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 (sb->s_flags & MS_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_dev, Opt_journal_checksum, Opt_journal_async_commit,
1340 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1341 Opt_data_err_abort, Opt_data_err_ignore,
1342 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1343 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1344 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1345 Opt_usrquota, Opt_grpquota, Opt_i_version,
1346 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1347 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1348 Opt_inode_readahead_blks, Opt_journal_ioprio,
1349 Opt_dioread_nolock, Opt_dioread_lock,
1350 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1353 static const match_table_t tokens = {
1354 {Opt_bsd_df, "bsddf"},
1355 {Opt_minix_df, "minixdf"},
1356 {Opt_grpid, "grpid"},
1357 {Opt_grpid, "bsdgroups"},
1358 {Opt_nogrpid, "nogrpid"},
1359 {Opt_nogrpid, "sysvgroups"},
1360 {Opt_resgid, "resgid=%u"},
1361 {Opt_resuid, "resuid=%u"},
1363 {Opt_err_cont, "errors=continue"},
1364 {Opt_err_panic, "errors=panic"},
1365 {Opt_err_ro, "errors=remount-ro"},
1366 {Opt_nouid32, "nouid32"},
1367 {Opt_debug, "debug"},
1368 {Opt_oldalloc, "oldalloc"},
1369 {Opt_orlov, "orlov"},
1370 {Opt_user_xattr, "user_xattr"},
1371 {Opt_nouser_xattr, "nouser_xattr"},
1373 {Opt_noacl, "noacl"},
1374 {Opt_noload, "noload"},
1375 {Opt_noload, "norecovery"},
1378 {Opt_commit, "commit=%u"},
1379 {Opt_min_batch_time, "min_batch_time=%u"},
1380 {Opt_max_batch_time, "max_batch_time=%u"},
1381 {Opt_journal_dev, "journal_dev=%u"},
1382 {Opt_journal_checksum, "journal_checksum"},
1383 {Opt_journal_async_commit, "journal_async_commit"},
1384 {Opt_abort, "abort"},
1385 {Opt_data_journal, "data=journal"},
1386 {Opt_data_ordered, "data=ordered"},
1387 {Opt_data_writeback, "data=writeback"},
1388 {Opt_data_err_abort, "data_err=abort"},
1389 {Opt_data_err_ignore, "data_err=ignore"},
1390 {Opt_offusrjquota, "usrjquota="},
1391 {Opt_usrjquota, "usrjquota=%s"},
1392 {Opt_offgrpjquota, "grpjquota="},
1393 {Opt_grpjquota, "grpjquota=%s"},
1394 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1395 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1396 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1397 {Opt_grpquota, "grpquota"},
1398 {Opt_noquota, "noquota"},
1399 {Opt_quota, "quota"},
1400 {Opt_usrquota, "usrquota"},
1401 {Opt_barrier, "barrier=%u"},
1402 {Opt_barrier, "barrier"},
1403 {Opt_nobarrier, "nobarrier"},
1404 {Opt_i_version, "i_version"},
1405 {Opt_stripe, "stripe=%u"},
1406 {Opt_delalloc, "delalloc"},
1407 {Opt_nodelalloc, "nodelalloc"},
1408 {Opt_mblk_io_submit, "mblk_io_submit"},
1409 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1410 {Opt_block_validity, "block_validity"},
1411 {Opt_noblock_validity, "noblock_validity"},
1412 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1413 {Opt_journal_ioprio, "journal_ioprio=%u"},
1414 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1415 {Opt_auto_da_alloc, "auto_da_alloc"},
1416 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1417 {Opt_dioread_nolock, "dioread_nolock"},
1418 {Opt_dioread_lock, "dioread_lock"},
1419 {Opt_discard, "discard"},
1420 {Opt_nodiscard, "nodiscard"},
1421 {Opt_init_itable, "init_itable=%u"},
1422 {Opt_init_itable, "init_itable"},
1423 {Opt_noinit_itable, "noinit_itable"},
1427 static ext4_fsblk_t get_sb_block(void **data)
1429 ext4_fsblk_t sb_block;
1430 char *options = (char *) *data;
1432 if (!options || strncmp(options, "sb=", 3) != 0)
1433 return 1; /* Default location */
1436 /* TODO: use simple_strtoll with >32bit ext4 */
1437 sb_block = simple_strtoul(options, &options, 0);
1438 if (*options && *options != ',') {
1439 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1443 if (*options == ',')
1445 *data = (void *) options;
1450 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1451 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1452 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1455 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1457 struct ext4_sb_info *sbi = EXT4_SB(sb);
1460 if (sb_any_quota_loaded(sb) &&
1461 !sbi->s_qf_names[qtype]) {
1462 ext4_msg(sb, KERN_ERR,
1463 "Cannot change journaled "
1464 "quota options when quota turned on");
1467 qname = match_strdup(args);
1469 ext4_msg(sb, KERN_ERR,
1470 "Not enough memory for storing quotafile name");
1473 if (sbi->s_qf_names[qtype] &&
1474 strcmp(sbi->s_qf_names[qtype], qname)) {
1475 ext4_msg(sb, KERN_ERR,
1476 "%s quota file already specified", QTYPE2NAME(qtype));
1480 sbi->s_qf_names[qtype] = qname;
1481 if (strchr(sbi->s_qf_names[qtype], '/')) {
1482 ext4_msg(sb, KERN_ERR,
1483 "quotafile must be on filesystem root");
1484 kfree(sbi->s_qf_names[qtype]);
1485 sbi->s_qf_names[qtype] = NULL;
1492 static int clear_qf_name(struct super_block *sb, int qtype)
1495 struct ext4_sb_info *sbi = EXT4_SB(sb);
1497 if (sb_any_quota_loaded(sb) &&
1498 sbi->s_qf_names[qtype]) {
1499 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1500 " when quota turned on");
1504 * The space will be released later when all options are confirmed
1507 sbi->s_qf_names[qtype] = NULL;
1512 static int parse_options(char *options, struct super_block *sb,
1513 unsigned long *journal_devnum,
1514 unsigned int *journal_ioprio,
1517 struct ext4_sb_info *sbi = EXT4_SB(sb);
1519 substring_t args[MAX_OPT_ARGS];
1529 while ((p = strsep(&options, ",")) != NULL) {
1535 * Initialize args struct so we know whether arg was
1536 * found; some options take optional arguments.
1538 args[0].to = args[0].from = NULL;
1539 token = match_token(p, tokens, args);
1542 clear_opt(sb, MINIX_DF);
1545 set_opt(sb, MINIX_DF);
1551 clear_opt(sb, GRPID);
1554 if (match_int(&args[0], &option))
1556 sbi->s_resuid = option;
1559 if (match_int(&args[0], &option))
1561 sbi->s_resgid = option;
1564 /* handled by get_sb_block() instead of here */
1565 /* *sb_block = match_int(&args[0]); */
1568 clear_opt(sb, ERRORS_MASK);
1569 set_opt(sb, ERRORS_PANIC);
1572 clear_opt(sb, ERRORS_MASK);
1573 set_opt(sb, ERRORS_RO);
1576 clear_opt(sb, ERRORS_MASK);
1577 set_opt(sb, ERRORS_CONT);
1580 set_opt(sb, NO_UID32);
1586 ext4_msg(sb, KERN_WARNING,
1587 "Ignoring deprecated oldalloc option");
1590 ext4_msg(sb, KERN_WARNING,
1591 "Ignoring deprecated orlov option");
1593 #ifdef CONFIG_EXT4_FS_XATTR
1594 case Opt_user_xattr:
1595 set_opt(sb, XATTR_USER);
1597 case Opt_nouser_xattr:
1598 clear_opt(sb, XATTR_USER);
1601 case Opt_user_xattr:
1602 case Opt_nouser_xattr:
1603 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1606 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1608 set_opt(sb, POSIX_ACL);
1611 clear_opt(sb, POSIX_ACL);
1616 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1619 case Opt_journal_dev:
1621 ext4_msg(sb, KERN_ERR,
1622 "Cannot specify journal on remount");
1625 if (match_int(&args[0], &option))
1627 *journal_devnum = option;
1629 case Opt_journal_checksum:
1630 set_opt(sb, JOURNAL_CHECKSUM);
1632 case Opt_journal_async_commit:
1633 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1634 set_opt(sb, JOURNAL_CHECKSUM);
1637 set_opt(sb, NOLOAD);
1640 if (match_int(&args[0], &option))
1645 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1646 sbi->s_commit_interval = HZ * option;
1648 case Opt_max_batch_time:
1649 if (match_int(&args[0], &option))
1654 option = EXT4_DEF_MAX_BATCH_TIME;
1655 sbi->s_max_batch_time = option;
1657 case Opt_min_batch_time:
1658 if (match_int(&args[0], &option))
1662 sbi->s_min_batch_time = option;
1664 case Opt_data_journal:
1665 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1667 case Opt_data_ordered:
1668 data_opt = EXT4_MOUNT_ORDERED_DATA;
1670 case Opt_data_writeback:
1671 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1674 if (!sbi->s_journal)
1675 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1676 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1677 ext4_msg(sb, KERN_ERR,
1678 "Cannot change data mode on remount");
1682 clear_opt(sb, DATA_FLAGS);
1683 sbi->s_mount_opt |= data_opt;
1686 case Opt_data_err_abort:
1687 set_opt(sb, DATA_ERR_ABORT);
1689 case Opt_data_err_ignore:
1690 clear_opt(sb, DATA_ERR_ABORT);
1694 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1698 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1701 case Opt_offusrjquota:
1702 if (!clear_qf_name(sb, USRQUOTA))
1705 case Opt_offgrpjquota:
1706 if (!clear_qf_name(sb, GRPQUOTA))
1710 case Opt_jqfmt_vfsold:
1711 qfmt = QFMT_VFS_OLD;
1713 case Opt_jqfmt_vfsv0:
1716 case Opt_jqfmt_vfsv1:
1719 if (sb_any_quota_loaded(sb) &&
1720 sbi->s_jquota_fmt != qfmt) {
1721 ext4_msg(sb, KERN_ERR, "Cannot change "
1722 "journaled quota options when "
1726 sbi->s_jquota_fmt = qfmt;
1731 set_opt(sb, USRQUOTA);
1735 set_opt(sb, GRPQUOTA);
1738 if (sb_any_quota_loaded(sb)) {
1739 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1740 "options when quota turned on");
1743 clear_opt(sb, QUOTA);
1744 clear_opt(sb, USRQUOTA);
1745 clear_opt(sb, GRPQUOTA);
1751 ext4_msg(sb, KERN_ERR,
1752 "quota options not supported");
1756 case Opt_offusrjquota:
1757 case Opt_offgrpjquota:
1758 case Opt_jqfmt_vfsold:
1759 case Opt_jqfmt_vfsv0:
1760 case Opt_jqfmt_vfsv1:
1761 ext4_msg(sb, KERN_ERR,
1762 "journaled quota options not supported");
1768 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1771 clear_opt(sb, BARRIER);
1775 if (match_int(&args[0], &option))
1778 option = 1; /* No argument, default to 1 */
1780 set_opt(sb, BARRIER);
1782 clear_opt(sb, BARRIER);
1785 ext4_msg(sb, KERN_WARNING,
1786 "Ignoring deprecated nobh option");
1789 ext4_msg(sb, KERN_WARNING,
1790 "Ignoring deprecated bh option");
1793 sb->s_flags |= MS_I_VERSION;
1795 case Opt_nodelalloc:
1796 clear_opt(sb, DELALLOC);
1797 clear_opt2(sb, EXPLICIT_DELALLOC);
1799 case Opt_mblk_io_submit:
1800 set_opt(sb, MBLK_IO_SUBMIT);
1802 case Opt_nomblk_io_submit:
1803 clear_opt(sb, MBLK_IO_SUBMIT);
1806 if (match_int(&args[0], &option))
1810 sbi->s_stripe = option;
1813 set_opt(sb, DELALLOC);
1814 set_opt2(sb, EXPLICIT_DELALLOC);
1816 case Opt_block_validity:
1817 set_opt(sb, BLOCK_VALIDITY);
1819 case Opt_noblock_validity:
1820 clear_opt(sb, BLOCK_VALIDITY);
1822 case Opt_inode_readahead_blks:
1823 if (match_int(&args[0], &option))
1825 if (option < 0 || option > (1 << 30))
1827 if (option && !is_power_of_2(option)) {
1828 ext4_msg(sb, KERN_ERR,
1829 "EXT4-fs: inode_readahead_blks"
1830 " must be a power of 2");
1833 sbi->s_inode_readahead_blks = option;
1835 case Opt_journal_ioprio:
1836 if (match_int(&args[0], &option))
1838 if (option < 0 || option > 7)
1840 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1843 case Opt_noauto_da_alloc:
1844 set_opt(sb, NO_AUTO_DA_ALLOC);
1846 case Opt_auto_da_alloc:
1848 if (match_int(&args[0], &option))
1851 option = 1; /* No argument, default to 1 */
1853 clear_opt(sb, NO_AUTO_DA_ALLOC);
1855 set_opt(sb,NO_AUTO_DA_ALLOC);
1858 set_opt(sb, DISCARD);
1861 clear_opt(sb, DISCARD);
1863 case Opt_dioread_nolock:
1864 set_opt(sb, DIOREAD_NOLOCK);
1866 case Opt_dioread_lock:
