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"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 void *ext4_kvmalloc(size_t size, gfp_t flags)
118 ret = kmalloc(size, flags);
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
124 void *ext4_kvzalloc(size_t size, gfp_t flags)
128 ret = kzalloc(size, flags);
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
134 void ext4_kvfree(void *ptr)
136 if (is_vmalloc_addr(ptr))
143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
159 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
162 return le32_to_cpu(bg->bg_inode_table_lo) |
163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
167 __u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
175 __u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
183 __u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
191 __u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
199 void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
207 void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
215 void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
223 void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
231 void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
239 void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
247 void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
256 /* Just increment the non-pointer handle value */
257 static handle_t *ext4_get_nojournal(void)
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
265 handle = (handle_t *)ref_cnt;
267 current->journal_info = handle;
272 /* Decrement the non-pointer handle value */
273 static void ext4_put_nojournal(handle_t *handle)
275 unsigned long ref_cnt = (unsigned long)handle;
277 BUG_ON(ref_cnt == 0);
280 handle = (handle_t *)ref_cnt;
282 current->journal_info = handle;
286 * Wrappers for jbd2_journal_start/end.
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
306 journal = EXT4_SB(sb)->s_journal;
307 handle = ext4_journal_current_handle();
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
320 return ext4_get_nojournal();
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
330 return jbd2_journal_start(journal, nblocks);
334 * The only special thing we need to do here is to make sure that all
335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
336 * that sync() will call the filesystem's write_super callback if
339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
341 struct super_block *sb;
345 if (!ext4_handle_valid(handle)) {
346 ext4_put_nojournal(handle);
349 sb = handle->h_transaction->t_journal->j_private;
351 rc = jbd2_journal_stop(handle);
356 __ext4_std_error(sb, where, line, err);
360 void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
367 BUG_ON(!ext4_handle_valid(handle));
370 BUFFER_TRACE(bh, "abort");
375 if (is_handle_aborted(handle))
378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
381 jbd2_journal_abort_handle(handle);
384 static void __save_error_info(struct super_block *sb, const char *func,
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
403 * Start the daily error reporting function if it hasn't been
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
411 static void save_error_info(struct super_block *sb, const char *func,
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
426 static int block_device_ejected(struct super_block *sb)
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431 return bdi->dev == NULL;
435 /* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
438 * On ext2, we can store the error state of the filesystem in the
439 * superblock. That is not possible on ext4, because we may have other
440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
445 * We'll just use the jbd2_journal_abort() error code to record an error in
446 * the journal instead. On recovery, the journal will complain about
447 * that error until we've noted it down and cleared it.
450 static void ext4_handle_error(struct super_block *sb)
452 if (sb->s_flags & MS_RDONLY)
455 if (!test_opt(sb, ERRORS_CONT)) {
456 journal_t *journal = EXT4_SB(sb)->s_journal;
458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
460 jbd2_journal_abort(journal, -EIO);
462 if (test_opt(sb, ERRORS_RO)) {
463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
464 sb->s_flags |= MS_RDONLY;
466 if (test_opt(sb, ERRORS_PANIC))
467 panic("EXT4-fs (device %s): panic forced after error\n",
471 void __ext4_error(struct super_block *sb, const char *function,
472 unsigned int line, const char *fmt, ...)
474 struct va_format vaf;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
484 ext4_handle_error(sb);
487 void ext4_error_inode(struct inode *inode, const char *function,
488 unsigned int line, ext4_fsblk_t block,
489 const char *fmt, ...)
492 struct va_format vaf;
493 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 es->s_last_error_block = cpu_to_le64(block);
497 save_error_info(inode->i_sb, function, line);
501 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
502 inode->i_sb->s_id, function, line, inode->i_ino);
504 printk(KERN_CONT "block %llu: ", block);
505 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
508 ext4_handle_error(inode->i_sb);
511 void ext4_error_file(struct file *file, const char *function,
512 unsigned int line, ext4_fsblk_t block,
513 const char *fmt, ...)
516 struct va_format vaf;
517 struct ext4_super_block *es;
518 struct inode *inode = file->f_dentry->d_inode;
519 char pathname[80], *path;
521 es = EXT4_SB(inode->i_sb)->s_es;
522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 save_error_info(inode->i_sb, function, line);
524 path = d_path(&(file->f_path), pathname, sizeof(pathname));
528 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
529 inode->i_sb->s_id, function, line, inode->i_ino);
531 printk(KERN_CONT "block %llu: ", block);
535 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
538 ext4_handle_error(inode->i_sb);
541 static const char *ext4_decode_error(struct super_block *sb, int errno,
548 errstr = "IO failure";
551 errstr = "Out of memory";
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
558 errstr = "Readonly filesystem";
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
578 void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
584 /* Special case: if the error is EROFS, and we're not already
585 * inside a transaction, then there's really no point in logging
587 if (errno == -EROFS && journal_current_handle() == NULL &&
588 (sb->s_flags & MS_RDONLY))
591 errstr = ext4_decode_error(sb, errno, nbuf);
592 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
593 sb->s_id, function, line, errstr);
594 save_error_info(sb, function, line);
596 ext4_handle_error(sb);
600 * ext4_abort is a much stronger failure handler than ext4_error. The
601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
609 void __ext4_abort(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
614 save_error_info(sb, function, line);
616 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
622 if ((sb->s_flags & MS_RDONLY) == 0) {
623 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
624 sb->s_flags |= MS_RDONLY;
625 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
626 if (EXT4_SB(sb)->s_journal)
627 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
628 save_error_info(sb, function, line);
630 if (test_opt(sb, ERRORS_PANIC))
631 panic("EXT4-fs panic from previous error\n");
634 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
636 struct va_format vaf;
642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
646 void __ext4_warning(struct super_block *sb, const char *function,
647 unsigned int line, const char *fmt, ...)
649 struct va_format vaf;
655 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
656 sb->s_id, function, line, &vaf);
660 void __ext4_grp_locked_error(const char *function, unsigned int line,
661 struct super_block *sb, ext4_group_t grp,
662 unsigned long ino, ext4_fsblk_t block,
663 const char *fmt, ...)
667 struct va_format vaf;
669 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
671 es->s_last_error_ino = cpu_to_le32(ino);
672 es->s_last_error_block = cpu_to_le64(block);
673 __save_error_info(sb, function, line);
679 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
680 sb->s_id, function, line, grp);
682 printk(KERN_CONT "inode %lu: ", ino);
684 printk(KERN_CONT "block %llu:", (unsigned long long) block);
685 printk(KERN_CONT "%pV\n", &vaf);
688 if (test_opt(sb, ERRORS_CONT)) {
689 ext4_commit_super(sb, 0);
693 ext4_unlock_group(sb, grp);
694 ext4_handle_error(sb);
696 * We only get here in the ERRORS_RO case; relocking the group
697 * may be dangerous, but nothing bad will happen since the
698 * filesystem will have already been marked read/only and the
699 * journal has been aborted. We return 1 as a hint to callers
700 * who might what to use the return value from
701 * ext4_grp_locked_error() to distinguish between the
702 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
703 * aggressively from the ext4 function in question, with a
704 * more appropriate error code.
706 ext4_lock_group(sb, grp);
710 void ext4_update_dynamic_rev(struct super_block *sb)
712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
714 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
718 "updating to rev %d because of new feature flag, "
719 "running e2fsck is recommended",
722 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
723 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
724 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
725 /* leave es->s_feature_*compat flags alone */
726 /* es->s_uuid will be set by e2fsck if empty */
729 * The rest of the superblock fields should be zero, and if not it
730 * means they are likely already in use, so leave them alone. We
731 * can leave it up to e2fsck to clean up any inconsistencies there.
736 * Open the external journal device
738 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
740 struct block_device *bdev;
741 char b[BDEVNAME_SIZE];
743 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
749 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
750 __bdevname(dev, b), PTR_ERR(bdev));
755 * Release the journal device
757 static int ext4_blkdev_put(struct block_device *bdev)
759 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
762 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
764 struct block_device *bdev;
767 bdev = sbi->journal_bdev;
769 ret = ext4_blkdev_put(bdev);
770 sbi->journal_bdev = NULL;
775 static inline struct inode *orphan_list_entry(struct list_head *l)
777 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
780 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
784 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
785 le32_to_cpu(sbi->s_es->s_last_orphan));
787 printk(KERN_ERR "sb_info orphan list:\n");
788 list_for_each(l, &sbi->s_orphan) {
789 struct inode *inode = orphan_list_entry(l);
791 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
792 inode->i_sb->s_id, inode->i_ino, inode,
793 inode->i_mode, inode->i_nlink,
798 static void ext4_put_super(struct super_block *sb)
800 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 struct ext4_super_block *es = sbi->s_es;
804 ext4_unregister_li_request(sb);
805 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
807 flush_workqueue(sbi->dio_unwritten_wq);
808 destroy_workqueue(sbi->dio_unwritten_wq);
812 ext4_commit_super(sb, 1);
814 if (sbi->s_journal) {
815 err = jbd2_journal_destroy(sbi->s_journal);
816 sbi->s_journal = NULL;
818 ext4_abort(sb, "Couldn't clean up the journal");
821 del_timer(&sbi->s_err_report);
822 ext4_release_system_zone(sb);
824 ext4_ext_release(sb);
825 ext4_xattr_put_super(sb);
827 if (!(sb->s_flags & MS_RDONLY)) {
828 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
829 es->s_state = cpu_to_le16(sbi->s_mount_state);
830 ext4_commit_super(sb, 1);
833 remove_proc_entry(sb->s_id, ext4_proc_root);
835 kobject_del(&sbi->s_kobj);
837 for (i = 0; i < sbi->s_gdb_count; i++)
838 brelse(sbi->s_group_desc[i]);
839 ext4_kvfree(sbi->s_group_desc);
840 ext4_kvfree(sbi->s_flex_groups);
841 percpu_counter_destroy(&sbi->s_freeclusters_counter);
842 percpu_counter_destroy(&sbi->s_freeinodes_counter);
843 percpu_counter_destroy(&sbi->s_dirs_counter);
844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
847 for (i = 0; i < MAXQUOTAS; i++)
848 kfree(sbi->s_qf_names[i]);
851 /* Debugging code just in case the in-memory inode orphan list
852 * isn't empty. The on-disk one can be non-empty if we've
853 * detected an error and taken the fs readonly, but the
854 * in-memory list had better be clean by this point. */
855 if (!list_empty(&sbi->s_orphan))
856 dump_orphan_list(sb, sbi);
857 J_ASSERT(list_empty(&sbi->s_orphan));
859 invalidate_bdev(sb->s_bdev);
860 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
862 * Invalidate the journal device's buffers. We don't want them
863 * floating about in memory - the physical journal device may
864 * hotswapped, and it breaks the `ro-after' testing code.
866 sync_blockdev(sbi->journal_bdev);
867 invalidate_bdev(sbi->journal_bdev);
868 ext4_blkdev_remove(sbi);
871 kthread_stop(sbi->s_mmp_tsk);
872 sb->s_fs_info = NULL;
874 * Now that we are completely done shutting down the
875 * superblock, we need to actually destroy the kobject.
