4159 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4160 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4163 if (!sbh || block_device_ejected(sb))
4165 if (buffer_write_io_error(sbh)) {
4167 * Oh, dear. A previous attempt to write the
4168 * superblock failed. This could happen because the
4169 * USB device was yanked out. Or it could happen to
4170 * be a transient write error and maybe the block will
4171 * be remapped. Nothing we can do but to retry the
4172 * write and hope for the best.
4174 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4175 "superblock detected");
4176 clear_buffer_write_io_error(sbh);
4177 set_buffer_uptodate(sbh);
4180 * If the file system is mounted read-only, don't update the
4181 * superblock write time. This avoids updating the superblock
4182 * write time when we are mounting the root file system
4183 * read/only but we need to replay the journal; at that point,
4184 * for people who are east of GMT and who make their clock
4185 * tick in localtime for Windows bug-for-bug compatibility,
4186 * the clock is set in the future, and this will cause e2fsck
4187 * to complain and force a full file system check.
4189 if (!(sb->s_flags & MS_RDONLY))
4190 es->s_wtime = cpu_to_le32(get_seconds());
4191 if (sb->s_bdev->bd_part)
4192 es->s_kbytes_written =
4193 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4194 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4195 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4197 es->s_kbytes_written =
4198 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4199 ext4_free_blocks_count_set(es,
4200 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4201 &EXT4_SB(sb)->s_freeclusters_counter)));
4202 es->s_free_inodes_count =
4203 cpu_to_le32(percpu_counter_sum_positive(
4204 &EXT4_SB(sb)->s_freeinodes_counter));
4206 BUFFER_TRACE(sbh, "marking dirty");
4207 mark_buffer_dirty(sbh);
4209 error = sync_dirty_buffer(sbh);
4213 error = buffer_write_io_error(sbh);
4215 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4217 clear_buffer_write_io_error(sbh);
4218 set_buffer_uptodate(sbh);
4225 * Have we just finished recovery? If so, and if we are mounting (or
4226 * remounting) the filesystem readonly, then we will end up with a
4227 * consistent fs on disk. Record that fact.
4229 static void ext4_mark_recovery_complete(struct super_block *sb,
4230 struct ext4_super_block *es)
4232 journal_t *journal = EXT4_SB(sb)->s_journal;
4234 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4235 BUG_ON(journal != NULL);
4238 jbd2_journal_lock_updates(journal);
4239 if (jbd2_journal_flush(journal) < 0)
4242 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4243 sb->s_flags & MS_RDONLY) {
4244 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4245 ext4_commit_super(sb, 1);
4249 jbd2_journal_unlock_updates(journal);
4253 * If we are mounting (or read-write remounting) a filesystem whose journal
4254 * has recorded an error from a previous lifetime, move that error to the
4255 * main filesystem now.
4257 static void ext4_clear_journal_err(struct super_block *sb,
4258 struct ext4_super_block *es)
4264 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4266 journal = EXT4_SB(sb)->s_journal;
4269 * Now check for any error status which may have been recorded in the
4270 * journal by a prior ext4_error() or ext4_abort()
4273 j_errno = jbd2_journal_errno(journal);
4277 errstr = ext4_decode_error(sb, j_errno, nbuf);
4278 ext4_warning(sb, "Filesystem error recorded "
4279 "from previous mount: %s", errstr);
4280 ext4_warning(sb, "Marking fs in need of filesystem check.");
4282 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4283 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4284 ext4_commit_super(sb, 1);
4286 jbd2_journal_clear_err(journal);
4291 * Force the running and committing transactions to commit,
4292 * and wait on the commit.
4294 int ext4_force_commit(struct super_block *sb)
4299 if (sb->s_flags & MS_RDONLY)
4302 journal = EXT4_SB(sb)->s_journal;
4304 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4305 ret = ext4_journal_force_commit(journal);
4311 static void ext4_write_super(struct super_block *sb)
4314 ext4_commit_super(sb, 1);
4318 static int ext4_sync_fs(struct super_block *sb, int wait)
4322 struct ext4_sb_info *sbi = EXT4_SB(sb);
4324 trace_ext4_sync_fs(sb, wait);
4325 flush_workqueue(sbi->dio_unwritten_wq);
4326 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4328 jbd2_log_wait_commit(sbi->s_journal, target);
4334 * LVM calls this function before a (read-only) snapshot is created. This
4335 * gives us a chance to flush the journal completely and mark the fs clean.
