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;
434 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
436 struct super_block *sb = journal->j_private;
437 struct ext4_sb_info *sbi = EXT4_SB(sb);
438 int error = is_journal_aborted(journal);
439 struct ext4_journal_cb_entry *jce, *tmp;
441 spin_lock(&sbi->s_md_lock);
442 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
443 list_del_init(&jce->jce_list);
444 spin_unlock(&sbi->s_md_lock);
445 jce->jce_func(sb, jce, error);
446 spin_lock(&sbi->s_md_lock);
448 spin_unlock(&sbi->s_md_lock);
451 /* Deal with the reporting of failure conditions on a filesystem such as
452 * inconsistencies detected or read IO failures.
454 * On ext2, we can store the error state of the filesystem in the
455 * superblock. That is not possible on ext4, because we may have other
456 * write ordering constraints on the superblock which prevent us from
457 * writing it out straight away; and given that the journal is about to
458 * be aborted, we can't rely on the current, or future, transactions to
459 * write out the superblock safely.
461 * We'll just use the jbd2_journal_abort() error code to record an error in
462 * the journal instead. On recovery, the journal will complain about
463 * that error until we've noted it down and cleared it.
466 static void ext4_handle_error(struct super_block *sb)
468 if (sb->s_flags & MS_RDONLY)
471 if (!test_opt(sb, ERRORS_CONT)) {
472 journal_t *journal = EXT4_SB(sb)->s_journal;
474 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
476 jbd2_journal_abort(journal, -EIO);
478 if (test_opt(sb, ERRORS_RO)) {
479 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
480 sb->s_flags |= MS_RDONLY;
482 if (test_opt(sb, ERRORS_PANIC))
483 panic("EXT4-fs (device %s): panic forced after error\n",
487 void __ext4_error(struct super_block *sb, const char *function,
488 unsigned int line, const char *fmt, ...)
490 struct va_format vaf;
496 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
497 sb->s_id, function, line, current->comm, &vaf);
500 ext4_handle_error(sb);
503 void ext4_error_inode(struct inode *inode, const char *function,
504 unsigned int line, ext4_fsblk_t block,
505 const char *fmt, ...)
508 struct va_format vaf;
509 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
511 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
512 es->s_last_error_block = cpu_to_le64(block);
513 save_error_info(inode->i_sb, function, line);
517 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
518 inode->i_sb->s_id, function, line, inode->i_ino);
520 printk(KERN_CONT "block %llu: ", block);
521 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
524 ext4_handle_error(inode->i_sb);
527 void ext4_error_file(struct file *file, const char *function,
528 unsigned int line, ext4_fsblk_t block,
529 const char *fmt, ...)
532 struct va_format vaf;
533 struct ext4_super_block *es;
534 struct inode *inode = file->f_dentry->d_inode;
535 char pathname[80], *path;
537 es = EXT4_SB(inode->i_sb)->s_es;
538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 save_error_info(inode->i_sb, function, line);
540 path = d_path(&(file->f_path), pathname, sizeof(pathname));
544 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
545 inode->i_sb->s_id, function, line, inode->i_ino);
547 printk(KERN_CONT "block %llu: ", block);
551 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
554 ext4_handle_error(inode->i_sb);
557 static const char *ext4_decode_error(struct super_block *sb, int errno,
564 errstr = "IO failure";
567 errstr = "Out of memory";
570 if (!sb || (EXT4_SB(sb)->s_journal &&
571 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
572 errstr = "Journal has aborted";
574 errstr = "Readonly filesystem";
577 /* If the caller passed in an extra buffer for unknown
578 * errors, textualise them now. Else we just return
581 /* Check for truncated error codes... */
582 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
591 /* __ext4_std_error decodes expected errors from journaling functions
592 * automatically and invokes the appropriate error response. */
594 void __ext4_std_error(struct super_block *sb, const char *function,
595 unsigned int line, int errno)
600 /* Special case: if the error is EROFS, and we're not already
601 * inside a transaction, then there's really no point in logging
603 if (errno == -EROFS && journal_current_handle() == NULL &&
604 (sb->s_flags & MS_RDONLY))
607 errstr = ext4_decode_error(sb, errno, nbuf);
608 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
609 sb->s_id, function, line, errstr);
610 save_error_info(sb, function, line);
612 ext4_handle_error(sb);
616 * ext4_abort is a much stronger failure handler than ext4_error. The
617 * abort function may be used to deal with unrecoverable failures such
618 * as journal IO errors or ENOMEM at a critical moment in log management.
620 * We unconditionally force the filesystem into an ABORT|READONLY state,
621 * unless the error response on the fs has been set to panic in which
622 * case we take the easy way out and panic immediately.
625 void __ext4_abort(struct super_block *sb, const char *function,
626 unsigned int line, const char *fmt, ...)
630 save_error_info(sb, function, line);
632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
638 if ((sb->s_flags & MS_RDONLY) == 0) {
639 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
640 sb->s_flags |= MS_RDONLY;
641 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
642 if (EXT4_SB(sb)->s_journal)
643 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
644 save_error_info(sb, function, line);
646 if (test_opt(sb, ERRORS_PANIC))
647 panic("EXT4-fs panic from previous error\n");
650 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
652 struct va_format vaf;
658 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
662 void __ext4_warning(struct super_block *sb, const char *function,
663 unsigned int line, const char *fmt, ...)
665 struct va_format vaf;
671 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
672 sb->s_id, function, line, &vaf);
676 void __ext4_grp_locked_error(const char *function, unsigned int line,
677 struct super_block *sb, ext4_group_t grp,
678 unsigned long ino, ext4_fsblk_t block,
679 const char *fmt, ...)
683 struct va_format vaf;
685 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
687 es->s_last_error_ino = cpu_to_le32(ino);
688 es->s_last_error_block = cpu_to_le64(block);
689 __save_error_info(sb, function, line);
695 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
696 sb->s_id, function, line, grp);
698 printk(KERN_CONT "inode %lu: ", ino);
700 printk(KERN_CONT "block %llu:", (unsigned long long) block);
701 printk(KERN_CONT "%pV\n", &vaf);
704 if (test_opt(sb, ERRORS_CONT)) {
705 ext4_commit_super(sb, 0);
709 ext4_unlock_group(sb, grp);
710 ext4_handle_error(sb);
712 * We only get here in the ERRORS_RO case; relocking the group
713 * may be dangerous, but nothing bad will happen since the
714 * filesystem will have already been marked read/only and the
715 * journal has been aborted. We return 1 as a hint to callers
716 * who might what to use the return value from
717 * ext4_grp_locked_error() to distinguish between the
718 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
719 * aggressively from the ext4 function in question, with a
720 * more appropriate error code.
722 ext4_lock_group(sb, grp);
726 void ext4_update_dynamic_rev(struct super_block *sb)
728 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
730 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
734 "updating to rev %d because of new feature flag, "
735 "running e2fsck is recommended",
738 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
739 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
740 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
741 /* leave es->s_feature_*compat flags alone */
742 /* es->s_uuid will be set by e2fsck if empty */
745 * The rest of the superblock fields should be zero, and if not it
746 * means they are likely already in use, so leave them alone. We
747 * can leave it up to e2fsck to clean up any inconsistencies there.
752 * Open the external journal device
754 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
756 struct block_device *bdev;
757 char b[BDEVNAME_SIZE];
759 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
765 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
766 __bdevname(dev, b), PTR_ERR(bdev));
771 * Release the journal device
773 static int ext4_blkdev_put(struct block_device *bdev)
775 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
778 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
780 struct block_device *bdev;
783 bdev = sbi->journal_bdev;
785 ret = ext4_blkdev_put(bdev);
786 sbi->journal_bdev = NULL;
791 static inline struct inode *orphan_list_entry(struct list_head *l)
793 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
796 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
800 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
801 le32_to_cpu(sbi->s_es->s_last_orphan));
803 printk(KERN_ERR "sb_info orphan list:\n");
804 list_for_each(l, &sbi->s_orphan) {
805 struct inode *inode = orphan_list_entry(l);
807 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
808 inode->i_sb->s_id, inode->i_ino, inode,
809 inode->i_mode, inode->i_nlink,
814 static void ext4_put_super(struct super_block *sb)
816 struct ext4_sb_info *sbi = EXT4_SB(sb);
817 struct ext4_super_block *es = sbi->s_es;
820 ext4_unregister_li_request(sb);
821 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
823 flush_workqueue(sbi->dio_unwritten_wq);
824 destroy_workqueue(sbi->dio_unwritten_wq);
828 ext4_commit_super(sb, 1);
830 if (sbi->s_journal) {
831 err = jbd2_journal_destroy(sbi->s_journal);
832 sbi->s_journal = NULL;
834 ext4_abort(sb, "Couldn't clean up the journal");
837 del_timer(&sbi->s_err_report);
838 ext4_release_system_zone(sb);
840 ext4_ext_release(sb);
841 ext4_xattr_put_super(sb);
843 if (!(sb->s_flags & MS_RDONLY)) {
844 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
845 es->s_state = cpu_to_le16(sbi->s_mount_state);
846 ext4_commit_super(sb, 1);
849 remove_proc_entry(sb->s_id, ext4_proc_root);
851 kobject_del(&sbi->s_kobj);
853 for (i = 0; i < sbi->s_gdb_count; i++)
854 brelse(sbi->s_group_desc[i]);
855 ext4_kvfree(sbi->s_group_desc);
856 ext4_kvfree(sbi->s_flex_groups);
857 percpu_counter_destroy(&sbi->s_freeclusters_counter);
858 percpu_counter_destroy(&sbi->s_freeinodes_counter);
859 percpu_counter_destroy(&sbi->s_dirs_counter);
860 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
863 for (i = 0; i < MAXQUOTAS; i++)
864 kfree(sbi->s_qf_names[i]);
867 /* Debugging code just in case the in-memory inode orphan list
868 * isn't empty. The on-disk one can be non-empty if we've
869 * detected an error and taken the fs readonly, but the
870 * in-memory list had better be clean by this point. */
871 if (!list_empty(&sbi->s_orphan))
872 dump_orphan_list(sb, sbi);
873 J_ASSERT(list_empty(&sbi->s_orphan));
875 invalidate_bdev(sb->s_bdev);
876 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
878 * Invalidate the journal device's buffers. We don't want them
879 * floating about in memory - the physical journal device may
880 * hotswapped, and it breaks the `ro-after' testing code.
882 sync_blockdev(sbi->journal_bdev);
883 invalidate_bdev(sbi->journal_bdev);
884 ext4_blkdev_remove(sbi);
887 kthread_stop(sbi->s_mmp_tsk);
888 sb->s_fs_info = NULL;
890 * Now that we are completely done shutting down the
891 * superblock, we need to actually destroy the kobject.