878 kobject_put(&sbi->s_kobj);
879 wait_for_completion(&sbi->s_kobj_unregister);
880 kfree(sbi->s_blockgroup_lock);
884 static struct kmem_cache *ext4_inode_cachep;
887 * Called inside transaction, so use GFP_NOFS
889 static struct inode *ext4_alloc_inode(struct super_block *sb)
891 struct ext4_inode_info *ei;
893 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
897 ei->vfs_inode.i_version = 1;
898 ei->vfs_inode.i_data.writeback_index = 0;
899 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
900 INIT_LIST_HEAD(&ei->i_prealloc_list);
901 spin_lock_init(&ei->i_prealloc_lock);
902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
905 ei->i_da_metadata_calc_len = 0;
906 spin_lock_init(&(ei->i_block_reservation_lock));
908 ei->i_reserved_quota = 0;
911 INIT_LIST_HEAD(&ei->i_completed_io_list);
912 spin_lock_init(&ei->i_completed_io_lock);
913 ei->cur_aio_dio = NULL;
915 ei->i_datasync_tid = 0;
916 atomic_set(&ei->i_ioend_count, 0);
917 atomic_set(&ei->i_aiodio_unwritten, 0);
919 return &ei->vfs_inode;
922 static int ext4_drop_inode(struct inode *inode)
924 int drop = generic_drop_inode(inode);
926 trace_ext4_drop_inode(inode, drop);
930 static void ext4_i_callback(struct rcu_head *head)
932 struct inode *inode = container_of(head, struct inode, i_rcu);
933 INIT_LIST_HEAD(&inode->i_dentry);
934 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
937 static void ext4_destroy_inode(struct inode *inode)
939 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
940 ext4_msg(inode->i_sb, KERN_ERR,
941 "Inode %lu (%p): orphan list check failed!",
942 inode->i_ino, EXT4_I(inode));
943 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
944 EXT4_I(inode), sizeof(struct ext4_inode_info),
948 call_rcu(&inode->i_rcu, ext4_i_callback);
951 static void init_once(void *foo)
953 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
955 INIT_LIST_HEAD(&ei->i_orphan);
956 #ifdef CONFIG_EXT4_FS_XATTR
957 init_rwsem(&ei->xattr_sem);
959 init_rwsem(&ei->i_data_sem);
960 inode_init_once(&ei->vfs_inode);
963 static int init_inodecache(void)
965 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
966 sizeof(struct ext4_inode_info),
967 0, (SLAB_RECLAIM_ACCOUNT|
970 if (ext4_inode_cachep == NULL)
975 static void destroy_inodecache(void)
977 kmem_cache_destroy(ext4_inode_cachep);
980 void ext4_clear_inode(struct inode *inode)
982 invalidate_inode_buffers(inode);
983 end_writeback(inode);
985 ext4_discard_preallocations(inode);
986 if (EXT4_I(inode)->jinode) {
987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
988 EXT4_I(inode)->jinode);
989 jbd2_free_inode(EXT4_I(inode)->jinode);
990 EXT4_I(inode)->jinode = NULL;
994 static inline void ext4_show_quota_options(struct seq_file *seq,
995 struct super_block *sb)
997 #if defined(CONFIG_QUOTA)
998 struct ext4_sb_info *sbi = EXT4_SB(sb);
1000 if (sbi->s_jquota_fmt) {
1003 switch (sbi->s_jquota_fmt) {
1014 seq_printf(seq, ",jqfmt=%s", fmtname);
1017 if (sbi->s_qf_names[USRQUOTA])
1018 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1020 if (sbi->s_qf_names[GRPQUOTA])
1021 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1023 if (test_opt(sb, USRQUOTA))
1024 seq_puts(seq, ",usrquota");
1026 if (test_opt(sb, GRPQUOTA))
1027 seq_puts(seq, ",grpquota");
1033 * - it's set to a non-default value OR
1034 * - if the per-sb default is different from the global default
1036 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1039 unsigned long def_mount_opts;
1040 struct super_block *sb = vfs->mnt_sb;
1041 struct ext4_sb_info *sbi = EXT4_SB(sb);
1042 struct ext4_super_block *es = sbi->s_es;
1044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1045 def_errors = le16_to_cpu(es->s_errors);
1047 if (sbi->s_sb_block != 1)
1048 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1049 if (test_opt(sb, MINIX_DF))
1050 seq_puts(seq, ",minixdf");
1051 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1052 seq_puts(seq, ",grpid");
1053 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1054 seq_puts(seq, ",nogrpid");
1055 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1056 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1057 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1059 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1060 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1061 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1063 if (test_opt(sb, ERRORS_RO)) {
1064 if (def_errors == EXT4_ERRORS_PANIC ||
1065 def_errors == EXT4_ERRORS_CONTINUE) {
1066 seq_puts(seq, ",errors=remount-ro");
1069 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1070 seq_puts(seq, ",errors=continue");
1071 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1072 seq_puts(seq, ",errors=panic");
1073 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1074 seq_puts(seq, ",nouid32");
1075 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1076 seq_puts(seq, ",debug");
1077 #ifdef CONFIG_EXT4_FS_XATTR
1078 if (test_opt(sb, XATTR_USER))
1079 seq_puts(seq, ",user_xattr");
1080 if (!test_opt(sb, XATTR_USER))
1081 seq_puts(seq, ",nouser_xattr");
1083 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1084 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1085 seq_puts(seq, ",acl");
1086 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1087 seq_puts(seq, ",noacl");
1089 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1090 seq_printf(seq, ",commit=%u",
1091 (unsigned) (sbi->s_commit_interval / HZ));
1093 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1094 seq_printf(seq, ",min_batch_time=%u",
1095 (unsigned) sbi->s_min_batch_time);
1097 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1098 seq_printf(seq, ",max_batch_time=%u",
1099 (unsigned) sbi->s_min_batch_time);
1103 * We're changing the default of barrier mount option, so
1104 * let's always display its mount state so it's clear what its
1107 seq_puts(seq, ",barrier=");
1108 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1109 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1110 seq_puts(seq, ",journal_async_commit");
1111 else if (test_opt(sb, JOURNAL_CHECKSUM))
1112 seq_puts(seq, ",journal_checksum");
1113 if (test_opt(sb, I_VERSION))
1114 seq_puts(seq, ",i_version");
1115 if (!test_opt(sb, DELALLOC) &&
1116 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1117 seq_puts(seq, ",nodelalloc");
1119 if (!test_opt(sb, MBLK_IO_SUBMIT))
1120 seq_puts(seq, ",nomblk_io_submit");
1122 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1124 * journal mode get enabled in different ways
1125 * So just print the value even if we didn't specify it
1127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1128 seq_puts(seq, ",data=journal");
1129 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1130 seq_puts(seq, ",data=ordered");
1131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1132 seq_puts(seq, ",data=writeback");
1134 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1135 seq_printf(seq, ",inode_readahead_blks=%u",
1136 sbi->s_inode_readahead_blks);
1138 if (test_opt(sb, DATA_ERR_ABORT))
1139 seq_puts(seq, ",data_err=abort");
1141 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1142 seq_puts(seq, ",noauto_da_alloc");
1144 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1145 seq_puts(seq, ",discard");
1147 if (test_opt(sb, NOLOAD))
1148 seq_puts(seq, ",norecovery");
1150 if (test_opt(sb, DIOREAD_NOLOCK))
1151 seq_puts(seq, ",dioread_nolock");
1153 if (test_opt(sb, BLOCK_VALIDITY) &&
1154 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1155 seq_puts(seq, ",block_validity");
1157 if (!test_opt(sb, INIT_INODE_TABLE))
1158 seq_puts(seq, ",noinit_itable");
1159 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1160 seq_printf(seq, ",init_itable=%u",
1161 (unsigned) sbi->s_li_wait_mult);
1163 ext4_show_quota_options(seq, sb);
1168 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1169 u64 ino, u32 generation)
1171 struct inode *inode;
1173 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1174 return ERR_PTR(-ESTALE);
1175 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1176 return ERR_PTR(-ESTALE);
1178 /* iget isn't really right if the inode is currently unallocated!!
1180 * ext4_read_inode will return a bad_inode if the inode had been
1181 * deleted, so we should be safe.
1183 * Currently we don't know the generation for parent directory, so
1184 * a generation of 0 means "accept any"
1186 inode = ext4_iget(sb, ino);
1188 return ERR_CAST(inode);
1189 if (generation && inode->i_generation != generation) {
1191 return ERR_PTR(-ESTALE);
1197 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1198 int fh_len, int fh_type)
1200 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1201 ext4_nfs_get_inode);
1204 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1205 int fh_len, int fh_type)
1207 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1208 ext4_nfs_get_inode);
1212 * Try to release metadata pages (indirect blocks, directories) which are
1213 * mapped via the block device. Since these pages could have journal heads
1214 * which would prevent try_to_free_buffers() from freeing them, we must use
1215 * jbd2 layer's try_to_free_buffers() function to release them.
1217 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1220 journal_t *journal = EXT4_SB(sb)->s_journal;
1222 WARN_ON(PageChecked(page));
1223 if (!page_has_buffers(page))
1226 return jbd2_journal_try_to_free_buffers(journal, page,
1227 wait & ~__GFP_WAIT);
1228 return try_to_free_buffers(page);
1232 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1233 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1235 static int ext4_write_dquot(struct dquot *dquot);
1236 static int ext4_acquire_dquot(struct dquot *dquot);
1237 static int ext4_release_dquot(struct dquot *dquot);
1238 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1239 static int ext4_write_info(struct super_block *sb, int type);
1240 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1242 static int ext4_quota_off(struct super_block *sb, int type);
1243 static int ext4_quota_on_mount(struct super_block *sb, int type);
1244 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1245 size_t len, loff_t off);
1246 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1247 const char *data, size_t len, loff_t off);
1249 static const struct dquot_operations ext4_quota_operations = {
1250 .get_reserved_space = ext4_get_reserved_space,
1251 .write_dquot = ext4_write_dquot,
1252 .acquire_dquot = ext4_acquire_dquot,
1253 .release_dquot = ext4_release_dquot,
1254 .mark_dirty = ext4_mark_dquot_dirty,
1255 .write_info = ext4_write_info,
1256 .alloc_dquot = dquot_alloc,
1257 .destroy_dquot = dquot_destroy,
1260 static const struct quotactl_ops ext4_qctl_operations = {
1261 .quota_on = ext4_quota_on,
1262 .quota_off = ext4_quota_off,
1263 .quota_sync = dquot_quota_sync,
1264 .get_info = dquot_get_dqinfo,
1265 .set_info = dquot_set_dqinfo,
1266 .get_dqblk = dquot_get_dqblk,
1267 .set_dqblk = dquot_set_dqblk
1271 static const struct super_operations ext4_sops = {
1272 .alloc_inode = ext4_alloc_inode,
1273 .destroy_inode = ext4_destroy_inode,
1274 .write_inode = ext4_write_inode,
1275 .dirty_inode = ext4_dirty_inode,
1276 .drop_inode = ext4_drop_inode,
1277 .evict_inode = ext4_evict_inode,
1278 .put_super = ext4_put_super,
1279 .sync_fs = ext4_sync_fs,
1280 .freeze_fs = ext4_freeze,
1281 .unfreeze_fs = ext4_unfreeze,
1282 .statfs = ext4_statfs,
1283 .remount_fs = ext4_remount,
1284 .show_options = ext4_show_options,
1286 .quota_read = ext4_quota_read,
1287 .quota_write = ext4_quota_write,
1289 .bdev_try_to_free_page = bdev_try_to_free_page,
1292 static const struct super_operations ext4_nojournal_sops = {
1293 .alloc_inode = ext4_alloc_inode,
1294 .destroy_inode = ext4_destroy_inode,
1295 .write_inode = ext4_write_inode,
1296 .dirty_inode = ext4_dirty_inode,
1297 .drop_inode = ext4_drop_inode,
1298 .evict_inode = ext4_evict_inode,
1299 .write_super = ext4_write_super,
1300 .put_super = ext4_put_super,
1301 .statfs = ext4_statfs,
1302 .remount_fs = ext4_remount,
1303 .show_options = ext4_show_options,
1305 .quota_read = ext4_quota_read,
1306 .quota_write = ext4_quota_write,
1308 .bdev_try_to_free_page = bdev_try_to_free_page,
1311 static const struct export_operations ext4_export_ops = {
1312 .fh_to_dentry = ext4_fh_to_dentry,
1313 .fh_to_parent = ext4_fh_to_parent,
1314 .get_parent = ext4_get_parent,
1318 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1319 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1320 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1321 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1322 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1323 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1324 Opt_journal_update, Opt_journal_dev,
1325 Opt_journal_checksum, Opt_journal_async_commit,
1326 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1327 Opt_data_err_abort, Opt_data_err_ignore,
1328 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1329 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1330 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1331 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1332 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1333 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1334 Opt_inode_readahead_blks, Opt_journal_ioprio,
1335 Opt_dioread_nolock, Opt_dioread_lock,
1336 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1339 static const match_table_t tokens = {
1340 {Opt_bsd_df, "bsddf"},
1341 {Opt_minix_df, "minixdf"},
1342 {Opt_grpid, "grpid"},
1343 {Opt_grpid, "bsdgroups"},
1344 {Opt_nogrpid, "nogrpid"},
1345 {Opt_nogrpid, "sysvgroups"},
1346 {Opt_resgid, "resgid=%u"},
1347 {Opt_resuid, "resuid=%u"},
1349 {Opt_err_cont, "errors=continue"},
1350 {Opt_err_panic, "errors=panic"},
1351 {Opt_err_ro, "errors=remount-ro"},
1352 {Opt_nouid32, "nouid32"},
1353 {Opt_debug, "debug"},
1354 {Opt_oldalloc, "oldalloc"},
1355 {Opt_orlov, "orlov"},
1356 {Opt_user_xattr, "user_xattr"},
1357 {Opt_nouser_xattr, "nouser_xattr"},
1359 {Opt_noacl, "noacl"},
1360 {Opt_noload, "noload"},
1361 {Opt_noload, "norecovery"},
1364 {Opt_commit, "commit=%u"},
1365 {Opt_min_batch_time, "min_batch_time=%u"},
1366 {Opt_max_batch_time, "max_batch_time=%u"},
1367 {Opt_journal_update, "journal=update"},
1368 {Opt_journal_dev, "journal_dev=%u"},
1369 {Opt_journal_checksum, "journal_checksum"},
1370 {Opt_journal_async_commit, "journal_async_commit"},
1371 {Opt_abort, "abort"},
1372 {Opt_data_journal, "data=journal"},
1373 {Opt_data_ordered, "data=ordered"},
1374 {Opt_data_writeback, "data=writeback"},
1375 {Opt_data_err_abort, "data_err=abort"},
1376 {Opt_data_err_ignore, "data_err=ignore"},
1377 {Opt_offusrjquota, "usrjquota="},
1378 {Opt_usrjquota, "usrjquota=%s"},
1379 {Opt_offgrpjquota, "grpjquota="},
1380 {Opt_grpjquota, "grpjquota=%s"},
1381 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1382 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1383 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1384 {Opt_grpquota, "grpquota"},
1385 {Opt_noquota, "noquota"},
1386 {Opt_quota, "quota"},
1387 {Opt_usrquota, "usrquota"},
1388 {Opt_barrier, "barrier=%u"},
1389 {Opt_barrier, "barrier"},
1390 {Opt_nobarrier, "nobarrier"},
1391 {Opt_i_version, "i_version"},
1392 {Opt_stripe, "stripe=%u"},
1393 {Opt_resize, "resize"},
1394 {Opt_delalloc, "delalloc"},
1395 {Opt_nodelalloc, "nodelalloc"},
1396 {Opt_mblk_io_submit, "mblk_io_submit"},
1397 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1398 {Opt_block_validity, "block_validity"},
1399 {Opt_noblock_validity, "noblock_validity"},
1400 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1401 {Opt_journal_ioprio, "journal_ioprio=%u"},
1402 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1403 {Opt_auto_da_alloc, "auto_da_alloc"},
1404 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1405 {Opt_dioread_nolock, "dioread_nolock"},
1406 {Opt_dioread_lock, "dioread_lock"},
1407 {Opt_discard, "discard"},
1408 {Opt_nodiscard, "nodiscard"},
1409 {Opt_init_itable, "init_itable=%u"},
1410 {Opt_init_itable, "init_itable"},
1411 {Opt_noinit_itable, "noinit_itable"},
1415 static ext4_fsblk_t get_sb_block(void **data)
1417 ext4_fsblk_t sb_block;
1418 char *options = (char *) *data;
1420 if (!options || strncmp(options, "sb=", 3) != 0)
1421 return 1; /* Default location */
1424 /* TODO: use simple_strtoll with >32bit ext4 */
1425 sb_block = simple_strtoul(options, &options, 0);
1426 if (*options && *options != ',') {
1427 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1431 if (*options == ',')
1433 *data = (void *) options;
1438 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1439 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1440 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1443 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1445 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448 if (sb_any_quota_loaded(sb) &&
1449 !sbi->s_qf_names[qtype]) {
1450 ext4_msg(sb, KERN_ERR,
1451 "Cannot change journaled "
1452 "quota options when quota turned on");
1455 qname = match_strdup(args);
1457 ext4_msg(sb, KERN_ERR,
1458 "Not enough memory for storing quotafile name");
1461 if (sbi->s_qf_names[qtype] &&
1462 strcmp(sbi->s_qf_names[qtype], qname)) {
1463 ext4_msg(sb, KERN_ERR,
1464 "%s quota file already specified", QTYPE2NAME(qtype));
1468 sbi->s_qf_names[qtype] = qname;
1469 if (strchr(sbi->s_qf_names[qtype], '/')) {
1470 ext4_msg(sb, KERN_ERR,
1471 "quotafile must be on filesystem root");
1472 kfree(sbi->s_qf_names[qtype]);
1473 sbi->s_qf_names[qtype] = NULL;
1480 static int clear_qf_name(struct super_block *sb, int qtype)
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1485 if (sb_any_quota_loaded(sb) &&
1486 sbi->s_qf_names[qtype]) {
1487 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1488 " when quota turned on");
1492 * The space will be released later when all options are confirmed
1495 sbi->s_qf_names[qtype] = NULL;
1500 static int parse_options(char *options, struct super_block *sb,
1501 unsigned long *journal_devnum,
1502 unsigned int *journal_ioprio,
1503 ext4_fsblk_t *n_blocks_count, int is_remount)
1505 struct ext4_sb_info *sbi = EXT4_SB(sb);
1507 substring_t args[MAX_OPT_ARGS];
1517 while ((p = strsep(&options, ",")) != NULL) {
1523 * Initialize args struct so we know whether arg was
1524 * found; some options take optional arguments.