4337 * Note that only this function cannot bring a filesystem to be in a clean
4338 * state independently, because ext4 prevents a new handle from being started
4339 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4342 static int ext4_freeze(struct super_block *sb)
4347 if (sb->s_flags & MS_RDONLY)
4350 journal = EXT4_SB(sb)->s_journal;
4352 /* Now we set up the journal barrier. */
4353 jbd2_journal_lock_updates(journal);
4356 * Don't clear the needs_recovery flag if we failed to flush
4359 error = jbd2_journal_flush(journal);
4363 /* Journal blocked and flushed, clear needs_recovery flag. */
4364 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4365 error = ext4_commit_super(sb, 1);
4367 /* we rely on s_frozen to stop further updates */
4368 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4373 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4374 * flag here, even though the filesystem is not technically dirty yet.
4376 static int ext4_unfreeze(struct super_block *sb)
4378 if (sb->s_flags & MS_RDONLY)
4382 /* Reset the needs_recovery flag before the fs is unlocked. */
4383 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4384 ext4_commit_super(sb, 1);
4390 * Structure to save mount options for ext4_remount's benefit
4392 struct ext4_mount_options {
4393 unsigned long s_mount_opt;
4394 unsigned long s_mount_opt2;
4397 unsigned long s_commit_interval;
4398 u32 s_min_batch_time, s_max_batch_time;
4401 char *s_qf_names[MAXQUOTAS];
4405 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4407 struct ext4_super_block *es;
4408 struct ext4_sb_info *sbi = EXT4_SB(sb);
4409 ext4_fsblk_t n_blocks_count = 0;
4410 unsigned long old_sb_flags;
4411 struct ext4_mount_options old_opts;
4412 int enable_quota = 0;
4414 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4419 char *orig_data = kstrdup(data, GFP_KERNEL);
4421 /* Store the original options */
4423 old_sb_flags = sb->s_flags;
4424 old_opts.s_mount_opt = sbi->s_mount_opt;
4425 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4426 old_opts.s_resuid = sbi->s_resuid;
4427 old_opts.s_resgid = sbi->s_resgid;
4428 old_opts.s_commit_interval = sbi->s_commit_interval;
4429 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4430 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4432 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4433 for (i = 0; i < MAXQUOTAS; i++)
4434 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4436 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4437 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4440 * Allow the "check" option to be passed as a remount option.
4442 if (!parse_options(data, sb, NULL, &journal_ioprio,
4443 &n_blocks_count, 1)) {
4448 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4449 ext4_abort(sb, "Abort forced by user");
4453 if (sbi->s_journal) {
4454 ext4_init_journal_params(sb, sbi->s_journal);
4455 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4458 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4459 n_blocks_count > ext4_blocks_count(es)) {
4460 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4465 if (*flags & MS_RDONLY) {
4466 err = dquot_suspend(sb, -1);
4471 * First of all, the unconditional stuff we have to do
4472 * to disable replay of the journal when we next remount
4474 sb->s_flags |= MS_RDONLY;
4477 * OK, test if we are remounting a valid rw partition
4478 * readonly, and if so set the rdonly flag and then
4479 * mark the partition as valid again.
4481 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4482 (sbi->s_mount_state & EXT4_VALID_FS))
4483 es->s_state = cpu_to_le16(sbi->s_mount_state);
4486 ext4_mark_recovery_complete(sb, es);
4488 /* Make sure we can mount this feature set readwrite */
4489 if (!ext4_feature_set_ok(sb, 0)) {
4494 * Make sure the group descriptor checksums
4495 * are sane. If they aren't, refuse to remount r/w.
4497 for (g = 0; g < sbi->s_groups_count; g++) {
4498 struct ext4_group_desc *gdp =
4499 ext4_get_group_desc(sb, g, NULL);
4501 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4502 ext4_msg(sb, KERN_ERR,
4503 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4504 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4505 le16_to_cpu(gdp->bg_checksum));
4512 * If we have an unprocessed orphan list hanging
4513 * around from a previously readonly bdev mount,
4514 * require a full umount/remount for now.