894 kobject_put(&sbi->s_kobj);
895 wait_for_completion(&sbi->s_kobj_unregister);
896 kfree(sbi->s_blockgroup_lock);
900 static struct kmem_cache *ext4_inode_cachep;
903 * Called inside transaction, so use GFP_NOFS
905 static struct inode *ext4_alloc_inode(struct super_block *sb)
907 struct ext4_inode_info *ei;
909 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
913 ei->vfs_inode.i_version = 1;
914 ei->vfs_inode.i_data.writeback_index = 0;
915 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
916 INIT_LIST_HEAD(&ei->i_prealloc_list);
917 spin_lock_init(&ei->i_prealloc_lock);
918 ei->i_reserved_data_blocks = 0;
919 ei->i_reserved_meta_blocks = 0;
920 ei->i_allocated_meta_blocks = 0;
921 ei->i_da_metadata_calc_len = 0;
922 spin_lock_init(&(ei->i_block_reservation_lock));
924 ei->i_reserved_quota = 0;
927 INIT_LIST_HEAD(&ei->i_completed_io_list);
928 spin_lock_init(&ei->i_completed_io_lock);
929 ei->cur_aio_dio = NULL;
931 ei->i_datasync_tid = 0;
932 atomic_set(&ei->i_ioend_count, 0);
933 atomic_set(&ei->i_aiodio_unwritten, 0);
935 return &ei->vfs_inode;
938 static int ext4_drop_inode(struct inode *inode)
940 int drop = generic_drop_inode(inode);
942 trace_ext4_drop_inode(inode, drop);
946 static void ext4_i_callback(struct rcu_head *head)
948 struct inode *inode = container_of(head, struct inode, i_rcu);
949 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
952 static void ext4_destroy_inode(struct inode *inode)
954 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
955 ext4_msg(inode->i_sb, KERN_ERR,
956 "Inode %lu (%p): orphan list check failed!",
957 inode->i_ino, EXT4_I(inode));
958 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
959 EXT4_I(inode), sizeof(struct ext4_inode_info),
963 call_rcu(&inode->i_rcu, ext4_i_callback);
966 static void init_once(void *foo)
968 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
970 INIT_LIST_HEAD(&ei->i_orphan);
971 #ifdef CONFIG_EXT4_FS_XATTR
972 init_rwsem(&ei->xattr_sem);
974 init_rwsem(&ei->i_data_sem);
975 inode_init_once(&ei->vfs_inode);
978 static int init_inodecache(void)
980 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
981 sizeof(struct ext4_inode_info),
982 0, (SLAB_RECLAIM_ACCOUNT|
985 if (ext4_inode_cachep == NULL)
990 static void destroy_inodecache(void)
992 kmem_cache_destroy(ext4_inode_cachep);
995 void ext4_clear_inode(struct inode *inode)
997 invalidate_inode_buffers(inode);
998 end_writeback(inode);
1000 ext4_discard_preallocations(inode);
1001 if (EXT4_I(inode)->jinode) {
1002 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1003 EXT4_I(inode)->jinode);
1004 jbd2_free_inode(EXT4_I(inode)->jinode);
1005 EXT4_I(inode)->jinode = NULL;
1009 static inline void ext4_show_quota_options(struct seq_file *seq,
1010 struct super_block *sb)
1012 #if defined(CONFIG_QUOTA)
1013 struct ext4_sb_info *sbi = EXT4_SB(sb);
1015 if (sbi->s_jquota_fmt) {
1018 switch (sbi->s_jquota_fmt) {
1029 seq_printf(seq, ",jqfmt=%s", fmtname);
1032 if (sbi->s_qf_names[USRQUOTA])
1033 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1035 if (sbi->s_qf_names[GRPQUOTA])
1036 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1038 if (test_opt(sb, USRQUOTA))
1039 seq_puts(seq, ",usrquota");
1041 if (test_opt(sb, GRPQUOTA))
1042 seq_puts(seq, ",grpquota");
1048 * - it's set to a non-default value OR
1049 * - if the per-sb default is different from the global default
1051 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1054 unsigned long def_mount_opts;
1055 struct super_block *sb = root->d_sb;
1056 struct ext4_sb_info *sbi = EXT4_SB(sb);
1057 struct ext4_super_block *es = sbi->s_es;
1059 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1060 def_errors = le16_to_cpu(es->s_errors);
1062 if (sbi->s_sb_block != 1)
1063 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1064 if (test_opt(sb, MINIX_DF))
1065 seq_puts(seq, ",minixdf");
1066 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1067 seq_puts(seq, ",grpid");
1068 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1069 seq_puts(seq, ",nogrpid");
1070 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1071 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1072 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1074 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1075 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1076 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1078 if (test_opt(sb, ERRORS_RO)) {
1079 if (def_errors == EXT4_ERRORS_PANIC ||
1080 def_errors == EXT4_ERRORS_CONTINUE) {
1081 seq_puts(seq, ",errors=remount-ro");
1084 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1085 seq_puts(seq, ",errors=continue");
1086 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1087 seq_puts(seq, ",errors=panic");
1088 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1089 seq_puts(seq, ",nouid32");
1090 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1091 seq_puts(seq, ",debug");
1092 #ifdef CONFIG_EXT4_FS_XATTR
1093 if (test_opt(sb, XATTR_USER))
1094 seq_puts(seq, ",user_xattr");
1095 if (!test_opt(sb, XATTR_USER))
1096 seq_puts(seq, ",nouser_xattr");
1098 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1099 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1100 seq_puts(seq, ",acl");
1101 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1102 seq_puts(seq, ",noacl");
1104 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1105 seq_printf(seq, ",commit=%u",
1106 (unsigned) (sbi->s_commit_interval / HZ));
1108 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1109 seq_printf(seq, ",min_batch_time=%u",
1110 (unsigned) sbi->s_min_batch_time);
1112 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1113 seq_printf(seq, ",max_batch_time=%u",
1114 (unsigned) sbi->s_max_batch_time);
1118 * We're changing the default of barrier mount option, so
1119 * let's always display its mount state so it's clear what its
1122 seq_puts(seq, ",barrier=");
1123 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1124 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1125 seq_puts(seq, ",journal_async_commit");
1126 else if (test_opt(sb, JOURNAL_CHECKSUM))
1127 seq_puts(seq, ",journal_checksum");
1128 if (test_opt(sb, I_VERSION))
1129 seq_puts(seq, ",i_version");
1130 if (!test_opt(sb, DELALLOC) &&
1131 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1132 seq_puts(seq, ",nodelalloc");
1134 if (!test_opt(sb, MBLK_IO_SUBMIT))
1135 seq_puts(seq, ",nomblk_io_submit");
1137 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1139 * journal mode get enabled in different ways
1140 * So just print the value even if we didn't specify it
1142 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1143 seq_puts(seq, ",data=journal");
1144 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1145 seq_puts(seq, ",data=ordered");
1146 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1147 seq_puts(seq, ",data=writeback");
1149 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1150 seq_printf(seq, ",inode_readahead_blks=%u",
1151 sbi->s_inode_readahead_blks);
1153 if (test_opt(sb, DATA_ERR_ABORT))
1154 seq_puts(seq, ",data_err=abort");
1156 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1157 seq_puts(seq, ",noauto_da_alloc");
1159 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1160 seq_puts(seq, ",discard");
1162 if (test_opt(sb, NOLOAD))
1163 seq_puts(seq, ",norecovery");
1165 if (test_opt(sb, DIOREAD_NOLOCK))
1166 seq_puts(seq, ",dioread_nolock");
1168 if (test_opt(sb, BLOCK_VALIDITY) &&
1169 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1170 seq_puts(seq, ",block_validity");
1172 if (!test_opt(sb, INIT_INODE_TABLE))
1173 seq_puts(seq, ",noinit_itable");
1174 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1175 seq_printf(seq, ",init_itable=%u",
1176 (unsigned) sbi->s_li_wait_mult);
1178 ext4_show_quota_options(seq, sb);
1183 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1184 u64 ino, u32 generation)
1186 struct inode *inode;
1188 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1189 return ERR_PTR(-ESTALE);
1190 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1191 return ERR_PTR(-ESTALE);
1193 /* iget isn't really right if the inode is currently unallocated!!
1195 * ext4_read_inode will return a bad_inode if the inode had been
1196 * deleted, so we should be safe.
1198 * Currently we don't know the generation for parent directory, so
1199 * a generation of 0 means "accept any"
1201 inode = ext4_iget(sb, ino);
1203 return ERR_CAST(inode);
1204 if (generation && inode->i_generation != generation) {
1206 return ERR_PTR(-ESTALE);
1212 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1213 int fh_len, int fh_type)
1215 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1216 ext4_nfs_get_inode);
1219 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1220 int fh_len, int fh_type)
1222 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1223 ext4_nfs_get_inode);
1227 * Try to release metadata pages (indirect blocks, directories) which are
1228 * mapped via the block device. Since these pages could have journal heads
1229 * which would prevent try_to_free_buffers() from freeing them, we must use
1230 * jbd2 layer's try_to_free_buffers() function to release them.
1232 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1235 journal_t *journal = EXT4_SB(sb)->s_journal;
1237 WARN_ON(PageChecked(page));
1238 if (!page_has_buffers(page))
1241 return jbd2_journal_try_to_free_buffers(journal, page,
1242 wait & ~__GFP_WAIT);
1243 return try_to_free_buffers(page);
1247 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1248 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1250 static int ext4_write_dquot(struct dquot *dquot);
1251 static int ext4_acquire_dquot(struct dquot *dquot);
1252 static int ext4_release_dquot(struct dquot *dquot);
1253 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1254 static int ext4_write_info(struct super_block *sb, int type);
1255 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1257 static int ext4_quota_off(struct super_block *sb, int type);
1258 static int ext4_quota_on_mount(struct super_block *sb, int type);
1259 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1260 size_t len, loff_t off);
1261 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1262 const char *data, size_t len, loff_t off);
1264 static const struct dquot_operations ext4_quota_operations = {
1265 .get_reserved_space = ext4_get_reserved_space,
1266 .write_dquot = ext4_write_dquot,
1267 .acquire_dquot = ext4_acquire_dquot,
1268 .release_dquot = ext4_release_dquot,
1269 .mark_dirty = ext4_mark_dquot_dirty,
1270 .write_info = ext4_write_info,
1271 .alloc_dquot = dquot_alloc,
1272 .destroy_dquot = dquot_destroy,
1275 static const struct quotactl_ops ext4_qctl_operations = {
1276 .quota_on = ext4_quota_on,
1277 .quota_off = ext4_quota_off,
1278 .quota_sync = dquot_quota_sync,
1279 .get_info = dquot_get_dqinfo,
1280 .set_info = dquot_set_dqinfo,
1281 .get_dqblk = dquot_get_dqblk,
1282 .set_dqblk = dquot_set_dqblk
1286 static const struct super_operations ext4_sops = {
1287 .alloc_inode = ext4_alloc_inode,
1288 .destroy_inode = ext4_destroy_inode,
1289 .write_inode = ext4_write_inode,
1290 .dirty_inode = ext4_dirty_inode,
1291 .drop_inode = ext4_drop_inode,
1292 .evict_inode = ext4_evict_inode,
1293 .put_super = ext4_put_super,
1294 .sync_fs = ext4_sync_fs,
1295 .freeze_fs = ext4_freeze,
1296 .unfreeze_fs = ext4_unfreeze,
1297 .statfs = ext4_statfs,
1298 .remount_fs = ext4_remount,
1299 .show_options = ext4_show_options,
1301 .quota_read = ext4_quota_read,
1302 .quota_write = ext4_quota_write,
1304 .bdev_try_to_free_page = bdev_try_to_free_page,
1307 static const struct super_operations ext4_nojournal_sops = {
1308 .alloc_inode = ext4_alloc_inode,
1309 .destroy_inode = ext4_destroy_inode,
1310 .write_inode = ext4_write_inode,
1311 .dirty_inode = ext4_dirty_inode,
1312 .drop_inode = ext4_drop_inode,
1313 .evict_inode = ext4_evict_inode,
1314 .write_super = ext4_write_super,
1315 .put_super = ext4_put_super,
1316 .statfs = ext4_statfs,
1317 .remount_fs = ext4_remount,
1318 .show_options = ext4_show_options,
1320 .quota_read = ext4_quota_read,
1321 .quota_write = ext4_quota_write,
1323 .bdev_try_to_free_page = bdev_try_to_free_page,
1326 static const struct export_operations ext4_export_ops = {
1327 .fh_to_dentry = ext4_fh_to_dentry,
1328 .fh_to_parent = ext4_fh_to_parent,
1329 .get_parent = ext4_get_parent,
1333 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1334 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1335 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1336 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1337 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1338 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1339 Opt_journal_update, Opt_journal_dev,
1340 Opt_journal_checksum, Opt_journal_async_commit,
1341 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1342 Opt_data_err_abort, Opt_data_err_ignore,
1343 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1344 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1345 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1346 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1347 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1348 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1349 Opt_inode_readahead_blks, Opt_journal_ioprio,
1350 Opt_dioread_nolock, Opt_dioread_lock,
1351 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1354 static const match_table_t tokens = {
1355 {Opt_bsd_df, "bsddf"},
1356 {Opt_minix_df, "minixdf"},
1357 {Opt_grpid, "grpid"},
1358 {Opt_grpid, "bsdgroups"},
1359 {Opt_nogrpid, "nogrpid"},
1360 {Opt_nogrpid, "sysvgroups"},
1361 {Opt_resgid, "resgid=%u"},
1362 {Opt_resuid, "resuid=%u"},
1364 {Opt_err_cont, "errors=continue"},
1365 {Opt_err_panic, "errors=panic"},
1366 {Opt_err_ro, "errors=remount-ro"},
1367 {Opt_nouid32, "nouid32"},
1368 {Opt_debug, "debug"},
1369 {Opt_oldalloc, "oldalloc"},
1370 {Opt_orlov, "orlov"},
1371 {Opt_user_xattr, "user_xattr"},
1372 {Opt_nouser_xattr, "nouser_xattr"},
1374 {Opt_noacl, "noacl"},
1375 {Opt_noload, "noload"},
1376 {Opt_noload, "norecovery"},
1379 {Opt_commit, "commit=%u"},
1380 {Opt_min_batch_time, "min_batch_time=%u"},
1381 {Opt_max_batch_time, "max_batch_time=%u"},
1382 {Opt_journal_update, "journal=update"},
1383 {Opt_journal_dev, "journal_dev=%u"},
1384 {Opt_journal_checksum, "journal_checksum"},
1385 {Opt_journal_async_commit, "journal_async_commit"},
1386 {Opt_abort, "abort"},
1387 {Opt_data_journal, "data=journal"},
1388 {Opt_data_ordered, "data=ordered"},
1389 {Opt_data_writeback, "data=writeback"},
1390 {Opt_data_err_abort, "data_err=abort"},
1391 {Opt_data_err_ignore, "data_err=ignore"},
1392 {Opt_offusrjquota, "usrjquota="},
1393 {Opt_usrjquota, "usrjquota=%s"},
1394 {Opt_offgrpjquota, "grpjquota="},
1395 {Opt_grpjquota, "grpjquota=%s"},
1396 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1397 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1398 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1399 {Opt_grpquota, "grpquota"},
1400 {Opt_noquota, "noquota"},
1401 {Opt_quota, "quota"},
1402 {Opt_usrquota, "usrquota"},
1403 {Opt_barrier, "barrier=%u"},
1404 {Opt_barrier, "barrier"},
1405 {Opt_nobarrier, "nobarrier"},
1406 {Opt_i_version, "i_version"},
1407 {Opt_stripe, "stripe=%u"},
1408 {Opt_resize, "resize"},
1409 {Opt_delalloc, "delalloc"},
1410 {Opt_nodelalloc, "nodelalloc"},
1411 {Opt_mblk_io_submit, "mblk_io_submit"},
1412 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1413 {Opt_block_validity, "block_validity"},
1414 {Opt_noblock_validity, "noblock_validity"},
1415 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1416 {Opt_journal_ioprio, "journal_ioprio=%u"},
1417 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1418 {Opt_auto_da_alloc, "auto_da_alloc"},
1419 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1420 {Opt_dioread_nolock, "dioread_nolock"},
1421 {Opt_dioread_lock, "dioread_lock"},
1422 {Opt_discard, "discard"},
1423 {Opt_nodiscard, "nodiscard"},
1424 {Opt_init_itable, "init_itable=%u"},
1425 {Opt_init_itable, "init_itable"},
1426 {Opt_noinit_itable, "noinit_itable"},
1430 static ext4_fsblk_t get_sb_block(void **data)
1432 ext4_fsblk_t sb_block;
1433 char *options = (char *) *data;
1435 if (!options || strncmp(options, "sb=", 3) != 0)
1436 return 1; /* Default location */
1439 /* TODO: use simple_strtoll with >32bit ext4 */
1440 sb_block = simple_strtoul(options, &options, 0);
1441 if (*options && *options != ',') {
1442 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1446 if (*options == ',')
1448 *data = (void *) options;
1453 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1454 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1455 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1458 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1460 struct ext4_sb_info *sbi = EXT4_SB(sb);
1463 if (sb_any_quota_loaded(sb) &&
1464 !sbi->s_qf_names[qtype]) {
1465 ext4_msg(sb, KERN_ERR,
1466 "Cannot change journaled "
1467 "quota options when quota turned on");
1470 qname = match_strdup(args);
1472 ext4_msg(sb, KERN_ERR,
1473 "Not enough memory for storing quotafile name");
1476 if (sbi->s_qf_names[qtype] &&
1477 strcmp(sbi->s_qf_names[qtype], qname)) {
1478 ext4_msg(sb, KERN_ERR,
1479 "%s quota file already specified", QTYPE2NAME(qtype));
1483 sbi->s_qf_names[qtype] = qname;
1484 if (strchr(sbi->s_qf_names[qtype], '/')) {
1485 ext4_msg(sb, KERN_ERR,
1486 "quotafile must be on filesystem root");
1487 kfree(sbi->s_qf_names[qtype]);
1488 sbi->s_qf_names[qtype] = NULL;
1495 static int clear_qf_name(struct super_block *sb, int qtype)
1498 struct ext4_sb_info *sbi = EXT4_SB(sb);
1500 if (sb_any_quota_loaded(sb) &&
1501 sbi->s_qf_names[qtype]) {
1502 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1503 " when quota turned on");
1507 * The space will be released later when all options are confirmed
1510 sbi->s_qf_names[qtype] = NULL;
1515 static int parse_options(char *options, struct super_block *sb,
1516 unsigned long *journal_devnum,
1517 unsigned int *journal_ioprio,
1518 ext4_fsblk_t *n_blocks_count, int is_remount)
1520 struct ext4_sb_info *sbi = EXT4_SB(sb);
1522 substring_t args[MAX_OPT_ARGS];
1532 while ((p = strsep(&options, ",")) != NULL) {
1538 * Initialize args struct so we know whether arg was
1539 * found; some options take optional arguments.