1526 args[0].to = args[0].from = NULL;
1527 token = match_token(p, tokens, args);
1530 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1531 clear_opt(sb, MINIX_DF);
1534 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1535 set_opt(sb, MINIX_DF);
1539 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1544 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1545 clear_opt(sb, GRPID);
1549 if (match_int(&args[0], &option))
1551 sbi->s_resuid = option;
1554 if (match_int(&args[0], &option))
1556 sbi->s_resgid = option;
1559 /* handled by get_sb_block() instead of here */
1560 /* *sb_block = match_int(&args[0]); */
1563 clear_opt(sb, ERRORS_CONT);
1564 clear_opt(sb, ERRORS_RO);
1565 set_opt(sb, ERRORS_PANIC);
1568 clear_opt(sb, ERRORS_CONT);
1569 clear_opt(sb, ERRORS_PANIC);
1570 set_opt(sb, ERRORS_RO);
1573 clear_opt(sb, ERRORS_RO);
1574 clear_opt(sb, ERRORS_PANIC);
1575 set_opt(sb, ERRORS_CONT);
1578 set_opt(sb, NO_UID32);
1584 ext4_msg(sb, KERN_WARNING,
1585 "Ignoring deprecated oldalloc option");
1588 ext4_msg(sb, KERN_WARNING,
1589 "Ignoring deprecated orlov option");
1591 #ifdef CONFIG_EXT4_FS_XATTR
1592 case Opt_user_xattr:
1593 set_opt(sb, XATTR_USER);
1595 case Opt_nouser_xattr:
1596 clear_opt(sb, XATTR_USER);
1599 case Opt_user_xattr:
1600 case Opt_nouser_xattr:
1601 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1604 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1606 set_opt(sb, POSIX_ACL);
1609 clear_opt(sb, POSIX_ACL);
1614 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1617 case Opt_journal_update:
1619 /* Eventually we will want to be able to create
1620 a journal file here. For now, only allow the
1621 user to specify an existing inode to be the
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot specify journal on remount");
1628 set_opt(sb, UPDATE_JOURNAL);
1630 case Opt_journal_dev:
1632 ext4_msg(sb, KERN_ERR,
1633 "Cannot specify journal on remount");
1636 if (match_int(&args[0], &option))
1638 *journal_devnum = option;
1640 case Opt_journal_checksum:
1641 set_opt(sb, JOURNAL_CHECKSUM);
1643 case Opt_journal_async_commit:
1644 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1645 set_opt(sb, JOURNAL_CHECKSUM);
1648 set_opt(sb, NOLOAD);
1651 if (match_int(&args[0], &option))
1656 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1657 sbi->s_commit_interval = HZ * option;
1659 case Opt_max_batch_time:
1660 if (match_int(&args[0], &option))
1665 option = EXT4_DEF_MAX_BATCH_TIME;
1666 sbi->s_max_batch_time = option;
1668 case Opt_min_batch_time:
1669 if (match_int(&args[0], &option))
1673 sbi->s_min_batch_time = option;
1675 case Opt_data_journal:
1676 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1678 case Opt_data_ordered:
1679 data_opt = EXT4_MOUNT_ORDERED_DATA;
1681 case Opt_data_writeback:
1682 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1685 if (!sbi->s_journal)
1686 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1687 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1688 ext4_msg(sb, KERN_ERR,
1689 "Cannot change data mode on remount");
1693 clear_opt(sb, DATA_FLAGS);
1694 sbi->s_mount_opt |= data_opt;
1697 case Opt_data_err_abort:
1698 set_opt(sb, DATA_ERR_ABORT);
1700 case Opt_data_err_ignore:
1701 clear_opt(sb, DATA_ERR_ABORT);
1705 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1709 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1712 case Opt_offusrjquota:
1713 if (!clear_qf_name(sb, USRQUOTA))
1716 case Opt_offgrpjquota:
1717 if (!clear_qf_name(sb, GRPQUOTA))
1721 case Opt_jqfmt_vfsold:
1722 qfmt = QFMT_VFS_OLD;
1724 case Opt_jqfmt_vfsv0:
1727 case Opt_jqfmt_vfsv1:
1730 if (sb_any_quota_loaded(sb) &&
1731 sbi->s_jquota_fmt != qfmt) {
1732 ext4_msg(sb, KERN_ERR, "Cannot change "
1733 "journaled quota options when "
1737 sbi->s_jquota_fmt = qfmt;
1742 set_opt(sb, USRQUOTA);
1746 set_opt(sb, GRPQUOTA);
1749 if (sb_any_quota_loaded(sb)) {
1750 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1751 "options when quota turned on");
1754 clear_opt(sb, QUOTA);
1755 clear_opt(sb, USRQUOTA);
1756 clear_opt(sb, GRPQUOTA);
1762 ext4_msg(sb, KERN_ERR,
1763 "quota options not supported");
1767 case Opt_offusrjquota:
1768 case Opt_offgrpjquota:
1769 case Opt_jqfmt_vfsold:
1770 case Opt_jqfmt_vfsv0:
1771 case Opt_jqfmt_vfsv1:
1772 ext4_msg(sb, KERN_ERR,
1773 "journaled quota options not supported");
1779 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1782 clear_opt(sb, BARRIER);
1786 if (match_int(&args[0], &option))
1789 option = 1; /* No argument, default to 1 */
1791 set_opt(sb, BARRIER);
1793 clear_opt(sb, BARRIER);
1799 ext4_msg(sb, KERN_ERR,
1800 "resize option only available "
1804 if (match_int(&args[0], &option) != 0)
1806 *n_blocks_count = option;
1809 ext4_msg(sb, KERN_WARNING,
1810 "Ignoring deprecated nobh option");
1813 ext4_msg(sb, KERN_WARNING,
1814 "Ignoring deprecated bh option");
1817 set_opt(sb, I_VERSION);
1818 sb->s_flags |= MS_I_VERSION;
1820 case Opt_nodelalloc:
1821 clear_opt(sb, DELALLOC);
1822 clear_opt2(sb, EXPLICIT_DELALLOC);
1824 case Opt_mblk_io_submit:
1825 set_opt(sb, MBLK_IO_SUBMIT);
1827 case Opt_nomblk_io_submit:
1828 clear_opt(sb, MBLK_IO_SUBMIT);
1831 if (match_int(&args[0], &option))
1835 sbi->s_stripe = option;
1838 set_opt(sb, DELALLOC);
1839 set_opt2(sb, EXPLICIT_DELALLOC);
1841 case Opt_block_validity:
1842 set_opt(sb, BLOCK_VALIDITY);
1844 case Opt_noblock_validity:
1845 clear_opt(sb, BLOCK_VALIDITY);
1847 case Opt_inode_readahead_blks:
1848 if (match_int(&args[0], &option))
1850 if (option < 0 || option > (1 << 30))
1852 if (option && !is_power_of_2(option)) {
1853 ext4_msg(sb, KERN_ERR,
1854 "EXT4-fs: inode_readahead_blks"
1855 " must be a power of 2");
1858 sbi->s_inode_readahead_blks = option;
1860 case Opt_journal_ioprio:
1861 if (match_int(&args[0], &option))
1863 if (option < 0 || option > 7)
1865 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1868 case Opt_noauto_da_alloc:
1869 set_opt(sb, NO_AUTO_DA_ALLOC);
1871 case Opt_auto_da_alloc:
1873 if (match_int(&args[0], &option))
1876 option = 1; /* No argument, default to 1 */
1878 clear_opt(sb, NO_AUTO_DA_ALLOC);
1880 set_opt(sb,NO_AUTO_DA_ALLOC);
1883 set_opt(sb, DISCARD);
1886 clear_opt(sb, DISCARD);
1888 case Opt_dioread_nolock:
1889 set_opt(sb, DIOREAD_NOLOCK);
1891 case Opt_dioread_lock:
1892 clear_opt(sb, DIOREAD_NOLOCK);
1894 case Opt_init_itable:
1895 set_opt(sb, INIT_INODE_TABLE);
1897 if (match_int(&args[0], &option))
1900 option = EXT4_DEF_LI_WAIT_MULT;
1903 sbi->s_li_wait_mult = option;
1905 case Opt_noinit_itable:
1906 clear_opt(sb, INIT_INODE_TABLE);
1909 ext4_msg(sb, KERN_ERR,
1910 "Unrecognized mount option \"%s\" "
1911 "or missing value", p);
1916 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1917 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1918 clear_opt(sb, USRQUOTA);
1920 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1921 clear_opt(sb, GRPQUOTA);
1923 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1924 ext4_msg(sb, KERN_ERR, "old and new quota "
1929 if (!sbi->s_jquota_fmt) {
1930 ext4_msg(sb, KERN_ERR, "journaled quota format "
1935 if (sbi->s_jquota_fmt) {
1936 ext4_msg(sb, KERN_ERR, "journaled quota format "
1937 "specified with no journaling "
1946 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1953 ext4_msg(sb, KERN_ERR, "revision level too high, "
1954 "forcing read-only mode");
1959 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1960 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1961 "running e2fsck is recommended");
1962 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1963 ext4_msg(sb, KERN_WARNING,
1964 "warning: mounting fs with errors, "
1965 "running e2fsck is recommended");
1966 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1967 le16_to_cpu(es->s_mnt_count) >=
1968 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1969 ext4_msg(sb, KERN_WARNING,
1970 "warning: maximal mount count reached, "
1971 "running e2fsck is recommended");
1972 else if (le32_to_cpu(es->s_checkinterval) &&
1973 (le32_to_cpu(es->s_lastcheck) +
1974 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1975 ext4_msg(sb, KERN_WARNING,
1976 "warning: checktime reached, "
1977 "running e2fsck is recommended");
1978 if (!sbi->s_journal)
1979 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1980 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1981 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1982 le16_add_cpu(&es->s_mnt_count, 1);
1983 es->s_mtime = cpu_to_le32(get_seconds());
1984 ext4_update_dynamic_rev(sb);
1986 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1988 ext4_commit_super(sb, 1);
1990 if (test_opt(sb, DEBUG))
1991 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1992 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1994 sbi->s_groups_count,
1995 EXT4_BLOCKS_PER_GROUP(sb),
1996 EXT4_INODES_PER_GROUP(sb),
1997 sbi->s_mount_opt, sbi->s_mount_opt2);
1999 cleancache_init_fs(sb);
2003 static int ext4_fill_flex_info(struct super_block *sb)
2005 struct ext4_sb_info *sbi = EXT4_SB(sb);
2006 struct ext4_group_desc *gdp = NULL;
2007 ext4_group_t flex_group_count;
2008 ext4_group_t flex_group;
2009 int groups_per_flex = 0;
2013 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2014 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2016 if (groups_per_flex < 2) {
2017 sbi->s_log_groups_per_flex = 0;
2021 /* We allocate both existing and potentially added groups */
2022 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2023 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2024 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2025 size = flex_group_count * sizeof(struct flex_groups);
2026 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2027 if (sbi->s_flex_groups == NULL) {
2028 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2033 for (i = 0; i < sbi->s_groups_count; i++) {
2034 gdp = ext4_get_group_desc(sb, i, NULL);
2036 flex_group = ext4_flex_group(sbi, i);
2037 atomic_add(ext4_free_inodes_count(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_inodes);
2039 atomic_add(ext4_free_group_clusters(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].free_clusters);
2041 atomic_add(ext4_used_dirs_count(sb, gdp),
2042 &sbi->s_flex_groups[flex_group].used_dirs);
2050 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2055 if (sbi->s_es->s_feature_ro_compat &
2056 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2057 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2058 __le32 le_group = cpu_to_le32(block_group);
2060 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2061 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2062 crc = crc16(crc, (__u8 *)gdp, offset);
2063 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2064 /* for checksum of struct ext4_group_desc do the rest...*/
2065 if ((sbi->s_es->s_feature_incompat &
2066 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2067 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2068 crc = crc16(crc, (__u8 *)gdp + offset,
2069 le16_to_cpu(sbi->s_es->s_desc_size) -
2073 return cpu_to_le16(crc);
2076 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2077 struct ext4_group_desc *gdp)
2079 if ((sbi->s_es->s_feature_ro_compat &
2080 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2081 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2087 /* Called at mount-time, super-block is locked */
2088 static int ext4_check_descriptors(struct super_block *sb,
2089 ext4_group_t *first_not_zeroed)
2091 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2093 ext4_fsblk_t last_block;
2094 ext4_fsblk_t block_bitmap;
2095 ext4_fsblk_t inode_bitmap;
2096 ext4_fsblk_t inode_table;
2097 int flexbg_flag = 0;
2098 ext4_group_t i, grp = sbi->s_groups_count;
2100 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2103 ext4_debug("Checking group descriptors");
2105 for (i = 0; i < sbi->s_groups_count; i++) {
2106 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2108 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2109 last_block = ext4_blocks_count(sbi->s_es) - 1;
2111 last_block = first_block +
2112 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2114 if ((grp == sbi->s_groups_count) &&
2115 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2118 block_bitmap = ext4_block_bitmap(sb, gdp);
2119 if (block_bitmap < first_block || block_bitmap > last_block) {
2120 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2121 "Block bitmap for group %u not in group "
2122 "(block %llu)!", i, block_bitmap);
2125 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2126 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2127 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2128 "Inode bitmap for group %u not in group "
2129 "(block %llu)!", i, inode_bitmap);
2132 inode_table = ext4_inode_table(sb, gdp);
2133 if (inode_table < first_block ||
2134 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2135 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2136 "Inode table for group %u not in group "
2137 "(block %llu)!", i, inode_table);
2140 ext4_lock_group(sb, i);
2141 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2142 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2143 "Checksum for group %u failed (%u!=%u)",
2144 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2145 gdp)), le16_to_cpu(gdp->bg_checksum));
2146 if (!(sb->s_flags & MS_RDONLY)) {
2147 ext4_unlock_group(sb, i);
2151 ext4_unlock_group(sb, i);
2153 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2155 if (NULL != first_not_zeroed)
2156 *first_not_zeroed = grp;
2158 ext4_free_blocks_count_set(sbi->s_es,
2159 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2160 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2164 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2165 * the superblock) which were deleted from all directories, but held open by
2166 * a process at the time of a crash. We walk the list and try to delete these
2167 * inodes at recovery time (only with a read-write filesystem).