4516 if (es->s_last_orphan) {
4517 ext4_msg(sb, KERN_WARNING, "Couldn't "
4518 "remount RDWR because of unprocessed "
4519 "orphan inode list. Please "
4520 "umount/remount instead");
4526 * Mounting a RDONLY partition read-write, so reread
4527 * and store the current valid flag. (It may have
4528 * been changed by e2fsck since we originally mounted
4532 ext4_clear_journal_err(sb, es);
4533 sbi->s_mount_state = le16_to_cpu(es->s_state);
4534 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4536 if (!ext4_setup_super(sb, es, 0))
4537 sb->s_flags &= ~MS_RDONLY;
4538 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4539 EXT4_FEATURE_INCOMPAT_MMP))
4540 if (ext4_multi_mount_protect(sb,
4541 le64_to_cpu(es->s_mmp_block))) {
4550 * Reinitialize lazy itable initialization thread based on
4553 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4554 ext4_unregister_li_request(sb);
4556 ext4_group_t first_not_zeroed;
4557 first_not_zeroed = ext4_has_uninit_itable(sb);
4558 ext4_register_li_request(sb, first_not_zeroed);
4561 ext4_setup_system_zone(sb);
4562 if (sbi->s_journal == NULL)
4563 ext4_commit_super(sb, 1);
4566 /* Release old quota file names */
4567 for (i = 0; i < MAXQUOTAS; i++)
4568 if (old_opts.s_qf_names[i] &&
4569 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4570 kfree(old_opts.s_qf_names[i]);
4574 dquot_resume(sb, -1);
4576 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4581 sb->s_flags = old_sb_flags;
4582 sbi->s_mount_opt = old_opts.s_mount_opt;
4583 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4584 sbi->s_resuid = old_opts.s_resuid;
4585 sbi->s_resgid = old_opts.s_resgid;
4586 sbi->s_commit_interval = old_opts.s_commit_interval;
4587 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4588 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4590 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4591 for (i = 0; i < MAXQUOTAS; i++) {
4592 if (sbi->s_qf_names[i] &&
4593 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4594 kfree(sbi->s_qf_names[i]);
4595 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4604 * Note: calculating the overhead so we can be compatible with
4605 * historical BSD practice is quite difficult in the face of
4606 * clusters/bigalloc. This is because multiple metadata blocks from
4607 * different block group can end up in the same allocation cluster.
4608 * Calculating the exact overhead in the face of clustered allocation
4609 * requires either O(all block bitmaps) in memory or O(number of block
4610 * groups**2) in time. We will still calculate the superblock for
4611 * older file systems --- and if we come across with a bigalloc file
4612 * system with zero in s_overhead_clusters the estimate will be close to
4613 * correct especially for very large cluster sizes --- but for newer
4614 * file systems, it's better to calculate this figure once at mkfs
4615 * time, and store it in the superblock. If the superblock value is
4616 * present (even for non-bigalloc file systems), we will use it.
4618 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4620 struct super_block *sb = dentry->d_sb;
4621 struct ext4_sb_info *sbi = EXT4_SB(sb);
4622 struct ext4_super_block *es = sbi->s_es;
4623 struct ext4_group_desc *gdp;
4627 if (test_opt(sb, MINIX_DF)) {
4628 sbi->s_overhead_last = 0;
4629 } else if (es->s_overhead_clusters) {
4630 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4631 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4632 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4633 ext4_fsblk_t overhead = 0;
4636 * Compute the overhead (FS structures). This is constant
4637 * for a given filesystem unless the number of block groups
4638 * changes so we cache the previous value until it does.
4642 * All of the blocks before first_data_block are
4645 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4648 * Add the overhead found in each block group
4650 for (i = 0; i < ngroups; i++) {
4651 gdp = ext4_get_group_desc(sb, i, NULL);
4652 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4655 sbi->s_overhead_last = overhead;
4657 sbi->s_blocks_last = ext4_blocks_count(es);
4660 buf->f_type = EXT4_SUPER_MAGIC;
4661 buf->f_bsize = sb->s_blocksize;
4662 buf->f_blocks = (ext4_blocks_count(es) -
4663 EXT4_C2B(sbi, sbi->s_overhead_last));
4664 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4665 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4666 /* prevent underflow in case that few free space is available */
4667 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4668 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4669 if (buf->f_bfree < ext4_r_blocks_count(es))
4671 buf->f_files = le32_to_cpu(es->s_inodes_count);
4672 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4673 buf->f_namelen = EXT4_NAME_LEN;
4674 fsid = le64_to_cpup((void *)es->s_uuid) ^
4675 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4676 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4677 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4682 /* Helper function for writing quotas on sync - we need to start transaction
4683 * before quota file is locked for write. Otherwise the are possible deadlocks:
4684 * Process 1 Process 2
4685 * ext4_create() quota_sync()
4686 * jbd2_journal_start() write_dquot()
4687 * dquot_initialize() down(dqio_mutex)