1867 clear_opt(sb, DIOREAD_NOLOCK);
1869 case Opt_init_itable:
1870 set_opt(sb, INIT_INODE_TABLE);
1872 if (match_int(&args[0], &option))
1875 option = EXT4_DEF_LI_WAIT_MULT;
1878 sbi->s_li_wait_mult = option;
1880 case Opt_noinit_itable:
1881 clear_opt(sb, INIT_INODE_TABLE);
1884 ext4_msg(sb, KERN_ERR,
1885 "Unrecognized mount option \"%s\" "
1886 "or missing value", p);
1891 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1892 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1893 clear_opt(sb, USRQUOTA);
1895 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1896 clear_opt(sb, GRPQUOTA);
1898 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1899 ext4_msg(sb, KERN_ERR, "old and new quota "
1904 if (!sbi->s_jquota_fmt) {
1905 ext4_msg(sb, KERN_ERR, "journaled quota format "
1910 if (sbi->s_jquota_fmt) {
1911 ext4_msg(sb, KERN_ERR, "journaled quota format "
1912 "specified with no journaling "
1921 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1924 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1928 ext4_msg(sb, KERN_ERR, "revision level too high, "
1929 "forcing read-only mode");
1934 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1935 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1936 "running e2fsck is recommended");
1937 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1938 ext4_msg(sb, KERN_WARNING,
1939 "warning: mounting fs with errors, "
1940 "running e2fsck is recommended");
1941 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1942 le16_to_cpu(es->s_mnt_count) >=
1943 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1944 ext4_msg(sb, KERN_WARNING,
1945 "warning: maximal mount count reached, "
1946 "running e2fsck is recommended");
1947 else if (le32_to_cpu(es->s_checkinterval) &&
1948 (le32_to_cpu(es->s_lastcheck) +
1949 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1950 ext4_msg(sb, KERN_WARNING,
1951 "warning: checktime reached, "
1952 "running e2fsck is recommended");
1953 if (!sbi->s_journal)
1954 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1955 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1956 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1957 le16_add_cpu(&es->s_mnt_count, 1);
1958 es->s_mtime = cpu_to_le32(get_seconds());
1959 ext4_update_dynamic_rev(sb);
1961 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1963 ext4_commit_super(sb, 1);
1965 if (test_opt(sb, DEBUG))
1966 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1967 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1969 sbi->s_groups_count,
1970 EXT4_BLOCKS_PER_GROUP(sb),
1971 EXT4_INODES_PER_GROUP(sb),
1972 sbi->s_mount_opt, sbi->s_mount_opt2);
1974 cleancache_init_fs(sb);
1978 static int ext4_fill_flex_info(struct super_block *sb)
1980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1981 struct ext4_group_desc *gdp = NULL;
1982 ext4_group_t flex_group_count;
1983 ext4_group_t flex_group;
1984 unsigned int groups_per_flex = 0;
1988 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1989 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1990 sbi->s_log_groups_per_flex = 0;
1993 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1995 /* We allocate both existing and potentially added groups */
1996 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1997 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1998 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1999 size = flex_group_count * sizeof(struct flex_groups);
2000 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2001 if (sbi->s_flex_groups == NULL) {
2002 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2007 for (i = 0; i < sbi->s_groups_count; i++) {
2008 gdp = ext4_get_group_desc(sb, i, NULL);
2010 flex_group = ext4_flex_group(sbi, i);
2011 atomic_add(ext4_free_inodes_count(sb, gdp),
2012 &sbi->s_flex_groups[flex_group].free_inodes);
2013 atomic_add(ext4_free_group_clusters(sb, gdp),
2014 &sbi->s_flex_groups[flex_group].free_clusters);
2015 atomic_add(ext4_used_dirs_count(sb, gdp),
2016 &sbi->s_flex_groups[flex_group].used_dirs);
2024 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2025 struct ext4_group_desc *gdp)
2029 if (sbi->s_es->s_feature_ro_compat &
2030 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2031 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2032 __le32 le_group = cpu_to_le32(block_group);
2034 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2035 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2036 crc = crc16(crc, (__u8 *)gdp, offset);
2037 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2038 /* for checksum of struct ext4_group_desc do the rest...*/
2039 if ((sbi->s_es->s_feature_incompat &
2040 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2041 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2042 crc = crc16(crc, (__u8 *)gdp + offset,
2043 le16_to_cpu(sbi->s_es->s_desc_size) -
2047 return cpu_to_le16(crc);
2050 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2053 if ((sbi->s_es->s_feature_ro_compat &
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2055 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2061 /* Called at mount-time, super-block is locked */
2062 static int ext4_check_descriptors(struct super_block *sb,
2063 ext4_group_t *first_not_zeroed)
2065 struct ext4_sb_info *sbi = EXT4_SB(sb);
2066 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2067 ext4_fsblk_t last_block;
2068 ext4_fsblk_t block_bitmap;
2069 ext4_fsblk_t inode_bitmap;
2070 ext4_fsblk_t inode_table;
2071 int flexbg_flag = 0;
2072 ext4_group_t i, grp = sbi->s_groups_count;
2074 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2077 ext4_debug("Checking group descriptors");
2079 for (i = 0; i < sbi->s_groups_count; i++) {
2080 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2082 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2083 last_block = ext4_blocks_count(sbi->s_es) - 1;
2085 last_block = first_block +
2086 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2088 if ((grp == sbi->s_groups_count) &&
2089 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2092 block_bitmap = ext4_block_bitmap(sb, gdp);
2093 if (block_bitmap < first_block || block_bitmap > last_block) {
2094 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2095 "Block bitmap for group %u not in group "
2096 "(block %llu)!", i, block_bitmap);
2099 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2100 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2101 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2102 "Inode bitmap for group %u not in group "
2103 "(block %llu)!", i, inode_bitmap);
2106 inode_table = ext4_inode_table(sb, gdp);
2107 if (inode_table < first_block ||
2108 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2109 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2110 "Inode table for group %u not in group "
2111 "(block %llu)!", i, inode_table);
2114 ext4_lock_group(sb, i);
2115 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2117 "Checksum for group %u failed (%u!=%u)",
2118 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2119 gdp)), le16_to_cpu(gdp->bg_checksum));
2120 if (!(sb->s_flags & MS_RDONLY)) {
2121 ext4_unlock_group(sb, i);
2125 ext4_unlock_group(sb, i);
2127 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2129 if (NULL != first_not_zeroed)
2130 *first_not_zeroed = grp;
2132 ext4_free_blocks_count_set(sbi->s_es,
2133 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2134 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2138 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2139 * the superblock) which were deleted from all directories, but held open by
2140 * a process at the time of a crash. We walk the list and try to delete these
2141 * inodes at recovery time (only with a read-write filesystem).
2143 * In order to keep the orphan inode chain consistent during traversal (in
2144 * case of crash during recovery), we link each inode into the superblock
2145 * orphan list_head and handle it the same way as an inode deletion during
2146 * normal operation (which journals the operations for us).
2148 * We only do an iget() and an iput() on each inode, which is very safe if we
2149 * accidentally point at an in-use or already deleted inode. The worst that
2150 * can happen in this case is that we get a "bit already cleared" message from
2151 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2152 * e2fsck was run on this filesystem, and it must have already done the orphan
2153 * inode cleanup for us, so we can safely abort without any further action.
2155 static void ext4_orphan_cleanup(struct super_block *sb,
2156 struct ext4_super_block *es)
2158 unsigned int s_flags = sb->s_flags;
2159 int nr_orphans = 0, nr_truncates = 0;
2163 if (!es->s_last_orphan) {
2164 jbd_debug(4, "no orphan inodes to clean up\n");
2168 if (bdev_read_only(sb->s_bdev)) {
2169 ext4_msg(sb, KERN_ERR, "write access "
2170 "unavailable, skipping orphan cleanup");
2174 /* Check if feature set would not allow a r/w mount */
2175 if (!ext4_feature_set_ok(sb, 0)) {
2176 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2177 "unknown ROCOMPAT features");
2181 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2182 if (es->s_last_orphan)
2183 jbd_debug(1, "Errors on filesystem, "
2184 "clearing orphan list.\n");
2185 es->s_last_orphan = 0;
2186 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2190 if (s_flags & MS_RDONLY) {
2191 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2192 sb->s_flags &= ~MS_RDONLY;
2195 /* Needed for iput() to work correctly and not trash data */
2196 sb->s_flags |= MS_ACTIVE;
2197 /* Turn on quotas so that they are updated correctly */
2198 for (i = 0; i < MAXQUOTAS; i++) {
2199 if (EXT4_SB(sb)->s_qf_names[i]) {
2200 int ret = ext4_quota_on_mount(sb, i);
2202 ext4_msg(sb, KERN_ERR,
2203 "Cannot turn on journaled "
2204 "quota: error %d", ret);
2209 while (es->s_last_orphan) {
2210 struct inode *inode;
2212 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2213 if (IS_ERR(inode)) {
2214 es->s_last_orphan = 0;
2218 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2219 dquot_initialize(inode);
2220 if (inode->i_nlink) {
2221 ext4_msg(sb, KERN_DEBUG,
2222 "%s: truncating inode %lu to %lld bytes",
2223 __func__, inode->i_ino, inode->i_size);
2224 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2225 inode->i_ino, inode->i_size);
2226 ext4_truncate(inode);
2229 ext4_msg(sb, KERN_DEBUG,
2230 "%s: deleting unreferenced inode %lu",
2231 __func__, inode->i_ino);
2232 jbd_debug(2, "deleting unreferenced inode %lu\n",
2236 iput(inode); /* The delete magic happens here! */
2239 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2242 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2243 PLURAL(nr_orphans));
2245 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2246 PLURAL(nr_truncates));
2248 /* Turn quotas off */
2249 for (i = 0; i < MAXQUOTAS; i++) {
2250 if (sb_dqopt(sb)->files[i])
2251 dquot_quota_off(sb, i);
2254 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2258 * Maximal extent format file size.