2169 * In order to keep the orphan inode chain consistent during traversal (in
2170 * case of crash during recovery), we link each inode into the superblock
2171 * orphan list_head and handle it the same way as an inode deletion during
2172 * normal operation (which journals the operations for us).
2174 * We only do an iget() and an iput() on each inode, which is very safe if we
2175 * accidentally point at an in-use or already deleted inode. The worst that
2176 * can happen in this case is that we get a "bit already cleared" message from
2177 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2178 * e2fsck was run on this filesystem, and it must have already done the orphan
2179 * inode cleanup for us, so we can safely abort without any further action.
2181 static void ext4_orphan_cleanup(struct super_block *sb,
2182 struct ext4_super_block *es)
2184 unsigned int s_flags = sb->s_flags;
2185 int nr_orphans = 0, nr_truncates = 0;
2189 if (!es->s_last_orphan) {
2190 jbd_debug(4, "no orphan inodes to clean up\n");
2194 if (bdev_read_only(sb->s_bdev)) {
2195 ext4_msg(sb, KERN_ERR, "write access "
2196 "unavailable, skipping orphan cleanup");
2200 /* Check if feature set would not allow a r/w mount */
2201 if (!ext4_feature_set_ok(sb, 0)) {
2202 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2203 "unknown ROCOMPAT features");
2207 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2208 if (es->s_last_orphan)
2209 jbd_debug(1, "Errors on filesystem, "
2210 "clearing orphan list.\n");
2211 es->s_last_orphan = 0;
2212 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2216 if (s_flags & MS_RDONLY) {
2217 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2218 sb->s_flags &= ~MS_RDONLY;
2221 /* Needed for iput() to work correctly and not trash data */
2222 sb->s_flags |= MS_ACTIVE;
2223 /* Turn on quotas so that they are updated correctly */
2224 for (i = 0; i < MAXQUOTAS; i++) {
2225 if (EXT4_SB(sb)->s_qf_names[i]) {
2226 int ret = ext4_quota_on_mount(sb, i);
2228 ext4_msg(sb, KERN_ERR,
2229 "Cannot turn on journaled "
2230 "quota: error %d", ret);
2235 while (es->s_last_orphan) {
2236 struct inode *inode;
2238 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2239 if (IS_ERR(inode)) {
2240 es->s_last_orphan = 0;
2244 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2245 dquot_initialize(inode);
2246 if (inode->i_nlink) {
2247 ext4_msg(sb, KERN_DEBUG,
2248 "%s: truncating inode %lu to %lld bytes",
2249 __func__, inode->i_ino, inode->i_size);
2250 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2251 inode->i_ino, inode->i_size);
2252 ext4_truncate(inode);
2255 ext4_msg(sb, KERN_DEBUG,
2256 "%s: deleting unreferenced inode %lu",
2257 __func__, inode->i_ino);
2258 jbd_debug(2, "deleting unreferenced inode %lu\n",
2262 iput(inode); /* The delete magic happens here! */
2265 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2268 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2269 PLURAL(nr_orphans));
2271 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2272 PLURAL(nr_truncates));
2274 /* Turn quotas off */
2275 for (i = 0; i < MAXQUOTAS; i++) {
2276 if (sb_dqopt(sb)->files[i])
2277 dquot_quota_off(sb, i);
2280 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2284 * Maximal extent format file size.
2285 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2286 * extent format containers, within a sector_t, and within i_blocks
2287 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2288 * so that won't be a limiting factor.
2290 * However there is other limiting factor. We do store extents in the form
2291 * of starting block and length, hence the resulting length of the extent
2292 * covering maximum file size must fit into on-disk format containers as
2293 * well. Given that length is always by 1 unit bigger than max unit (because
2294 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2296 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2298 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2301 loff_t upper_limit = MAX_LFS_FILESIZE;
2303 /* small i_blocks in vfs inode? */
2304 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2306 * CONFIG_LBDAF is not enabled implies the inode
2307 * i_block represent total blocks in 512 bytes
2308 * 32 == size of vfs inode i_blocks * 8
2310 upper_limit = (1LL << 32) - 1;
2312 /* total blocks in file system block size */
2313 upper_limit >>= (blkbits - 9);
2314 upper_limit <<= blkbits;
2318 * 32-bit extent-start container, ee_block. We lower the maxbytes
2319 * by one fs block, so ee_len can cover the extent of maximum file
2322 res = (1LL << 32) - 1;
2325 /* Sanity check against vm- & vfs- imposed limits */
2326 if (res > upper_limit)
2333 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2334 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2335 * We need to be 1 filesystem block less than the 2^48 sector limit.
2337 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2339 loff_t res = EXT4_NDIR_BLOCKS;
2342 /* This is calculated to be the largest file size for a dense, block
2343 * mapped file such that the file's total number of 512-byte sectors,
2344 * including data and all indirect blocks, does not exceed (2^48 - 1).
2346 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2347 * number of 512-byte sectors of the file.
2350 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2352 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2353 * the inode i_block field represents total file blocks in
2354 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2356 upper_limit = (1LL << 32) - 1;
2358 /* total blocks in file system block size */
2359 upper_limit >>= (bits - 9);
2363 * We use 48 bit ext4_inode i_blocks
2364 * With EXT4_HUGE_FILE_FL set the i_blocks
2365 * represent total number of blocks in
2366 * file system block size
2368 upper_limit = (1LL << 48) - 1;
2372 /* indirect blocks */
2374 /* double indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2));
2376 /* tripple indirect blocks */
2377 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2379 upper_limit -= meta_blocks;
2380 upper_limit <<= bits;
2382 res += 1LL << (bits-2);
2383 res += 1LL << (2*(bits-2));
2384 res += 1LL << (3*(bits-2));
2386 if (res > upper_limit)
2389 if (res > MAX_LFS_FILESIZE)
2390 res = MAX_LFS_FILESIZE;
2395 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2396 ext4_fsblk_t logical_sb_block, int nr)
2398 struct ext4_sb_info *sbi = EXT4_SB(sb);
2399 ext4_group_t bg, first_meta_bg;
2402 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2404 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2406 return logical_sb_block + nr + 1;
2407 bg = sbi->s_desc_per_block * nr;
2408 if (ext4_bg_has_super(sb, bg))
2411 return (has_super + ext4_group_first_block_no(sb, bg));
2415 * ext4_get_stripe_size: Get the stripe size.
2416 * @sbi: In memory super block info
2418 * If we have specified it via mount option, then
2419 * use the mount option value. If the value specified at mount time is
2420 * greater than the blocks per group use the super block value.
2421 * If the super block value is greater than blocks per group return 0.
2422 * Allocator needs it be less than blocks per group.
2425 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2427 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2428 unsigned long stripe_width =
2429 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2432 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2433 ret = sbi->s_stripe;
2434 else if (stripe_width <= sbi->s_blocks_per_group)
2436 else if (stride <= sbi->s_blocks_per_group)
2442 * If the stripe width is 1, this makes no sense and
2443 * we set it to 0 to turn off stripe handling code.
2454 struct attribute attr;
2455 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2456 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2457 const char *, size_t);
2461 static int parse_strtoul(const char *buf,
2462 unsigned long max, unsigned long *value)
2466 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2467 endp = skip_spaces(endp);
2468 if (*endp || *value > max)
2474 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi,
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2480 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2483 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2484 struct ext4_sb_info *sbi, char *buf)
2486 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2488 if (!sb->s_bdev->bd_part)
2489 return snprintf(buf, PAGE_SIZE, "0\n");
2490 return snprintf(buf, PAGE_SIZE, "%lu\n",
2491 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2492 sbi->s_sectors_written_start) >> 1);
2495 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2498 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2500 if (!sb->s_bdev->bd_part)
2501 return snprintf(buf, PAGE_SIZE, "0\n");
2502 return snprintf(buf, PAGE_SIZE, "%llu\n",
2503 (unsigned long long)(sbi->s_kbytes_written +
2504 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2505 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2508 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2509 struct ext4_sb_info *sbi, char *buf)
2511 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2514 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2515 struct ext4_sb_info *sbi, char *buf)
2517 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2520 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi,
2522 const char *buf, size_t count)
2526 if (parse_strtoul(buf, 0x40000000, &t))
2529 if (t && !is_power_of_2(t))
2532 sbi->s_inode_readahead_blks = t;
2536 static ssize_t sbi_ui_show(struct ext4_attr *a,
2537 struct ext4_sb_info *sbi, char *buf)
2539 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2541 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2544 static ssize_t sbi_ui_store(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi,
2546 const char *buf, size_t count)
2548 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2551 if (parse_strtoul(buf, 0xffffffff, &t))
2557 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2558 static struct ext4_attr ext4_attr_##_name = { \
2559 .attr = {.name = __stringify(_name), .mode = _mode }, \
2562 .offset = offsetof(struct ext4_sb_info, _elname), \
2564 #define EXT4_ATTR(name, mode, show, store) \
2565 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2567 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2568 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2569 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2570 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2571 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2572 #define ATTR_LIST(name) &ext4_attr_##name.attr
2574 EXT4_RO_ATTR(delayed_allocation_blocks);
2575 EXT4_RO_ATTR(session_write_kbytes);
2576 EXT4_RO_ATTR(lifetime_write_kbytes);
2577 EXT4_RO_ATTR(extent_cache_hits);
2578 EXT4_RO_ATTR(extent_cache_misses);
2579 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2580 inode_readahead_blks_store, s_inode_readahead_blks);
2581 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2582 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2583 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2584 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2585 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2586 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2587 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2588 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2590 static struct attribute *ext4_attrs[] = {
2591 ATTR_LIST(delayed_allocation_blocks),
2592 ATTR_LIST(session_write_kbytes),
2593 ATTR_LIST(lifetime_write_kbytes),
2594 ATTR_LIST(extent_cache_hits),
2595 ATTR_LIST(extent_cache_misses),
2596 ATTR_LIST(inode_readahead_blks),
2597 ATTR_LIST(inode_goal),
2598 ATTR_LIST(mb_stats),
2599 ATTR_LIST(mb_max_to_scan),
2600 ATTR_LIST(mb_min_to_scan),
2601 ATTR_LIST(mb_order2_req),
2602 ATTR_LIST(mb_stream_req),
2603 ATTR_LIST(mb_group_prealloc),
2604 ATTR_LIST(max_writeback_mb_bump),
2608 /* Features this copy of ext4 supports */
2609 EXT4_INFO_ATTR(lazy_itable_init);
2610 EXT4_INFO_ATTR(batched_discard);
2612 static struct attribute *ext4_feat_attrs[] = {
2613 ATTR_LIST(lazy_itable_init),
2614 ATTR_LIST(batched_discard),
2618 static ssize_t ext4_attr_show(struct kobject *kobj,
2619 struct attribute *attr, char *buf)
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2623 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2625 return a->show ? a->show(a, sbi, buf) : 0;
2628 static ssize_t ext4_attr_store(struct kobject *kobj,
2629 struct attribute *attr,
2630 const char *buf, size_t len)
2632 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2634 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2636 return a->store ? a->store(a, sbi, buf, len) : 0;
2639 static void ext4_sb_release(struct kobject *kobj)
2641 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2643 complete(&sbi->s_kobj_unregister);
2646 static const struct sysfs_ops ext4_attr_ops = {
2647 .show = ext4_attr_show,
2648 .store = ext4_attr_store,
2651 static struct kobj_type ext4_ktype = {
2652 .default_attrs = ext4_attrs,
2653 .sysfs_ops = &ext4_attr_ops,
2654 .release = ext4_sb_release,
2657 static void ext4_feat_release(struct kobject *kobj)
2659 complete(&ext4_feat->f_kobj_unregister);
2662 static struct kobj_type ext4_feat_ktype = {
2663 .default_attrs = ext4_feat_attrs,
2664 .sysfs_ops = &ext4_attr_ops,
2665 .release = ext4_feat_release,
2669 * Check whether this filesystem can be mounted based on
2670 * the features present and the RDONLY/RDWR mount requested.
2671 * Returns 1 if this filesystem can be mounted as requested,
2672 * 0 if it cannot be.