4688 * down(dqio_mutex) jbd2_journal_start()
4694 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4696 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4699 static int ext4_write_dquot(struct dquot *dquot)
4703 struct inode *inode;
4705 inode = dquot_to_inode(dquot);
4706 handle = ext4_journal_start(inode,
4707 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4709 return PTR_ERR(handle);
4710 ret = dquot_commit(dquot);
4711 err = ext4_journal_stop(handle);
4717 static int ext4_acquire_dquot(struct dquot *dquot)
4722 handle = ext4_journal_start(dquot_to_inode(dquot),
4723 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4725 return PTR_ERR(handle);
4726 ret = dquot_acquire(dquot);
4727 err = ext4_journal_stop(handle);
4733 static int ext4_release_dquot(struct dquot *dquot)
4738 handle = ext4_journal_start(dquot_to_inode(dquot),
4739 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4740 if (IS_ERR(handle)) {
4741 /* Release dquot anyway to avoid endless cycle in dqput() */
4742 dquot_release(dquot);
4743 return PTR_ERR(handle);
4745 ret = dquot_release(dquot);
4746 err = ext4_journal_stop(handle);
4752 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4754 /* Are we journaling quotas? */
4755 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4756 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4757 dquot_mark_dquot_dirty(dquot);
4758 return ext4_write_dquot(dquot);
4760 return dquot_mark_dquot_dirty(dquot);
4764 static int ext4_write_info(struct super_block *sb, int type)
4769 /* Data block + inode block */
4770 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4772 return PTR_ERR(handle);
4773 ret = dquot_commit_info(sb, type);
4774 err = ext4_journal_stop(handle);
4781 * Turn on quotas during mount time - we need to find
4782 * the quota file and such...
4784 static int ext4_quota_on_mount(struct super_block *sb, int type)
4786 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4787 EXT4_SB(sb)->s_jquota_fmt, type);
4791 * Standard function to be called on quota_on
4793 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4798 if (!test_opt(sb, QUOTA))
4801 /* Quotafile not on the same filesystem? */
4802 if (path->dentry->d_sb != sb)
4804 /* Journaling quota? */
4805 if (EXT4_SB(sb)->s_qf_names[type]) {
4806 /* Quotafile not in fs root? */
4807 if (path->dentry->d_parent != sb->s_root)
4808 ext4_msg(sb, KERN_WARNING,
4809 "Quota file not on filesystem root. "
4810 "Journaled quota will not work");
4814 * When we journal data on quota file, we have to flush journal to see
4815 * all updates to the file when we bypass pagecache...
4817 if (EXT4_SB(sb)->s_journal &&
4818 ext4_should_journal_data(path->dentry->d_inode)) {
4820 * We don't need to lock updates but journal_flush() could
4821 * otherwise be livelocked...
4823 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4824 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4825 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4830 return dquot_quota_on(sb, type, format_id, path);
4833 static int ext4_quota_off(struct super_block *sb, int type)
4835 struct inode *inode = sb_dqopt(sb)->files[type];
4838 /* Force all delayed allocation blocks to be allocated.
4839 * Caller already holds s_umount sem */
4840 if (test_opt(sb, DELALLOC))
4841 sync_filesystem(sb);
4846 /* Update modification times of quota files when userspace can
4847 * start looking at them */
4848 handle = ext4_journal_start(inode, 1);
4851 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4852 ext4_mark_inode_dirty(handle, inode);
4853 ext4_journal_stop(handle);
4856 return dquot_quota_off(sb, type);
4859 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4860 * acquiring the locks... As quota files are never truncated and quota code
4861 * itself serializes the operations (and no one else should touch the files)
4862 * we don't have to be afraid of races */
4863 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4864 size_t len, loff_t off)
4866 struct inode *inode = sb_dqopt(sb)->files[type];
4867 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4869 int offset = off & (sb->s_blocksize - 1);
4872 struct buffer_head *bh;
4873 loff_t i_size = i_size_read(inode);
4877 if (off+len > i_size)
4880 while (toread > 0) {
4881 tocopy = sb->s_blocksize - offset < toread ?
4882 sb->s_blocksize - offset : toread;
4883 bh = ext4_bread(NULL, inode, blk, 0, &err);
4886 if (!bh) /* A hole? */
4887 memset(data, 0, tocopy);
4889 memcpy(data, bh->b_data+offset, tocopy);
4899 /* Write to quotafile (we know the transaction is already started and has
4900 * enough credits) */
4901 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4902 const char *data, size_t len, loff_t off)
4904 struct inode *inode = sb_dqopt(sb)->files[type];
4905 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4907 int offset = off & (sb->s_blocksize - 1);
4908 struct buffer_head *bh;
4909 handle_t *handle = journal_current_handle();
4911 if (EXT4_SB(sb)->s_journal && !handle) {
4912 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4913 " cancelled because transaction is not started",
4914 (unsigned long long)off, (unsigned long long)len);
4918 * Since we account only one data block in transaction credits,
4919 * then it is impossible to cross a block boundary.