2259 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2260 * extent format containers, within a sector_t, and within i_blocks
2261 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2262 * so that won't be a limiting factor.
2264 * However there is other limiting factor. We do store extents in the form
2265 * of starting block and length, hence the resulting length of the extent
2266 * covering maximum file size must fit into on-disk format containers as
2267 * well. Given that length is always by 1 unit bigger than max unit (because
2268 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2270 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2272 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2275 loff_t upper_limit = MAX_LFS_FILESIZE;
2277 /* small i_blocks in vfs inode? */
2278 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2280 * CONFIG_LBDAF is not enabled implies the inode
2281 * i_block represent total blocks in 512 bytes
2282 * 32 == size of vfs inode i_blocks * 8
2284 upper_limit = (1LL << 32) - 1;
2286 /* total blocks in file system block size */
2287 upper_limit >>= (blkbits - 9);
2288 upper_limit <<= blkbits;
2292 * 32-bit extent-start container, ee_block. We lower the maxbytes
2293 * by one fs block, so ee_len can cover the extent of maximum file
2296 res = (1LL << 32) - 1;
2299 /* Sanity check against vm- & vfs- imposed limits */
2300 if (res > upper_limit)
2307 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2308 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2309 * We need to be 1 filesystem block less than the 2^48 sector limit.
2311 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2313 loff_t res = EXT4_NDIR_BLOCKS;
2316 /* This is calculated to be the largest file size for a dense, block
2317 * mapped file such that the file's total number of 512-byte sectors,
2318 * including data and all indirect blocks, does not exceed (2^48 - 1).
2320 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2321 * number of 512-byte sectors of the file.
2324 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2326 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2327 * the inode i_block field represents total file blocks in
2328 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2330 upper_limit = (1LL << 32) - 1;
2332 /* total blocks in file system block size */
2333 upper_limit >>= (bits - 9);
2337 * We use 48 bit ext4_inode i_blocks
2338 * With EXT4_HUGE_FILE_FL set the i_blocks
2339 * represent total number of blocks in
2340 * file system block size
2342 upper_limit = (1LL << 48) - 1;
2346 /* indirect blocks */
2348 /* double indirect blocks */
2349 meta_blocks += 1 + (1LL << (bits-2));
2350 /* tripple indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2353 upper_limit -= meta_blocks;
2354 upper_limit <<= bits;
2356 res += 1LL << (bits-2);
2357 res += 1LL << (2*(bits-2));
2358 res += 1LL << (3*(bits-2));
2360 if (res > upper_limit)
2363 if (res > MAX_LFS_FILESIZE)
2364 res = MAX_LFS_FILESIZE;
2369 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2370 ext4_fsblk_t logical_sb_block, int nr)
2372 struct ext4_sb_info *sbi = EXT4_SB(sb);
2373 ext4_group_t bg, first_meta_bg;
2376 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2378 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2380 return logical_sb_block + nr + 1;
2381 bg = sbi->s_desc_per_block * nr;
2382 if (ext4_bg_has_super(sb, bg))
2385 return (has_super + ext4_group_first_block_no(sb, bg));
2389 * ext4_get_stripe_size: Get the stripe size.
2390 * @sbi: In memory super block info
2392 * If we have specified it via mount option, then
2393 * use the mount option value. If the value specified at mount time is
2394 * greater than the blocks per group use the super block value.
2395 * If the super block value is greater than blocks per group return 0.
2396 * Allocator needs it be less than blocks per group.
2399 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2401 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2402 unsigned long stripe_width =
2403 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2406 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2407 ret = sbi->s_stripe;
2408 else if (stripe_width <= sbi->s_blocks_per_group)
2410 else if (stride <= sbi->s_blocks_per_group)
2416 * If the stripe width is 1, this makes no sense and
2417 * we set it to 0 to turn off stripe handling code.
2428 struct attribute attr;
2429 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2430 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2431 const char *, size_t);
2435 static int parse_strtoul(const char *buf,
2436 unsigned long max, unsigned long *value)
2440 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2441 endp = skip_spaces(endp);
2442 if (*endp || *value > max)
2448 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2449 struct ext4_sb_info *sbi,
2452 return snprintf(buf, PAGE_SIZE, "%llu\n",
2454 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2457 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2458 struct ext4_sb_info *sbi, char *buf)
2460 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2462 if (!sb->s_bdev->bd_part)
2463 return snprintf(buf, PAGE_SIZE, "0\n");
2464 return snprintf(buf, PAGE_SIZE, "%lu\n",
2465 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2466 sbi->s_sectors_written_start) >> 1);
2469 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi, char *buf)
2472 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2474 if (!sb->s_bdev->bd_part)
2475 return snprintf(buf, PAGE_SIZE, "0\n");
2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
2477 (unsigned long long)(sbi->s_kbytes_written +
2478 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2479 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2482 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi, char *buf)
2485 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2488 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2489 struct ext4_sb_info *sbi, char *buf)
2491 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2494 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2495 struct ext4_sb_info *sbi,
2496 const char *buf, size_t count)
2500 if (parse_strtoul(buf, 0x40000000, &t))
2503 if (t && !is_power_of_2(t))
2506 sbi->s_inode_readahead_blks = t;
2510 static ssize_t sbi_ui_show(struct ext4_attr *a,
2511 struct ext4_sb_info *sbi, char *buf)
2513 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2515 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2518 static ssize_t sbi_ui_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2522 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2525 if (parse_strtoul(buf, 0xffffffff, &t))
2531 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2532 static struct ext4_attr ext4_attr_##_name = { \
2533 .attr = {.name = __stringify(_name), .mode = _mode }, \
2536 .offset = offsetof(struct ext4_sb_info, _elname), \
2538 #define EXT4_ATTR(name, mode, show, store) \
2539 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2541 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2542 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2543 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2544 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2545 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2546 #define ATTR_LIST(name) &ext4_attr_##name.attr
2548 EXT4_RO_ATTR(delayed_allocation_blocks);
2549 EXT4_RO_ATTR(session_write_kbytes);
2550 EXT4_RO_ATTR(lifetime_write_kbytes);
2551 EXT4_RO_ATTR(extent_cache_hits);
2552 EXT4_RO_ATTR(extent_cache_misses);
2553 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2554 inode_readahead_blks_store, s_inode_readahead_blks);
2555 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2556 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2557 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2558 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2559 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2560 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2561 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2562 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2564 static struct attribute *ext4_attrs[] = {
2565 ATTR_LIST(delayed_allocation_blocks),
2566 ATTR_LIST(session_write_kbytes),
2567 ATTR_LIST(lifetime_write_kbytes),
2568 ATTR_LIST(extent_cache_hits),
2569 ATTR_LIST(extent_cache_misses),
2570 ATTR_LIST(inode_readahead_blks),
2571 ATTR_LIST(inode_goal),
2572 ATTR_LIST(mb_stats),
2573 ATTR_LIST(mb_max_to_scan),
2574 ATTR_LIST(mb_min_to_scan),
2575 ATTR_LIST(mb_order2_req),
2576 ATTR_LIST(mb_stream_req),
2577 ATTR_LIST(mb_group_prealloc),
2578 ATTR_LIST(max_writeback_mb_bump),
2582 /* Features this copy of ext4 supports */
2583 EXT4_INFO_ATTR(lazy_itable_init);
2584 EXT4_INFO_ATTR(batched_discard);
2586 static struct attribute *ext4_feat_attrs[] = {
2587 ATTR_LIST(lazy_itable_init),
2588 ATTR_LIST(batched_discard),
2592 static ssize_t ext4_attr_show(struct kobject *kobj,
2593 struct attribute *attr, char *buf)
2595 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2597 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2599 return a->show ? a->show(a, sbi, buf) : 0;
2602 static ssize_t ext4_attr_store(struct kobject *kobj,
2603 struct attribute *attr,
2604 const char *buf, size_t len)
2606 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2608 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2610 return a->store ? a->store(a, sbi, buf, len) : 0;
2613 static void ext4_sb_release(struct kobject *kobj)
2615 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2617 complete(&sbi->s_kobj_unregister);
2620 static const struct sysfs_ops ext4_attr_ops = {
2621 .show = ext4_attr_show,
2622 .store = ext4_attr_store,
2625 static struct kobj_type ext4_ktype = {
2626 .default_attrs = ext4_attrs,
2627 .sysfs_ops = &ext4_attr_ops,
2628 .release = ext4_sb_release,
2631 static void ext4_feat_release(struct kobject *kobj)
2633 complete(&ext4_feat->f_kobj_unregister);
2636 static struct kobj_type ext4_feat_ktype = {
2637 .default_attrs = ext4_feat_attrs,
2638 .sysfs_ops = &ext4_attr_ops,
2639 .release = ext4_feat_release,
2643 * Check whether this filesystem can be mounted based on
2644 * the features present and the RDONLY/RDWR mount requested.
2645 * Returns 1 if this filesystem can be mounted as requested,
2646 * 0 if it cannot be.
