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_CONT);
1569 clear_opt(sb, ERRORS_RO);
1570 set_opt(sb, ERRORS_PANIC);
1573 clear_opt(sb, ERRORS_CONT);
1574 clear_opt(sb, ERRORS_PANIC);
1575 set_opt(sb, ERRORS_RO);
1578 clear_opt(sb, ERRORS_RO);
1579 clear_opt(sb, ERRORS_PANIC);
1580 set_opt(sb, ERRORS_CONT);
1583 set_opt(sb, NO_UID32);
1589 ext4_msg(sb, KERN_WARNING,
1590 "Ignoring deprecated oldalloc option");
1593 ext4_msg(sb, KERN_WARNING,
1594 "Ignoring deprecated orlov option");
1596 #ifdef CONFIG_EXT4_FS_XATTR
1597 case Opt_user_xattr:
1598 set_opt(sb, XATTR_USER);
1600 case Opt_nouser_xattr:
1601 clear_opt(sb, XATTR_USER);
1604 case Opt_user_xattr:
1605 case Opt_nouser_xattr:
1606 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1609 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1611 set_opt(sb, POSIX_ACL);
1614 clear_opt(sb, POSIX_ACL);
1619 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1622 case Opt_journal_dev:
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot specify journal on remount");
1628 if (match_int(&args[0], &option))
1630 *journal_devnum = option;
1632 case Opt_journal_checksum:
1633 set_opt(sb, JOURNAL_CHECKSUM);
1635 case Opt_journal_async_commit:
1636 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1637 set_opt(sb, JOURNAL_CHECKSUM);
1640 set_opt(sb, NOLOAD);
1643 if (match_int(&args[0], &option))
1648 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1649 sbi->s_commit_interval = HZ * option;
1651 case Opt_max_batch_time:
1652 if (match_int(&args[0], &option))
1657 option = EXT4_DEF_MAX_BATCH_TIME;
1658 sbi->s_max_batch_time = option;
1660 case Opt_min_batch_time:
1661 if (match_int(&args[0], &option))
1665 sbi->s_min_batch_time = option;
1667 case Opt_data_journal:
1668 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1670 case Opt_data_ordered:
1671 data_opt = EXT4_MOUNT_ORDERED_DATA;
1673 case Opt_data_writeback:
1674 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1677 if (!sbi->s_journal)
1678 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1679 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1680 ext4_msg(sb, KERN_ERR,
1681 "Cannot change data mode on remount");
1685 clear_opt(sb, DATA_FLAGS);
1686 sbi->s_mount_opt |= data_opt;
1689 case Opt_data_err_abort:
1690 set_opt(sb, DATA_ERR_ABORT);
1692 case Opt_data_err_ignore:
1693 clear_opt(sb, DATA_ERR_ABORT);
1697 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1701 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1704 case Opt_offusrjquota:
1705 if (!clear_qf_name(sb, USRQUOTA))
1708 case Opt_offgrpjquota:
1709 if (!clear_qf_name(sb, GRPQUOTA))
1713 case Opt_jqfmt_vfsold:
1714 qfmt = QFMT_VFS_OLD;
1716 case Opt_jqfmt_vfsv0:
1719 case Opt_jqfmt_vfsv1:
1722 if (sb_any_quota_loaded(sb) &&
1723 sbi->s_jquota_fmt != qfmt) {
1724 ext4_msg(sb, KERN_ERR, "Cannot change "
1725 "journaled quota options when "
1729 sbi->s_jquota_fmt = qfmt;
1734 set_opt(sb, USRQUOTA);
1738 set_opt(sb, GRPQUOTA);
1741 if (sb_any_quota_loaded(sb)) {
1742 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1743 "options when quota turned on");
1746 clear_opt(sb, QUOTA);
1747 clear_opt(sb, USRQUOTA);
1748 clear_opt(sb, GRPQUOTA);
1754 ext4_msg(sb, KERN_ERR,
1755 "quota options not supported");
1759 case Opt_offusrjquota:
1760 case Opt_offgrpjquota:
1761 case Opt_jqfmt_vfsold:
1762 case Opt_jqfmt_vfsv0:
1763 case Opt_jqfmt_vfsv1:
1764 ext4_msg(sb, KERN_ERR,
1765 "journaled quota options not supported");
1771 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1774 clear_opt(sb, BARRIER);
1778 if (match_int(&args[0], &option))
1781 option = 1; /* No argument, default to 1 */
1783 set_opt(sb, BARRIER);
1785 clear_opt(sb, BARRIER);
1788 ext4_msg(sb, KERN_WARNING,
1789 "Ignoring deprecated nobh option");
1792 ext4_msg(sb, KERN_WARNING,
1793 "Ignoring deprecated bh option");
1796 sb->s_flags |= MS_I_VERSION;
1798 case Opt_nodelalloc:
1799 clear_opt(sb, DELALLOC);
1800 clear_opt2(sb, EXPLICIT_DELALLOC);
1802 case Opt_mblk_io_submit:
1803 set_opt(sb, MBLK_IO_SUBMIT);
1805 case Opt_nomblk_io_submit:
1806 clear_opt(sb, MBLK_IO_SUBMIT);
1809 if (match_int(&args[0], &option))
1813 sbi->s_stripe = option;
1816 set_opt(sb, DELALLOC);
1817 set_opt2(sb, EXPLICIT_DELALLOC);
1819 case Opt_block_validity:
1820 set_opt(sb, BLOCK_VALIDITY);
1822 case Opt_noblock_validity:
1823 clear_opt(sb, BLOCK_VALIDITY);
1825 case Opt_inode_readahead_blks:
1826 if (match_int(&args[0], &option))
1828 if (option < 0 || option > (1 << 30))
1830 if (option && !is_power_of_2(option)) {
1831 ext4_msg(sb, KERN_ERR,
1832 "EXT4-fs: inode_readahead_blks"
1833 " must be a power of 2");
1836 sbi->s_inode_readahead_blks = option;
1838 case Opt_journal_ioprio:
1839 if (match_int(&args[0], &option))
1841 if (option < 0 || option > 7)
1843 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1846 case Opt_noauto_da_alloc:
1847 set_opt(sb, NO_AUTO_DA_ALLOC);
1849 case Opt_auto_da_alloc:
1851 if (match_int(&args[0], &option))
1854 option = 1; /* No argument, default to 1 */
1856 clear_opt(sb, NO_AUTO_DA_ALLOC);
1858 set_opt(sb,NO_AUTO_DA_ALLOC);
1861 set_opt(sb, DISCARD);
1864 clear_opt(sb, DISCARD);
1866 case Opt_dioread_nolock:
1867 set_opt(sb, DIOREAD_NOLOCK);
1869 case Opt_dioread_lock:
1870 clear_opt(sb, DIOREAD_NOLOCK);
1872 case Opt_init_itable:
1873 set_opt(sb, INIT_INODE_TABLE);
1875 if (match_int(&args[0], &option))
1878 option = EXT4_DEF_LI_WAIT_MULT;