4921 if (sb->s_blocksize - offset < len) {
4922 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4923 " cancelled because not block aligned",
4924 (unsigned long long)off, (unsigned long long)len);
4928 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4929 bh = ext4_bread(handle, inode, blk, 1, &err);
4932 err = ext4_journal_get_write_access(handle, bh);
4938 memcpy(bh->b_data+offset, data, len);
4939 flush_dcache_page(bh->b_page);
4941 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4945 mutex_unlock(&inode->i_mutex);
4948 if (inode->i_size < off + len) {
4949 i_size_write(inode, off + len);
4950 EXT4_I(inode)->i_disksize = inode->i_size;
4951 ext4_mark_inode_dirty(handle, inode);
4953 mutex_unlock(&inode->i_mutex);
4959 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4960 const char *dev_name, void *data)
4962 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4965 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4966 static inline void register_as_ext2(void)
4968 int err = register_filesystem(&ext2_fs_type);
4971 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4974 static inline void unregister_as_ext2(void)
4976 unregister_filesystem(&ext2_fs_type);
4979 static inline int ext2_feature_set_ok(struct super_block *sb)
4981 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4983 if (sb->s_flags & MS_RDONLY)
4985 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4989 MODULE_ALIAS("ext2");
4991 static inline void register_as_ext2(void) { }
4992 static inline void unregister_as_ext2(void) { }
4993 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4996 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4997 static inline void register_as_ext3(void)
4999 int err = register_filesystem(&ext3_fs_type);
5002 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5005 static inline void unregister_as_ext3(void)
5007 unregister_filesystem(&ext3_fs_type);
5010 static inline int ext3_feature_set_ok(struct super_block *sb)
5012 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5014 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5016 if (sb->s_flags & MS_RDONLY)
5018 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5022 MODULE_ALIAS("ext3");
5024 static inline void register_as_ext3(void) { }
5025 static inline void unregister_as_ext3(void) { }
5026 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5029 static struct file_system_type ext4_fs_type = {
5030 .owner = THIS_MODULE,
5032 .mount = ext4_mount,
5033 .kill_sb = kill_block_super,
5034 .fs_flags = FS_REQUIRES_DEV,
5037 static int __init ext4_init_feat_adverts(void)
5039 struct ext4_features *ef;
5042 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5046 ef->f_kobj.kset = ext4_kset;
5047 init_completion(&ef->f_kobj_unregister);
5048 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5061 static void ext4_exit_feat_adverts(void)
5063 kobject_put(&ext4_feat->f_kobj);
5064 wait_for_completion(&ext4_feat->f_kobj_unregister);
5068 /* Shared across all ext4 file systems */
5069 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5070 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5072 static int __init ext4_init_fs(void)
5076 ext4_check_flag_values();
5078 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5079 mutex_init(&ext4__aio_mutex[i]);
5080 init_waitqueue_head(&ext4__ioend_wq[i]);
5083 err = ext4_init_pageio();
5086 err = ext4_init_system_zone();
5089 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5092 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5094 err = ext4_init_feat_adverts();
5098 err = ext4_init_mballoc();
5102 err = ext4_init_xattr();
5105 err = init_inodecache();
5110 err = register_filesystem(&ext4_fs_type);
5114 ext4_li_info = NULL;
5115 mutex_init(&ext4_li_mtx);
5118 unregister_as_ext2();
5119 unregister_as_ext3();
5120 destroy_inodecache();
5124 ext4_exit_mballoc();
5126 ext4_exit_feat_adverts();
5129 remove_proc_entry("fs/ext4", NULL);
5130 kset_unregister(ext4_kset);
5132 ext4_exit_system_zone();
5138 static void __exit ext4_exit_fs(void)
5140 ext4_destroy_lazyinit_thread();
5141 unregister_as_ext2();
5142 unregister_as_ext3();
5143 unregister_filesystem(&ext4_fs_type);
5144 destroy_inodecache();
5146 ext4_exit_mballoc();
5147 ext4_exit_feat_adverts();
5148 remove_proc_entry("fs/ext4", NULL);
5149 kset_unregister(ext4_kset);
5150 ext4_exit_system_zone();
5154 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5155 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5156 MODULE_LICENSE("GPL");
5157 module_init(ext4_init_fs)
5158 module_exit(ext4_exit_fs)