1541 args[0].to = args[0].from = NULL;
1542 token = match_token(p, tokens, args);
1545 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1546 clear_opt(sb, MINIX_DF);
1549 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1550 set_opt(sb, MINIX_DF);
1554 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1559 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1560 clear_opt(sb, GRPID);
1564 if (match_int(&args[0], &option))
1566 sbi->s_resuid = option;
1569 if (match_int(&args[0], &option))
1571 sbi->s_resgid = option;
1574 /* handled by get_sb_block() instead of here */
1575 /* *sb_block = match_int(&args[0]); */
1578 clear_opt(sb, ERRORS_CONT);
1579 clear_opt(sb, ERRORS_RO);
1580 set_opt(sb, ERRORS_PANIC);
1583 clear_opt(sb, ERRORS_CONT);
1584 clear_opt(sb, ERRORS_PANIC);
1585 set_opt(sb, ERRORS_RO);
1588 clear_opt(sb, ERRORS_RO);
1589 clear_opt(sb, ERRORS_PANIC);
1590 set_opt(sb, ERRORS_CONT);
1593 set_opt(sb, NO_UID32);
1599 ext4_msg(sb, KERN_WARNING,
1600 "Ignoring deprecated oldalloc option");
1603 ext4_msg(sb, KERN_WARNING,
1604 "Ignoring deprecated orlov option");
1606 #ifdef CONFIG_EXT4_FS_XATTR
1607 case Opt_user_xattr:
1608 set_opt(sb, XATTR_USER);
1610 case Opt_nouser_xattr:
1611 clear_opt(sb, XATTR_USER);
1614 case Opt_user_xattr:
1615 case Opt_nouser_xattr:
1616 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1619 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1621 set_opt(sb, POSIX_ACL);
1624 clear_opt(sb, POSIX_ACL);
1629 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1632 case Opt_journal_update:
1634 /* Eventually we will want to be able to create
1635 a journal file here. For now, only allow the
1636 user to specify an existing inode to be the
1639 ext4_msg(sb, KERN_ERR,
1640 "Cannot specify journal on remount");
1643 set_opt(sb, UPDATE_JOURNAL);
1645 case Opt_journal_dev:
1647 ext4_msg(sb, KERN_ERR,
1648 "Cannot specify journal on remount");
1651 if (match_int(&args[0], &option))
1653 *journal_devnum = option;
1655 case Opt_journal_checksum:
1656 set_opt(sb, JOURNAL_CHECKSUM);
1658 case Opt_journal_async_commit:
1659 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1660 set_opt(sb, JOURNAL_CHECKSUM);
1663 set_opt(sb, NOLOAD);
1666 if (match_int(&args[0], &option))
1671 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1672 sbi->s_commit_interval = HZ * option;
1674 case Opt_max_batch_time:
1675 if (match_int(&args[0], &option))
1680 option = EXT4_DEF_MAX_BATCH_TIME;
1681 sbi->s_max_batch_time = option;
1683 case Opt_min_batch_time:
1684 if (match_int(&args[0], &option))
1688 sbi->s_min_batch_time = option;
1690 case Opt_data_journal:
1691 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1693 case Opt_data_ordered:
1694 data_opt = EXT4_MOUNT_ORDERED_DATA;
1696 case Opt_data_writeback:
1697 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1700 if (!sbi->s_journal)
1701 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1702 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1703 ext4_msg(sb, KERN_ERR,
1704 "Cannot change data mode on remount");
1708 clear_opt(sb, DATA_FLAGS);
1709 sbi->s_mount_opt |= data_opt;
1712 case Opt_data_err_abort:
1713 set_opt(sb, DATA_ERR_ABORT);
1715 case Opt_data_err_ignore:
1716 clear_opt(sb, DATA_ERR_ABORT);
1720 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1724 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1727 case Opt_offusrjquota:
1728 if (!clear_qf_name(sb, USRQUOTA))
1731 case Opt_offgrpjquota:
1732 if (!clear_qf_name(sb, GRPQUOTA))
1736 case Opt_jqfmt_vfsold:
1737 qfmt = QFMT_VFS_OLD;
1739 case Opt_jqfmt_vfsv0:
1742 case Opt_jqfmt_vfsv1:
1745 if (sb_any_quota_loaded(sb) &&
1746 sbi->s_jquota_fmt != qfmt) {
1747 ext4_msg(sb, KERN_ERR, "Cannot change "
1748 "journaled quota options when "
1752 sbi->s_jquota_fmt = qfmt;
1757 set_opt(sb, USRQUOTA);
1761 set_opt(sb, GRPQUOTA);
1764 if (sb_any_quota_loaded(sb)) {
1765 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1766 "options when quota turned on");
1769 clear_opt(sb, QUOTA);
1770 clear_opt(sb, USRQUOTA);
1771 clear_opt(sb, GRPQUOTA);
1777 ext4_msg(sb, KERN_ERR,
1778 "quota options not supported");
1782 case Opt_offusrjquota:
1783 case Opt_offgrpjquota:
1784 case Opt_jqfmt_vfsold:
1785 case Opt_jqfmt_vfsv0:
1786 case Opt_jqfmt_vfsv1:
1787 ext4_msg(sb, KERN_ERR,
1788 "journaled quota options not supported");
1794 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1797 clear_opt(sb, BARRIER);
1801 if (match_int(&args[0], &option))
1804 option = 1; /* No argument, default to 1 */
1806 set_opt(sb, BARRIER);
1808 clear_opt(sb, BARRIER);
1814 ext4_msg(sb, KERN_ERR,
1815 "resize option only available "
1819 if (match_int(&args[0], &option) != 0)
1821 *n_blocks_count = option;
1824 ext4_msg(sb, KERN_WARNING,
1825 "Ignoring deprecated nobh option");
1828 ext4_msg(sb, KERN_WARNING,
1829 "Ignoring deprecated bh option");
1832 set_opt(sb, I_VERSION);
1833 sb->s_flags |= MS_I_VERSION;
1835 case Opt_nodelalloc:
1836 clear_opt(sb, DELALLOC);
1837 clear_opt2(sb, EXPLICIT_DELALLOC);
1839 case Opt_mblk_io_submit:
1840 set_opt(sb, MBLK_IO_SUBMIT);
1842 case Opt_nomblk_io_submit:
1843 clear_opt(sb, MBLK_IO_SUBMIT);
1846 if (match_int(&args[0], &option))
1850 sbi->s_stripe = option;
1853 set_opt(sb, DELALLOC);
1854 set_opt2(sb, EXPLICIT_DELALLOC);
1856 case Opt_block_validity:
1857 set_opt(sb, BLOCK_VALIDITY);
1859 case Opt_noblock_validity:
1860 clear_opt(sb, BLOCK_VALIDITY);
1862 case Opt_inode_readahead_blks:
1863 if (match_int(&args[0], &option))
1865 if (option < 0 || option > (1 << 30))
1867 if (option && !is_power_of_2(option)) {
1868 ext4_msg(sb, KERN_ERR,
1869 "EXT4-fs: inode_readahead_blks"
1870 " must be a power of 2");
1873 sbi->s_inode_readahead_blks = option;
1875 case Opt_journal_ioprio:
1876 if (match_int(&args[0], &option))
1878 if (option < 0 || option > 7)
1880 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1883 case Opt_noauto_da_alloc:
1884 set_opt(sb, NO_AUTO_DA_ALLOC);
1886 case Opt_auto_da_alloc:
1888 if (match_int(&args[0], &option))
1891 option = 1; /* No argument, default to 1 */
1893 clear_opt(sb, NO_AUTO_DA_ALLOC);
1895 set_opt(sb,NO_AUTO_DA_ALLOC);
1898 set_opt(sb, DISCARD);
1901 clear_opt(sb, DISCARD);
1903 case Opt_dioread_nolock:
1904 set_opt(sb, DIOREAD_NOLOCK);
1906 case Opt_dioread_lock:
1907 clear_opt(sb, DIOREAD_NOLOCK);
1909 case Opt_init_itable:
1910 set_opt(sb, INIT_INODE_TABLE);
1912 if (match_int(&args[0], &option))
1915 option = EXT4_DEF_LI_WAIT_MULT;
1918 sbi->s_li_wait_mult = option;
1920 case Opt_noinit_itable:
1921 clear_opt(sb, INIT_INODE_TABLE);
1924 ext4_msg(sb, KERN_ERR,
1925 "Unrecognized mount option \"%s\" "
1926 "or missing value", p);
1931 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1932 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1933 clear_opt(sb, USRQUOTA);
1935 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1936 clear_opt(sb, GRPQUOTA);
1938 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1939 ext4_msg(sb, KERN_ERR, "old and new quota "
1944 if (!sbi->s_jquota_fmt) {
1945 ext4_msg(sb, KERN_ERR, "journaled quota format "
1950 if (sbi->s_jquota_fmt) {
1951 ext4_msg(sb, KERN_ERR, "journaled quota format "
1952 "specified with no journaling "
1961 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1964 struct ext4_sb_info *sbi = EXT4_SB(sb);
1967 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1968 ext4_msg(sb, KERN_ERR, "revision level too high, "
1969 "forcing read-only mode");
1974 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1975 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1976 "running e2fsck is recommended");
1977 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1978 ext4_msg(sb, KERN_WARNING,
1979 "warning: mounting fs with errors, "
1980 "running e2fsck is recommended");
1981 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1982 le16_to_cpu(es->s_mnt_count) >=
1983 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1984 ext4_msg(sb, KERN_WARNING,
1985 "warning: maximal mount count reached, "
1986 "running e2fsck is recommended");
1987 else if (le32_to_cpu(es->s_checkinterval) &&
1988 (le32_to_cpu(es->s_lastcheck) +
1989 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1990 ext4_msg(sb, KERN_WARNING,
1991 "warning: checktime reached, "
1992 "running e2fsck is recommended");
1993 if (!sbi->s_journal)
1994 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1995 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1996 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1997 le16_add_cpu(&es->s_mnt_count, 1);
1998 es->s_mtime = cpu_to_le32(get_seconds());
1999 ext4_update_dynamic_rev(sb);
2001 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2003 ext4_commit_super(sb, 1);
2005 if (test_opt(sb, DEBUG))
2006 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2007 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2009 sbi->s_groups_count,
2010 EXT4_BLOCKS_PER_GROUP(sb),
2011 EXT4_INODES_PER_GROUP(sb),
2012 sbi->s_mount_opt, sbi->s_mount_opt2);
2014 cleancache_init_fs(sb);
2018 static int ext4_fill_flex_info(struct super_block *sb)
2020 struct ext4_sb_info *sbi = EXT4_SB(sb);
2021 struct ext4_group_desc *gdp = NULL;
2022 ext4_group_t flex_group_count;
2023 ext4_group_t flex_group;
2024 unsigned int groups_per_flex = 0;
2028 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2029 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2030 sbi->s_log_groups_per_flex = 0;
2033 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2035 /* We allocate both existing and potentially added groups */
2036 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2037 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2038 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2039 size = flex_group_count * sizeof(struct flex_groups);
2040 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2041 if (sbi->s_flex_groups == NULL) {
2042 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2047 for (i = 0; i < sbi->s_groups_count; i++) {
2048 gdp = ext4_get_group_desc(sb, i, NULL);
2050 flex_group = ext4_flex_group(sbi, i);
2051 atomic_add(ext4_free_inodes_count(sb, gdp),
2052 &sbi->s_flex_groups[flex_group].free_inodes);
2053 atomic_add(ext4_free_group_clusters(sb, gdp),
2054 &sbi->s_flex_groups[flex_group].free_clusters);
2055 atomic_add(ext4_used_dirs_count(sb, gdp),
2056 &sbi->s_flex_groups[flex_group].used_dirs);
2064 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2065 struct ext4_group_desc *gdp)
2069 if (sbi->s_es->s_feature_ro_compat &
2070 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2071 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2072 __le32 le_group = cpu_to_le32(block_group);
2074 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2075 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2076 crc = crc16(crc, (__u8 *)gdp, offset);
2077 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2078 /* for checksum of struct ext4_group_desc do the rest...*/
2079 if ((sbi->s_es->s_feature_incompat &
2080 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2081 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2082 crc = crc16(crc, (__u8 *)gdp + offset,
2083 le16_to_cpu(sbi->s_es->s_desc_size) -
2087 return cpu_to_le16(crc);
2090 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2091 struct ext4_group_desc *gdp)
2093 if ((sbi->s_es->s_feature_ro_compat &
2094 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2095 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2101 /* Called at mount-time, super-block is locked */
2102 static int ext4_check_descriptors(struct super_block *sb,
2103 ext4_group_t *first_not_zeroed)
2105 struct ext4_sb_info *sbi = EXT4_SB(sb);
2106 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2107 ext4_fsblk_t last_block;
2108 ext4_fsblk_t block_bitmap;
2109 ext4_fsblk_t inode_bitmap;
2110 ext4_fsblk_t inode_table;
2111 int flexbg_flag = 0;
2112 ext4_group_t i, grp = sbi->s_groups_count;
2114 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2117 ext4_debug("Checking group descriptors");
2119 for (i = 0; i < sbi->s_groups_count; i++) {
2120 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2122 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2123 last_block = ext4_blocks_count(sbi->s_es) - 1;
2125 last_block = first_block +
2126 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2128 if ((grp == sbi->s_groups_count) &&
2129 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2132 block_bitmap = ext4_block_bitmap(sb, gdp);
2133 if (block_bitmap < first_block || block_bitmap > last_block) {
2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2135 "Block bitmap for group %u not in group "
2136 "(block %llu)!", i, block_bitmap);
2139 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2140 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2142 "Inode bitmap for group %u not in group "
2143 "(block %llu)!", i, inode_bitmap);
2146 inode_table = ext4_inode_table(sb, gdp);
2147 if (inode_table < first_block ||
2148 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2149 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2150 "Inode table for group %u not in group "
2151 "(block %llu)!", i, inode_table);
2154 ext4_lock_group(sb, i);
2155 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2156 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2157 "Checksum for group %u failed (%u!=%u)",
2158 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2159 gdp)), le16_to_cpu(gdp->bg_checksum));
2160 if (!(sb->s_flags & MS_RDONLY)) {
2161 ext4_unlock_group(sb, i);
2165 ext4_unlock_group(sb, i);
2167 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2169 if (NULL != first_not_zeroed)
2170 *first_not_zeroed = grp;
2172 ext4_free_blocks_count_set(sbi->s_es,
2173 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2174 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2178 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2179 * the superblock) which were deleted from all directories, but held open by
2180 * a process at the time of a crash. We walk the list and try to delete these
2181 * inodes at recovery time (only with a read-write filesystem).