2648 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2650 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2651 ext4_msg(sb, KERN_ERR,
2652 "Couldn't mount because of "
2653 "unsupported optional features (%x)",
2654 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2655 ~EXT4_FEATURE_INCOMPAT_SUPP));
2662 /* Check that feature set is OK for a read-write mount */
2663 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2664 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2665 "unsupported optional features (%x)",
2666 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2667 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2671 * Large file size enabled file system can only be mounted
2672 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2674 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2675 if (sizeof(blkcnt_t) < sizeof(u64)) {
2676 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2677 "cannot be mounted RDWR without "
2682 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2683 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2684 ext4_msg(sb, KERN_ERR,
2685 "Can't support bigalloc feature without "
2686 "extents feature\n");
2693 * This function is called once a day if we have errors logged
2694 * on the file system
2696 static void print_daily_error_info(unsigned long arg)
2698 struct super_block *sb = (struct super_block *) arg;
2699 struct ext4_sb_info *sbi;
2700 struct ext4_super_block *es;
2705 if (es->s_error_count)
2706 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2707 le32_to_cpu(es->s_error_count));
2708 if (es->s_first_error_time) {
2709 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2710 sb->s_id, le32_to_cpu(es->s_first_error_time),
2711 (int) sizeof(es->s_first_error_func),
2712 es->s_first_error_func,
2713 le32_to_cpu(es->s_first_error_line));
2714 if (es->s_first_error_ino)
2715 printk(": inode %u",
2716 le32_to_cpu(es->s_first_error_ino));
2717 if (es->s_first_error_block)
2718 printk(": block %llu", (unsigned long long)
2719 le64_to_cpu(es->s_first_error_block));
2722 if (es->s_last_error_time) {
2723 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2724 sb->s_id, le32_to_cpu(es->s_last_error_time),
2725 (int) sizeof(es->s_last_error_func),
2726 es->s_last_error_func,
2727 le32_to_cpu(es->s_last_error_line));
2728 if (es->s_last_error_ino)
2729 printk(": inode %u",
2730 le32_to_cpu(es->s_last_error_ino));
2731 if (es->s_last_error_block)
2732 printk(": block %llu", (unsigned long long)
2733 le64_to_cpu(es->s_last_error_block));
2736 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2739 /* Find next suitable group and run ext4_init_inode_table */
2740 static int ext4_run_li_request(struct ext4_li_request *elr)
2742 struct ext4_group_desc *gdp = NULL;
2743 ext4_group_t group, ngroups;
2744 struct super_block *sb;
2745 unsigned long timeout = 0;
2749 ngroups = EXT4_SB(sb)->s_groups_count;
2751 for (group = elr->lr_next_group; group < ngroups; group++) {
2752 gdp = ext4_get_group_desc(sb, group, NULL);
2758 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2762 if (group == ngroups)
2767 ret = ext4_init_inode_table(sb, group,
2768 elr->lr_timeout ? 0 : 1);
2769 if (elr->lr_timeout == 0) {
2770 timeout = (jiffies - timeout) *
2771 elr->lr_sbi->s_li_wait_mult;
2772 elr->lr_timeout = timeout;
2774 elr->lr_next_sched = jiffies + elr->lr_timeout;
2775 elr->lr_next_group = group + 1;
2782 * Remove lr_request from the list_request and free the
2783 * request structure. Should be called with li_list_mtx held
2785 static void ext4_remove_li_request(struct ext4_li_request *elr)
2787 struct ext4_sb_info *sbi;
2794 list_del(&elr->lr_request);
2795 sbi->s_li_request = NULL;
2799 static void ext4_unregister_li_request(struct super_block *sb)
2801 mutex_lock(&ext4_li_mtx);
2802 if (!ext4_li_info) {
2803 mutex_unlock(&ext4_li_mtx);
2807 mutex_lock(&ext4_li_info->li_list_mtx);
2808 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2809 mutex_unlock(&ext4_li_info->li_list_mtx);
2810 mutex_unlock(&ext4_li_mtx);
2813 static struct task_struct *ext4_lazyinit_task;
2816 * This is the function where ext4lazyinit thread lives. It walks
2817 * through the request list searching for next scheduled filesystem.
2818 * When such a fs is found, run the lazy initialization request
2819 * (ext4_rn_li_request) and keep track of the time spend in this
2820 * function. Based on that time we compute next schedule time of
2821 * the request. When walking through the list is complete, compute
2822 * next waking time and put itself into sleep.
2824 static int ext4_lazyinit_thread(void *arg)
2826 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2827 struct list_head *pos, *n;
2828 struct ext4_li_request *elr;
2829 unsigned long next_wakeup, cur;
2831 BUG_ON(NULL == eli);
2835 next_wakeup = MAX_JIFFY_OFFSET;
2837 mutex_lock(&eli->li_list_mtx);
2838 if (list_empty(&eli->li_request_list)) {
2839 mutex_unlock(&eli->li_list_mtx);
2843 list_for_each_safe(pos, n, &eli->li_request_list) {
2844 elr = list_entry(pos, struct ext4_li_request,
2847 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2848 if (ext4_run_li_request(elr) != 0) {
2849 /* error, remove the lazy_init job */
2850 ext4_remove_li_request(elr);
2855 if (time_before(elr->lr_next_sched, next_wakeup))
2856 next_wakeup = elr->lr_next_sched;
2858 mutex_unlock(&eli->li_list_mtx);
2863 if ((time_after_eq(cur, next_wakeup)) ||
2864 (MAX_JIFFY_OFFSET == next_wakeup)) {
2869 schedule_timeout_interruptible(next_wakeup - cur);
2871 if (kthread_should_stop()) {
2872 ext4_clear_request_list();
2879 * It looks like the request list is empty, but we need
2880 * to check it under the li_list_mtx lock, to prevent any
2881 * additions into it, and of course we should lock ext4_li_mtx
2882 * to atomically free the list and ext4_li_info, because at
2883 * this point another ext4 filesystem could be registering
2886 mutex_lock(&ext4_li_mtx);
2887 mutex_lock(&eli->li_list_mtx);
2888 if (!list_empty(&eli->li_request_list)) {
2889 mutex_unlock(&eli->li_list_mtx);
2890 mutex_unlock(&ext4_li_mtx);
2893 mutex_unlock(&eli->li_list_mtx);
2894 kfree(ext4_li_info);
2895 ext4_li_info = NULL;
2896 mutex_unlock(&ext4_li_mtx);
2901 static void ext4_clear_request_list(void)
2903 struct list_head *pos, *n;
2904 struct ext4_li_request *elr;
2906 mutex_lock(&ext4_li_info->li_list_mtx);
2907 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2908 elr = list_entry(pos, struct ext4_li_request,
2910 ext4_remove_li_request(elr);
2912 mutex_unlock(&ext4_li_info->li_list_mtx);
2915 static int ext4_run_lazyinit_thread(void)
2917 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2918 ext4_li_info, "ext4lazyinit");
2919 if (IS_ERR(ext4_lazyinit_task)) {
2920 int err = PTR_ERR(ext4_lazyinit_task);
2921 ext4_clear_request_list();
2922 kfree(ext4_li_info);
2923 ext4_li_info = NULL;
2924 printk(KERN_CRIT "EXT4: error %d creating inode table "
2925 "initialization thread\n",
2929 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2934 * Check whether it make sense to run itable init. thread or not.
2935 * If there is at least one uninitialized inode table, return
2936 * corresponding group number, else the loop goes through all
2937 * groups and return total number of groups.
2939 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2941 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2942 struct ext4_group_desc *gdp = NULL;
2944 for (group = 0; group < ngroups; group++) {
2945 gdp = ext4_get_group_desc(sb, group, NULL);
2949 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2956 static int ext4_li_info_new(void)
2958 struct ext4_lazy_init *eli = NULL;
2960 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2964 INIT_LIST_HEAD(&eli->li_request_list);
2965 mutex_init(&eli->li_list_mtx);
2967 eli->li_state |= EXT4_LAZYINIT_QUIT;
2974 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2977 struct ext4_sb_info *sbi = EXT4_SB(sb);
2978 struct ext4_li_request *elr;
2981 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2987 elr->lr_next_group = start;
2990 * Randomize first schedule time of the request to
2991 * spread the inode table initialization requests
2994 get_random_bytes(&rnd, sizeof(rnd));
2995 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2996 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3001 static int ext4_register_li_request(struct super_block *sb,
3002 ext4_group_t first_not_zeroed)
3004 struct ext4_sb_info *sbi = EXT4_SB(sb);
3005 struct ext4_li_request *elr;
3006 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3009 if (sbi->s_li_request != NULL) {
3011 * Reset timeout so it can be computed again, because
3012 * s_li_wait_mult might have changed.
3014 sbi->s_li_request->lr_timeout = 0;
3018 if (first_not_zeroed == ngroups ||
3019 (sb->s_flags & MS_RDONLY) ||
3020 !test_opt(sb, INIT_INODE_TABLE))
3023 elr = ext4_li_request_new(sb, first_not_zeroed);
3027 mutex_lock(&ext4_li_mtx);
3029 if (NULL == ext4_li_info) {
3030 ret = ext4_li_info_new();
3035 mutex_lock(&ext4_li_info->li_list_mtx);
3036 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3037 mutex_unlock(&ext4_li_info->li_list_mtx);
3039 sbi->s_li_request = elr;
3041 * set elr to NULL here since it has been inserted to
3042 * the request_list and the removal and free of it is
3043 * handled by ext4_clear_request_list from now on.
3047 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3048 ret = ext4_run_lazyinit_thread();
3053 mutex_unlock(&ext4_li_mtx);
3060 * We do not need to lock anything since this is called on
3063 static void ext4_destroy_lazyinit_thread(void)
3066 * If thread exited earlier
3067 * there's nothing to be done.
3069 if (!ext4_li_info || !ext4_lazyinit_task)
3072 kthread_stop(ext4_lazyinit_task);
3075 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3077 char *orig_data = kstrdup(data, GFP_KERNEL);
3078 struct buffer_head *bh;
3079 struct ext4_super_block *es = NULL;
3080 struct ext4_sb_info *sbi;
3082 ext4_fsblk_t sb_block = get_sb_block(&data);
3083 ext4_fsblk_t logical_sb_block;
3084 unsigned long offset = 0;
3085 unsigned long journal_devnum = 0;
3086 unsigned long def_mount_opts;
3091 int blocksize, clustersize;
3092 unsigned int db_count;
3094 int needs_recovery, has_huge_files, has_bigalloc;
3097 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3098 ext4_group_t first_not_zeroed;
3100 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3104 sbi->s_blockgroup_lock =
3105 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3106 if (!sbi->s_blockgroup_lock) {
3110 sb->s_fs_info = sbi;
3111 sbi->s_mount_opt = 0;
3112 sbi->s_resuid = EXT4_DEF_RESUID;
3113 sbi->s_resgid = EXT4_DEF_RESGID;
3114 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3115 sbi->s_sb_block = sb_block;
3116 if (sb->s_bdev->bd_part)
3117 sbi->s_sectors_written_start =
3118 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3120 /* Cleanup superblock name */
3121 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3125 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3127 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3132 * The ext4 superblock will not be buffer aligned for other than 1kB
3133 * block sizes. We need to calculate the offset from buffer start.