1881 sbi->s_li_wait_mult = option;
1883 case Opt_noinit_itable:
1884 clear_opt(sb, INIT_INODE_TABLE);
1887 ext4_msg(sb, KERN_ERR,
1888 "Unrecognized mount option \"%s\" "
1889 "or missing value", p);
1894 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1895 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1896 clear_opt(sb, USRQUOTA);
1898 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1899 clear_opt(sb, GRPQUOTA);
1901 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1902 ext4_msg(sb, KERN_ERR, "old and new quota "
1907 if (!sbi->s_jquota_fmt) {
1908 ext4_msg(sb, KERN_ERR, "journaled quota format "
1913 if (sbi->s_jquota_fmt) {
1914 ext4_msg(sb, KERN_ERR, "journaled quota format "
1915 "specified with no journaling "
1924 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1930 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1931 ext4_msg(sb, KERN_ERR, "revision level too high, "
1932 "forcing read-only mode");
1937 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1938 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1939 "running e2fsck is recommended");
1940 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1941 ext4_msg(sb, KERN_WARNING,
1942 "warning: mounting fs with errors, "
1943 "running e2fsck is recommended");
1944 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1945 le16_to_cpu(es->s_mnt_count) >=
1946 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1947 ext4_msg(sb, KERN_WARNING,
1948 "warning: maximal mount count reached, "
1949 "running e2fsck is recommended");
1950 else if (le32_to_cpu(es->s_checkinterval) &&
1951 (le32_to_cpu(es->s_lastcheck) +
1952 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1953 ext4_msg(sb, KERN_WARNING,
1954 "warning: checktime reached, "
1955 "running e2fsck is recommended");
1956 if (!sbi->s_journal)
1957 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1958 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1959 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1960 le16_add_cpu(&es->s_mnt_count, 1);
1961 es->s_mtime = cpu_to_le32(get_seconds());
1962 ext4_update_dynamic_rev(sb);
1964 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1966 ext4_commit_super(sb, 1);
1968 if (test_opt(sb, DEBUG))
1969 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1970 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1972 sbi->s_groups_count,
1973 EXT4_BLOCKS_PER_GROUP(sb),
1974 EXT4_INODES_PER_GROUP(sb),
1975 sbi->s_mount_opt, sbi->s_mount_opt2);
1977 cleancache_init_fs(sb);
1981 static int ext4_fill_flex_info(struct super_block *sb)
1983 struct ext4_sb_info *sbi = EXT4_SB(sb);
1984 struct ext4_group_desc *gdp = NULL;
1985 ext4_group_t flex_group_count;
1986 ext4_group_t flex_group;
1987 unsigned int groups_per_flex = 0;
1991 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1992 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1993 sbi->s_log_groups_per_flex = 0;
1996 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1998 /* We allocate both existing and potentially added groups */
1999 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2000 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2001 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2002 size = flex_group_count * sizeof(struct flex_groups);
2003 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2004 if (sbi->s_flex_groups == NULL) {
2005 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2010 for (i = 0; i < sbi->s_groups_count; i++) {
2011 gdp = ext4_get_group_desc(sb, i, NULL);
2013 flex_group = ext4_flex_group(sbi, i);
2014 atomic_add(ext4_free_inodes_count(sb, gdp),
2015 &sbi->s_flex_groups[flex_group].free_inodes);
2016 atomic_add(ext4_free_group_clusters(sb, gdp),
2017 &sbi->s_flex_groups[flex_group].free_clusters);
2018 atomic_add(ext4_used_dirs_count(sb, gdp),
2019 &sbi->s_flex_groups[flex_group].used_dirs);
2027 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2028 struct ext4_group_desc *gdp)
2032 if (sbi->s_es->s_feature_ro_compat &
2033 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2034 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2035 __le32 le_group = cpu_to_le32(block_group);
2037 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2038 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2039 crc = crc16(crc, (__u8 *)gdp, offset);
2040 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2041 /* for checksum of struct ext4_group_desc do the rest...*/
2042 if ((sbi->s_es->s_feature_incompat &
2043 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2044 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2045 crc = crc16(crc, (__u8 *)gdp + offset,
2046 le16_to_cpu(sbi->s_es->s_desc_size) -
2050 return cpu_to_le16(crc);
2053 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2054 struct ext4_group_desc *gdp)
2056 if ((sbi->s_es->s_feature_ro_compat &
2057 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2058 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2064 /* Called at mount-time, super-block is locked */
2065 static int ext4_check_descriptors(struct super_block *sb,
2066 ext4_group_t *first_not_zeroed)
2068 struct ext4_sb_info *sbi = EXT4_SB(sb);
2069 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2070 ext4_fsblk_t last_block;
2071 ext4_fsblk_t block_bitmap;
2072 ext4_fsblk_t inode_bitmap;
2073 ext4_fsblk_t inode_table;
2074 int flexbg_flag = 0;