2674 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2676 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2677 ext4_msg(sb, KERN_ERR,
2678 "Couldn't mount because of "
2679 "unsupported optional features (%x)",
2680 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2681 ~EXT4_FEATURE_INCOMPAT_SUPP));
2688 /* Check that feature set is OK for a read-write mount */
2689 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2690 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2691 "unsupported optional features (%x)",
2692 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2693 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2697 * Large file size enabled file system can only be mounted
2698 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2701 if (sizeof(blkcnt_t) < sizeof(u64)) {
2702 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2703 "cannot be mounted RDWR without "
2708 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2709 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2710 ext4_msg(sb, KERN_ERR,
2711 "Can't support bigalloc feature without "
2712 "extents feature\n");
2719 * This function is called once a day if we have errors logged
2720 * on the file system
2722 static void print_daily_error_info(unsigned long arg)
2724 struct super_block *sb = (struct super_block *) arg;
2725 struct ext4_sb_info *sbi;
2726 struct ext4_super_block *es;
2731 if (es->s_error_count)
2732 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2733 le32_to_cpu(es->s_error_count));
2734 if (es->s_first_error_time) {
2735 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2736 sb->s_id, le32_to_cpu(es->s_first_error_time),
2737 (int) sizeof(es->s_first_error_func),
2738 es->s_first_error_func,
2739 le32_to_cpu(es->s_first_error_line));
2740 if (es->s_first_error_ino)
2741 printk(": inode %u",
2742 le32_to_cpu(es->s_first_error_ino));
2743 if (es->s_first_error_block)
2744 printk(": block %llu", (unsigned long long)
2745 le64_to_cpu(es->s_first_error_block));
2748 if (es->s_last_error_time) {
2749 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2750 sb->s_id, le32_to_cpu(es->s_last_error_time),
2751 (int) sizeof(es->s_last_error_func),
2752 es->s_last_error_func,
2753 le32_to_cpu(es->s_last_error_line));
2754 if (es->s_last_error_ino)
2755 printk(": inode %u",
2756 le32_to_cpu(es->s_last_error_ino));
2757 if (es->s_last_error_block)
2758 printk(": block %llu", (unsigned long long)
2759 le64_to_cpu(es->s_last_error_block));
2762 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2765 /* Find next suitable group and run ext4_init_inode_table */
2766 static int ext4_run_li_request(struct ext4_li_request *elr)
2768 struct ext4_group_desc *gdp = NULL;
2769 ext4_group_t group, ngroups;
2770 struct super_block *sb;
2771 unsigned long timeout = 0;
2775 ngroups = EXT4_SB(sb)->s_groups_count;
2777 for (group = elr->lr_next_group; group < ngroups; group++) {
2778 gdp = ext4_get_group_desc(sb, group, NULL);
2784 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2788 if (group == ngroups)
2793 ret = ext4_init_inode_table(sb, group,
2794 elr->lr_timeout ? 0 : 1);
2795 if (elr->lr_timeout == 0) {
2796 timeout = (jiffies - timeout) *
2797 elr->lr_sbi->s_li_wait_mult;
2798 elr->lr_timeout = timeout;
2800 elr->lr_next_sched = jiffies + elr->lr_timeout;
2801 elr->lr_next_group = group + 1;
2808 * Remove lr_request from the list_request and free the
2809 * request structure. Should be called with li_list_mtx held
2811 static void ext4_remove_li_request(struct ext4_li_request *elr)
2813 struct ext4_sb_info *sbi;
2820 list_del(&elr->lr_request);
2821 sbi->s_li_request = NULL;
2825 static void ext4_unregister_li_request(struct super_block *sb)
2827 mutex_lock(&ext4_li_mtx);
2828 if (!ext4_li_info) {
2829 mutex_unlock(&ext4_li_mtx);
2833 mutex_lock(&ext4_li_info->li_list_mtx);
2834 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2835 mutex_unlock(&ext4_li_info->li_list_mtx);
2836 mutex_unlock(&ext4_li_mtx);
2839 static struct task_struct *ext4_lazyinit_task;
2842 * This is the function where ext4lazyinit thread lives. It walks
2843 * through the request list searching for next scheduled filesystem.
2844 * When such a fs is found, run the lazy initialization request
2845 * (ext4_rn_li_request) and keep track of the time spend in this
2846 * function. Based on that time we compute next schedule time of
2847 * the request. When walking through the list is complete, compute
2848 * next waking time and put itself into sleep.
2850 static int ext4_lazyinit_thread(void *arg)
2852 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2853 struct list_head *pos, *n;
2854 struct ext4_li_request *elr;
2855 unsigned long next_wakeup, cur;
2857 BUG_ON(NULL == eli);
2861 next_wakeup = MAX_JIFFY_OFFSET;
2863 mutex_lock(&eli->li_list_mtx);
2864 if (list_empty(&eli->li_request_list)) {
2865 mutex_unlock(&eli->li_list_mtx);
2869 list_for_each_safe(pos, n, &eli->li_request_list) {
2870 elr = list_entry(pos, struct ext4_li_request,
2873 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2874 if (ext4_run_li_request(elr) != 0) {
2875 /* error, remove the lazy_init job */
2876 ext4_remove_li_request(elr);
2881 if (time_before(elr->lr_next_sched, next_wakeup))
2882 next_wakeup = elr->lr_next_sched;
2884 mutex_unlock(&eli->li_list_mtx);
2889 if ((time_after_eq(cur, next_wakeup)) ||
2890 (MAX_JIFFY_OFFSET == next_wakeup)) {
2895 schedule_timeout_interruptible(next_wakeup - cur);
2897 if (kthread_should_stop()) {
2898 ext4_clear_request_list();
2905 * It looks like the request list is empty, but we need
2906 * to check it under the li_list_mtx lock, to prevent any
2907 * additions into it, and of course we should lock ext4_li_mtx
2908 * to atomically free the list and ext4_li_info, because at
2909 * this point another ext4 filesystem could be registering
2912 mutex_lock(&ext4_li_mtx);
2913 mutex_lock(&eli->li_list_mtx);
2914 if (!list_empty(&eli->li_request_list)) {
2915 mutex_unlock(&eli->li_list_mtx);
2916 mutex_unlock(&ext4_li_mtx);
2919 mutex_unlock(&eli->li_list_mtx);
2920 kfree(ext4_li_info);
2921 ext4_li_info = NULL;
2922 mutex_unlock(&ext4_li_mtx);
2927 static void ext4_clear_request_list(void)
2929 struct list_head *pos, *n;
2930 struct ext4_li_request *elr;
2932 mutex_lock(&ext4_li_info->li_list_mtx);
2933 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2934 elr = list_entry(pos, struct ext4_li_request,
2936 ext4_remove_li_request(elr);
2938 mutex_unlock(&ext4_li_info->li_list_mtx);
2941 static int ext4_run_lazyinit_thread(void)
2943 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2944 ext4_li_info, "ext4lazyinit");
2945 if (IS_ERR(ext4_lazyinit_task)) {
2946 int err = PTR_ERR(ext4_lazyinit_task);
2947 ext4_clear_request_list();
2948 kfree(ext4_li_info);
2949 ext4_li_info = NULL;
2950 printk(KERN_CRIT "EXT4: error %d creating inode table "
2951 "initialization thread\n",
2955 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2960 * Check whether it make sense to run itable init. thread or not.
2961 * If there is at least one uninitialized inode table, return
2962 * corresponding group number, else the loop goes through all
2963 * groups and return total number of groups.
2965 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2967 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2968 struct ext4_group_desc *gdp = NULL;
2970 for (group = 0; group < ngroups; group++) {
2971 gdp = ext4_get_group_desc(sb, group, NULL);
2975 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2982 static int ext4_li_info_new(void)
2984 struct ext4_lazy_init *eli = NULL;
2986 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2990 INIT_LIST_HEAD(&eli->li_request_list);
2991 mutex_init(&eli->li_list_mtx);
2993 eli->li_state |= EXT4_LAZYINIT_QUIT;
3000 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3003 struct ext4_sb_info *sbi = EXT4_SB(sb);
3004 struct ext4_li_request *elr;
3007 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3013 elr->lr_next_group = start;
3016 * Randomize first schedule time of the request to
3017 * spread the inode table initialization requests
3020 get_random_bytes(&rnd, sizeof(rnd));
3021 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3022 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3027 static int ext4_register_li_request(struct super_block *sb,
3028 ext4_group_t first_not_zeroed)
3030 struct ext4_sb_info *sbi = EXT4_SB(sb);
3031 struct ext4_li_request *elr;
3032 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3035 if (sbi->s_li_request != NULL) {
3037 * Reset timeout so it can be computed again, because
3038 * s_li_wait_mult might have changed.
3040 sbi->s_li_request->lr_timeout = 0;
3044 if (first_not_zeroed == ngroups ||
3045 (sb->s_flags & MS_RDONLY) ||
3046 !test_opt(sb, INIT_INODE_TABLE))
3049 elr = ext4_li_request_new(sb, first_not_zeroed);
3053 mutex_lock(&ext4_li_mtx);
3055 if (NULL == ext4_li_info) {
3056 ret = ext4_li_info_new();
3061 mutex_lock(&ext4_li_info->li_list_mtx);
3062 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3063 mutex_unlock(&ext4_li_info->li_list_mtx);
3065 sbi->s_li_request = elr;
3067 * set elr to NULL here since it has been inserted to
3068 * the request_list and the removal and free of it is
3069 * handled by ext4_clear_request_list from now on.
3073 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3074 ret = ext4_run_lazyinit_thread();
3079 mutex_unlock(&ext4_li_mtx);
3086 * We do not need to lock anything since this is called on
3089 static void ext4_destroy_lazyinit_thread(void)
3092 * If thread exited earlier
3093 * there's nothing to be done.
3095 if (!ext4_li_info || !ext4_lazyinit_task)
3098 kthread_stop(ext4_lazyinit_task);
3101 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3103 char *orig_data = kstrdup(data, GFP_KERNEL);
3104 struct buffer_head *bh;
3105 struct ext4_super_block *es = NULL;
3106 struct ext4_sb_info *sbi;
3108 ext4_fsblk_t sb_block = get_sb_block(&data);
3109 ext4_fsblk_t logical_sb_block;
3110 unsigned long offset = 0;
3111 unsigned long journal_devnum = 0;
3112 unsigned long def_mount_opts;
3117 int blocksize, clustersize;
3118 unsigned int db_count;
3120 int needs_recovery, has_huge_files, has_bigalloc;
3123 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3124 ext4_group_t first_not_zeroed;
3126 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3130 sbi->s_blockgroup_lock =
3131 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3132 if (!sbi->s_blockgroup_lock) {
3136 sb->s_fs_info = sbi;
3137 sbi->s_mount_opt = 0;
3138 sbi->s_resuid = EXT4_DEF_RESUID;
3139 sbi->s_resgid = EXT4_DEF_RESGID;
3140 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3141 sbi->s_sb_block = sb_block;
3142 if (sb->s_bdev->bd_part)
3143 sbi->s_sectors_written_start =
3144 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3146 /* Cleanup superblock name */
3147 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3151 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3153 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3158 * The ext4 superblock will not be buffer aligned for other than 1kB
3159 * block sizes. We need to calculate the offset from buffer start.
3161 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3162 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3163 offset = do_div(logical_sb_block, blocksize);
3165 logical_sb_block = sb_block;
3168 if (!(bh = sb_bread(sb, logical_sb_block))) {
3169 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3173 * Note: s_es must be initialized as soon as possible because
3174 * some ext4 macro-instructions depend on its value
3176 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3178 sb->s_magic = le16_to_cpu(es->s_magic);
3179 if (sb->s_magic != EXT4_SUPER_MAGIC)
3181 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3183 /* Set defaults before we parse the mount options */
3184 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3185 set_opt(sb, INIT_INODE_TABLE);
3186 if (def_mount_opts & EXT4_DEFM_DEBUG)
3188 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3189 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3193 if (def_mount_opts & EXT4_DEFM_UID16)
3194 set_opt(sb, NO_UID32);
3195 /* xattr user namespace & acls are now defaulted on */
3196 #ifdef CONFIG_EXT4_FS_XATTR
3197 set_opt(sb, XATTR_USER);
3199 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3200 set_opt(sb, POSIX_ACL);
3202 set_opt(sb, MBLK_IO_SUBMIT);
3203 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3204 set_opt(sb, JOURNAL_DATA);
3205 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3206 set_opt(sb, ORDERED_DATA);
3207 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3208 set_opt(sb, WRITEBACK_DATA);
3210 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3211 set_opt(sb, ERRORS_PANIC);
3212 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3213 set_opt(sb, ERRORS_CONT);
3215 set_opt(sb, ERRORS_RO);
3216 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3217 set_opt(sb, BLOCK_VALIDITY);
3218 if (def_mount_opts & EXT4_DEFM_DISCARD)
3219 set_opt(sb, DISCARD);
3221 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3222 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3223 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3224 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3225 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3227 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3228 set_opt(sb, BARRIER);
3231 * enable delayed allocation by default
3232 * Use -o nodelalloc to turn it off
3234 if (!IS_EXT3_SB(sb) &&
3235 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3236 set_opt(sb, DELALLOC);
3239 * set default s_li_wait_mult for lazyinit, for the case there is
3240 * no mount option specified.