2183 * In order to keep the orphan inode chain consistent during traversal (in
2184 * case of crash during recovery), we link each inode into the superblock
2185 * orphan list_head and handle it the same way as an inode deletion during
2186 * normal operation (which journals the operations for us).
2188 * We only do an iget() and an iput() on each inode, which is very safe if we
2189 * accidentally point at an in-use or already deleted inode. The worst that
2190 * can happen in this case is that we get a "bit already cleared" message from
2191 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2192 * e2fsck was run on this filesystem, and it must have already done the orphan
2193 * inode cleanup for us, so we can safely abort without any further action.
2195 static void ext4_orphan_cleanup(struct super_block *sb,
2196 struct ext4_super_block *es)
2198 unsigned int s_flags = sb->s_flags;
2199 int nr_orphans = 0, nr_truncates = 0;
2203 if (!es->s_last_orphan) {
2204 jbd_debug(4, "no orphan inodes to clean up\n");
2208 if (bdev_read_only(sb->s_bdev)) {
2209 ext4_msg(sb, KERN_ERR, "write access "
2210 "unavailable, skipping orphan cleanup");
2214 /* Check if feature set would not allow a r/w mount */
2215 if (!ext4_feature_set_ok(sb, 0)) {
2216 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2217 "unknown ROCOMPAT features");
2221 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2222 if (es->s_last_orphan)
2223 jbd_debug(1, "Errors on filesystem, "
2224 "clearing orphan list.\n");
2225 es->s_last_orphan = 0;
2226 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2230 if (s_flags & MS_RDONLY) {
2231 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2232 sb->s_flags &= ~MS_RDONLY;
2235 /* Needed for iput() to work correctly and not trash data */
2236 sb->s_flags |= MS_ACTIVE;
2237 /* Turn on quotas so that they are updated correctly */
2238 for (i = 0; i < MAXQUOTAS; i++) {
2239 if (EXT4_SB(sb)->s_qf_names[i]) {
2240 int ret = ext4_quota_on_mount(sb, i);
2242 ext4_msg(sb, KERN_ERR,
2243 "Cannot turn on journaled "
2244 "quota: error %d", ret);
2249 while (es->s_last_orphan) {
2250 struct inode *inode;
2252 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2253 if (IS_ERR(inode)) {
2254 es->s_last_orphan = 0;
2258 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2259 dquot_initialize(inode);
2260 if (inode->i_nlink) {
2261 ext4_msg(sb, KERN_DEBUG,
2262 "%s: truncating inode %lu to %lld bytes",
2263 __func__, inode->i_ino, inode->i_size);
2264 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2265 inode->i_ino, inode->i_size);
2266 ext4_truncate(inode);
2269 ext4_msg(sb, KERN_DEBUG,
2270 "%s: deleting unreferenced inode %lu",
2271 __func__, inode->i_ino);
2272 jbd_debug(2, "deleting unreferenced inode %lu\n",
2276 iput(inode); /* The delete magic happens here! */
2279 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2282 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2283 PLURAL(nr_orphans));
2285 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2286 PLURAL(nr_truncates));
2288 /* Turn quotas off */
2289 for (i = 0; i < MAXQUOTAS; i++) {
2290 if (sb_dqopt(sb)->files[i])
2291 dquot_quota_off(sb, i);
2294 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2298 * Maximal extent format file size.
2299 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2300 * extent format containers, within a sector_t, and within i_blocks
2301 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2302 * so that won't be a limiting factor.
2304 * However there is other limiting factor. We do store extents in the form
2305 * of starting block and length, hence the resulting length of the extent
2306 * covering maximum file size must fit into on-disk format containers as
2307 * well. Given that length is always by 1 unit bigger than max unit (because
2308 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2310 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2312 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2315 loff_t upper_limit = MAX_LFS_FILESIZE;
2317 /* small i_blocks in vfs inode? */
2318 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2320 * CONFIG_LBDAF is not enabled implies the inode
2321 * i_block represent total blocks in 512 bytes
2322 * 32 == size of vfs inode i_blocks * 8
2324 upper_limit = (1LL << 32) - 1;
2326 /* total blocks in file system block size */
2327 upper_limit >>= (blkbits - 9);
2328 upper_limit <<= blkbits;
2332 * 32-bit extent-start container, ee_block. We lower the maxbytes
2333 * by one fs block, so ee_len can cover the extent of maximum file
2336 res = (1LL << 32) - 1;
2339 /* Sanity check against vm- & vfs- imposed limits */
2340 if (res > upper_limit)
2347 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2348 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2349 * We need to be 1 filesystem block less than the 2^48 sector limit.
2351 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2353 loff_t res = EXT4_NDIR_BLOCKS;
2356 /* This is calculated to be the largest file size for a dense, block
2357 * mapped file such that the file's total number of 512-byte sectors,
2358 * including data and all indirect blocks, does not exceed (2^48 - 1).
2360 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2361 * number of 512-byte sectors of the file.
2364 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2366 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2367 * the inode i_block field represents total file blocks in
2368 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2370 upper_limit = (1LL << 32) - 1;
2372 /* total blocks in file system block size */
2373 upper_limit >>= (bits - 9);
2377 * We use 48 bit ext4_inode i_blocks
2378 * With EXT4_HUGE_FILE_FL set the i_blocks
2379 * represent total number of blocks in
2380 * file system block size
2382 upper_limit = (1LL << 48) - 1;
2386 /* indirect blocks */
2388 /* double indirect blocks */
2389 meta_blocks += 1 + (1LL << (bits-2));
2390 /* tripple indirect blocks */
2391 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2393 upper_limit -= meta_blocks;
2394 upper_limit <<= bits;
2396 res += 1LL << (bits-2);
2397 res += 1LL << (2*(bits-2));
2398 res += 1LL << (3*(bits-2));
2400 if (res > upper_limit)
2403 if (res > MAX_LFS_FILESIZE)
2404 res = MAX_LFS_FILESIZE;
2409 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2410 ext4_fsblk_t logical_sb_block, int nr)
2412 struct ext4_sb_info *sbi = EXT4_SB(sb);
2413 ext4_group_t bg, first_meta_bg;
2416 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2418 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2420 return logical_sb_block + nr + 1;
2421 bg = sbi->s_desc_per_block * nr;
2422 if (ext4_bg_has_super(sb, bg))
2425 return (has_super + ext4_group_first_block_no(sb, bg));
2429 * ext4_get_stripe_size: Get the stripe size.
2430 * @sbi: In memory super block info
2432 * If we have specified it via mount option, then
2433 * use the mount option value. If the value specified at mount time is
2434 * greater than the blocks per group use the super block value.
2435 * If the super block value is greater than blocks per group return 0.
2436 * Allocator needs it be less than blocks per group.
2439 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2441 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2442 unsigned long stripe_width =
2443 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2446 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2447 ret = sbi->s_stripe;
2448 else if (stripe_width <= sbi->s_blocks_per_group)
2450 else if (stride <= sbi->s_blocks_per_group)
2456 * If the stripe width is 1, this makes no sense and
2457 * we set it to 0 to turn off stripe handling code.
2468 struct attribute attr;
2469 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2470 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2471 const char *, size_t);
2475 static int parse_strtoul(const char *buf,
2476 unsigned long max, unsigned long *value)
2480 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2481 endp = skip_spaces(endp);
2482 if (*endp || *value > max)
2488 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2489 struct ext4_sb_info *sbi,
2492 return snprintf(buf, PAGE_SIZE, "%llu\n",
2494 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2497 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi, char *buf)
2500 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2502 if (!sb->s_bdev->bd_part)
2503 return snprintf(buf, PAGE_SIZE, "0\n");
2504 return snprintf(buf, PAGE_SIZE, "%lu\n",
2505 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2506 sbi->s_sectors_written_start) >> 1);
2509 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2510 struct ext4_sb_info *sbi, char *buf)
2512 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2514 if (!sb->s_bdev->bd_part)
2515 return snprintf(buf, PAGE_SIZE, "0\n");
2516 return snprintf(buf, PAGE_SIZE, "%llu\n",
2517 (unsigned long long)(sbi->s_kbytes_written +
2518 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2519 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2522 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2525 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2528 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2529 struct ext4_sb_info *sbi, char *buf)
2531 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2534 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi,
2536 const char *buf, size_t count)
2540 if (parse_strtoul(buf, 0x40000000, &t))
2543 if (t && !is_power_of_2(t))
2546 sbi->s_inode_readahead_blks = t;
2550 static ssize_t sbi_ui_show(struct ext4_attr *a,
2551 struct ext4_sb_info *sbi, char *buf)
2553 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2555 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2558 static ssize_t sbi_ui_store(struct ext4_attr *a,
2559 struct ext4_sb_info *sbi,
2560 const char *buf, size_t count)
2562 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2565 if (parse_strtoul(buf, 0xffffffff, &t))
2571 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2572 static struct ext4_attr ext4_attr_##_name = { \
2573 .attr = {.name = __stringify(_name), .mode = _mode }, \
2576 .offset = offsetof(struct ext4_sb_info, _elname), \
2578 #define EXT4_ATTR(name, mode, show, store) \
2579 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2581 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2582 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2583 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2584 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2585 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2586 #define ATTR_LIST(name) &ext4_attr_##name.attr
2588 EXT4_RO_ATTR(delayed_allocation_blocks);
2589 EXT4_RO_ATTR(session_write_kbytes);
2590 EXT4_RO_ATTR(lifetime_write_kbytes);
2591 EXT4_RO_ATTR(extent_cache_hits);
2592 EXT4_RO_ATTR(extent_cache_misses);
2593 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2594 inode_readahead_blks_store, s_inode_readahead_blks);
2595 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2596 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2597 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2598 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2599 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2600 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2601 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2602 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2604 static struct attribute *ext4_attrs[] = {
2605 ATTR_LIST(delayed_allocation_blocks),
2606 ATTR_LIST(session_write_kbytes),
2607 ATTR_LIST(lifetime_write_kbytes),
2608 ATTR_LIST(extent_cache_hits),
2609 ATTR_LIST(extent_cache_misses),
2610 ATTR_LIST(inode_readahead_blks),
2611 ATTR_LIST(inode_goal),
2612 ATTR_LIST(mb_stats),
2613 ATTR_LIST(mb_max_to_scan),
2614 ATTR_LIST(mb_min_to_scan),
2615 ATTR_LIST(mb_order2_req),
2616 ATTR_LIST(mb_stream_req),
2617 ATTR_LIST(mb_group_prealloc),
2618 ATTR_LIST(max_writeback_mb_bump),
2622 /* Features this copy of ext4 supports */
2623 EXT4_INFO_ATTR(lazy_itable_init);
2624 EXT4_INFO_ATTR(batched_discard);
2626 static struct attribute *ext4_feat_attrs[] = {
2627 ATTR_LIST(lazy_itable_init),
2628 ATTR_LIST(batched_discard),
2632 static ssize_t ext4_attr_show(struct kobject *kobj,
2633 struct attribute *attr, char *buf)
2635 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2637 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2639 return a->show ? a->show(a, sbi, buf) : 0;
2642 static ssize_t ext4_attr_store(struct kobject *kobj,
2643 struct attribute *attr,
2644 const char *buf, size_t len)
2646 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2648 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2650 return a->store ? a->store(a, sbi, buf, len) : 0;
2653 static void ext4_sb_release(struct kobject *kobj)
2655 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2657 complete(&sbi->s_kobj_unregister);
2660 static const struct sysfs_ops ext4_attr_ops = {
2661 .show = ext4_attr_show,
2662 .store = ext4_attr_store,
2665 static struct kobj_type ext4_ktype = {
2666 .default_attrs = ext4_attrs,
2667 .sysfs_ops = &ext4_attr_ops,
2668 .release = ext4_sb_release,
2671 static void ext4_feat_release(struct kobject *kobj)
2673 complete(&ext4_feat->f_kobj_unregister);
2676 static struct kobj_type ext4_feat_ktype = {
2677 .default_attrs = ext4_feat_attrs,
2678 .sysfs_ops = &ext4_attr_ops,
2679 .release = ext4_feat_release,
2683 * Check whether this filesystem can be mounted based on
2684 * the features present and the RDONLY/RDWR mount requested.