2075 ext4_group_t i, grp = sbi->s_groups_count;
2077 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2080 ext4_debug("Checking group descriptors");
2082 for (i = 0; i < sbi->s_groups_count; i++) {
2083 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2085 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2086 last_block = ext4_blocks_count(sbi->s_es) - 1;
2088 last_block = first_block +
2089 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2091 if ((grp == sbi->s_groups_count) &&
2092 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2095 block_bitmap = ext4_block_bitmap(sb, gdp);
2096 if (block_bitmap < first_block || block_bitmap > last_block) {
2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2098 "Block bitmap for group %u not in group "
2099 "(block %llu)!", i, block_bitmap);
2102 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2103 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2105 "Inode bitmap for group %u not in group "
2106 "(block %llu)!", i, inode_bitmap);
2109 inode_table = ext4_inode_table(sb, gdp);
2110 if (inode_table < first_block ||
2111 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 "Inode table for group %u not in group "
2114 "(block %llu)!", i, inode_table);
2117 ext4_lock_group(sb, i);
2118 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2120 "Checksum for group %u failed (%u!=%u)",
2121 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2122 gdp)), le16_to_cpu(gdp->bg_checksum));
2123 if (!(sb->s_flags & MS_RDONLY)) {
2124 ext4_unlock_group(sb, i);
2128 ext4_unlock_group(sb, i);
2130 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2132 if (NULL != first_not_zeroed)
2133 *first_not_zeroed = grp;
2135 ext4_free_blocks_count_set(sbi->s_es,
2136 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2137 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2141 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2142 * the superblock) which were deleted from all directories, but held open by
2143 * a process at the time of a crash. We walk the list and try to delete these
2144 * inodes at recovery time (only with a read-write filesystem).
2146 * In order to keep the orphan inode chain consistent during traversal (in
2147 * case of crash during recovery), we link each inode into the superblock
2148 * orphan list_head and handle it the same way as an inode deletion during
2149 * normal operation (which journals the operations for us).
2151 * We only do an iget() and an iput() on each inode, which is very safe if we
2152 * accidentally point at an in-use or already deleted inode. The worst that
2153 * can happen in this case is that we get a "bit already cleared" message from
2154 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2155 * e2fsck was run on this filesystem, and it must have already done the orphan
2156 * inode cleanup for us, so we can safely abort without any further action.
2158 static void ext4_orphan_cleanup(struct super_block *sb,
2159 struct ext4_super_block *es)
2161 unsigned int s_flags = sb->s_flags;
2162 int nr_orphans = 0, nr_truncates = 0;
2166 if (!es->s_last_orphan) {
2167 jbd_debug(4, "no orphan inodes to clean up\n");
2171 if (bdev_read_only(sb->s_bdev)) {
2172 ext4_msg(sb, KERN_ERR, "write access "
2173 "unavailable, skipping orphan cleanup");
2177 /* Check if feature set would not allow a r/w mount */
2178 if (!ext4_feature_set_ok(sb, 0)) {
2179 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2180 "unknown ROCOMPAT features");
2184 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2185 if (es->s_last_orphan)
2186 jbd_debug(1, "Errors on filesystem, "
2187 "clearing orphan list.\n");
2188 es->s_last_orphan = 0;
2189 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2193 if (s_flags & MS_RDONLY) {
2194 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2195 sb->s_flags &= ~MS_RDONLY;
2198 /* Needed for iput() to work correctly and not trash data */
2199 sb->s_flags |= MS_ACTIVE;
2200 /* Turn on quotas so that they are updated correctly */
2201 for (i = 0; i < MAXQUOTAS; i++) {
2202 if (EXT4_SB(sb)->s_qf_names[i]) {
2203 int ret = ext4_quota_on_mount(sb, i);
2205 ext4_msg(sb, KERN_ERR,
2206 "Cannot turn on journaled "
2207 "quota: error %d", ret);
2212 while (es->s_last_orphan) {
2213 struct inode *inode;
2215 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2216 if (IS_ERR(inode)) {
2217 es->s_last_orphan = 0;
2221 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2222 dquot_initialize(inode);
2223 if (inode->i_nlink) {
2224 ext4_msg(sb, KERN_DEBUG,
2225 "%s: truncating inode %lu to %lld bytes",
2226 __func__, inode->i_ino, inode->i_size);
2227 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2228 inode->i_ino, inode->i_size);
2229 ext4_truncate(inode);
2232 ext4_msg(sb, KERN_DEBUG,
2233 "%s: deleting unreferenced inode %lu",
2234 __func__, inode->i_ino);
2235 jbd_debug(2, "deleting unreferenced inode %lu\n",
2239 iput(inode); /* The delete magic happens here! */
2242 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2245 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2246 PLURAL(nr_orphans));
2248 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2249 PLURAL(nr_truncates));
2251 /* Turn quotas off */
2252 for (i = 0; i < MAXQUOTAS; i++) {
2253 if (sb_dqopt(sb)->files[i])
2254 dquot_quota_off(sb, i);
2257 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2261 * Maximal extent format file size.