3242 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3244 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3245 &journal_devnum, &journal_ioprio, NULL, 0)) {
3246 ext4_msg(sb, KERN_WARNING,
3247 "failed to parse options in superblock: %s",
3248 sbi->s_es->s_mount_opts);
3250 if (!parse_options((char *) data, sb, &journal_devnum,
3251 &journal_ioprio, NULL, 0))
3254 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3255 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3256 "with data=journal disables delayed "
3257 "allocation and O_DIRECT support!\n");
3258 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3259 ext4_msg(sb, KERN_ERR, "can't mount with "
3260 "both data=journal and delalloc");
3263 if (test_opt(sb, DIOREAD_NOLOCK)) {
3264 ext4_msg(sb, KERN_ERR, "can't mount with "
3265 "both data=journal and delalloc");
3268 if (test_opt(sb, DELALLOC))
3269 clear_opt(sb, DELALLOC);
3272 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3273 if (test_opt(sb, DIOREAD_NOLOCK)) {
3274 if (blocksize < PAGE_SIZE) {
3275 ext4_msg(sb, KERN_ERR, "can't mount with "
3276 "dioread_nolock if block size != PAGE_SIZE");
3281 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3282 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3284 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3285 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3286 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3287 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3288 ext4_msg(sb, KERN_WARNING,
3289 "feature flags set on rev 0 fs, "
3290 "running e2fsck is recommended");
3292 if (IS_EXT2_SB(sb)) {
3293 if (ext2_feature_set_ok(sb))
3294 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3295 "using the ext4 subsystem");
3297 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3298 "to feature incompatibilities");
3303 if (IS_EXT3_SB(sb)) {
3304 if (ext3_feature_set_ok(sb))
3305 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3306 "using the ext4 subsystem");
3308 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3309 "to feature incompatibilities");
3315 * Check feature flags regardless of the revision level, since we
3316 * previously didn't change the revision level when setting the flags,
3317 * so there is a chance incompat flags are set on a rev 0 filesystem.
3319 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3322 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3323 blocksize > EXT4_MAX_BLOCK_SIZE) {
3324 ext4_msg(sb, KERN_ERR,
3325 "Unsupported filesystem blocksize %d", blocksize);
3329 if (sb->s_blocksize != blocksize) {
3330 /* Validate the filesystem blocksize */
3331 if (!sb_set_blocksize(sb, blocksize)) {
3332 ext4_msg(sb, KERN_ERR, "bad block size %d",
3338 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3339 offset = do_div(logical_sb_block, blocksize);
3340 bh = sb_bread(sb, logical_sb_block);
3342 ext4_msg(sb, KERN_ERR,
3343 "Can't read superblock on 2nd try");
3346 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3348 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3349 ext4_msg(sb, KERN_ERR,
3350 "Magic mismatch, very weird!");
3355 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3356 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3357 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3359 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3361 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3362 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3363 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3365 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3366 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3367 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3368 (!is_power_of_2(sbi->s_inode_size)) ||
3369 (sbi->s_inode_size > blocksize)) {
3370 ext4_msg(sb, KERN_ERR,
3371 "unsupported inode size: %d",
3375 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3376 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3379 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3380 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3381 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3382 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3383 !is_power_of_2(sbi->s_desc_size)) {
3384 ext4_msg(sb, KERN_ERR,
3385 "unsupported descriptor size %lu",
3390 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3392 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3393 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3394 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3397 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3398 if (sbi->s_inodes_per_block == 0)
3400 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3401 sbi->s_inodes_per_block;
3402 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3404 sbi->s_mount_state = le16_to_cpu(es->s_state);
3405 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3406 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3408 for (i = 0; i < 4; i++)
3409 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3410 sbi->s_def_hash_version = es->s_def_hash_version;
3411 i = le32_to_cpu(es->s_flags);
3412 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3413 sbi->s_hash_unsigned = 3;
3414 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3415 #ifdef __CHAR_UNSIGNED__
3416 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3417 sbi->s_hash_unsigned = 3;
3419 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3424 /* Handle clustersize */
3425 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3426 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3427 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3429 if (clustersize < blocksize) {
3430 ext4_msg(sb, KERN_ERR,
3431 "cluster size (%d) smaller than "
3432 "block size (%d)", clustersize, blocksize);
3435 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3436 le32_to_cpu(es->s_log_block_size);
3437 sbi->s_clusters_per_group =
3438 le32_to_cpu(es->s_clusters_per_group);
3439 if (sbi->s_clusters_per_group > blocksize * 8) {
3440 ext4_msg(sb, KERN_ERR,
3441 "#clusters per group too big: %lu",
3442 sbi->s_clusters_per_group);
3445 if (sbi->s_blocks_per_group !=
3446 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3447 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3448 "clusters per group (%lu) inconsistent",
3449 sbi->s_blocks_per_group,
3450 sbi->s_clusters_per_group);
3454 if (clustersize != blocksize) {
3455 ext4_warning(sb, "fragment/cluster size (%d) != "
3456 "block size (%d)", clustersize,
3458 clustersize = blocksize;
3460 if (sbi->s_blocks_per_group > blocksize * 8) {
3461 ext4_msg(sb, KERN_ERR,
3462 "#blocks per group too big: %lu",
3463 sbi->s_blocks_per_group);
3466 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3467 sbi->s_cluster_bits = 0;
3469 sbi->s_cluster_ratio = clustersize / blocksize;
3471 if (sbi->s_inodes_per_group > blocksize * 8) {
3472 ext4_msg(sb, KERN_ERR,
3473 "#inodes per group too big: %lu",
3474 sbi->s_inodes_per_group);
3479 * Test whether we have more sectors than will fit in sector_t,
3480 * and whether the max offset is addressable by the page cache.
3482 err = generic_check_addressable(sb->s_blocksize_bits,
3483 ext4_blocks_count(es));
3485 ext4_msg(sb, KERN_ERR, "filesystem"
3486 " too large to mount safely on this system");
3487 if (sizeof(sector_t) < 8)
3488 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3493 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3496 /* check blocks count against device size */
3497 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3498 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3499 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3500 "exceeds size of device (%llu blocks)",
3501 ext4_blocks_count(es), blocks_count);
3506 * It makes no sense for the first data block to be beyond the end
3507 * of the filesystem.
3509 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3510 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3511 "block %u is beyond end of filesystem (%llu)",
3512 le32_to_cpu(es->s_first_data_block),
3513 ext4_blocks_count(es));
3516 blocks_count = (ext4_blocks_count(es) -
3517 le32_to_cpu(es->s_first_data_block) +
3518 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3519 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3520 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3521 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3522 "(block count %llu, first data block %u, "
3523 "blocks per group %lu)", sbi->s_groups_count,
3524 ext4_blocks_count(es),
3525 le32_to_cpu(es->s_first_data_block),
3526 EXT4_BLOCKS_PER_GROUP(sb));
3529 sbi->s_groups_count = blocks_count;
3530 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3531 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3532 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3533 EXT4_DESC_PER_BLOCK(sb);
3534 sbi->s_group_desc = ext4_kvmalloc(db_count *
3535 sizeof(struct buffer_head *),
3537 if (sbi->s_group_desc == NULL) {
3538 ext4_msg(sb, KERN_ERR, "not enough memory");
3543 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3545 bgl_lock_init(sbi->s_blockgroup_lock);
3547 for (i = 0; i < db_count; i++) {
3548 block = descriptor_loc(sb, logical_sb_block, i);
3549 sbi->s_group_desc[i] = sb_bread(sb, block);
3550 if (!sbi->s_group_desc[i]) {
3551 ext4_msg(sb, KERN_ERR,
3552 "can't read group descriptor %d", i);
3557 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3558 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3561 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3562 if (!ext4_fill_flex_info(sb)) {
3563 ext4_msg(sb, KERN_ERR,
3564 "unable to initialize "
3565 "flex_bg meta info!");
3569 sbi->s_gdb_count = db_count;
3570 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3571 spin_lock_init(&sbi->s_next_gen_lock);
3573 init_timer(&sbi->s_err_report);
3574 sbi->s_err_report.function = print_daily_error_info;
3575 sbi->s_err_report.data = (unsigned long) sb;
3577 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3578 ext4_count_free_clusters(sb));
3580 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3581 ext4_count_free_inodes(sb));
3584 err = percpu_counter_init(&sbi->s_dirs_counter,
3585 ext4_count_dirs(sb));
3588 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3591 ext4_msg(sb, KERN_ERR, "insufficient memory");
3595 sbi->s_stripe = ext4_get_stripe_size(sbi);
3596 sbi->s_max_writeback_mb_bump = 128;
3599 * set up enough so that it can read an inode
3601 if (!test_opt(sb, NOLOAD) &&
3602 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3603 sb->s_op = &ext4_sops;
3605 sb->s_op = &ext4_nojournal_sops;
3606 sb->s_export_op = &ext4_export_ops;
3607 sb->s_xattr = ext4_xattr_handlers;
3609 sb->s_qcop = &ext4_qctl_operations;
3610 sb->dq_op = &ext4_quota_operations;
3612 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3614 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3615 mutex_init(&sbi->s_orphan_lock);
3616 sbi->s_resize_flags = 0;
3620 needs_recovery = (es->s_last_orphan != 0 ||
3621 EXT4_HAS_INCOMPAT_FEATURE(sb,
3622 EXT4_FEATURE_INCOMPAT_RECOVER));
3624 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3625 !(sb->s_flags & MS_RDONLY))
3626 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3630 * The first inode we look at is the journal inode. Don't try
3631 * root first: it may be modified in the journal!
3633 if (!test_opt(sb, NOLOAD) &&
3634 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3635 if (ext4_load_journal(sb, es, journal_devnum))
3637 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3638 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3639 ext4_msg(sb, KERN_ERR, "required journal recovery "
3640 "suppressed and not mounted read-only");
3641 goto failed_mount_wq;
3643 clear_opt(sb, DATA_FLAGS);
3644 sbi->s_journal = NULL;
3649 if (ext4_blocks_count(es) > 0xffffffffULL &&
3650 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3651 JBD2_FEATURE_INCOMPAT_64BIT)) {
3652 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3653 goto failed_mount_wq;
3656 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3657 jbd2_journal_set_features(sbi->s_journal,
3658 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3659 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3660 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3661 jbd2_journal_set_features(sbi->s_journal,
3662 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3663 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3664 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3666 jbd2_journal_clear_features(sbi->s_journal,
3667 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3668 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3671 /* We have now updated the journal if required, so we can
3672 * validate the data journaling mode. */
3673 switch (test_opt(sb, DATA_FLAGS)) {
3675 /* No mode set, assume a default based on the journal
3676 * capabilities: ORDERED_DATA if the journal can
3677 * cope, else JOURNAL_DATA
3679 if (jbd2_journal_check_available_features
3680 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3681 set_opt(sb, ORDERED_DATA);
3683 set_opt(sb, JOURNAL_DATA);
3686 case EXT4_MOUNT_ORDERED_DATA:
3687 case EXT4_MOUNT_WRITEBACK_DATA:
3688 if (!jbd2_journal_check_available_features
3689 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3690 ext4_msg(sb, KERN_ERR, "Journal does not support "
3691 "requested data journaling mode");
3692 goto failed_mount_wq;
3697 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3700 * The journal may have updated the bg summary counts, so we
3701 * need to update the global counters.
3703 percpu_counter_set(&sbi->s_freeclusters_counter,
3704 ext4_count_free_clusters(sb));
3705 percpu_counter_set(&sbi->s_freeinodes_counter,
3706 ext4_count_free_inodes(sb));
3707 percpu_counter_set(&sbi->s_dirs_counter,
3708 ext4_count_dirs(sb));
3709 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3713 * The maximum number of concurrent works can be high and
3714 * concurrency isn't really necessary. Limit it to 1.
3716 EXT4_SB(sb)->dio_unwritten_wq =
3717 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3718 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3719 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3720 goto failed_mount_wq;
3724 * The jbd2_journal_load will have done any necessary log recovery,
3725 * so we can safely mount the rest of the filesystem now.
3728 root = ext4_iget(sb, EXT4_ROOT_INO);
3730 ext4_msg(sb, KERN_ERR, "get root inode failed");
3731 ret = PTR_ERR(root);
3735 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3736 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3739 sb->s_root = d_alloc_root(root);
3741 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3746 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3748 /* determine the minimum size of new large inodes, if present */
3749 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3750 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3751 EXT4_GOOD_OLD_INODE_SIZE;
3752 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3753 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3754 if (sbi->s_want_extra_isize <
3755 le16_to_cpu(es->s_want_extra_isize))
3756 sbi->s_want_extra_isize =
3757 le16_to_cpu(es->s_want_extra_isize);
3758 if (sbi->s_want_extra_isize <
3759 le16_to_cpu(es->s_min_extra_isize))
3760 sbi->s_want_extra_isize =
3761 le16_to_cpu(es->s_min_extra_isize);
3764 /* Check if enough inode space is available */
3765 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3766 sbi->s_inode_size) {
3767 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3768 EXT4_GOOD_OLD_INODE_SIZE;
3769 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3773 err = ext4_setup_system_zone(sb);
3775 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3781 err = ext4_mb_init(sb, needs_recovery);
3783 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3788 err = ext4_register_li_request(sb, first_not_zeroed);
3792 sbi->s_kobj.kset = ext4_kset;
3793 init_completion(&sbi->s_kobj_unregister);
3794 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3799 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3800 ext4_orphan_cleanup(sb, es);
3801 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3802 if (needs_recovery) {
3803 ext4_msg(sb, KERN_INFO, "recovery complete");
3804 ext4_mark_recovery_complete(sb, es);
3806 if (EXT4_SB(sb)->s_journal) {
3807 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3808 descr = " journalled data mode";
3809 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3810 descr = " ordered data mode";
3812 descr = " writeback data mode";
3814 descr = "out journal";
3816 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3817 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3818 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3820 if (es->s_error_count)
3821 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3828 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3832 ext4_unregister_li_request(sb);
3834 ext4_ext_release(sb);
3836 ext4_mb_release(sb);
3837 ext4_release_system_zone(sb);
3841 ext4_msg(sb, KERN_ERR, "mount failed");
3842 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3844 if (sbi->s_journal) {
3845 jbd2_journal_destroy(sbi->s_journal);
3846 sbi->s_journal = NULL;
3849 del_timer(&sbi->s_err_report);
3850 if (sbi->s_flex_groups)
3851 ext4_kvfree(sbi->s_flex_groups);
3852 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3853 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3854 percpu_counter_destroy(&sbi->s_dirs_counter);
3855 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3857 kthread_stop(sbi->s_mmp_tsk);
3859 for (i = 0; i < db_count; i++)
3860 brelse(sbi->s_group_desc[i]);
3861 ext4_kvfree(sbi->s_group_desc);
3864 remove_proc_entry(sb->s_id, ext4_proc_root);
3867 for (i = 0; i < MAXQUOTAS; i++)
3868 kfree(sbi->s_qf_names[i]);
3870 ext4_blkdev_remove(sbi);
3873 sb->s_fs_info = NULL;
3874 kfree(sbi->s_blockgroup_lock);
3882 * Setup any per-fs journal parameters now. We'll do this both on
3883 * initial mount, once the journal has been initialised but before we've
3884 * done any recovery; and again on any subsequent remount.