2685 * Returns 1 if this filesystem can be mounted as requested,
2686 * 0 if it cannot be.
2688 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2690 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2691 ext4_msg(sb, KERN_ERR,
2692 "Couldn't mount because of "
2693 "unsupported optional features (%x)",
2694 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2695 ~EXT4_FEATURE_INCOMPAT_SUPP));
2702 /* Check that feature set is OK for a read-write mount */
2703 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2704 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2705 "unsupported optional features (%x)",
2706 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2707 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2711 * Large file size enabled file system can only be mounted
2712 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2714 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2715 if (sizeof(blkcnt_t) < sizeof(u64)) {
2716 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2717 "cannot be mounted RDWR without "
2722 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2723 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2724 ext4_msg(sb, KERN_ERR,
2725 "Can't support bigalloc feature without "
2726 "extents feature\n");
2733 * This function is called once a day if we have errors logged
2734 * on the file system
2736 static void print_daily_error_info(unsigned long arg)
2738 struct super_block *sb = (struct super_block *) arg;
2739 struct ext4_sb_info *sbi;
2740 struct ext4_super_block *es;
2745 if (es->s_error_count)
2746 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2747 le32_to_cpu(es->s_error_count));
2748 if (es->s_first_error_time) {
2749 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2750 sb->s_id, le32_to_cpu(es->s_first_error_time),
2751 (int) sizeof(es->s_first_error_func),
2752 es->s_first_error_func,
2753 le32_to_cpu(es->s_first_error_line));
2754 if (es->s_first_error_ino)
2755 printk(": inode %u",
2756 le32_to_cpu(es->s_first_error_ino));
2757 if (es->s_first_error_block)
2758 printk(": block %llu", (unsigned long long)
2759 le64_to_cpu(es->s_first_error_block));
2762 if (es->s_last_error_time) {
2763 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2764 sb->s_id, le32_to_cpu(es->s_last_error_time),
2765 (int) sizeof(es->s_last_error_func),
2766 es->s_last_error_func,
2767 le32_to_cpu(es->s_last_error_line));
2768 if (es->s_last_error_ino)
2769 printk(": inode %u",
2770 le32_to_cpu(es->s_last_error_ino));
2771 if (es->s_last_error_block)
2772 printk(": block %llu", (unsigned long long)
2773 le64_to_cpu(es->s_last_error_block));
2776 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2779 /* Find next suitable group and run ext4_init_inode_table */
2780 static int ext4_run_li_request(struct ext4_li_request *elr)
2782 struct ext4_group_desc *gdp = NULL;
2783 ext4_group_t group, ngroups;
2784 struct super_block *sb;
2785 unsigned long timeout = 0;
2789 ngroups = EXT4_SB(sb)->s_groups_count;
2791 for (group = elr->lr_next_group; group < ngroups; group++) {
2792 gdp = ext4_get_group_desc(sb, group, NULL);
2798 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2802 if (group == ngroups)
2807 ret = ext4_init_inode_table(sb, group,
2808 elr->lr_timeout ? 0 : 1);
2809 if (elr->lr_timeout == 0) {
2810 timeout = (jiffies - timeout) *
2811 elr->lr_sbi->s_li_wait_mult;
2812 elr->lr_timeout = timeout;
2814 elr->lr_next_sched = jiffies + elr->lr_timeout;
2815 elr->lr_next_group = group + 1;
2822 * Remove lr_request from the list_request and free the
2823 * request structure. Should be called with li_list_mtx held
2825 static void ext4_remove_li_request(struct ext4_li_request *elr)
2827 struct ext4_sb_info *sbi;
2834 list_del(&elr->lr_request);
2835 sbi->s_li_request = NULL;
2839 static void ext4_unregister_li_request(struct super_block *sb)
2841 mutex_lock(&ext4_li_mtx);
2842 if (!ext4_li_info) {
2843 mutex_unlock(&ext4_li_mtx);
2847 mutex_lock(&ext4_li_info->li_list_mtx);
2848 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2849 mutex_unlock(&ext4_li_info->li_list_mtx);
2850 mutex_unlock(&ext4_li_mtx);
2853 static struct task_struct *ext4_lazyinit_task;
2856 * This is the function where ext4lazyinit thread lives. It walks
2857 * through the request list searching for next scheduled filesystem.
2858 * When such a fs is found, run the lazy initialization request
2859 * (ext4_rn_li_request) and keep track of the time spend in this
2860 * function. Based on that time we compute next schedule time of
2861 * the request. When walking through the list is complete, compute
2862 * next waking time and put itself into sleep.
2864 static int ext4_lazyinit_thread(void *arg)
2866 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2867 struct list_head *pos, *n;
2868 struct ext4_li_request *elr;
2869 unsigned long next_wakeup, cur;
2871 BUG_ON(NULL == eli);
2875 next_wakeup = MAX_JIFFY_OFFSET;
2877 mutex_lock(&eli->li_list_mtx);
2878 if (list_empty(&eli->li_request_list)) {
2879 mutex_unlock(&eli->li_list_mtx);
2883 list_for_each_safe(pos, n, &eli->li_request_list) {
2884 elr = list_entry(pos, struct ext4_li_request,
2887 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2888 if (ext4_run_li_request(elr) != 0) {
2889 /* error, remove the lazy_init job */
2890 ext4_remove_li_request(elr);
2895 if (time_before(elr->lr_next_sched, next_wakeup))
2896 next_wakeup = elr->lr_next_sched;
2898 mutex_unlock(&eli->li_list_mtx);
2903 if ((time_after_eq(cur, next_wakeup)) ||
2904 (MAX_JIFFY_OFFSET == next_wakeup)) {
2909 schedule_timeout_interruptible(next_wakeup - cur);
2911 if (kthread_should_stop()) {
2912 ext4_clear_request_list();
2919 * It looks like the request list is empty, but we need
2920 * to check it under the li_list_mtx lock, to prevent any
2921 * additions into it, and of course we should lock ext4_li_mtx
2922 * to atomically free the list and ext4_li_info, because at
2923 * this point another ext4 filesystem could be registering
2926 mutex_lock(&ext4_li_mtx);
2927 mutex_lock(&eli->li_list_mtx);
2928 if (!list_empty(&eli->li_request_list)) {
2929 mutex_unlock(&eli->li_list_mtx);
2930 mutex_unlock(&ext4_li_mtx);
2933 mutex_unlock(&eli->li_list_mtx);
2934 kfree(ext4_li_info);
2935 ext4_li_info = NULL;
2936 mutex_unlock(&ext4_li_mtx);
2941 static void ext4_clear_request_list(void)
2943 struct list_head *pos, *n;
2944 struct ext4_li_request *elr;
2946 mutex_lock(&ext4_li_info->li_list_mtx);
2947 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2948 elr = list_entry(pos, struct ext4_li_request,
2950 ext4_remove_li_request(elr);
2952 mutex_unlock(&ext4_li_info->li_list_mtx);
2955 static int ext4_run_lazyinit_thread(void)
2957 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2958 ext4_li_info, "ext4lazyinit");
2959 if (IS_ERR(ext4_lazyinit_task)) {
2960 int err = PTR_ERR(ext4_lazyinit_task);
2961 ext4_clear_request_list();
2962 kfree(ext4_li_info);
2963 ext4_li_info = NULL;
2964 printk(KERN_CRIT "EXT4: error %d creating inode table "
2965 "initialization thread\n",
2969 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2974 * Check whether it make sense to run itable init. thread or not.
2975 * If there is at least one uninitialized inode table, return
2976 * corresponding group number, else the loop goes through all
2977 * groups and return total number of groups.
2979 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2981 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2982 struct ext4_group_desc *gdp = NULL;
2984 for (group = 0; group < ngroups; group++) {
2985 gdp = ext4_get_group_desc(sb, group, NULL);
2989 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2996 static int ext4_li_info_new(void)
2998 struct ext4_lazy_init *eli = NULL;
3000 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3004 INIT_LIST_HEAD(&eli->li_request_list);
3005 mutex_init(&eli->li_list_mtx);
3007 eli->li_state |= EXT4_LAZYINIT_QUIT;
3014 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3017 struct ext4_sb_info *sbi = EXT4_SB(sb);
3018 struct ext4_li_request *elr;
3021 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3027 elr->lr_next_group = start;
3030 * Randomize first schedule time of the request to
3031 * spread the inode table initialization requests
3034 get_random_bytes(&rnd, sizeof(rnd));
3035 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3036 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3041 static int ext4_register_li_request(struct super_block *sb,
3042 ext4_group_t first_not_zeroed)
3044 struct ext4_sb_info *sbi = EXT4_SB(sb);
3045 struct ext4_li_request *elr;
3046 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3049 if (sbi->s_li_request != NULL) {
3051 * Reset timeout so it can be computed again, because
3052 * s_li_wait_mult might have changed.
3054 sbi->s_li_request->lr_timeout = 0;
3058 if (first_not_zeroed == ngroups ||
3059 (sb->s_flags & MS_RDONLY) ||
3060 !test_opt(sb, INIT_INODE_TABLE))
3063 elr = ext4_li_request_new(sb, first_not_zeroed);
3067 mutex_lock(&ext4_li_mtx);
3069 if (NULL == ext4_li_info) {
3070 ret = ext4_li_info_new();
3075 mutex_lock(&ext4_li_info->li_list_mtx);
3076 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3077 mutex_unlock(&ext4_li_info->li_list_mtx);
3079 sbi->s_li_request = elr;
3081 * set elr to NULL here since it has been inserted to
3082 * the request_list and the removal and free of it is
3083 * handled by ext4_clear_request_list from now on.
3087 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3088 ret = ext4_run_lazyinit_thread();
3093 mutex_unlock(&ext4_li_mtx);
3100 * We do not need to lock anything since this is called on
3103 static void ext4_destroy_lazyinit_thread(void)
3106 * If thread exited earlier
3107 * there's nothing to be done.
3109 if (!ext4_li_info || !ext4_lazyinit_task)
3112 kthread_stop(ext4_lazyinit_task);
3115 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3117 char *orig_data = kstrdup(data, GFP_KERNEL);
3118 struct buffer_head *bh;
3119 struct ext4_super_block *es = NULL;
3120 struct ext4_sb_info *sbi;
3122 ext4_fsblk_t sb_block = get_sb_block(&data);
3123 ext4_fsblk_t logical_sb_block;
3124 unsigned long offset = 0;
3125 unsigned long journal_devnum = 0;
3126 unsigned long def_mount_opts;
3131 int blocksize, clustersize;
3132 unsigned int db_count;
3134 int needs_recovery, has_huge_files, has_bigalloc;
3137 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3138 ext4_group_t first_not_zeroed;
3140 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3144 sbi->s_blockgroup_lock =
3145 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3146 if (!sbi->s_blockgroup_lock) {
3150 sb->s_fs_info = sbi;
3151 sbi->s_mount_opt = 0;
3152 sbi->s_resuid = EXT4_DEF_RESUID;
3153 sbi->s_resgid = EXT4_DEF_RESGID;
3154 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3155 sbi->s_sb_block = sb_block;
3156 if (sb->s_bdev->bd_part)
3157 sbi->s_sectors_written_start =
3158 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3160 /* Cleanup superblock name */
3161 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3165 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3167 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3172 * The ext4 superblock will not be buffer aligned for other than 1kB
3173 * block sizes. We need to calculate the offset from buffer start.
3175 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3176 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3177 offset = do_div(logical_sb_block, blocksize);
3179 logical_sb_block = sb_block;
3182 if (!(bh = sb_bread(sb, logical_sb_block))) {
3183 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3187 * Note: s_es must be initialized as soon as possible because
3188 * some ext4 macro-instructions depend on its value
3190 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3192 sb->s_magic = le16_to_cpu(es->s_magic);
3193 if (sb->s_magic != EXT4_SUPER_MAGIC)
3195 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3197 /* Set defaults before we parse the mount options */
3198 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3199 set_opt(sb, INIT_INODE_TABLE);
3200 if (def_mount_opts & EXT4_DEFM_DEBUG)
3202 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3203 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3207 if (def_mount_opts & EXT4_DEFM_UID16)
3208 set_opt(sb, NO_UID32);
3209 /* xattr user namespace & acls are now defaulted on */
3210 #ifdef CONFIG_EXT4_FS_XATTR
3211 set_opt(sb, XATTR_USER);
3213 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3214 set_opt(sb, POSIX_ACL);
3216 set_opt(sb, MBLK_IO_SUBMIT);
3217 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3218 set_opt(sb, JOURNAL_DATA);
3219 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3220 set_opt(sb, ORDERED_DATA);
3221 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3222 set_opt(sb, WRITEBACK_DATA);
3224 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3225 set_opt(sb, ERRORS_PANIC);
3226 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3227 set_opt(sb, ERRORS_CONT);
3229 set_opt(sb, ERRORS_RO);
3230 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3231 set_opt(sb, BLOCK_VALIDITY);
3232 if (def_mount_opts & EXT4_DEFM_DISCARD)
3233 set_opt(sb, DISCARD);
3235 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3236 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3237 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3238 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3239 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3241 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3242 set_opt(sb, BARRIER);
3245 * enable delayed allocation by default
3246 * Use -o nodelalloc to turn it off
3248 if (!IS_EXT3_SB(sb) &&
3249 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3250 set_opt(sb, DELALLOC);
3253 * set default s_li_wait_mult for lazyinit, for the case there is
3254 * no mount option specified.