3135 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3136 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3137 offset = do_div(logical_sb_block, blocksize);
3139 logical_sb_block = sb_block;
3142 if (!(bh = sb_bread(sb, logical_sb_block))) {
3143 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3147 * Note: s_es must be initialized as soon as possible because
3148 * some ext4 macro-instructions depend on its value
3150 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3152 sb->s_magic = le16_to_cpu(es->s_magic);
3153 if (sb->s_magic != EXT4_SUPER_MAGIC)
3155 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3157 /* Set defaults before we parse the mount options */
3158 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3159 set_opt(sb, INIT_INODE_TABLE);
3160 if (def_mount_opts & EXT4_DEFM_DEBUG)
3162 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3164 if (def_mount_opts & EXT4_DEFM_UID16)
3165 set_opt(sb, NO_UID32);
3166 /* xattr user namespace & acls are now defaulted on */
3167 #ifdef CONFIG_EXT4_FS_XATTR
3168 set_opt(sb, XATTR_USER);
3170 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3171 set_opt(sb, POSIX_ACL);
3173 set_opt(sb, MBLK_IO_SUBMIT);
3174 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3175 set_opt(sb, JOURNAL_DATA);
3176 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3177 set_opt(sb, ORDERED_DATA);
3178 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3179 set_opt(sb, WRITEBACK_DATA);
3181 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3182 set_opt(sb, ERRORS_PANIC);
3183 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3184 set_opt(sb, ERRORS_CONT);
3186 set_opt(sb, ERRORS_RO);
3187 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3188 set_opt(sb, BLOCK_VALIDITY);
3189 if (def_mount_opts & EXT4_DEFM_DISCARD)
3190 set_opt(sb, DISCARD);
3192 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3193 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3194 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3195 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3196 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3198 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3199 set_opt(sb, BARRIER);
3202 * enable delayed allocation by default
3203 * Use -o nodelalloc to turn it off
3205 if (!IS_EXT3_SB(sb) &&
3206 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3207 set_opt(sb, DELALLOC);
3210 * set default s_li_wait_mult for lazyinit, for the case there is
3211 * no mount option specified.
3213 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3215 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3216 &journal_devnum, &journal_ioprio, 0)) {
3217 ext4_msg(sb, KERN_WARNING,
3218 "failed to parse options in superblock: %s",
3219 sbi->s_es->s_mount_opts);
3221 if (!parse_options((char *) data, sb, &journal_devnum,
3222 &journal_ioprio, 0))
3225 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3226 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3227 "with data=journal disables delayed "
3228 "allocation and O_DIRECT support!\n");
3229 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3230 ext4_msg(sb, KERN_ERR, "can't mount with "
3231 "both data=journal and delalloc");
3234 if (test_opt(sb, DIOREAD_NOLOCK)) {
3235 ext4_msg(sb, KERN_ERR, "can't mount with "
3236 "both data=journal and delalloc");
3239 if (test_opt(sb, DELALLOC))
3240 clear_opt(sb, DELALLOC);
3243 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3244 if (test_opt(sb, DIOREAD_NOLOCK)) {
3245 if (blocksize < PAGE_SIZE) {
3246 ext4_msg(sb, KERN_ERR, "can't mount with "
3247 "dioread_nolock if block size != PAGE_SIZE");
3252 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3253 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3255 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3256 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3257 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3258 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3259 ext4_msg(sb, KERN_WARNING,
3260 "feature flags set on rev 0 fs, "
3261 "running e2fsck is recommended");
3263 if (IS_EXT2_SB(sb)) {
3264 if (ext2_feature_set_ok(sb))
3265 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3266 "using the ext4 subsystem");
3268 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3269 "to feature incompatibilities");
3274 if (IS_EXT3_SB(sb)) {
3275 if (ext3_feature_set_ok(sb))
3276 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3277 "using the ext4 subsystem");
3279 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3280 "to feature incompatibilities");
3286 * Check feature flags regardless of the revision level, since we
3287 * previously didn't change the revision level when setting the flags,
3288 * so there is a chance incompat flags are set on a rev 0 filesystem.
3290 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3293 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3294 blocksize > EXT4_MAX_BLOCK_SIZE) {
3295 ext4_msg(sb, KERN_ERR,
3296 "Unsupported filesystem blocksize %d", blocksize);
3300 if (sb->s_blocksize != blocksize) {
3301 /* Validate the filesystem blocksize */
3302 if (!sb_set_blocksize(sb, blocksize)) {
3303 ext4_msg(sb, KERN_ERR, "bad block size %d",
3309 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3310 offset = do_div(logical_sb_block, blocksize);
3311 bh = sb_bread(sb, logical_sb_block);
3313 ext4_msg(sb, KERN_ERR,
3314 "Can't read superblock on 2nd try");
3317 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3319 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3320 ext4_msg(sb, KERN_ERR,
3321 "Magic mismatch, very weird!");
3326 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3327 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3328 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3330 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3332 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3333 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3334 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3336 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3337 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3338 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3339 (!is_power_of_2(sbi->s_inode_size)) ||
3340 (sbi->s_inode_size > blocksize)) {
3341 ext4_msg(sb, KERN_ERR,
3342 "unsupported inode size: %d",
3346 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3347 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3350 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3351 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3352 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3353 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3354 !is_power_of_2(sbi->s_desc_size)) {
3355 ext4_msg(sb, KERN_ERR,
3356 "unsupported descriptor size %lu",
3361 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3363 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3364 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3365 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3368 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3369 if (sbi->s_inodes_per_block == 0)
3371 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3372 sbi->s_inodes_per_block;
3373 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3375 sbi->s_mount_state = le16_to_cpu(es->s_state);
3376 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3377 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3379 for (i = 0; i < 4; i++)
3380 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3381 sbi->s_def_hash_version = es->s_def_hash_version;
3382 i = le32_to_cpu(es->s_flags);
3383 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3384 sbi->s_hash_unsigned = 3;
3385 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3386 #ifdef __CHAR_UNSIGNED__
3387 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3388 sbi->s_hash_unsigned = 3;
3390 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3395 /* Handle clustersize */
3396 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3397 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3398 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3400 if (clustersize < blocksize) {
3401 ext4_msg(sb, KERN_ERR,
3402 "cluster size (%d) smaller than "
3403 "block size (%d)", clustersize, blocksize);
3406 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3407 le32_to_cpu(es->s_log_block_size);
3408 sbi->s_clusters_per_group =
3409 le32_to_cpu(es->s_clusters_per_group);
3410 if (sbi->s_clusters_per_group > blocksize * 8) {
3411 ext4_msg(sb, KERN_ERR,
3412 "#clusters per group too big: %lu",
3413 sbi->s_clusters_per_group);
3416 if (sbi->s_blocks_per_group !=
3417 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3418 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3419 "clusters per group (%lu) inconsistent",
3420 sbi->s_blocks_per_group,
3421 sbi->s_clusters_per_group);
3425 if (clustersize != blocksize) {
3426 ext4_warning(sb, "fragment/cluster size (%d) != "
3427 "block size (%d)", clustersize,
3429 clustersize = blocksize;
3431 if (sbi->s_blocks_per_group > blocksize * 8) {
3432 ext4_msg(sb, KERN_ERR,
3433 "#blocks per group too big: %lu",
3434 sbi->s_blocks_per_group);
3437 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3438 sbi->s_cluster_bits = 0;
3440 sbi->s_cluster_ratio = clustersize / blocksize;
3442 if (sbi->s_inodes_per_group > blocksize * 8) {
3443 ext4_msg(sb, KERN_ERR,
3444 "#inodes per group too big: %lu",
3445 sbi->s_inodes_per_group);
3450 * Test whether we have more sectors than will fit in sector_t,
3451 * and whether the max offset is addressable by the page cache.
3453 err = generic_check_addressable(sb->s_blocksize_bits,
3454 ext4_blocks_count(es));
3456 ext4_msg(sb, KERN_ERR, "filesystem"
3457 " too large to mount safely on this system");
3458 if (sizeof(sector_t) < 8)
3459 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3464 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3467 /* check blocks count against device size */
3468 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3469 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3470 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3471 "exceeds size of device (%llu blocks)",
3472 ext4_blocks_count(es), blocks_count);
3477 * It makes no sense for the first data block to be beyond the end
3478 * of the filesystem.
3480 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3481 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3482 "block %u is beyond end of filesystem (%llu)",
3483 le32_to_cpu(es->s_first_data_block),
3484 ext4_blocks_count(es));
3487 blocks_count = (ext4_blocks_count(es) -
3488 le32_to_cpu(es->s_first_data_block) +
3489 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3490 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3491 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3492 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3493 "(block count %llu, first data block %u, "
3494 "blocks per group %lu)", sbi->s_groups_count,
3495 ext4_blocks_count(es),
3496 le32_to_cpu(es->s_first_data_block),
3497 EXT4_BLOCKS_PER_GROUP(sb));
3500 sbi->s_groups_count = blocks_count;
3501 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3502 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3503 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3504 EXT4_DESC_PER_BLOCK(sb);
3505 sbi->s_group_desc = ext4_kvmalloc(db_count *
3506 sizeof(struct buffer_head *),
3508 if (sbi->s_group_desc == NULL) {
3509 ext4_msg(sb, KERN_ERR, "not enough memory");
3514 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3516 bgl_lock_init(sbi->s_blockgroup_lock);
3518 for (i = 0; i < db_count; i++) {
3519 block = descriptor_loc(sb, logical_sb_block, i);
3520 sbi->s_group_desc[i] = sb_bread(sb, block);
3521 if (!sbi->s_group_desc[i]) {
3522 ext4_msg(sb, KERN_ERR,
3523 "can't read group descriptor %d", i);
3528 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3529 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3532 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3533 if (!ext4_fill_flex_info(sb)) {
3534 ext4_msg(sb, KERN_ERR,
3535 "unable to initialize "
3536 "flex_bg meta info!");
3540 sbi->s_gdb_count = db_count;
3541 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3542 spin_lock_init(&sbi->s_next_gen_lock);
3544 init_timer(&sbi->s_err_report);
3545 sbi->s_err_report.function = print_daily_error_info;
3546 sbi->s_err_report.data = (unsigned long) sb;
3548 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3549 ext4_count_free_clusters(sb));
3551 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3552 ext4_count_free_inodes(sb));
3555 err = percpu_counter_init(&sbi->s_dirs_counter,
3556 ext4_count_dirs(sb));
3559 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3562 ext4_msg(sb, KERN_ERR, "insufficient memory");
3566 sbi->s_stripe = ext4_get_stripe_size(sbi);
3567 sbi->s_max_writeback_mb_bump = 128;
3570 * set up enough so that it can read an inode
3572 if (!test_opt(sb, NOLOAD) &&
3573 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3574 sb->s_op = &ext4_sops;
3576 sb->s_op = &ext4_nojournal_sops;
3577 sb->s_export_op = &ext4_export_ops;
3578 sb->s_xattr = ext4_xattr_handlers;
3580 sb->s_qcop = &ext4_qctl_operations;
3581 sb->dq_op = &ext4_quota_operations;
3583 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3585 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3586 mutex_init(&sbi->s_orphan_lock);
3587 sbi->s_resize_flags = 0;
3591 needs_recovery = (es->s_last_orphan != 0 ||
3592 EXT4_HAS_INCOMPAT_FEATURE(sb,
3593 EXT4_FEATURE_INCOMPAT_RECOVER));
3595 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3596 !(sb->s_flags & MS_RDONLY))
3597 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3601 * The first inode we look at is the journal inode. Don't try
3602 * root first: it may be modified in the journal!