2262 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2263 * extent format containers, within a sector_t, and within i_blocks
2264 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2265 * so that won't be a limiting factor.
2267 * However there is other limiting factor. We do store extents in the form
2268 * of starting block and length, hence the resulting length of the extent
2269 * covering maximum file size must fit into on-disk format containers as
2270 * well. Given that length is always by 1 unit bigger than max unit (because
2271 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2273 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2275 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2278 loff_t upper_limit = MAX_LFS_FILESIZE;
2280 /* small i_blocks in vfs inode? */
2281 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2283 * CONFIG_LBDAF is not enabled implies the inode
2284 * i_block represent total blocks in 512 bytes
2285 * 32 == size of vfs inode i_blocks * 8
2287 upper_limit = (1LL << 32) - 1;
2289 /* total blocks in file system block size */
2290 upper_limit >>= (blkbits - 9);
2291 upper_limit <<= blkbits;
2295 * 32-bit extent-start container, ee_block. We lower the maxbytes
2296 * by one fs block, so ee_len can cover the extent of maximum file
2299 res = (1LL << 32) - 1;
2302 /* Sanity check against vm- & vfs- imposed limits */
2303 if (res > upper_limit)
2310 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2311 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2312 * We need to be 1 filesystem block less than the 2^48 sector limit.
2314 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2316 loff_t res = EXT4_NDIR_BLOCKS;
2319 /* This is calculated to be the largest file size for a dense, block
2320 * mapped file such that the file's total number of 512-byte sectors,
2321 * including data and all indirect blocks, does not exceed (2^48 - 1).
2323 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2324 * number of 512-byte sectors of the file.
2327 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2329 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2330 * the inode i_block field represents total file blocks in
2331 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2333 upper_limit = (1LL << 32) - 1;
2335 /* total blocks in file system block size */
2336 upper_limit >>= (bits - 9);
2340 * We use 48 bit ext4_inode i_blocks
2341 * With EXT4_HUGE_FILE_FL set the i_blocks
2342 * represent total number of blocks in
2343 * file system block size
2345 upper_limit = (1LL << 48) - 1;
2349 /* indirect blocks */
2351 /* double indirect blocks */
2352 meta_blocks += 1 + (1LL << (bits-2));
2353 /* tripple indirect blocks */
2354 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2356 upper_limit -= meta_blocks;
2357 upper_limit <<= bits;
2359 res += 1LL << (bits-2);
2360 res += 1LL << (2*(bits-2));
2361 res += 1LL << (3*(bits-2));
2363 if (res > upper_limit)
2366 if (res > MAX_LFS_FILESIZE)
2367 res = MAX_LFS_FILESIZE;
2372 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2373 ext4_fsblk_t logical_sb_block, int nr)
2375 struct ext4_sb_info *sbi = EXT4_SB(sb);
2376 ext4_group_t bg, first_meta_bg;
2379 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2381 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2383 return logical_sb_block + nr + 1;
2384 bg = sbi->s_desc_per_block * nr;
2385 if (ext4_bg_has_super(sb, bg))
2388 return (has_super + ext4_group_first_block_no(sb, bg));
2392 * ext4_get_stripe_size: Get the stripe size.
2393 * @sbi: In memory super block info
2395 * If we have specified it via mount option, then
2396 * use the mount option value. If the value specified at mount time is
2397 * greater than the blocks per group use the super block value.
2398 * If the super block value is greater than blocks per group return 0.
2399 * Allocator needs it be less than blocks per group.
2402 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2404 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2405 unsigned long stripe_width =
2406 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2409 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2410 ret = sbi->s_stripe;
2411 else if (stripe_width <= sbi->s_blocks_per_group)
2413 else if (stride <= sbi->s_blocks_per_group)
2419 * If the stripe width is 1, this makes no sense and
2420 * we set it to 0 to turn off stripe handling code.