3256 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3258 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3259 &journal_devnum, &journal_ioprio, NULL, 0)) {
3260 ext4_msg(sb, KERN_WARNING,
3261 "failed to parse options in superblock: %s",
3262 sbi->s_es->s_mount_opts);
3264 if (!parse_options((char *) data, sb, &journal_devnum,
3265 &journal_ioprio, NULL, 0))
3268 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3269 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3270 "with data=journal disables delayed "
3271 "allocation and O_DIRECT support!\n");
3272 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3273 ext4_msg(sb, KERN_ERR, "can't mount with "
3274 "both data=journal and delalloc");
3277 if (test_opt(sb, DIOREAD_NOLOCK)) {
3278 ext4_msg(sb, KERN_ERR, "can't mount with "
3279 "both data=journal and delalloc");
3282 if (test_opt(sb, DELALLOC))
3283 clear_opt(sb, DELALLOC);
3286 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3287 if (test_opt(sb, DIOREAD_NOLOCK)) {
3288 if (blocksize < PAGE_SIZE) {
3289 ext4_msg(sb, KERN_ERR, "can't mount with "
3290 "dioread_nolock if block size != PAGE_SIZE");
3295 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3296 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3298 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3299 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3300 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3301 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3302 ext4_msg(sb, KERN_WARNING,
3303 "feature flags set on rev 0 fs, "
3304 "running e2fsck is recommended");
3306 if (IS_EXT2_SB(sb)) {
3307 if (ext2_feature_set_ok(sb))
3308 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3309 "using the ext4 subsystem");
3311 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3312 "to feature incompatibilities");
3317 if (IS_EXT3_SB(sb)) {
3318 if (ext3_feature_set_ok(sb))
3319 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3320 "using the ext4 subsystem");
3322 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3323 "to feature incompatibilities");
3329 * Check feature flags regardless of the revision level, since we
3330 * previously didn't change the revision level when setting the flags,
3331 * so there is a chance incompat flags are set on a rev 0 filesystem.
3333 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3336 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3337 blocksize > EXT4_MAX_BLOCK_SIZE) {
3338 ext4_msg(sb, KERN_ERR,
3339 "Unsupported filesystem blocksize %d", blocksize);
3343 if (sb->s_blocksize != blocksize) {
3344 /* Validate the filesystem blocksize */
3345 if (!sb_set_blocksize(sb, blocksize)) {
3346 ext4_msg(sb, KERN_ERR, "bad block size %d",
3352 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3353 offset = do_div(logical_sb_block, blocksize);
3354 bh = sb_bread(sb, logical_sb_block);
3356 ext4_msg(sb, KERN_ERR,
3357 "Can't read superblock on 2nd try");
3360 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3362 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3363 ext4_msg(sb, KERN_ERR,
3364 "Magic mismatch, very weird!");
3369 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3370 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3371 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3373 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3375 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3376 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3377 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3379 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3380 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3381 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3382 (!is_power_of_2(sbi->s_inode_size)) ||
3383 (sbi->s_inode_size > blocksize)) {
3384 ext4_msg(sb, KERN_ERR,
3385 "unsupported inode size: %d",
3389 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3390 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3393 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3394 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3395 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3396 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3397 !is_power_of_2(sbi->s_desc_size)) {
3398 ext4_msg(sb, KERN_ERR,
3399 "unsupported descriptor size %lu",
3404 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3406 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3407 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3408 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3411 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3412 if (sbi->s_inodes_per_block == 0)
3414 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3415 sbi->s_inodes_per_block;
3416 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3418 sbi->s_mount_state = le16_to_cpu(es->s_state);
3419 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3420 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3422 for (i = 0; i < 4; i++)
3423 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3424 sbi->s_def_hash_version = es->s_def_hash_version;
3425 i = le32_to_cpu(es->s_flags);
3426 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3427 sbi->s_hash_unsigned = 3;
3428 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3429 #ifdef __CHAR_UNSIGNED__
3430 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3431 sbi->s_hash_unsigned = 3;
3433 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3438 /* Handle clustersize */
3439 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3440 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3441 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3443 if (clustersize < blocksize) {
3444 ext4_msg(sb, KERN_ERR,
3445 "cluster size (%d) smaller than "
3446 "block size (%d)", clustersize, blocksize);
3449 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3450 le32_to_cpu(es->s_log_block_size);
3451 sbi->s_clusters_per_group =
3452 le32_to_cpu(es->s_clusters_per_group);
3453 if (sbi->s_clusters_per_group > blocksize * 8) {
3454 ext4_msg(sb, KERN_ERR,
3455 "#clusters per group too big: %lu",
3456 sbi->s_clusters_per_group);
3459 if (sbi->s_blocks_per_group !=
3460 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3461 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3462 "clusters per group (%lu) inconsistent",
3463 sbi->s_blocks_per_group,
3464 sbi->s_clusters_per_group);
3468 if (clustersize != blocksize) {
3469 ext4_warning(sb, "fragment/cluster size (%d) != "
3470 "block size (%d)", clustersize,
3472 clustersize = blocksize;
3474 if (sbi->s_blocks_per_group > blocksize * 8) {
3475 ext4_msg(sb, KERN_ERR,
3476 "#blocks per group too big: %lu",
3477 sbi->s_blocks_per_group);
3480 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3481 sbi->s_cluster_bits = 0;
3483 sbi->s_cluster_ratio = clustersize / blocksize;
3485 if (sbi->s_inodes_per_group > blocksize * 8) {
3486 ext4_msg(sb, KERN_ERR,
3487 "#inodes per group too big: %lu",
3488 sbi->s_inodes_per_group);
3493 * Test whether we have more sectors than will fit in sector_t,
3494 * and whether the max offset is addressable by the page cache.
3496 err = generic_check_addressable(sb->s_blocksize_bits,
3497 ext4_blocks_count(es));
3499 ext4_msg(sb, KERN_ERR, "filesystem"
3500 " too large to mount safely on this system");
3501 if (sizeof(sector_t) < 8)
3502 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3507 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3510 /* check blocks count against device size */
3511 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3512 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3513 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3514 "exceeds size of device (%llu blocks)",
3515 ext4_blocks_count(es), blocks_count);
3520 * It makes no sense for the first data block to be beyond the end
3521 * of the filesystem.
3523 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3524 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3525 "block %u is beyond end of filesystem (%llu)",
3526 le32_to_cpu(es->s_first_data_block),
3527 ext4_blocks_count(es));
3530 blocks_count = (ext4_blocks_count(es) -
3531 le32_to_cpu(es->s_first_data_block) +
3532 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3533 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3534 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3535 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3536 "(block count %llu, first data block %u, "
3537 "blocks per group %lu)", sbi->s_groups_count,
3538 ext4_blocks_count(es),
3539 le32_to_cpu(es->s_first_data_block),
3540 EXT4_BLOCKS_PER_GROUP(sb));
3543 sbi->s_groups_count = blocks_count;
3544 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3545 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3546 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3547 EXT4_DESC_PER_BLOCK(sb);
3548 sbi->s_group_desc = ext4_kvmalloc(db_count *
3549 sizeof(struct buffer_head *),
3551 if (sbi->s_group_desc == NULL) {
3552 ext4_msg(sb, KERN_ERR, "not enough memory");
3557 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3559 bgl_lock_init(sbi->s_blockgroup_lock);
3561 for (i = 0; i < db_count; i++) {
3562 block = descriptor_loc(sb, logical_sb_block, i);
3563 sbi->s_group_desc[i] = sb_bread(sb, block);
3564 if (!sbi->s_group_desc[i]) {
3565 ext4_msg(sb, KERN_ERR,
3566 "can't read group descriptor %d", i);
3571 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3572 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3575 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3576 if (!ext4_fill_flex_info(sb)) {
3577 ext4_msg(sb, KERN_ERR,
3578 "unable to initialize "
3579 "flex_bg meta info!");
3583 sbi->s_gdb_count = db_count;
3584 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3585 spin_lock_init(&sbi->s_next_gen_lock);
3587 init_timer(&sbi->s_err_report);
3588 sbi->s_err_report.function = print_daily_error_info;
3589 sbi->s_err_report.data = (unsigned long) sb;
3591 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3592 ext4_count_free_clusters(sb));
3594 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3595 ext4_count_free_inodes(sb));
3598 err = percpu_counter_init(&sbi->s_dirs_counter,
3599 ext4_count_dirs(sb));
3602 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3605 ext4_msg(sb, KERN_ERR, "insufficient memory");
3609 sbi->s_stripe = ext4_get_stripe_size(sbi);
3610 sbi->s_max_writeback_mb_bump = 128;
3613 * set up enough so that it can read an inode
3615 if (!test_opt(sb, NOLOAD) &&
3616 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3617 sb->s_op = &ext4_sops;
3619 sb->s_op = &ext4_nojournal_sops;
3620 sb->s_export_op = &ext4_export_ops;
3621 sb->s_xattr = ext4_xattr_handlers;
3623 sb->s_qcop = &ext4_qctl_operations;
3624 sb->dq_op = &ext4_quota_operations;
3626 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3628 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3629 mutex_init(&sbi->s_orphan_lock);
3630 sbi->s_resize_flags = 0;
3634 needs_recovery = (es->s_last_orphan != 0 ||
3635 EXT4_HAS_INCOMPAT_FEATURE(sb,
3636 EXT4_FEATURE_INCOMPAT_RECOVER));
3638 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3639 !(sb->s_flags & MS_RDONLY))
3640 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3644 * The first inode we look at is the journal inode. Don't try
3645 * root first: it may be modified in the journal!
3647 if (!test_opt(sb, NOLOAD) &&
3648 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3649 if (ext4_load_journal(sb, es, journal_devnum))
3651 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3652 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3653 ext4_msg(sb, KERN_ERR, "required journal recovery "
3654 "suppressed and not mounted read-only");
3655 goto failed_mount_wq;
3657 clear_opt(sb, DATA_FLAGS);
3658 sbi->s_journal = NULL;
3663 if (ext4_blocks_count(es) > 0xffffffffULL &&
3664 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3665 JBD2_FEATURE_INCOMPAT_64BIT)) {
3666 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3667 goto failed_mount_wq;
3670 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3671 jbd2_journal_set_features(sbi->s_journal,
3672 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3673 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3674 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3675 jbd2_journal_set_features(sbi->s_journal,
3676 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3677 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3678 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3680 jbd2_journal_clear_features(sbi->s_journal,
3681 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3682 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3685 /* We have now updated the journal if required, so we can
3686 * validate the data journaling mode. */
3687 switch (test_opt(sb, DATA_FLAGS)) {
3689 /* No mode set, assume a default based on the journal
3690 * capabilities: ORDERED_DATA if the journal can
3691 * cope, else JOURNAL_DATA
3693 if (jbd2_journal_check_available_features
3694 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3695 set_opt(sb, ORDERED_DATA);
3697 set_opt(sb, JOURNAL_DATA);
3700 case EXT4_MOUNT_ORDERED_DATA:
3701 case EXT4_MOUNT_WRITEBACK_DATA:
3702 if (!jbd2_journal_check_available_features
3703 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3704 ext4_msg(sb, KERN_ERR, "Journal does not support "
3705 "requested data journaling mode");
3706 goto failed_mount_wq;
3711 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3713 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3716 * The journal may have updated the bg summary counts, so we
3717 * need to update the global counters.
3719 percpu_counter_set(&sbi->s_freeclusters_counter,
3720 ext4_count_free_clusters(sb));
3721 percpu_counter_set(&sbi->s_freeinodes_counter,
3722 ext4_count_free_inodes(sb));
3723 percpu_counter_set(&sbi->s_dirs_counter,
3724 ext4_count_dirs(sb));
3725 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3729 * The maximum number of concurrent works can be high and
3730 * concurrency isn't really necessary. Limit it to 1.
3732 EXT4_SB(sb)->dio_unwritten_wq =
3733 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3734 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3735 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3736 goto failed_mount_wq;
3740 * The jbd2_journal_load will have done any necessary log recovery,
3741 * so we can safely mount the rest of the filesystem now.
3744 root = ext4_iget(sb, EXT4_ROOT_INO);
3746 ext4_msg(sb, KERN_ERR, "get root inode failed");
3747 ret = PTR_ERR(root);
3751 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3752 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3756 sb->s_root = d_alloc_root(root);
3759 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3764 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3766 /* determine the minimum size of new large inodes, if present */
3767 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3768 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3769 EXT4_GOOD_OLD_INODE_SIZE;
3770 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3771 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3772 if (sbi->s_want_extra_isize <
3773 le16_to_cpu(es->s_want_extra_isize))
3774 sbi->s_want_extra_isize =
3775 le16_to_cpu(es->s_want_extra_isize);
3776 if (sbi->s_want_extra_isize <
3777 le16_to_cpu(es->s_min_extra_isize))
3778 sbi->s_want_extra_isize =
3779 le16_to_cpu(es->s_min_extra_isize);
3782 /* Check if enough inode space is available */
3783 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3784 sbi->s_inode_size) {
3785 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3786 EXT4_GOOD_OLD_INODE_SIZE;
3787 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3791 err = ext4_setup_system_zone(sb);
3793 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3795 goto failed_mount4a;
3799 err = ext4_mb_init(sb, needs_recovery);
3801 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3806 err = ext4_register_li_request(sb, first_not_zeroed);
3810 sbi->s_kobj.kset = ext4_kset;
3811 init_completion(&sbi->s_kobj_unregister);
3812 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3817 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3818 ext4_orphan_cleanup(sb, es);
3819 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3820 if (needs_recovery) {
3821 ext4_msg(sb, KERN_INFO, "recovery complete");
3822 ext4_mark_recovery_complete(sb, es);
3824 if (EXT4_SB(sb)->s_journal) {
3825 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3826 descr = " journalled data mode";
3827 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3828 descr = " ordered data mode";
3830 descr = " writeback data mode";
3832 descr = "out journal";
3834 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3835 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3836 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3838 if (es->s_error_count)
3839 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3846 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3850 ext4_unregister_li_request(sb);
3852 ext4_mb_release(sb);
3854 ext4_ext_release(sb);
3855 ext4_release_system_zone(sb);
3860 ext4_msg(sb, KERN_ERR, "mount failed");
3861 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3863 if (sbi->s_journal) {
3864 jbd2_journal_destroy(sbi->s_journal);
3865 sbi->s_journal = NULL;
3868 del_timer(&sbi->s_err_report);
3869 if (sbi->s_flex_groups)
3870 ext4_kvfree(sbi->s_flex_groups);
3871 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3872 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3873 percpu_counter_destroy(&sbi->s_dirs_counter);
3874 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3876 kthread_stop(sbi->s_mmp_tsk);
3878 for (i = 0; i < db_count; i++)
3879 brelse(sbi->s_group_desc[i]);
3880 ext4_kvfree(sbi->s_group_desc);
3883 remove_proc_entry(sb->s_id, ext4_proc_root);
3886 for (i = 0; i < MAXQUOTAS; i++)
3887 kfree(sbi->s_qf_names[i]);
3889 ext4_blkdev_remove(sbi);
3892 sb->s_fs_info = NULL;
3893 kfree(sbi->s_blockgroup_lock);
3901 * Setup any per-fs journal parameters now. We'll do this both on
3902 * initial mount, once the journal has been initialised but before we've
3903 * done any recovery; and again on any subsequent remount.