3604 if (!test_opt(sb, NOLOAD) &&
3605 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3606 if (ext4_load_journal(sb, es, journal_devnum))
3608 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3609 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3610 ext4_msg(sb, KERN_ERR, "required journal recovery "
3611 "suppressed and not mounted read-only");
3612 goto failed_mount_wq;
3614 clear_opt(sb, DATA_FLAGS);
3615 sbi->s_journal = NULL;
3620 if (ext4_blocks_count(es) > 0xffffffffULL &&
3621 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3622 JBD2_FEATURE_INCOMPAT_64BIT)) {
3623 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3624 goto failed_mount_wq;
3627 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3628 jbd2_journal_set_features(sbi->s_journal,
3629 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3630 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3631 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3632 jbd2_journal_set_features(sbi->s_journal,
3633 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3634 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3635 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3637 jbd2_journal_clear_features(sbi->s_journal,
3638 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3639 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3642 /* We have now updated the journal if required, so we can
3643 * validate the data journaling mode. */
3644 switch (test_opt(sb, DATA_FLAGS)) {
3646 /* No mode set, assume a default based on the journal
3647 * capabilities: ORDERED_DATA if the journal can
3648 * cope, else JOURNAL_DATA
3650 if (jbd2_journal_check_available_features
3651 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3652 set_opt(sb, ORDERED_DATA);
3654 set_opt(sb, JOURNAL_DATA);
3657 case EXT4_MOUNT_ORDERED_DATA:
3658 case EXT4_MOUNT_WRITEBACK_DATA:
3659 if (!jbd2_journal_check_available_features
3660 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3661 ext4_msg(sb, KERN_ERR, "Journal does not support "
3662 "requested data journaling mode");
3663 goto failed_mount_wq;
3668 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3670 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3673 * The journal may have updated the bg summary counts, so we
3674 * need to update the global counters.
3676 percpu_counter_set(&sbi->s_freeclusters_counter,
3677 ext4_count_free_clusters(sb));
3678 percpu_counter_set(&sbi->s_freeinodes_counter,
3679 ext4_count_free_inodes(sb));
3680 percpu_counter_set(&sbi->s_dirs_counter,
3681 ext4_count_dirs(sb));
3682 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3686 * The maximum number of concurrent works can be high and
3687 * concurrency isn't really necessary. Limit it to 1.
3689 EXT4_SB(sb)->dio_unwritten_wq =
3690 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3691 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3692 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3693 goto failed_mount_wq;
3697 * The jbd2_journal_load will have done any necessary log recovery,
3698 * so we can safely mount the rest of the filesystem now.
3701 root = ext4_iget(sb, EXT4_ROOT_INO);
3703 ext4_msg(sb, KERN_ERR, "get root inode failed");
3704 ret = PTR_ERR(root);
3708 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3709 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3713 sb->s_root = d_alloc_root(root);
3716 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3721 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3723 /* determine the minimum size of new large inodes, if present */
3724 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3725 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3726 EXT4_GOOD_OLD_INODE_SIZE;
3727 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3728 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3729 if (sbi->s_want_extra_isize <
3730 le16_to_cpu(es->s_want_extra_isize))
3731 sbi->s_want_extra_isize =
3732 le16_to_cpu(es->s_want_extra_isize);
3733 if (sbi->s_want_extra_isize <
3734 le16_to_cpu(es->s_min_extra_isize))
3735 sbi->s_want_extra_isize =
3736 le16_to_cpu(es->s_min_extra_isize);
3739 /* Check if enough inode space is available */
3740 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3741 sbi->s_inode_size) {
3742 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3743 EXT4_GOOD_OLD_INODE_SIZE;
3744 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3748 err = ext4_setup_system_zone(sb);
3750 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3752 goto failed_mount4a;
3756 err = ext4_mb_init(sb, needs_recovery);
3758 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3763 err = ext4_register_li_request(sb, first_not_zeroed);
3767 sbi->s_kobj.kset = ext4_kset;
3768 init_completion(&sbi->s_kobj_unregister);
3769 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3774 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3775 ext4_orphan_cleanup(sb, es);
3776 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3777 if (needs_recovery) {
3778 ext4_msg(sb, KERN_INFO, "recovery complete");
3779 ext4_mark_recovery_complete(sb, es);
3781 if (EXT4_SB(sb)->s_journal) {
3782 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3783 descr = " journalled data mode";
3784 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3785 descr = " ordered data mode";
3787 descr = " writeback data mode";
3789 descr = "out journal";
3791 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3792 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3793 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3795 if (es->s_error_count)
3796 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3803 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3807 ext4_unregister_li_request(sb);
3809 ext4_mb_release(sb);
3811 ext4_ext_release(sb);
3812 ext4_release_system_zone(sb);
3817 ext4_msg(sb, KERN_ERR, "mount failed");
3818 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3820 if (sbi->s_journal) {
3821 jbd2_journal_destroy(sbi->s_journal);
3822 sbi->s_journal = NULL;
3825 del_timer(&sbi->s_err_report);
3826 if (sbi->s_flex_groups)
3827 ext4_kvfree(sbi->s_flex_groups);
3828 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3829 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3830 percpu_counter_destroy(&sbi->s_dirs_counter);
3831 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3833 kthread_stop(sbi->s_mmp_tsk);
3835 for (i = 0; i < db_count; i++)
3836 brelse(sbi->s_group_desc[i]);
3837 ext4_kvfree(sbi->s_group_desc);
3840 remove_proc_entry(sb->s_id, ext4_proc_root);
3843 for (i = 0; i < MAXQUOTAS; i++)
3844 kfree(sbi->s_qf_names[i]);
3846 ext4_blkdev_remove(sbi);
3849 sb->s_fs_info = NULL;
3850 kfree(sbi->s_blockgroup_lock);
3858 * Setup any per-fs journal parameters now. We'll do this both on
3859 * initial mount, once the journal has been initialised but before we've
3860 * done any recovery; and again on any subsequent remount.
3862 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3864 struct ext4_sb_info *sbi = EXT4_SB(sb);
3866 journal->j_commit_interval = sbi->s_commit_interval;
3867 journal->j_min_batch_time = sbi->s_min_batch_time;
3868 journal->j_max_batch_time = sbi->s_max_batch_time;
3870 write_lock(&journal->j_state_lock);
3871 if (test_opt(sb, BARRIER))
3872 journal->j_flags |= JBD2_BARRIER;
3874 journal->j_flags &= ~JBD2_BARRIER;
3875 if (test_opt(sb, DATA_ERR_ABORT))
3876 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3878 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3879 write_unlock(&journal->j_state_lock);
3882 static journal_t *ext4_get_journal(struct super_block *sb,
3883 unsigned int journal_inum)
3885 struct inode *journal_inode;
3888 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3890 /* First, test for the existence of a valid inode on disk. Bad
3891 * things happen if we iget() an unused inode, as the subsequent
3892 * iput() will try to delete it. */
3894 journal_inode = ext4_iget(sb, journal_inum);
3895 if (IS_ERR(journal_inode)) {
3896 ext4_msg(sb, KERN_ERR, "no journal found");
3899 if (!journal_inode->i_nlink) {
3900 make_bad_inode(journal_inode);
3901 iput(journal_inode);
3902 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3906 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3907 journal_inode, journal_inode->i_size);
3908 if (!S_ISREG(journal_inode->i_mode)) {
3909 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3910 iput(journal_inode);
3914 journal = jbd2_journal_init_inode(journal_inode);
3916 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3917 iput(journal_inode);
3920 journal->j_private = sb;
3921 ext4_init_journal_params(sb, journal);
3925 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3928 struct buffer_head *bh;
3932 int hblock, blocksize;
3933 ext4_fsblk_t sb_block;
3934 unsigned long offset;
3935 struct ext4_super_block *es;
3936 struct block_device *bdev;
3938 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3940 bdev = ext4_blkdev_get(j_dev, sb);
3944 blocksize = sb->s_blocksize;
3945 hblock = bdev_logical_block_size(bdev);
3946 if (blocksize < hblock) {
3947 ext4_msg(sb, KERN_ERR,
3948 "blocksize too small for journal device");
3952 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3953 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3954 set_blocksize(bdev, blocksize);
3955 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3956 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3957 "external journal");
3961 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3962 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3963 !(le32_to_cpu(es->s_feature_incompat) &
3964 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3965 ext4_msg(sb, KERN_ERR, "external journal has "
3971 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3972 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3977 len = ext4_blocks_count(es);
3978 start = sb_block + 1;
3979 brelse(bh); /* we're done with the superblock */
3981 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3982 start, len, blocksize);
3984 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3987 journal->j_private = sb;
3988 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3989 wait_on_buffer(journal->j_sb_buffer);
3990 if (!buffer_uptodate(journal->j_sb_buffer)) {
3991 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3994 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3995 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3996 "user (unsupported) - %d",
3997 be32_to_cpu(journal->j_superblock->s_nr_users));
4000 EXT4_SB(sb)->journal_bdev = bdev;
4001 ext4_init_journal_params(sb, journal);
4005 jbd2_journal_destroy(journal);
4007 ext4_blkdev_put(bdev);
4011 static int ext4_load_journal(struct super_block *sb,
4012 struct ext4_super_block *es,
4013 unsigned long journal_devnum)
4016 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4019 int really_read_only;
4021 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4023 if (journal_devnum &&
4024 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4025 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4026 "numbers have changed");
4027 journal_dev = new_decode_dev(journal_devnum);
4029 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4031 really_read_only = bdev_read_only(sb->s_bdev);
4034 * Are we loading a blank journal or performing recovery after a
4035 * crash? For recovery, we need to check in advance whether we
4036 * can get read-write access to the device.
4038 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4039 if (sb->s_flags & MS_RDONLY) {
4040 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4041 "required on readonly filesystem");
4042 if (really_read_only) {
4043 ext4_msg(sb, KERN_ERR, "write access "
4044 "unavailable, cannot proceed");
4047 ext4_msg(sb, KERN_INFO, "write access will "
4048 "be enabled during recovery");
4052 if (journal_inum && journal_dev) {
4053 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4054 "and inode journals!");
4059 if (!(journal = ext4_get_journal(sb, journal_inum)))
4062 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4066 if (!(journal->j_flags & JBD2_BARRIER))
4067 ext4_msg(sb, KERN_INFO, "barriers disabled");
4069 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4070 err = jbd2_journal_wipe(journal, !really_read_only);
4072 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4074 memcpy(save, ((char *) es) +
4075 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4076 err = jbd2_journal_load(journal);
4078 memcpy(((char *) es) + EXT4_S_ERR_START,
4079 save, EXT4_S_ERR_LEN);
4084 ext4_msg(sb, KERN_ERR, "error loading journal");
4085 jbd2_journal_destroy(journal);
4089 EXT4_SB(sb)->s_journal = journal;
4090 ext4_clear_journal_err(sb, es);
4092 if (!really_read_only && journal_devnum &&
4093 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4094 es->s_journal_dev = cpu_to_le32(journal_devnum);
4096 /* Make sure we flush the recovery flag to disk. */
4097 ext4_commit_super(sb, 1);
4103 static int ext4_commit_super(struct super_block *sb, int sync)
4105 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4106 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4109 if (!sbh || block_device_ejected(sb))
4111 if (buffer_write_io_error(sbh)) {
4113 * Oh, dear. A previous attempt to write the
4114 * superblock failed. This could happen because the
4115 * USB device was yanked out. Or it could happen to
4116 * be a transient write error and maybe the block will
4117 * be remapped. Nothing we can do but to retry the
4118 * write and hope for the best.