2431 struct attribute attr;
2432 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2433 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2434 const char *, size_t);
2438 static int parse_strtoul(const char *buf,
2439 unsigned long max, unsigned long *value)
2443 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2444 endp = skip_spaces(endp);
2445 if (*endp || *value > max)
2451 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2452 struct ext4_sb_info *sbi,
2455 return snprintf(buf, PAGE_SIZE, "%llu\n",
2457 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2460 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2461 struct ext4_sb_info *sbi, char *buf)
2463 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465 if (!sb->s_bdev->bd_part)
2466 return snprintf(buf, PAGE_SIZE, "0\n");
2467 return snprintf(buf, PAGE_SIZE, "%lu\n",
2468 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2469 sbi->s_sectors_written_start) >> 1);
2472 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi, char *buf)
2475 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2477 if (!sb->s_bdev->bd_part)
2478 return snprintf(buf, PAGE_SIZE, "0\n");
2479 return snprintf(buf, PAGE_SIZE, "%llu\n",
2480 (unsigned long long)(sbi->s_kbytes_written +
2481 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2482 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2485 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2486 struct ext4_sb_info *sbi, char *buf)
2488 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2491 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2494 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2497 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi,
2499 const char *buf, size_t count)
2503 if (parse_strtoul(buf, 0x40000000, &t))
2506 if (t && !is_power_of_2(t))
2509 sbi->s_inode_readahead_blks = t;
2513 static ssize_t sbi_ui_show(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi, char *buf)
2516 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2518 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2521 static ssize_t sbi_ui_store(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi,
2523 const char *buf, size_t count)
2525 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2528 if (parse_strtoul(buf, 0xffffffff, &t))
2534 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2535 static struct ext4_attr ext4_attr_##_name = { \
2536 .attr = {.name = __stringify(_name), .mode = _mode }, \
2539 .offset = offsetof(struct ext4_sb_info, _elname), \
2541 #define EXT4_ATTR(name, mode, show, store) \
2542 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2544 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2545 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2546 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2547 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2548 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2549 #define ATTR_LIST(name) &ext4_attr_##name.attr
2551 EXT4_RO_ATTR(delayed_allocation_blocks);
2552 EXT4_RO_ATTR(session_write_kbytes);
2553 EXT4_RO_ATTR(lifetime_write_kbytes);
2554 EXT4_RO_ATTR(extent_cache_hits);
2555 EXT4_RO_ATTR(extent_cache_misses);
2556 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2557 inode_readahead_blks_store, s_inode_readahead_blks);
2558 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2559 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2560 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2561 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2562 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2563 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2564 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2565 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2567 static struct attribute *ext4_attrs[] = {
2568 ATTR_LIST(delayed_allocation_blocks),
2569 ATTR_LIST(session_write_kbytes),
2570 ATTR_LIST(lifetime_write_kbytes),
2571 ATTR_LIST(extent_cache_hits),
2572 ATTR_LIST(extent_cache_misses),
2573 ATTR_LIST(inode_readahead_blks),
2574 ATTR_LIST(inode_goal),
2575 ATTR_LIST(mb_stats),
2576 ATTR_LIST(mb_max_to_scan),
2577 ATTR_LIST(mb_min_to_scan),
2578 ATTR_LIST(mb_order2_req),
2579 ATTR_LIST(mb_stream_req),
2580 ATTR_LIST(mb_group_prealloc),
2581 ATTR_LIST(max_writeback_mb_bump),
2585 /* Features this copy of ext4 supports */
2586 EXT4_INFO_ATTR(lazy_itable_init);
2587 EXT4_INFO_ATTR(batched_discard);
2589 static struct attribute *ext4_feat_attrs[] = {
2590 ATTR_LIST(lazy_itable_init),
2591 ATTR_LIST(batched_discard),
2595 static ssize_t ext4_attr_show(struct kobject *kobj,
2596 struct attribute *attr, char *buf)
2598 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2600 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2602 return a->show ? a->show(a, sbi, buf) : 0;
2605 static ssize_t ext4_attr_store(struct kobject *kobj,
2606 struct attribute *attr,
2607 const char *buf, size_t len)
2609 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2611 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2613 return a->store ? a->store(a, sbi, buf, len) : 0;
2616 static void ext4_sb_release(struct kobject *kobj)
2618 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 complete(&sbi->s_kobj_unregister);
2623 static const struct sysfs_ops ext4_attr_ops = {
2624 .show = ext4_attr_show,
2625 .store = ext4_attr_store,
2628 static struct kobj_type ext4_ktype = {
2629 .default_attrs = ext4_attrs,
2630 .sysfs_ops = &ext4_attr_ops,
2631 .release = ext4_sb_release,
2634 static void ext4_feat_release(struct kobject *kobj)
2636 complete(&ext4_feat->f_kobj_unregister);
2639 static struct kobj_type ext4_feat_ktype = {
2640 .default_attrs = ext4_feat_attrs,
2641 .sysfs_ops = &ext4_attr_ops,
2642 .release = ext4_feat_release,
2646 * Check whether this filesystem can be mounted based on
2647 * the features present and the RDONLY/RDWR mount requested.
2648 * Returns 1 if this filesystem can be mounted as requested,
2649 * 0 if it cannot be.