3886 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3888 struct ext4_sb_info *sbi = EXT4_SB(sb);
3890 journal->j_commit_interval = sbi->s_commit_interval;
3891 journal->j_min_batch_time = sbi->s_min_batch_time;
3892 journal->j_max_batch_time = sbi->s_max_batch_time;
3894 write_lock(&journal->j_state_lock);
3895 if (test_opt(sb, BARRIER))
3896 journal->j_flags |= JBD2_BARRIER;
3898 journal->j_flags &= ~JBD2_BARRIER;
3899 if (test_opt(sb, DATA_ERR_ABORT))
3900 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3902 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3903 write_unlock(&journal->j_state_lock);
3906 static journal_t *ext4_get_journal(struct super_block *sb,
3907 unsigned int journal_inum)
3909 struct inode *journal_inode;
3912 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3914 /* First, test for the existence of a valid inode on disk. Bad
3915 * things happen if we iget() an unused inode, as the subsequent
3916 * iput() will try to delete it. */
3918 journal_inode = ext4_iget(sb, journal_inum);
3919 if (IS_ERR(journal_inode)) {
3920 ext4_msg(sb, KERN_ERR, "no journal found");
3923 if (!journal_inode->i_nlink) {
3924 make_bad_inode(journal_inode);
3925 iput(journal_inode);
3926 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3930 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3931 journal_inode, journal_inode->i_size);
3932 if (!S_ISREG(journal_inode->i_mode)) {
3933 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3934 iput(journal_inode);
3938 journal = jbd2_journal_init_inode(journal_inode);
3940 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3941 iput(journal_inode);
3944 journal->j_private = sb;
3945 ext4_init_journal_params(sb, journal);
3949 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3952 struct buffer_head *bh;
3956 int hblock, blocksize;
3957 ext4_fsblk_t sb_block;
3958 unsigned long offset;
3959 struct ext4_super_block *es;
3960 struct block_device *bdev;
3962 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3964 bdev = ext4_blkdev_get(j_dev, sb);
3968 blocksize = sb->s_blocksize;
3969 hblock = bdev_logical_block_size(bdev);
3970 if (blocksize < hblock) {
3971 ext4_msg(sb, KERN_ERR,
3972 "blocksize too small for journal device");
3976 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3977 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3978 set_blocksize(bdev, blocksize);
3979 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3980 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3981 "external journal");
3985 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3986 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3987 !(le32_to_cpu(es->s_feature_incompat) &
3988 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3989 ext4_msg(sb, KERN_ERR, "external journal has "
3995 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3996 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4001 len = ext4_blocks_count(es);
4002 start = sb_block + 1;
4003 brelse(bh); /* we're done with the superblock */
4005 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4006 start, len, blocksize);
4008 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4011 journal->j_private = sb;
4012 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4013 wait_on_buffer(journal->j_sb_buffer);
4014 if (!buffer_uptodate(journal->j_sb_buffer)) {
4015 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4018 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4019 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4020 "user (unsupported) - %d",
4021 be32_to_cpu(journal->j_superblock->s_nr_users));
4024 EXT4_SB(sb)->journal_bdev = bdev;
4025 ext4_init_journal_params(sb, journal);
4029 jbd2_journal_destroy(journal);
4031 ext4_blkdev_put(bdev);
4035 static int ext4_load_journal(struct super_block *sb,
4036 struct ext4_super_block *es,
4037 unsigned long journal_devnum)
4040 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4043 int really_read_only;
4045 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4047 if (journal_devnum &&
4048 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4049 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4050 "numbers have changed");
4051 journal_dev = new_decode_dev(journal_devnum);
4053 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4055 really_read_only = bdev_read_only(sb->s_bdev);
4058 * Are we loading a blank journal or performing recovery after a
4059 * crash? For recovery, we need to check in advance whether we
4060 * can get read-write access to the device.
4062 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4063 if (sb->s_flags & MS_RDONLY) {
4064 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4065 "required on readonly filesystem");
4066 if (really_read_only) {
4067 ext4_msg(sb, KERN_ERR, "write access "
4068 "unavailable, cannot proceed");
4071 ext4_msg(sb, KERN_INFO, "write access will "
4072 "be enabled during recovery");
4076 if (journal_inum && journal_dev) {
4077 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4078 "and inode journals!");
4083 if (!(journal = ext4_get_journal(sb, journal_inum)))
4086 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4090 if (!(journal->j_flags & JBD2_BARRIER))
4091 ext4_msg(sb, KERN_INFO, "barriers disabled");
4093 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4094 err = jbd2_journal_update_format(journal);
4096 ext4_msg(sb, KERN_ERR, "error updating journal");
4097 jbd2_journal_destroy(journal);
4102 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4103 err = jbd2_journal_wipe(journal, !really_read_only);
4105 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4107 memcpy(save, ((char *) es) +
4108 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4109 err = jbd2_journal_load(journal);
4111 memcpy(((char *) es) + EXT4_S_ERR_START,
4112 save, EXT4_S_ERR_LEN);
4117 ext4_msg(sb, KERN_ERR, "error loading journal");
4118 jbd2_journal_destroy(journal);
4122 EXT4_SB(sb)->s_journal = journal;
4123 ext4_clear_journal_err(sb, es);
4125 if (!really_read_only && journal_devnum &&
4126 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4127 es->s_journal_dev = cpu_to_le32(journal_devnum);
4129 /* Make sure we flush the recovery flag to disk. */
4130 ext4_commit_super(sb, 1);
4136 static int ext4_commit_super(struct super_block *sb, int sync)
4138 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4139 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4142 if (!sbh || block_device_ejected(sb))
4144 if (buffer_write_io_error(sbh)) {
4146 * Oh, dear. A previous attempt to write the
4147 * superblock failed. This could happen because the
4148 * USB device was yanked out. Or it could happen to
4149 * be a transient write error and maybe the block will
4150 * be remapped. Nothing we can do but to retry the
4151 * write and hope for the best.
4153 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4154 "superblock detected");
4155 clear_buffer_write_io_error(sbh);
4156 set_buffer_uptodate(sbh);
4159 * If the file system is mounted read-only, don't update the
4160 * superblock write time. This avoids updating the superblock
4161 * write time when we are mounting the root file system
4162 * read/only but we need to replay the journal; at that point,
4163 * for people who are east of GMT and who make their clock
4164 * tick in localtime for Windows bug-for-bug compatibility,
4165 * the clock is set in the future, and this will cause e2fsck
4166 * to complain and force a full file system check.
4168 if (!(sb->s_flags & MS_RDONLY))
4169 es->s_wtime = cpu_to_le32(get_seconds());
4170 if (sb->s_bdev->bd_part)
4171 es->s_kbytes_written =
4172 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4176 es->s_kbytes_written =
4177 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4178 ext4_free_blocks_count_set(es,
4179 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4180 &EXT4_SB(sb)->s_freeclusters_counter)));
4181 es->s_free_inodes_count =
4182 cpu_to_le32(percpu_counter_sum_positive(
4183 &EXT4_SB(sb)->s_freeinodes_counter));
4185 BUFFER_TRACE(sbh, "marking dirty");
4186 mark_buffer_dirty(sbh);
4188 error = sync_dirty_buffer(sbh);
4192 error = buffer_write_io_error(sbh);
4194 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4196 clear_buffer_write_io_error(sbh);
4197 set_buffer_uptodate(sbh);
4204 * Have we just finished recovery? If so, and if we are mounting (or
4205 * remounting) the filesystem readonly, then we will end up with a
4206 * consistent fs on disk. Record that fact.
4208 static void ext4_mark_recovery_complete(struct super_block *sb,
4209 struct ext4_super_block *es)
4211 journal_t *journal = EXT4_SB(sb)->s_journal;
4213 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4214 BUG_ON(journal != NULL);
4217 jbd2_journal_lock_updates(journal);
4218 if (jbd2_journal_flush(journal) < 0)
4221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4222 sb->s_flags & MS_RDONLY) {
4223 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4224 ext4_commit_super(sb, 1);
4228 jbd2_journal_unlock_updates(journal);
4232 * If we are mounting (or read-write remounting) a filesystem whose journal
4233 * has recorded an error from a previous lifetime, move that error to the
4234 * main filesystem now.
4236 static void ext4_clear_journal_err(struct super_block *sb,
4237 struct ext4_super_block *es)
4243 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4245 journal = EXT4_SB(sb)->s_journal;
4248 * Now check for any error status which may have been recorded in the
4249 * journal by a prior ext4_error() or ext4_abort()
4252 j_errno = jbd2_journal_errno(journal);
4256 errstr = ext4_decode_error(sb, j_errno, nbuf);
4257 ext4_warning(sb, "Filesystem error recorded "
4258 "from previous mount: %s", errstr);
4259 ext4_warning(sb, "Marking fs in need of filesystem check.");
4261 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4262 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4263 ext4_commit_super(sb, 1);
4265 jbd2_journal_clear_err(journal);
4270 * Force the running and committing transactions to commit,
4271 * and wait on the commit.
4273 int ext4_force_commit(struct super_block *sb)
4278 if (sb->s_flags & MS_RDONLY)
4281 journal = EXT4_SB(sb)->s_journal;
4283 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4284 ret = ext4_journal_force_commit(journal);
4290 static void ext4_write_super(struct super_block *sb)
4293 ext4_commit_super(sb, 1);
4297 static int ext4_sync_fs(struct super_block *sb, int wait)
4301 struct ext4_sb_info *sbi = EXT4_SB(sb);
4303 trace_ext4_sync_fs(sb, wait);
4304 flush_workqueue(sbi->dio_unwritten_wq);
4305 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4307 jbd2_log_wait_commit(sbi->s_journal, target);
4313 * LVM calls this function before a (read-only) snapshot is created. This
4314 * gives us a chance to flush the journal completely and mark the fs clean.
4316 * Note that only this function cannot bring a filesystem to be in a clean
4317 * state independently, because ext4 prevents a new handle from being started
4318 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4321 static int ext4_freeze(struct super_block *sb)
4326 if (sb->s_flags & MS_RDONLY)
4329 journal = EXT4_SB(sb)->s_journal;
4331 /* Now we set up the journal barrier. */
4332 jbd2_journal_lock_updates(journal);
4335 * Don't clear the needs_recovery flag if we failed to flush
4338 error = jbd2_journal_flush(journal);
4342 /* Journal blocked and flushed, clear needs_recovery flag. */
4343 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4344 error = ext4_commit_super(sb, 1);
4346 /* we rely on s_frozen to stop further updates */
4347 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4352 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4353 * flag here, even though the filesystem is not technically dirty yet.
4355 static int ext4_unfreeze(struct super_block *sb)
4357 if (sb->s_flags & MS_RDONLY)
4361 /* Reset the needs_recovery flag before the fs is unlocked. */
4362 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4363 ext4_commit_super(sb, 1);
4369 * Structure to save mount options for ext4_remount's benefit
4371 struct ext4_mount_options {
4372 unsigned long s_mount_opt;
4373 unsigned long s_mount_opt2;
4376 unsigned long s_commit_interval;
4377 u32 s_min_batch_time, s_max_batch_time;
4380 char *s_qf_names[MAXQUOTAS];
4384 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4386 struct ext4_super_block *es;
4387 struct ext4_sb_info *sbi = EXT4_SB(sb);
4388 ext4_fsblk_t n_blocks_count = 0;
4389 unsigned long old_sb_flags;
4390 struct ext4_mount_options old_opts;
4391 int enable_quota = 0;
4393 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4398 char *orig_data = kstrdup(data, GFP_KERNEL);
4400 /* Store the original options */
4402 old_sb_flags = sb->s_flags;
4403 old_opts.s_mount_opt = sbi->s_mount_opt;
4404 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4405 old_opts.s_resuid = sbi->s_resuid;
4406 old_opts.s_resgid = sbi->s_resgid;
4407 old_opts.s_commit_interval = sbi->s_commit_interval;
4408 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4409 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4411 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4412 for (i = 0; i < MAXQUOTAS; i++)
4413 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4415 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4416 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4419 * Allow the "check" option to be passed as a remount option.
4421 if (!parse_options(data, sb, NULL, &journal_ioprio,
4422 &n_blocks_count, 1)) {
4427 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4428 ext4_abort(sb, "Abort forced by user");
4430 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4431 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4435 if (sbi->s_journal) {
4436 ext4_init_journal_params(sb, sbi->s_journal);
4437 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4440 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4441 n_blocks_count > ext4_blocks_count(es)) {
4442 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4447 if (*flags & MS_RDONLY) {
4448 err = dquot_suspend(sb, -1);
4453 * First of all, the unconditional stuff we have to do
4454 * to disable replay of the journal when we next remount
4456 sb->s_flags |= MS_RDONLY;
4459 * OK, test if we are remounting a valid rw partition
4460 * readonly, and if so set the rdonly flag and then
4461 * mark the partition as valid again.
4463 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4464 (sbi->s_mount_state & EXT4_VALID_FS))
4465 es->s_state = cpu_to_le16(sbi->s_mount_state);
4468 ext4_mark_recovery_complete(sb, es);
4470 /* Make sure we can mount this feature set readwrite */
4471 if (!ext4_feature_set_ok(sb, 0)) {
4476 * Make sure the group descriptor checksums
4477 * are sane. If they aren't, refuse to remount r/w.
4479 for (g = 0; g < sbi->s_groups_count; g++) {
4480 struct ext4_group_desc *gdp =
4481 ext4_get_group_desc(sb, g, NULL);
4483 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4484 ext4_msg(sb, KERN_ERR,
4485 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4486 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4487 le16_to_cpu(gdp->bg_checksum));
4494 * If we have an unprocessed orphan list hanging
4495 * around from a previously readonly bdev mount,
4496 * require a full umount/remount for now.