4120 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4121 "superblock detected");
4122 clear_buffer_write_io_error(sbh);
4123 set_buffer_uptodate(sbh);
4126 * If the file system is mounted read-only, don't update the
4127 * superblock write time. This avoids updating the superblock
4128 * write time when we are mounting the root file system
4129 * read/only but we need to replay the journal; at that point,
4130 * for people who are east of GMT and who make their clock
4131 * tick in localtime for Windows bug-for-bug compatibility,
4132 * the clock is set in the future, and this will cause e2fsck
4133 * to complain and force a full file system check.
4135 if (!(sb->s_flags & MS_RDONLY))
4136 es->s_wtime = cpu_to_le32(get_seconds());
4137 if (sb->s_bdev->bd_part)
4138 es->s_kbytes_written =
4139 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4140 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4141 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4143 es->s_kbytes_written =
4144 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4145 ext4_free_blocks_count_set(es,
4146 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4147 &EXT4_SB(sb)->s_freeclusters_counter)));
4148 es->s_free_inodes_count =
4149 cpu_to_le32(percpu_counter_sum_positive(
4150 &EXT4_SB(sb)->s_freeinodes_counter));
4152 BUFFER_TRACE(sbh, "marking dirty");
4153 mark_buffer_dirty(sbh);
4155 error = sync_dirty_buffer(sbh);
4159 error = buffer_write_io_error(sbh);
4161 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4163 clear_buffer_write_io_error(sbh);
4164 set_buffer_uptodate(sbh);
4171 * Have we just finished recovery? If so, and if we are mounting (or
4172 * remounting) the filesystem readonly, then we will end up with a
4173 * consistent fs on disk. Record that fact.
4175 static void ext4_mark_recovery_complete(struct super_block *sb,
4176 struct ext4_super_block *es)
4178 journal_t *journal = EXT4_SB(sb)->s_journal;
4180 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4181 BUG_ON(journal != NULL);
4184 jbd2_journal_lock_updates(journal);
4185 if (jbd2_journal_flush(journal) < 0)
4188 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4189 sb->s_flags & MS_RDONLY) {
4190 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4191 ext4_commit_super(sb, 1);
4195 jbd2_journal_unlock_updates(journal);
4199 * If we are mounting (or read-write remounting) a filesystem whose journal
4200 * has recorded an error from a previous lifetime, move that error to the
4201 * main filesystem now.
4203 static void ext4_clear_journal_err(struct super_block *sb,
4204 struct ext4_super_block *es)
4210 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4212 journal = EXT4_SB(sb)->s_journal;
4215 * Now check for any error status which may have been recorded in the
4216 * journal by a prior ext4_error() or ext4_abort()
4219 j_errno = jbd2_journal_errno(journal);
4223 errstr = ext4_decode_error(sb, j_errno, nbuf);
4224 ext4_warning(sb, "Filesystem error recorded "
4225 "from previous mount: %s", errstr);
4226 ext4_warning(sb, "Marking fs in need of filesystem check.");
4228 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4229 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4230 ext4_commit_super(sb, 1);
4232 jbd2_journal_clear_err(journal);
4237 * Force the running and committing transactions to commit,
4238 * and wait on the commit.
4240 int ext4_force_commit(struct super_block *sb)
4245 if (sb->s_flags & MS_RDONLY)
4248 journal = EXT4_SB(sb)->s_journal;
4250 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4251 ret = ext4_journal_force_commit(journal);
4257 static void ext4_write_super(struct super_block *sb)
4260 ext4_commit_super(sb, 1);
4264 static int ext4_sync_fs(struct super_block *sb, int wait)
4268 struct ext4_sb_info *sbi = EXT4_SB(sb);
4270 trace_ext4_sync_fs(sb, wait);
4271 flush_workqueue(sbi->dio_unwritten_wq);
4272 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4274 jbd2_log_wait_commit(sbi->s_journal, target);
4280 * LVM calls this function before a (read-only) snapshot is created. This
4281 * gives us a chance to flush the journal completely and mark the fs clean.
4283 * Note that only this function cannot bring a filesystem to be in a clean
4284 * state independently, because ext4 prevents a new handle from being started
4285 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4288 static int ext4_freeze(struct super_block *sb)
4293 if (sb->s_flags & MS_RDONLY)
4296 journal = EXT4_SB(sb)->s_journal;
4298 /* Now we set up the journal barrier. */
4299 jbd2_journal_lock_updates(journal);
4302 * Don't clear the needs_recovery flag if we failed to flush
4305 error = jbd2_journal_flush(journal);
4309 /* Journal blocked and flushed, clear needs_recovery flag. */
4310 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4311 error = ext4_commit_super(sb, 1);
4313 /* we rely on s_frozen to stop further updates */
4314 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4319 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4320 * flag here, even though the filesystem is not technically dirty yet.
4322 static int ext4_unfreeze(struct super_block *sb)
4324 if (sb->s_flags & MS_RDONLY)
4328 /* Reset the needs_recovery flag before the fs is unlocked. */
4329 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4330 ext4_commit_super(sb, 1);
4336 * Structure to save mount options for ext4_remount's benefit
4338 struct ext4_mount_options {
4339 unsigned long s_mount_opt;
4340 unsigned long s_mount_opt2;
4343 unsigned long s_commit_interval;
4344 u32 s_min_batch_time, s_max_batch_time;
4347 char *s_qf_names[MAXQUOTAS];
4351 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4353 struct ext4_super_block *es;
4354 struct ext4_sb_info *sbi = EXT4_SB(sb);
4355 unsigned long old_sb_flags;
4356 struct ext4_mount_options old_opts;
4357 int enable_quota = 0;
4359 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4364 char *orig_data = kstrdup(data, GFP_KERNEL);
4366 /* Store the original options */
4368 old_sb_flags = sb->s_flags;
4369 old_opts.s_mount_opt = sbi->s_mount_opt;
4370 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4371 old_opts.s_resuid = sbi->s_resuid;
4372 old_opts.s_resgid = sbi->s_resgid;
4373 old_opts.s_commit_interval = sbi->s_commit_interval;
4374 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4375 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4377 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4378 for (i = 0; i < MAXQUOTAS; i++)
4379 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4381 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4382 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4385 * Allow the "check" option to be passed as a remount option.
4387 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4392 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4393 ext4_abort(sb, "Abort forced by user");
4395 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4396 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4400 if (sbi->s_journal) {
4401 ext4_init_journal_params(sb, sbi->s_journal);
4402 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4405 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4406 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4411 if (*flags & MS_RDONLY) {
4412 err = dquot_suspend(sb, -1);
4417 * First of all, the unconditional stuff we have to do
4418 * to disable replay of the journal when we next remount
4420 sb->s_flags |= MS_RDONLY;
4423 * OK, test if we are remounting a valid rw partition
4424 * readonly, and if so set the rdonly flag and then
4425 * mark the partition as valid again.
4427 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4428 (sbi->s_mount_state & EXT4_VALID_FS))
4429 es->s_state = cpu_to_le16(sbi->s_mount_state);
4432 ext4_mark_recovery_complete(sb, es);
4434 /* Make sure we can mount this feature set readwrite */
4435 if (!ext4_feature_set_ok(sb, 0)) {
4440 * Make sure the group descriptor checksums
4441 * are sane. If they aren't, refuse to remount r/w.
4443 for (g = 0; g < sbi->s_groups_count; g++) {
4444 struct ext4_group_desc *gdp =
4445 ext4_get_group_desc(sb, g, NULL);
4447 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4448 ext4_msg(sb, KERN_ERR,
4449 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4450 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4451 le16_to_cpu(gdp->bg_checksum));
4458 * If we have an unprocessed orphan list hanging
4459 * around from a previously readonly bdev mount,
4460 * require a full umount/remount for now.
4462 if (es->s_last_orphan) {
4463 ext4_msg(sb, KERN_WARNING, "Couldn't "
4464 "remount RDWR because of unprocessed "
4465 "orphan inode list. Please "
4466 "umount/remount instead");
4472 * Mounting a RDONLY partition read-write, so reread
4473 * and store the current valid flag. (It may have
4474 * been changed by e2fsck since we originally mounted
4478 ext4_clear_journal_err(sb, es);
4479 sbi->s_mount_state = le16_to_cpu(es->s_state);
4480 if (!ext4_setup_super(sb, es, 0))
4481 sb->s_flags &= ~MS_RDONLY;
4482 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4483 EXT4_FEATURE_INCOMPAT_MMP))
4484 if (ext4_multi_mount_protect(sb,
4485 le64_to_cpu(es->s_mmp_block))) {
4494 * Reinitialize lazy itable initialization thread based on
4497 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4498 ext4_unregister_li_request(sb);
4500 ext4_group_t first_not_zeroed;
4501 first_not_zeroed = ext4_has_uninit_itable(sb);
4502 ext4_register_li_request(sb, first_not_zeroed);
4505 ext4_setup_system_zone(sb);
4506 if (sbi->s_journal == NULL)
4507 ext4_commit_super(sb, 1);
4510 /* Release old quota file names */
4511 for (i = 0; i < MAXQUOTAS; i++)
4512 if (old_opts.s_qf_names[i] &&
4513 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4514 kfree(old_opts.s_qf_names[i]);
4518 dquot_resume(sb, -1);
4520 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4525 sb->s_flags = old_sb_flags;
4526 sbi->s_mount_opt = old_opts.s_mount_opt;
4527 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4528 sbi->s_resuid = old_opts.s_resuid;
4529 sbi->s_resgid = old_opts.s_resgid;
4530 sbi->s_commit_interval = old_opts.s_commit_interval;
4531 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4532 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4534 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4535 for (i = 0; i < MAXQUOTAS; i++) {
4536 if (sbi->s_qf_names[i] &&
4537 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4538 kfree(sbi->s_qf_names[i]);
4539 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4548 * Note: calculating the overhead so we can be compatible with
4549 * historical BSD practice is quite difficult in the face of
4550 * clusters/bigalloc. This is because multiple metadata blocks from
4551 * different block group can end up in the same allocation cluster.
4552 * Calculating the exact overhead in the face of clustered allocation
4553 * requires either O(all block bitmaps) in memory or O(number of block
4554 * groups**2) in time. We will still calculate the superblock for
4555 * older file systems --- and if we come across with a bigalloc file
4556 * system with zero in s_overhead_clusters the estimate will be close to
4557 * correct especially for very large cluster sizes --- but for newer
4558 * file systems, it's better to calculate this figure once at mkfs
4559 * time, and store it in the superblock. If the superblock value is
4560 * present (even for non-bigalloc file systems), we will use it.