2651 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2654 ext4_msg(sb, KERN_ERR,
2655 "Couldn't mount because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2658 ~EXT4_FEATURE_INCOMPAT_SUPP));
2665 /* Check that feature set is OK for a read-write mount */
2666 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2667 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2670 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674 * Large file size enabled file system can only be mounted
2675 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2677 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2678 if (sizeof(blkcnt_t) < sizeof(u64)) {
2679 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2680 "cannot be mounted RDWR without "
2685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2686 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2687 ext4_msg(sb, KERN_ERR,
2688 "Can't support bigalloc feature without "
2689 "extents feature\n");
2696 * This function is called once a day if we have errors logged
2697 * on the file system
2699 static void print_daily_error_info(unsigned long arg)
2701 struct super_block *sb = (struct super_block *) arg;
2702 struct ext4_sb_info *sbi;
2703 struct ext4_super_block *es;
2708 if (es->s_error_count)
2709 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2710 le32_to_cpu(es->s_error_count));
2711 if (es->s_first_error_time) {
2712 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2713 sb->s_id, le32_to_cpu(es->s_first_error_time),
2714 (int) sizeof(es->s_first_error_func),
2715 es->s_first_error_func,
2716 le32_to_cpu(es->s_first_error_line));
2717 if (es->s_first_error_ino)
2718 printk(": inode %u",
2719 le32_to_cpu(es->s_first_error_ino));
2720 if (es->s_first_error_block)
2721 printk(": block %llu", (unsigned long long)
2722 le64_to_cpu(es->s_first_error_block));
2725 if (es->s_last_error_time) {
2726 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2727 sb->s_id, le32_to_cpu(es->s_last_error_time),
2728 (int) sizeof(es->s_last_error_func),
2729 es->s_last_error_func,
2730 le32_to_cpu(es->s_last_error_line));
2731 if (es->s_last_error_ino)
2732 printk(": inode %u",
2733 le32_to_cpu(es->s_last_error_ino));
2734 if (es->s_last_error_block)
2735 printk(": block %llu", (unsigned long long)
2736 le64_to_cpu(es->s_last_error_block));
2739 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2742 /* Find next suitable group and run ext4_init_inode_table */
2743 static int ext4_run_li_request(struct ext4_li_request *elr)
2745 struct ext4_group_desc *gdp = NULL;
2746 ext4_group_t group, ngroups;
2747 struct super_block *sb;
2748 unsigned long timeout = 0;
2752 ngroups = EXT4_SB(sb)->s_groups_count;
2754 for (group = elr->lr_next_group; group < ngroups; group++) {
2755 gdp = ext4_get_group_desc(sb, group, NULL);
2761 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2765 if (group == ngroups)
2770 ret = ext4_init_inode_table(sb, group,
2771 elr->lr_timeout ? 0 : 1);
2772 if (elr->lr_timeout == 0) {
2773 timeout = (jiffies - timeout) *
2774 elr->lr_sbi->s_li_wait_mult;
2775 elr->lr_timeout = timeout;
2777 elr->lr_next_sched = jiffies + elr->lr_timeout;
2778 elr->lr_next_group = group + 1;
2785 * Remove lr_request from the list_request and free the
2786 * request structure. Should be called with li_list_mtx held
2788 static void ext4_remove_li_request(struct ext4_li_request *elr)
2790 struct ext4_sb_info *sbi;
2797 list_del(&elr->lr_request);
2798 sbi->s_li_request = NULL;
2802 static void ext4_unregister_li_request(struct super_block *sb)
2804 mutex_lock(&ext4_li_mtx);
2805 if (!ext4_li_info) {
2806 mutex_unlock(&ext4_li_mtx);
2810 mutex_lock(&ext4_li_info->li_list_mtx);
2811 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2812 mutex_unlock(&ext4_li_info->li_list_mtx);
2813 mutex_unlock(&ext4_li_mtx);
2816 static struct task_struct *ext4_lazyinit_task;
2819 * This is the function where ext4lazyinit thread lives. It walks
2820 * through the request list searching for next scheduled filesystem.
2821 * When such a fs is found, run the lazy initialization request
2822 * (ext4_rn_li_request) and keep track of the time spend in this
2823 * function. Based on that time we compute next schedule time of
2824 * the request. When walking through the list is complete, compute
2825 * next waking time and put itself into sleep.
2827 static int ext4_lazyinit_thread(void *arg)
2829 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2830 struct list_head *pos, *n;
2831 struct ext4_li_request *elr;
2832 unsigned long next_wakeup, cur;
2834 BUG_ON(NULL == eli);
2838 next_wakeup = MAX_JIFFY_OFFSET;
2840 mutex_lock(&eli->li_list_mtx);
2841 if (list_empty(&eli->li_request_list)) {
2842 mutex_unlock(&eli->li_list_mtx);
2846 list_for_each_safe(pos, n, &eli->li_request_list) {
2847 elr = list_entry(pos, struct ext4_li_request,
2850 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2851 if (ext4_run_li_request(elr) != 0) {
2852 /* error, remove the lazy_init job */
2853 ext4_remove_li_request(elr);
2858 if (time_before(elr->lr_next_sched, next_wakeup))
2859 next_wakeup = elr->lr_next_sched;
2861 mutex_unlock(&eli->li_list_mtx);
2866 if ((time_after_eq(cur, next_wakeup)) ||
2867 (MAX_JIFFY_OFFSET == next_wakeup)) {
2872 schedule_timeout_interruptible(next_wakeup - cur);
2874 if (kthread_should_stop()) {
2875 ext4_clear_request_list();
2882 * It looks like the request list is empty, but we need
2883 * to check it under the li_list_mtx lock, to prevent any
2884 * additions into it, and of course we should lock ext4_li_mtx
2885 * to atomically free the list and ext4_li_info, because at
2886 * this point another ext4 filesystem could be registering
2889 mutex_lock(&ext4_li_mtx);
2890 mutex_lock(&eli->li_list_mtx);
2891 if (!list_empty(&eli->li_request_list)) {
2892 mutex_unlock(&eli->li_list_mtx);
2893 mutex_unlock(&ext4_li_mtx);
2896 mutex_unlock(&eli->li_list_mtx);
2897 kfree(ext4_li_info);
2898 ext4_li_info = NULL;
2899 mutex_unlock(&ext4_li_mtx);
2904 static void ext4_clear_request_list(void)
2906 struct list_head *pos, *n;
2907 struct ext4_li_request *elr;
2909 mutex_lock(&ext4_li_info->li_list_mtx);
2910 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2911 elr = list_entry(pos, struct ext4_li_request,
2913 ext4_remove_li_request(elr);
2915 mutex_unlock(&ext4_li_info->li_list_mtx);
2918 static int ext4_run_lazyinit_thread(void)
2920 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2921 ext4_li_info, "ext4lazyinit");
2922 if (IS_ERR(ext4_lazyinit_task)) {
2923 int err = PTR_ERR(ext4_lazyinit_task);
2924 ext4_clear_request_list();
2925 kfree(ext4_li_info);
2926 ext4_li_info = NULL;
2927 printk(KERN_CRIT "EXT4: error %d creating inode table "
2928 "initialization thread\n",
2932 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2937 * Check whether it make sense to run itable init. thread or not.