4498 if (es->s_last_orphan) {
4499 ext4_msg(sb, KERN_WARNING, "Couldn't "
4500 "remount RDWR because of unprocessed "
4501 "orphan inode list. Please "
4502 "umount/remount instead");
4508 * Mounting a RDONLY partition read-write, so reread
4509 * and store the current valid flag. (It may have
4510 * been changed by e2fsck since we originally mounted
4514 ext4_clear_journal_err(sb, es);
4515 sbi->s_mount_state = le16_to_cpu(es->s_state);
4516 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4518 if (!ext4_setup_super(sb, es, 0))
4519 sb->s_flags &= ~MS_RDONLY;
4520 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4521 EXT4_FEATURE_INCOMPAT_MMP))
4522 if (ext4_multi_mount_protect(sb,
4523 le64_to_cpu(es->s_mmp_block))) {
4532 * Reinitialize lazy itable initialization thread based on
4535 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4536 ext4_unregister_li_request(sb);
4538 ext4_group_t first_not_zeroed;
4539 first_not_zeroed = ext4_has_uninit_itable(sb);
4540 ext4_register_li_request(sb, first_not_zeroed);
4543 ext4_setup_system_zone(sb);
4544 if (sbi->s_journal == NULL)
4545 ext4_commit_super(sb, 1);
4548 /* Release old quota file names */
4549 for (i = 0; i < MAXQUOTAS; i++)
4550 if (old_opts.s_qf_names[i] &&
4551 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4552 kfree(old_opts.s_qf_names[i]);
4556 dquot_resume(sb, -1);
4558 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4563 sb->s_flags = old_sb_flags;
4564 sbi->s_mount_opt = old_opts.s_mount_opt;
4565 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4566 sbi->s_resuid = old_opts.s_resuid;
4567 sbi->s_resgid = old_opts.s_resgid;
4568 sbi->s_commit_interval = old_opts.s_commit_interval;
4569 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4570 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4572 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4573 for (i = 0; i < MAXQUOTAS; i++) {
4574 if (sbi->s_qf_names[i] &&
4575 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4576 kfree(sbi->s_qf_names[i]);
4577 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4586 * Note: calculating the overhead so we can be compatible with
4587 * historical BSD practice is quite difficult in the face of
4588 * clusters/bigalloc. This is because multiple metadata blocks from
4589 * different block group can end up in the same allocation cluster.
4590 * Calculating the exact overhead in the face of clustered allocation
4591 * requires either O(all block bitmaps) in memory or O(number of block
4592 * groups**2) in time. We will still calculate the superblock for
4593 * older file systems --- and if we come across with a bigalloc file
4594 * system with zero in s_overhead_clusters the estimate will be close to
4595 * correct especially for very large cluster sizes --- but for newer
4596 * file systems, it's better to calculate this figure once at mkfs
4597 * time, and store it in the superblock. If the superblock value is
4598 * present (even for non-bigalloc file systems), we will use it.
4600 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4602 struct super_block *sb = dentry->d_sb;
4603 struct ext4_sb_info *sbi = EXT4_SB(sb);
4604 struct ext4_super_block *es = sbi->s_es;
4605 struct ext4_group_desc *gdp;
4609 if (test_opt(sb, MINIX_DF)) {
4610 sbi->s_overhead_last = 0;
4611 } else if (es->s_overhead_clusters) {
4612 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4613 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4614 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4615 ext4_fsblk_t overhead = 0;
4618 * Compute the overhead (FS structures). This is constant
4619 * for a given filesystem unless the number of block groups
4620 * changes so we cache the previous value until it does.
4624 * All of the blocks before first_data_block are
4627 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4630 * Add the overhead found in each block group
4632 for (i = 0; i < ngroups; i++) {
4633 gdp = ext4_get_group_desc(sb, i, NULL);
4634 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4637 sbi->s_overhead_last = overhead;
4639 sbi->s_blocks_last = ext4_blocks_count(es);
4642 buf->f_type = EXT4_SUPER_MAGIC;
4643 buf->f_bsize = sb->s_blocksize;
4644 buf->f_blocks = (ext4_blocks_count(es) -
4645 EXT4_C2B(sbi, sbi->s_overhead_last));
4646 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4647 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4648 /* prevent underflow in case that few free space is available */
4649 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4650 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4651 if (buf->f_bfree < ext4_r_blocks_count(es))
4653 buf->f_files = le32_to_cpu(es->s_inodes_count);
4654 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4655 buf->f_namelen = EXT4_NAME_LEN;
4656 fsid = le64_to_cpup((void *)es->s_uuid) ^
4657 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4658 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4659 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4664 /* Helper function for writing quotas on sync - we need to start transaction
4665 * before quota file is locked for write. Otherwise the are possible deadlocks:
4666 * Process 1 Process 2
4667 * ext4_create() quota_sync()
4668 * jbd2_journal_start() write_dquot()
4669 * dquot_initialize() down(dqio_mutex)
4670 * down(dqio_mutex) jbd2_journal_start()
4676 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4678 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4681 static int ext4_write_dquot(struct dquot *dquot)
4685 struct inode *inode;
4687 inode = dquot_to_inode(dquot);
4688 handle = ext4_journal_start(inode,
4689 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4691 return PTR_ERR(handle);
4692 ret = dquot_commit(dquot);
4693 err = ext4_journal_stop(handle);
4699 static int ext4_acquire_dquot(struct dquot *dquot)
4704 handle = ext4_journal_start(dquot_to_inode(dquot),
4705 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4707 return PTR_ERR(handle);
4708 ret = dquot_acquire(dquot);
4709 err = ext4_journal_stop(handle);
4715 static int ext4_release_dquot(struct dquot *dquot)
4720 handle = ext4_journal_start(dquot_to_inode(dquot),
4721 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4722 if (IS_ERR(handle)) {
4723 /* Release dquot anyway to avoid endless cycle in dqput() */
4724 dquot_release(dquot);
4725 return PTR_ERR(handle);
4727 ret = dquot_release(dquot);
4728 err = ext4_journal_stop(handle);
4734 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4736 /* Are we journaling quotas? */
4737 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4738 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4739 dquot_mark_dquot_dirty(dquot);
4740 return ext4_write_dquot(dquot);
4742 return dquot_mark_dquot_dirty(dquot);
4746 static int ext4_write_info(struct super_block *sb, int type)
4751 /* Data block + inode block */
4752 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4754 return PTR_ERR(handle);
4755 ret = dquot_commit_info(sb, type);
4756 err = ext4_journal_stop(handle);
4763 * Turn on quotas during mount time - we need to find
4764 * the quota file and such...
4766 static int ext4_quota_on_mount(struct super_block *sb, int type)
4768 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4769 EXT4_SB(sb)->s_jquota_fmt, type);
4773 * Standard function to be called on quota_on
4775 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4780 if (!test_opt(sb, QUOTA))
4783 /* Quotafile not on the same filesystem? */
4784 if (path->mnt->mnt_sb != sb)
4786 /* Journaling quota? */
4787 if (EXT4_SB(sb)->s_qf_names[type]) {
4788 /* Quotafile not in fs root? */
4789 if (path->dentry->d_parent != sb->s_root)
4790 ext4_msg(sb, KERN_WARNING,
4791 "Quota file not on filesystem root. "
4792 "Journaled quota will not work");
4796 * When we journal data on quota file, we have to flush journal to see
4797 * all updates to the file when we bypass pagecache...
4799 if (EXT4_SB(sb)->s_journal &&
4800 ext4_should_journal_data(path->dentry->d_inode)) {
4802 * We don't need to lock updates but journal_flush() could
4803 * otherwise be livelocked...
4805 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4806 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4807 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4812 return dquot_quota_on(sb, type, format_id, path);
4815 static int ext4_quota_off(struct super_block *sb, int type)
4817 struct inode *inode = sb_dqopt(sb)->files[type];
4820 /* Force all delayed allocation blocks to be allocated.
4821 * Caller already holds s_umount sem */
4822 if (test_opt(sb, DELALLOC))
4823 sync_filesystem(sb);
4828 /* Update modification times of quota files when userspace can
4829 * start looking at them */
4830 handle = ext4_journal_start(inode, 1);
4833 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4834 ext4_mark_inode_dirty(handle, inode);
4835 ext4_journal_stop(handle);
4838 return dquot_quota_off(sb, type);
4841 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4842 * acquiring the locks... As quota files are never truncated and quota code
4843 * itself serializes the operations (and no one else should touch the files)
4844 * we don't have to be afraid of races */
4845 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4846 size_t len, loff_t off)
4848 struct inode *inode = sb_dqopt(sb)->files[type];
4849 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4851 int offset = off & (sb->s_blocksize - 1);
4854 struct buffer_head *bh;
4855 loff_t i_size = i_size_read(inode);
4859 if (off+len > i_size)
4862 while (toread > 0) {
4863 tocopy = sb->s_blocksize - offset < toread ?
4864 sb->s_blocksize - offset : toread;
4865 bh = ext4_bread(NULL, inode, blk, 0, &err);
4868 if (!bh) /* A hole? */
4869 memset(data, 0, tocopy);
4871 memcpy(data, bh->b_data+offset, tocopy);
4881 /* Write to quotafile (we know the transaction is already started and has
4882 * enough credits) */
4883 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4884 const char *data, size_t len, loff_t off)
4886 struct inode *inode = sb_dqopt(sb)->files[type];
4887 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4889 int offset = off & (sb->s_blocksize - 1);
4890 struct buffer_head *bh;
4891 handle_t *handle = journal_current_handle();
4893 if (EXT4_SB(sb)->s_journal && !handle) {
4894 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4895 " cancelled because transaction is not started",
4896 (unsigned long long)off, (unsigned long long)len);
4900 * Since we account only one data block in transaction credits,
4901 * then it is impossible to cross a block boundary.
4903 if (sb->s_blocksize - offset < len) {
4904 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4905 " cancelled because not block aligned",
4906 (unsigned long long)off, (unsigned long long)len);
4910 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4911 bh = ext4_bread(handle, inode, blk, 1, &err);
4914 err = ext4_journal_get_write_access(handle, bh);
4920 memcpy(bh->b_data+offset, data, len);
4921 flush_dcache_page(bh->b_page);
4923 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4927 mutex_unlock(&inode->i_mutex);
4930 if (inode->i_size < off + len) {
4931 i_size_write(inode, off + len);
4932 EXT4_I(inode)->i_disksize = inode->i_size;
4933 ext4_mark_inode_dirty(handle, inode);
4935 mutex_unlock(&inode->i_mutex);
4941 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4942 const char *dev_name, void *data)
4944 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4947 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4948 static inline void register_as_ext2(void)
4950 int err = register_filesystem(&ext2_fs_type);
4953 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4956 static inline void unregister_as_ext2(void)
4958 unregister_filesystem(&ext2_fs_type);
4961 static inline int ext2_feature_set_ok(struct super_block *sb)
4963 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4965 if (sb->s_flags & MS_RDONLY)
4967 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4971 MODULE_ALIAS("ext2");
4973 static inline void register_as_ext2(void) { }
4974 static inline void unregister_as_ext2(void) { }
4975 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4978 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4979 static inline void register_as_ext3(void)
4981 int err = register_filesystem(&ext3_fs_type);
4984 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4987 static inline void unregister_as_ext3(void)
4989 unregister_filesystem(&ext3_fs_type);
4992 static inline int ext3_feature_set_ok(struct super_block *sb)
4994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4996 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4998 if (sb->s_flags & MS_RDONLY)
5000 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5004 MODULE_ALIAS("ext3");
5006 static inline void register_as_ext3(void) { }
5007 static inline void unregister_as_ext3(void) { }
5008 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5011 static struct file_system_type ext4_fs_type = {
5012 .owner = THIS_MODULE,
5014 .mount = ext4_mount,
5015 .kill_sb = kill_block_super,
5016 .fs_flags = FS_REQUIRES_DEV,
5019 static int __init ext4_init_feat_adverts(void)
5021 struct ext4_features *ef;
5024 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5028 ef->f_kobj.kset = ext4_kset;
5029 init_completion(&ef->f_kobj_unregister);
5030 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5043 static void ext4_exit_feat_adverts(void)
5045 kobject_put(&ext4_feat->f_kobj);
5046 wait_for_completion(&ext4_feat->f_kobj_unregister);
5050 /* Shared across all ext4 file systems */
5051 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5052 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5054 static int __init ext4_init_fs(void)
5058 ext4_check_flag_values();
5060 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5061 mutex_init(&ext4__aio_mutex[i]);
5062 init_waitqueue_head(&ext4__ioend_wq[i]);
5065 err = ext4_init_pageio();
5068 err = ext4_init_system_zone();
5071 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5074 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5076 err = ext4_init_feat_adverts();
5080 err = ext4_init_mballoc();
5084 err = ext4_init_xattr();
5087 err = init_inodecache();
5092 err = register_filesystem(&ext4_fs_type);
5096 ext4_li_info = NULL;
5097 mutex_init(&ext4_li_mtx);
5100 unregister_as_ext2();
5101 unregister_as_ext3();
5102 destroy_inodecache();
5106 ext4_exit_mballoc();
5108 ext4_exit_feat_adverts();
5111 remove_proc_entry("fs/ext4", NULL);
5112 kset_unregister(ext4_kset);
5114 ext4_exit_system_zone();
5120 static void __exit ext4_exit_fs(void)
5122 ext4_destroy_lazyinit_thread();
5123 unregister_as_ext2();
5124 unregister_as_ext3();
5125 unregister_filesystem(&ext4_fs_type);
5126 destroy_inodecache();
5128 ext4_exit_mballoc();
5129 ext4_exit_feat_adverts();
5130 remove_proc_entry("fs/ext4", NULL);
5131 kset_unregister(ext4_kset);
5132 ext4_exit_system_zone();
5136 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5137 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5138 MODULE_LICENSE("GPL");
5139 module_init(ext4_init_fs)
5140 module_exit(ext4_exit_fs)