3905 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3907 struct ext4_sb_info *sbi = EXT4_SB(sb);
3909 journal->j_commit_interval = sbi->s_commit_interval;
3910 journal->j_min_batch_time = sbi->s_min_batch_time;
3911 journal->j_max_batch_time = sbi->s_max_batch_time;
3913 write_lock(&journal->j_state_lock);
3914 if (test_opt(sb, BARRIER))
3915 journal->j_flags |= JBD2_BARRIER;
3917 journal->j_flags &= ~JBD2_BARRIER;
3918 if (test_opt(sb, DATA_ERR_ABORT))
3919 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3921 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3922 write_unlock(&journal->j_state_lock);
3925 static journal_t *ext4_get_journal(struct super_block *sb,
3926 unsigned int journal_inum)
3928 struct inode *journal_inode;
3931 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3933 /* First, test for the existence of a valid inode on disk. Bad
3934 * things happen if we iget() an unused inode, as the subsequent
3935 * iput() will try to delete it. */
3937 journal_inode = ext4_iget(sb, journal_inum);
3938 if (IS_ERR(journal_inode)) {
3939 ext4_msg(sb, KERN_ERR, "no journal found");
3942 if (!journal_inode->i_nlink) {
3943 make_bad_inode(journal_inode);
3944 iput(journal_inode);
3945 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3949 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3950 journal_inode, journal_inode->i_size);
3951 if (!S_ISREG(journal_inode->i_mode)) {
3952 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3953 iput(journal_inode);
3957 journal = jbd2_journal_init_inode(journal_inode);
3959 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3960 iput(journal_inode);
3963 journal->j_private = sb;
3964 ext4_init_journal_params(sb, journal);
3968 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3971 struct buffer_head *bh;
3975 int hblock, blocksize;
3976 ext4_fsblk_t sb_block;
3977 unsigned long offset;
3978 struct ext4_super_block *es;
3979 struct block_device *bdev;
3981 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3983 bdev = ext4_blkdev_get(j_dev, sb);
3987 blocksize = sb->s_blocksize;
3988 hblock = bdev_logical_block_size(bdev);
3989 if (blocksize < hblock) {
3990 ext4_msg(sb, KERN_ERR,
3991 "blocksize too small for journal device");
3995 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3996 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3997 set_blocksize(bdev, blocksize);
3998 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3999 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4000 "external journal");
4004 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4005 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4006 !(le32_to_cpu(es->s_feature_incompat) &
4007 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4008 ext4_msg(sb, KERN_ERR, "external journal has "
4014 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4015 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4020 len = ext4_blocks_count(es);
4021 start = sb_block + 1;
4022 brelse(bh); /* we're done with the superblock */
4024 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4025 start, len, blocksize);
4027 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4030 journal->j_private = sb;
4031 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4032 wait_on_buffer(journal->j_sb_buffer);
4033 if (!buffer_uptodate(journal->j_sb_buffer)) {
4034 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4037 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4038 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4039 "user (unsupported) - %d",
4040 be32_to_cpu(journal->j_superblock->s_nr_users));
4043 EXT4_SB(sb)->journal_bdev = bdev;
4044 ext4_init_journal_params(sb, journal);
4048 jbd2_journal_destroy(journal);
4050 ext4_blkdev_put(bdev);
4054 static int ext4_load_journal(struct super_block *sb,
4055 struct ext4_super_block *es,
4056 unsigned long journal_devnum)
4059 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4062 int really_read_only;
4064 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4066 if (journal_devnum &&
4067 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4068 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4069 "numbers have changed");
4070 journal_dev = new_decode_dev(journal_devnum);
4072 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4074 really_read_only = bdev_read_only(sb->s_bdev);
4077 * Are we loading a blank journal or performing recovery after a
4078 * crash? For recovery, we need to check in advance whether we
4079 * can get read-write access to the device.
4081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4082 if (sb->s_flags & MS_RDONLY) {
4083 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4084 "required on readonly filesystem");
4085 if (really_read_only) {
4086 ext4_msg(sb, KERN_ERR, "write access "
4087 "unavailable, cannot proceed");
4090 ext4_msg(sb, KERN_INFO, "write access will "
4091 "be enabled during recovery");
4095 if (journal_inum && journal_dev) {
4096 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4097 "and inode journals!");
4102 if (!(journal = ext4_get_journal(sb, journal_inum)))
4105 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4109 if (!(journal->j_flags & JBD2_BARRIER))
4110 ext4_msg(sb, KERN_INFO, "barriers disabled");
4112 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4113 err = jbd2_journal_update_format(journal);
4115 ext4_msg(sb, KERN_ERR, "error updating journal");
4116 jbd2_journal_destroy(journal);
4121 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4122 err = jbd2_journal_wipe(journal, !really_read_only);
4124 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4126 memcpy(save, ((char *) es) +
4127 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4128 err = jbd2_journal_load(journal);
4130 memcpy(((char *) es) + EXT4_S_ERR_START,
4131 save, EXT4_S_ERR_LEN);
4136 ext4_msg(sb, KERN_ERR, "error loading journal");
4137 jbd2_journal_destroy(journal);
4141 EXT4_SB(sb)->s_journal = journal;
4142 ext4_clear_journal_err(sb, es);
4144 if (!really_read_only && journal_devnum &&
4145 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4146 es->s_journal_dev = cpu_to_le32(journal_devnum);
4148 /* Make sure we flush the recovery flag to disk. */
4149 ext4_commit_super(sb, 1);
4155 static int ext4_commit_super(struct super_block *sb, int sync)
4157 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4158 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4161 if (!sbh || block_device_ejected(sb))
4163 if (buffer_write_io_error(sbh)) {
4165 * Oh, dear. A previous attempt to write the
4166 * superblock failed. This could happen because the
4167 * USB device was yanked out. Or it could happen to
4168 * be a transient write error and maybe the block will
4169 * be remapped. Nothing we can do but to retry the
4170 * write and hope for the best.
4172 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4173 "superblock detected");
4174 clear_buffer_write_io_error(sbh);
4175 set_buffer_uptodate(sbh);
4178 * If the file system is mounted read-only, don't update the
4179 * superblock write time. This avoids updating the superblock
4180 * write time when we are mounting the root file system
4181 * read/only but we need to replay the journal; at that point,
4182 * for people who are east of GMT and who make their clock
4183 * tick in localtime for Windows bug-for-bug compatibility,
4184 * the clock is set in the future, and this will cause e2fsck
4185 * to complain and force a full file system check.
4187 if (!(sb->s_flags & MS_RDONLY))
4188 es->s_wtime = cpu_to_le32(get_seconds());
4189 if (sb->s_bdev->bd_part)
4190 es->s_kbytes_written =
4191 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4192 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4193 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4195 es->s_kbytes_written =
4196 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4197 ext4_free_blocks_count_set(es,
4198 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4199 &EXT4_SB(sb)->s_freeclusters_counter)));
4200 es->s_free_inodes_count =
4201 cpu_to_le32(percpu_counter_sum_positive(
4202 &EXT4_SB(sb)->s_freeinodes_counter));
4204 BUFFER_TRACE(sbh, "marking dirty");
4205 mark_buffer_dirty(sbh);
4207 error = sync_dirty_buffer(sbh);
4211 error = buffer_write_io_error(sbh);
4213 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4215 clear_buffer_write_io_error(sbh);
4216 set_buffer_uptodate(sbh);
4223 * Have we just finished recovery? If so, and if we are mounting (or
4224 * remounting) the filesystem readonly, then we will end up with a
4225 * consistent fs on disk. Record that fact.
4227 static void ext4_mark_recovery_complete(struct super_block *sb,
4228 struct ext4_super_block *es)
4230 journal_t *journal = EXT4_SB(sb)->s_journal;
4232 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4233 BUG_ON(journal != NULL);
4236 jbd2_journal_lock_updates(journal);
4237 if (jbd2_journal_flush(journal) < 0)
4240 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4241 sb->s_flags & MS_RDONLY) {
4242 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4243 ext4_commit_super(sb, 1);
4247 jbd2_journal_unlock_updates(journal);
4251 * If we are mounting (or read-write remounting) a filesystem whose journal
4252 * has recorded an error from a previous lifetime, move that error to the
4253 * main filesystem now.
4255 static void ext4_clear_journal_err(struct super_block *sb,
4256 struct ext4_super_block *es)
4262 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4264 journal = EXT4_SB(sb)->s_journal;
4267 * Now check for any error status which may have been recorded in the
4268 * journal by a prior ext4_error() or ext4_abort()
4271 j_errno = jbd2_journal_errno(journal);
4275 errstr = ext4_decode_error(sb, j_errno, nbuf);
4276 ext4_warning(sb, "Filesystem error recorded "
4277 "from previous mount: %s", errstr);
4278 ext4_warning(sb, "Marking fs in need of filesystem check.");
4280 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4281 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4282 ext4_commit_super(sb, 1);
4284 jbd2_journal_clear_err(journal);
4289 * Force the running and committing transactions to commit,
4290 * and wait on the commit.
4292 int ext4_force_commit(struct super_block *sb)
4297 if (sb->s_flags & MS_RDONLY)
4300 journal = EXT4_SB(sb)->s_journal;
4302 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4303 ret = ext4_journal_force_commit(journal);
4309 static void ext4_write_super(struct super_block *sb)
4312 ext4_commit_super(sb, 1);
4316 static int ext4_sync_fs(struct super_block *sb, int wait)
4320 struct ext4_sb_info *sbi = EXT4_SB(sb);
4322 trace_ext4_sync_fs(sb, wait);
4323 flush_workqueue(sbi->dio_unwritten_wq);
4324 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4326 jbd2_log_wait_commit(sbi->s_journal, target);
4332 * LVM calls this function before a (read-only) snapshot is created. This
4333 * gives us a chance to flush the journal completely and mark the fs clean.
4335 * Note that only this function cannot bring a filesystem to be in a clean
4336 * state independently, because ext4 prevents a new handle from being started
4337 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4340 static int ext4_freeze(struct super_block *sb)
4345 if (sb->s_flags & MS_RDONLY)
4348 journal = EXT4_SB(sb)->s_journal;
4350 /* Now we set up the journal barrier. */
4351 jbd2_journal_lock_updates(journal);
4354 * Don't clear the needs_recovery flag if we failed to flush
4357 error = jbd2_journal_flush(journal);
4361 /* Journal blocked and flushed, clear needs_recovery flag. */
4362 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4363 error = ext4_commit_super(sb, 1);
4365 /* we rely on s_frozen to stop further updates */
4366 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4371 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4372 * flag here, even though the filesystem is not technically dirty yet.
4374 static int ext4_unfreeze(struct super_block *sb)
4376 if (sb->s_flags & MS_RDONLY)
4380 /* Reset the needs_recovery flag before the fs is unlocked. */
4381 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4382 ext4_commit_super(sb, 1);
4388 * Structure to save mount options for ext4_remount's benefit
4390 struct ext4_mount_options {
4391 unsigned long s_mount_opt;
4392 unsigned long s_mount_opt2;
4395 unsigned long s_commit_interval;
4396 u32 s_min_batch_time, s_max_batch_time;
4399 char *s_qf_names[MAXQUOTAS];
4403 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4405 struct ext4_super_block *es;
4406 struct ext4_sb_info *sbi = EXT4_SB(sb);
4407 ext4_fsblk_t n_blocks_count = 0;
4408 unsigned long old_sb_flags;
4409 struct ext4_mount_options old_opts;
4410 int enable_quota = 0;
4412 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4417 char *orig_data = kstrdup(data, GFP_KERNEL);
4419 /* Store the original options */
4421 old_sb_flags = sb->s_flags;
4422 old_opts.s_mount_opt = sbi->s_mount_opt;
4423 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4424 old_opts.s_resuid = sbi->s_resuid;
4425 old_opts.s_resgid = sbi->s_resgid;
4426 old_opts.s_commit_interval = sbi->s_commit_interval;
4427 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4428 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4430 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4431 for (i = 0; i < MAXQUOTAS; i++)
4432 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4434 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4435 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4438 * Allow the "check" option to be passed as a remount option.
4440 if (!parse_options(data, sb, NULL, &journal_ioprio,
4441 &n_blocks_count, 1)) {
4446 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4447 ext4_abort(sb, "Abort forced by user");
4449 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4450 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4454 if (sbi->s_journal) {
4455 ext4_init_journal_params(sb, sbi->s_journal);
4456 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4459 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4460 n_blocks_count > ext4_blocks_count(es)) {
4461 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4466 if (*flags & MS_RDONLY) {
4467 err = dquot_suspend(sb, -1);
4472 * First of all, the unconditional stuff we have to do
4473 * to disable replay of the journal when we next remount
4475 sb->s_flags |= MS_RDONLY;
4478 * OK, test if we are remounting a valid rw partition
4479 * readonly, and if so set the rdonly flag and then
4480 * mark the partition as valid again.
4482 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4483 (sbi->s_mount_state & EXT4_VALID_FS))
4484 es->s_state = cpu_to_le16(sbi->s_mount_state);
4487 ext4_mark_recovery_complete(sb, es);
4489 /* Make sure we can mount this feature set readwrite */
4490 if (!ext4_feature_set_ok(sb, 0)) {
4495 * Make sure the group descriptor checksums
4496 * are sane. If they aren't, refuse to remount r/w.
4498 for (g = 0; g < sbi->s_groups_count; g++) {
4499 struct ext4_group_desc *gdp =
4500 ext4_get_group_desc(sb, g, NULL);
4502 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4503 ext4_msg(sb, KERN_ERR,
4504 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4505 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4506 le16_to_cpu(gdp->bg_checksum));
4513 * If we have an unprocessed orphan list hanging
4514 * around from a previously readonly bdev mount,
4515 * require a full umount/remount for now.