2938 * If there is at least one uninitialized inode table, return
2939 * corresponding group number, else the loop goes through all
2940 * groups and return total number of groups.
2942 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2944 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2945 struct ext4_group_desc *gdp = NULL;
2947 for (group = 0; group < ngroups; group++) {
2948 gdp = ext4_get_group_desc(sb, group, NULL);
2952 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2959 static int ext4_li_info_new(void)
2961 struct ext4_lazy_init *eli = NULL;
2963 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2967 INIT_LIST_HEAD(&eli->li_request_list);
2968 mutex_init(&eli->li_list_mtx);
2970 eli->li_state |= EXT4_LAZYINIT_QUIT;
2977 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2980 struct ext4_sb_info *sbi = EXT4_SB(sb);
2981 struct ext4_li_request *elr;
2984 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2990 elr->lr_next_group = start;
2993 * Randomize first schedule time of the request to
2994 * spread the inode table initialization requests
2997 get_random_bytes(&rnd, sizeof(rnd));
2998 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2999 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3004 static int ext4_register_li_request(struct super_block *sb,
3005 ext4_group_t first_not_zeroed)
3007 struct ext4_sb_info *sbi = EXT4_SB(sb);
3008 struct ext4_li_request *elr;
3009 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3012 if (sbi->s_li_request != NULL) {
3014 * Reset timeout so it can be computed again, because
3015 * s_li_wait_mult might have changed.
3017 sbi->s_li_request->lr_timeout = 0;
3021 if (first_not_zeroed == ngroups ||
3022 (sb->s_flags & MS_RDONLY) ||
3023 !test_opt(sb, INIT_INODE_TABLE))
3026 elr = ext4_li_request_new(sb, first_not_zeroed);
3030 mutex_lock(&ext4_li_mtx);
3032 if (NULL == ext4_li_info) {
3033 ret = ext4_li_info_new();
3038 mutex_lock(&ext4_li_info->li_list_mtx);
3039 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3040 mutex_unlock(&ext4_li_info->li_list_mtx);
3042 sbi->s_li_request = elr;
3044 * set elr to NULL here since it has been inserted to
3045 * the request_list and the removal and free of it is
3046 * handled by ext4_clear_request_list from now on.
3050 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3051 ret = ext4_run_lazyinit_thread();
3056 mutex_unlock(&ext4_li_mtx);
3063 * We do not need to lock anything since this is called on
3066 static void ext4_destroy_lazyinit_thread(void)
3069 * If thread exited earlier
3070 * there's nothing to be done.
3072 if (!ext4_li_info || !ext4_lazyinit_task)
3075 kthread_stop(ext4_lazyinit_task);
3078 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3080 char *orig_data = kstrdup(data, GFP_KERNEL);
3081 struct buffer_head *bh;
3082 struct ext4_super_block *es = NULL;
3083 struct ext4_sb_info *sbi;
3085 ext4_fsblk_t sb_block = get_sb_block(&data);
3086 ext4_fsblk_t logical_sb_block;
3087 unsigned long offset = 0;
3088 unsigned long journal_devnum = 0;
3089 unsigned long def_mount_opts;
3094 int blocksize, clustersize;
3095 unsigned int db_count;
3097 int needs_recovery, has_huge_files, has_bigalloc;
3100 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3101 ext4_group_t first_not_zeroed;
3103 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3107 sbi->s_blockgroup_lock =
3108 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3109 if (!sbi->s_blockgroup_lock) {
3113 sb->s_fs_info = sbi;
3114 sbi->s_mount_opt = 0;
3115 sbi->s_resuid = EXT4_DEF_RESUID;
3116 sbi->s_resgid = EXT4_DEF_RESGID;
3117 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3118 sbi->s_sb_block = sb_block;
3119 if (sb->s_bdev->bd_part)
3120 sbi->s_sectors_written_start =
3121 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3123 /* Cleanup superblock name */
3124 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3128 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3130 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3135 * The ext4 superblock will not be buffer aligned for other than 1kB
3136 * block sizes. We need to calculate the offset from buffer start.