4562 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4564 struct super_block *sb = dentry->d_sb;
4565 struct ext4_sb_info *sbi = EXT4_SB(sb);
4566 struct ext4_super_block *es = sbi->s_es;
4567 struct ext4_group_desc *gdp;
4571 if (test_opt(sb, MINIX_DF)) {
4572 sbi->s_overhead_last = 0;
4573 } else if (es->s_overhead_clusters) {
4574 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4575 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4576 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4577 ext4_fsblk_t overhead = 0;
4580 * Compute the overhead (FS structures). This is constant
4581 * for a given filesystem unless the number of block groups
4582 * changes so we cache the previous value until it does.
4586 * All of the blocks before first_data_block are
4589 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4592 * Add the overhead found in each block group
4594 for (i = 0; i < ngroups; i++) {
4595 gdp = ext4_get_group_desc(sb, i, NULL);
4596 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4599 sbi->s_overhead_last = overhead;
4601 sbi->s_blocks_last = ext4_blocks_count(es);
4604 buf->f_type = EXT4_SUPER_MAGIC;
4605 buf->f_bsize = sb->s_blocksize;
4606 buf->f_blocks = (ext4_blocks_count(es) -
4607 EXT4_C2B(sbi, sbi->s_overhead_last));
4608 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4609 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4610 /* prevent underflow in case that few free space is available */
4611 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4612 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4613 if (buf->f_bfree < ext4_r_blocks_count(es))
4615 buf->f_files = le32_to_cpu(es->s_inodes_count);
4616 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4617 buf->f_namelen = EXT4_NAME_LEN;
4618 fsid = le64_to_cpup((void *)es->s_uuid) ^
4619 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4620 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4621 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4626 /* Helper function for writing quotas on sync - we need to start transaction
4627 * before quota file is locked for write. Otherwise the are possible deadlocks:
4628 * Process 1 Process 2
4629 * ext4_create() quota_sync()
4630 * jbd2_journal_start() write_dquot()
4631 * dquot_initialize() down(dqio_mutex)
4632 * down(dqio_mutex) jbd2_journal_start()
4638 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4640 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4643 static int ext4_write_dquot(struct dquot *dquot)
4647 struct inode *inode;
4649 inode = dquot_to_inode(dquot);
4650 handle = ext4_journal_start(inode,
4651 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4653 return PTR_ERR(handle);
4654 ret = dquot_commit(dquot);
4655 err = ext4_journal_stop(handle);
4661 static int ext4_acquire_dquot(struct dquot *dquot)
4666 handle = ext4_journal_start(dquot_to_inode(dquot),
4667 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4669 return PTR_ERR(handle);
4670 ret = dquot_acquire(dquot);
4671 err = ext4_journal_stop(handle);
4677 static int ext4_release_dquot(struct dquot *dquot)
4682 handle = ext4_journal_start(dquot_to_inode(dquot),
4683 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4684 if (IS_ERR(handle)) {
4685 /* Release dquot anyway to avoid endless cycle in dqput() */
4686 dquot_release(dquot);
4687 return PTR_ERR(handle);
4689 ret = dquot_release(dquot);
4690 err = ext4_journal_stop(handle);
4696 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4698 /* Are we journaling quotas? */
4699 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4700 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4701 dquot_mark_dquot_dirty(dquot);
4702 return ext4_write_dquot(dquot);
4704 return dquot_mark_dquot_dirty(dquot);
4708 static int ext4_write_info(struct super_block *sb, int type)
4713 /* Data block + inode block */
4714 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4716 return PTR_ERR(handle);
4717 ret = dquot_commit_info(sb, type);
4718 err = ext4_journal_stop(handle);
4725 * Turn on quotas during mount time - we need to find
4726 * the quota file and such...
4728 static int ext4_quota_on_mount(struct super_block *sb, int type)
4730 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4731 EXT4_SB(sb)->s_jquota_fmt, type);
4735 * Standard function to be called on quota_on
4737 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4742 if (!test_opt(sb, QUOTA))
4745 /* Quotafile not on the same filesystem? */
4746 if (path->dentry->d_sb != sb)
4748 /* Journaling quota? */
4749 if (EXT4_SB(sb)->s_qf_names[type]) {
4750 /* Quotafile not in fs root? */
4751 if (path->dentry->d_parent != sb->s_root)
4752 ext4_msg(sb, KERN_WARNING,
4753 "Quota file not on filesystem root. "
4754 "Journaled quota will not work");
4758 * When we journal data on quota file, we have to flush journal to see
4759 * all updates to the file when we bypass pagecache...
4761 if (EXT4_SB(sb)->s_journal &&
4762 ext4_should_journal_data(path->dentry->d_inode)) {
4764 * We don't need to lock updates but journal_flush() could
4765 * otherwise be livelocked...
4767 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4768 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4769 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4774 return dquot_quota_on(sb, type, format_id, path);
4777 static int ext4_quota_off(struct super_block *sb, int type)
4779 struct inode *inode = sb_dqopt(sb)->files[type];
4782 /* Force all delayed allocation blocks to be allocated.
4783 * Caller already holds s_umount sem */
4784 if (test_opt(sb, DELALLOC))
4785 sync_filesystem(sb);
4790 /* Update modification times of quota files when userspace can
4791 * start looking at them */
4792 handle = ext4_journal_start(inode, 1);
4795 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4796 ext4_mark_inode_dirty(handle, inode);
4797 ext4_journal_stop(handle);
4800 return dquot_quota_off(sb, type);
4803 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4804 * acquiring the locks... As quota files are never truncated and quota code
4805 * itself serializes the operations (and no one else should touch the files)
4806 * we don't have to be afraid of races */
4807 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4808 size_t len, loff_t off)
4810 struct inode *inode = sb_dqopt(sb)->files[type];
4811 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4813 int offset = off & (sb->s_blocksize - 1);
4816 struct buffer_head *bh;
4817 loff_t i_size = i_size_read(inode);
4821 if (off+len > i_size)
4824 while (toread > 0) {
4825 tocopy = sb->s_blocksize - offset < toread ?
4826 sb->s_blocksize - offset : toread;
4827 bh = ext4_bread(NULL, inode, blk, 0, &err);
4830 if (!bh) /* A hole? */
4831 memset(data, 0, tocopy);
4833 memcpy(data, bh->b_data+offset, tocopy);
4843 /* Write to quotafile (we know the transaction is already started and has
4844 * enough credits) */
4845 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4846 const char *data, size_t len, loff_t off)
4848 struct inode *inode = sb_dqopt(sb)->files[type];
4849 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4851 int offset = off & (sb->s_blocksize - 1);
4852 struct buffer_head *bh;
4853 handle_t *handle = journal_current_handle();
4855 if (EXT4_SB(sb)->s_journal && !handle) {
4856 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4857 " cancelled because transaction is not started",
4858 (unsigned long long)off, (unsigned long long)len);
4862 * Since we account only one data block in transaction credits,
4863 * then it is impossible to cross a block boundary.
4865 if (sb->s_blocksize - offset < len) {
4866 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4867 " cancelled because not block aligned",
4868 (unsigned long long)off, (unsigned long long)len);
4872 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4873 bh = ext4_bread(handle, inode, blk, 1, &err);
4876 err = ext4_journal_get_write_access(handle, bh);
4882 memcpy(bh->b_data+offset, data, len);
4883 flush_dcache_page(bh->b_page);
4885 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4889 mutex_unlock(&inode->i_mutex);
4892 if (inode->i_size < off + len) {
4893 i_size_write(inode, off + len);
4894 EXT4_I(inode)->i_disksize = inode->i_size;
4895 ext4_mark_inode_dirty(handle, inode);
4897 mutex_unlock(&inode->i_mutex);
4903 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4904 const char *dev_name, void *data)
4906 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4909 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4910 static inline void register_as_ext2(void)
4912 int err = register_filesystem(&ext2_fs_type);
4915 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4918 static inline void unregister_as_ext2(void)
4920 unregister_filesystem(&ext2_fs_type);
4923 static inline int ext2_feature_set_ok(struct super_block *sb)
4925 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4927 if (sb->s_flags & MS_RDONLY)
4929 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4933 MODULE_ALIAS("ext2");
4935 static inline void register_as_ext2(void) { }
4936 static inline void unregister_as_ext2(void) { }
4937 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4940 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4941 static inline void register_as_ext3(void)
4943 int err = register_filesystem(&ext3_fs_type);
4946 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4949 static inline void unregister_as_ext3(void)
4951 unregister_filesystem(&ext3_fs_type);
4954 static inline int ext3_feature_set_ok(struct super_block *sb)
4956 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4958 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4960 if (sb->s_flags & MS_RDONLY)
4962 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4966 MODULE_ALIAS("ext3");
4968 static inline void register_as_ext3(void) { }
4969 static inline void unregister_as_ext3(void) { }
4970 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4973 static struct file_system_type ext4_fs_type = {
4974 .owner = THIS_MODULE,
4976 .mount = ext4_mount,
4977 .kill_sb = kill_block_super,
4978 .fs_flags = FS_REQUIRES_DEV,
4981 static int __init ext4_init_feat_adverts(void)
4983 struct ext4_features *ef;
4986 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4990 ef->f_kobj.kset = ext4_kset;
4991 init_completion(&ef->f_kobj_unregister);
4992 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5005 static void ext4_exit_feat_adverts(void)
5007 kobject_put(&ext4_feat->f_kobj);
5008 wait_for_completion(&ext4_feat->f_kobj_unregister);
5012 /* Shared across all ext4 file systems */
5013 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5014 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5016 static int __init ext4_init_fs(void)
5020 ext4_check_flag_values();
5022 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5023 mutex_init(&ext4__aio_mutex[i]);
5024 init_waitqueue_head(&ext4__ioend_wq[i]);
5027 err = ext4_init_pageio();
5030 err = ext4_init_system_zone();
5033 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5036 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5038 err = ext4_init_feat_adverts();
5042 err = ext4_init_mballoc();
5046 err = ext4_init_xattr();
5049 err = init_inodecache();
5054 err = register_filesystem(&ext4_fs_type);
5058 ext4_li_info = NULL;
5059 mutex_init(&ext4_li_mtx);
5062 unregister_as_ext2();
5063 unregister_as_ext3();
5064 destroy_inodecache();
5068 ext4_exit_mballoc();
5070 ext4_exit_feat_adverts();
5073 remove_proc_entry("fs/ext4", NULL);
5074 kset_unregister(ext4_kset);
5076 ext4_exit_system_zone();
5082 static void __exit ext4_exit_fs(void)
5084 ext4_destroy_lazyinit_thread();
5085 unregister_as_ext2();
5086 unregister_as_ext3();
5087 unregister_filesystem(&ext4_fs_type);
5088 destroy_inodecache();
5090 ext4_exit_mballoc();
5091 ext4_exit_feat_adverts();
5092 remove_proc_entry("fs/ext4", NULL);
5093 kset_unregister(ext4_kset);
5094 ext4_exit_system_zone();
5098 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5099 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5100 MODULE_LICENSE("GPL");
5101 module_init(ext4_init_fs)
5102 module_exit(ext4_exit_fs)