4517 if (es->s_last_orphan) {
4518 ext4_msg(sb, KERN_WARNING, "Couldn't "
4519 "remount RDWR because of unprocessed "
4520 "orphan inode list. Please "
4521 "umount/remount instead");
4527 * Mounting a RDONLY partition read-write, so reread
4528 * and store the current valid flag. (It may have
4529 * been changed by e2fsck since we originally mounted
4533 ext4_clear_journal_err(sb, es);
4534 sbi->s_mount_state = le16_to_cpu(es->s_state);
4535 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4537 if (!ext4_setup_super(sb, es, 0))
4538 sb->s_flags &= ~MS_RDONLY;
4539 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4540 EXT4_FEATURE_INCOMPAT_MMP))
4541 if (ext4_multi_mount_protect(sb,
4542 le64_to_cpu(es->s_mmp_block))) {
4551 * Reinitialize lazy itable initialization thread based on
4554 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4555 ext4_unregister_li_request(sb);
4557 ext4_group_t first_not_zeroed;
4558 first_not_zeroed = ext4_has_uninit_itable(sb);
4559 ext4_register_li_request(sb, first_not_zeroed);
4562 ext4_setup_system_zone(sb);
4563 if (sbi->s_journal == NULL)
4564 ext4_commit_super(sb, 1);
4567 /* Release old quota file names */
4568 for (i = 0; i < MAXQUOTAS; i++)
4569 if (old_opts.s_qf_names[i] &&
4570 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4571 kfree(old_opts.s_qf_names[i]);
4575 dquot_resume(sb, -1);
4577 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4582 sb->s_flags = old_sb_flags;
4583 sbi->s_mount_opt = old_opts.s_mount_opt;
4584 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4585 sbi->s_resuid = old_opts.s_resuid;
4586 sbi->s_resgid = old_opts.s_resgid;
4587 sbi->s_commit_interval = old_opts.s_commit_interval;
4588 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4589 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4591 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4592 for (i = 0; i < MAXQUOTAS; i++) {
4593 if (sbi->s_qf_names[i] &&
4594 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4595 kfree(sbi->s_qf_names[i]);
4596 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4605 * Note: calculating the overhead so we can be compatible with
4606 * historical BSD practice is quite difficult in the face of
4607 * clusters/bigalloc. This is because multiple metadata blocks from
4608 * different block group can end up in the same allocation cluster.
4609 * Calculating the exact overhead in the face of clustered allocation
4610 * requires either O(all block bitmaps) in memory or O(number of block
4611 * groups**2) in time. We will still calculate the superblock for
4612 * older file systems --- and if we come across with a bigalloc file
4613 * system with zero in s_overhead_clusters the estimate will be close to
4614 * correct especially for very large cluster sizes --- but for newer
4615 * file systems, it's better to calculate this figure once at mkfs
4616 * time, and store it in the superblock. If the superblock value is
4617 * present (even for non-bigalloc file systems), we will use it.
4619 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4621 struct super_block *sb = dentry->d_sb;
4622 struct ext4_sb_info *sbi = EXT4_SB(sb);
4623 struct ext4_super_block *es = sbi->s_es;
4624 struct ext4_group_desc *gdp;
4628 if (test_opt(sb, MINIX_DF)) {
4629 sbi->s_overhead_last = 0;
4630 } else if (es->s_overhead_clusters) {
4631 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4632 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4633 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4634 ext4_fsblk_t overhead = 0;
4637 * Compute the overhead (FS structures). This is constant
4638 * for a given filesystem unless the number of block groups
4639 * changes so we cache the previous value until it does.
4643 * All of the blocks before first_data_block are
4646 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4649 * Add the overhead found in each block group
4651 for (i = 0; i < ngroups; i++) {
4652 gdp = ext4_get_group_desc(sb, i, NULL);
4653 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4656 sbi->s_overhead_last = overhead;
4658 sbi->s_blocks_last = ext4_blocks_count(es);
4661 buf->f_type = EXT4_SUPER_MAGIC;
4662 buf->f_bsize = sb->s_blocksize;
4663 buf->f_blocks = (ext4_blocks_count(es) -
4664 EXT4_C2B(sbi, sbi->s_overhead_last));
4665 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4666 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4667 /* prevent underflow in case that few free space is available */
4668 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4669 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4670 if (buf->f_bfree < ext4_r_blocks_count(es))
4672 buf->f_files = le32_to_cpu(es->s_inodes_count);
4673 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4674 buf->f_namelen = EXT4_NAME_LEN;
4675 fsid = le64_to_cpup((void *)es->s_uuid) ^
4676 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4677 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4678 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4683 /* Helper function for writing quotas on sync - we need to start transaction
4684 * before quota file is locked for write. Otherwise the are possible deadlocks:
4685 * Process 1 Process 2
4686 * ext4_create() quota_sync()
4687 * jbd2_journal_start() write_dquot()
4688 * dquot_initialize() down(dqio_mutex)
4689 * down(dqio_mutex) jbd2_journal_start()
4695 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4697 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4700 static int ext4_write_dquot(struct dquot *dquot)
4704 struct inode *inode;
4706 inode = dquot_to_inode(dquot);
4707 handle = ext4_journal_start(inode,
4708 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4710 return PTR_ERR(handle);
4711 ret = dquot_commit(dquot);
4712 err = ext4_journal_stop(handle);
4718 static int ext4_acquire_dquot(struct dquot *dquot)
4723 handle = ext4_journal_start(dquot_to_inode(dquot),
4724 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4726 return PTR_ERR(handle);
4727 ret = dquot_acquire(dquot);
4728 err = ext4_journal_stop(handle);
4734 static int ext4_release_dquot(struct dquot *dquot)
4739 handle = ext4_journal_start(dquot_to_inode(dquot),
4740 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4741 if (IS_ERR(handle)) {
4742 /* Release dquot anyway to avoid endless cycle in dqput() */
4743 dquot_release(dquot);
4744 return PTR_ERR(handle);
4746 ret = dquot_release(dquot);
4747 err = ext4_journal_stop(handle);
4753 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4755 /* Are we journaling quotas? */
4756 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4757 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4758 dquot_mark_dquot_dirty(dquot);
4759 return ext4_write_dquot(dquot);
4761 return dquot_mark_dquot_dirty(dquot);
4765 static int ext4_write_info(struct super_block *sb, int type)
4770 /* Data block + inode block */
4771 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4773 return PTR_ERR(handle);
4774 ret = dquot_commit_info(sb, type);
4775 err = ext4_journal_stop(handle);
4782 * Turn on quotas during mount time - we need to find
4783 * the quota file and such...
4785 static int ext4_quota_on_mount(struct super_block *sb, int type)
4787 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4788 EXT4_SB(sb)->s_jquota_fmt, type);
4792 * Standard function to be called on quota_on
4794 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4799 if (!test_opt(sb, QUOTA))
4802 /* Quotafile not on the same filesystem? */
4803 if (path->dentry->d_sb != sb)
4805 /* Journaling quota? */
4806 if (EXT4_SB(sb)->s_qf_names[type]) {
4807 /* Quotafile not in fs root? */
4808 if (path->dentry->d_parent != sb->s_root)
4809 ext4_msg(sb, KERN_WARNING,
4810 "Quota file not on filesystem root. "
4811 "Journaled quota will not work");
4815 * When we journal data on quota file, we have to flush journal to see
4816 * all updates to the file when we bypass pagecache...
4818 if (EXT4_SB(sb)->s_journal &&
4819 ext4_should_journal_data(path->dentry->d_inode)) {
4821 * We don't need to lock updates but journal_flush() could
4822 * otherwise be livelocked...
4824 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4825 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4826 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4831 return dquot_quota_on(sb, type, format_id, path);
4834 static int ext4_quota_off(struct super_block *sb, int type)
4836 struct inode *inode = sb_dqopt(sb)->files[type];
4839 /* Force all delayed allocation blocks to be allocated.
4840 * Caller already holds s_umount sem */
4841 if (test_opt(sb, DELALLOC))
4842 sync_filesystem(sb);
4847 /* Update modification times of quota files when userspace can
4848 * start looking at them */
4849 handle = ext4_journal_start(inode, 1);
4852 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4853 ext4_mark_inode_dirty(handle, inode);
4854 ext4_journal_stop(handle);
4857 return dquot_quota_off(sb, type);
4860 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4861 * acquiring the locks... As quota files are never truncated and quota code
4862 * itself serializes the operations (and no one else should touch the files)
4863 * we don't have to be afraid of races */
4864 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4865 size_t len, loff_t off)
4867 struct inode *inode = sb_dqopt(sb)->files[type];
4868 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4870 int offset = off & (sb->s_blocksize - 1);
4873 struct buffer_head *bh;
4874 loff_t i_size = i_size_read(inode);
4878 if (off+len > i_size)
4881 while (toread > 0) {
4882 tocopy = sb->s_blocksize - offset < toread ?
4883 sb->s_blocksize - offset : toread;
4884 bh = ext4_bread(NULL, inode, blk, 0, &err);
4887 if (!bh) /* A hole? */
4888 memset(data, 0, tocopy);
4890 memcpy(data, bh->b_data+offset, tocopy);
4900 /* Write to quotafile (we know the transaction is already started and has
4901 * enough credits) */
4902 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4903 const char *data, size_t len, loff_t off)
4905 struct inode *inode = sb_dqopt(sb)->files[type];
4906 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4908 int offset = off & (sb->s_blocksize - 1);
4909 struct buffer_head *bh;
4910 handle_t *handle = journal_current_handle();
4912 if (EXT4_SB(sb)->s_journal && !handle) {
4913 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4914 " cancelled because transaction is not started",
4915 (unsigned long long)off, (unsigned long long)len);
4919 * Since we account only one data block in transaction credits,
4920 * then it is impossible to cross a block boundary.
4922 if (sb->s_blocksize - offset < len) {
4923 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4924 " cancelled because not block aligned",
4925 (unsigned long long)off, (unsigned long long)len);
4929 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4930 bh = ext4_bread(handle, inode, blk, 1, &err);
4933 err = ext4_journal_get_write_access(handle, bh);
4939 memcpy(bh->b_data+offset, data, len);
4940 flush_dcache_page(bh->b_page);
4942 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4946 mutex_unlock(&inode->i_mutex);
4949 if (inode->i_size < off + len) {
4950 i_size_write(inode, off + len);
4951 EXT4_I(inode)->i_disksize = inode->i_size;
4952 ext4_mark_inode_dirty(handle, inode);
4954 mutex_unlock(&inode->i_mutex);
4960 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4961 const char *dev_name, void *data)
4963 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4966 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4967 static inline void register_as_ext2(void)
4969 int err = register_filesystem(&ext2_fs_type);
4972 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4975 static inline void unregister_as_ext2(void)
4977 unregister_filesystem(&ext2_fs_type);
4980 static inline int ext2_feature_set_ok(struct super_block *sb)
4982 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4984 if (sb->s_flags & MS_RDONLY)
4986 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4990 MODULE_ALIAS("ext2");
4992 static inline void register_as_ext2(void) { }
4993 static inline void unregister_as_ext2(void) { }
4994 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4997 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4998 static inline void register_as_ext3(void)
5000 int err = register_filesystem(&ext3_fs_type);
5003 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5006 static inline void unregister_as_ext3(void)
5008 unregister_filesystem(&ext3_fs_type);
5011 static inline int ext3_feature_set_ok(struct super_block *sb)
5013 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5015 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5017 if (sb->s_flags & MS_RDONLY)
5019 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5023 MODULE_ALIAS("ext3");
5025 static inline void register_as_ext3(void) { }
5026 static inline void unregister_as_ext3(void) { }
5027 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5030 static struct file_system_type ext4_fs_type = {
5031 .owner = THIS_MODULE,
5033 .mount = ext4_mount,
5034 .kill_sb = kill_block_super,
5035 .fs_flags = FS_REQUIRES_DEV,
5038 static int __init ext4_init_feat_adverts(void)
5040 struct ext4_features *ef;
5043 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5047 ef->f_kobj.kset = ext4_kset;
5048 init_completion(&ef->f_kobj_unregister);
5049 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5062 static void ext4_exit_feat_adverts(void)
5064 kobject_put(&ext4_feat->f_kobj);
5065 wait_for_completion(&ext4_feat->f_kobj_unregister);
5069 /* Shared across all ext4 file systems */
5070 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5071 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5073 static int __init ext4_init_fs(void)
5077 ext4_check_flag_values();
5079 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5080 mutex_init(&ext4__aio_mutex[i]);
5081 init_waitqueue_head(&ext4__ioend_wq[i]);
5084 err = ext4_init_pageio();
5087 err = ext4_init_system_zone();
5090 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5093 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5095 err = ext4_init_feat_adverts();
5099 err = ext4_init_mballoc();
5103 err = ext4_init_xattr();
5106 err = init_inodecache();
5111 err = register_filesystem(&ext4_fs_type);
5115 ext4_li_info = NULL;
5116 mutex_init(&ext4_li_mtx);
5119 unregister_as_ext2();
5120 unregister_as_ext3();
5121 destroy_inodecache();
5125 ext4_exit_mballoc();
5127 ext4_exit_feat_adverts();
5130 remove_proc_entry("fs/ext4", NULL);
5131 kset_unregister(ext4_kset);
5133 ext4_exit_system_zone();
5139 static void __exit ext4_exit_fs(void)
5141 ext4_destroy_lazyinit_thread();
5142 unregister_as_ext2();
5143 unregister_as_ext3();
5144 unregister_filesystem(&ext4_fs_type);
5145 destroy_inodecache();
5147 ext4_exit_mballoc();
5148 ext4_exit_feat_adverts();
5149 remove_proc_entry("fs/ext4", NULL);
5150 kset_unregister(ext4_kset);
5151 ext4_exit_system_zone();
5155 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5156 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5157 MODULE_LICENSE("GPL");
5158 module_init(ext4_init_fs)
5159 module_exit(ext4_exit_fs)