3138 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3139 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3140 offset = do_div(logical_sb_block, blocksize);
3142 logical_sb_block = sb_block;
3145 if (!(bh = sb_bread(sb, logical_sb_block))) {
3146 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3150 * Note: s_es must be initialized as soon as possible because
3151 * some ext4 macro-instructions depend on its value
3153 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3155 sb->s_magic = le16_to_cpu(es->s_magic);
3156 if (sb->s_magic != EXT4_SUPER_MAGIC)
3158 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3160 /* Set defaults before we parse the mount options */
3161 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3162 set_opt(sb, INIT_INODE_TABLE);
3163 if (def_mount_opts & EXT4_DEFM_DEBUG)
3165 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3167 if (def_mount_opts & EXT4_DEFM_UID16)
3168 set_opt(sb, NO_UID32);
3169 /* xattr user namespace & acls are now defaulted on */
3170 #ifdef CONFIG_EXT4_FS_XATTR
3171 set_opt(sb, XATTR_USER);
3173 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3174 set_opt(sb, POSIX_ACL);
3176 set_opt(sb, MBLK_IO_SUBMIT);
3177 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3178 set_opt(sb, JOURNAL_DATA);
3179 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3180 set_opt(sb, ORDERED_DATA);
3181 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3182 set_opt(sb, WRITEBACK_DATA);
3184 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3185 set_opt(sb, ERRORS_PANIC);
3186 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3187 set_opt(sb, ERRORS_CONT);
3189 set_opt(sb, ERRORS_RO);
3190 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3191 set_opt(sb, BLOCK_VALIDITY);
3192 if (def_mount_opts & EXT4_DEFM_DISCARD)
3193 set_opt(sb, DISCARD);
3195 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3196 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3197 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3198 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3199 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3201 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3202 set_opt(sb, BARRIER);
3205 * enable delayed allocation by default
3206 * Use -o nodelalloc to turn it off
3208 if (!IS_EXT3_SB(sb) &&
3209 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3210 set_opt(sb, DELALLOC);
3213 * set default s_li_wait_mult for lazyinit, for the case there is
3214 * no mount option specified.
3216 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3218 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3219 &journal_devnum, &journal_ioprio, 0)) {
3220 ext4_msg(sb, KERN_WARNING,
3221 "failed to parse options in superblock: %s",
3222 sbi->s_es->s_mount_opts);
3224 if (!parse_options((char *) data, sb, &journal_devnum,
3225 &journal_ioprio, 0))
3228 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3229 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3230 "with data=journal disables delayed "
3231 "allocation and O_DIRECT support!\n");
3232 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3233 ext4_msg(sb, KERN_ERR, "can't mount with "
3234 "both data=journal and delalloc");
3237 if (test_opt(sb, DIOREAD_NOLOCK)) {
3238 ext4_msg(sb, KERN_ERR, "can't mount with "
3239 "both data=journal and delalloc");
3242 if (test_opt(sb, DELALLOC))
3243 clear_opt(sb, DELALLOC);
3246 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3247 if (test_opt(sb, DIOREAD_NOLOCK)) {
3248 if (blocksize < PAGE_SIZE) {
3249 ext4_msg(sb, KERN_ERR, "can't mount with "
3250 "dioread_nolock if block size != PAGE_SIZE");
3255 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3256 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3258 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3259 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3260 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3261 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3262 ext4_msg(sb, KERN_WARNING,
3263 "feature flags set on rev 0 fs, "
3264 "running e2fsck is recommended");
3266 if (IS_EXT2_SB(sb)) {
3267 if (ext2_feature_set_ok(sb))
3268 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3269 "using the ext4 subsystem");
3271 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3272 "to feature incompatibilities");
3277 if (IS_EXT3_SB(sb)) {
3278 if (ext3_feature_set_ok(sb))
3279 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3280 "using the ext4 subsystem");
3282 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3283 "to feature incompatibilities");
3289 * Check feature flags regardless of the revision level, since we
3290 * previously didn't change the revision level when setting the flags,
3291 * so there is a chance incompat flags are set on a rev 0 filesystem.
3293 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3296 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3297 blocksize > EXT4_MAX_BLOCK_SIZE) {
3298 ext4_msg(sb, KERN_ERR,
3299 "Unsupported filesystem blocksize %d", blocksize);
3303 if (sb->s_blocksize != blocksize) {
3304 /* Validate the filesystem blocksize */
3305 if (!sb_set_blocksize(sb, blocksize)) {
3306 ext4_msg(sb, KERN_ERR, "bad block size %d",
3312 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3313 offset = do_div(logical_sb_block, blocksize);
3314 bh = sb_bread(sb, logical_sb_block);
3316 ext4_msg(sb, KERN_ERR,
3317 "Can't read superblock on 2nd try");
3320 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3322 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3323 ext4_msg(sb, KERN_ERR,
3324 "Magic mismatch, very weird!");
3329 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3330 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3331 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3333 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3335 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3336 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3337 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3339 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3340 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3341 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3342 (!is_power_of_2(sbi->s_inode_size)) ||
3343 (sbi->s_inode_size > blocksize)) {
3344 ext4_msg(sb, KERN_ERR,
3345 "unsupported inode size: %d",
3349 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3350 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3353 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3354 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3355 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3356 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3357 !is_power_of_2(sbi->s_desc_size)) {
3358 ext4_msg(sb, KERN_ERR,
3359 "unsupported descriptor size %lu",
3364 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3366 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3367 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3368 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3371 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3372 if (sbi->s_inodes_per_block == 0)
3374 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3375 sbi->s_inodes_per_block;
3376 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3378 sbi->s_mount_state = le16_to_cpu(es->s_state);
3379 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3380 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3382 for (i = 0; i < 4; i++)
3383 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3384 sbi->s_def_hash_version = es->s_def_hash_version;
3385 i = le32_to_cpu(es->s_flags);
3386 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3387 sbi->s_hash_unsigned = 3;
3388 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3389 #ifdef __CHAR_UNSIGNED__
3390 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3391 sbi->s_hash_unsigned = 3;
3393 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3398 /* Handle clustersize */
3399 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3400 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3401 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3403 if (clustersize < blocksize) {
3404 ext4_msg(sb, KERN_ERR,
3405 "cluster size (%d) smaller than "
3406 "block size (%d)", clustersize, blocksize);
3409 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3410 le32_to_cpu(es->s_log_block_size);
3411 sbi->s_clusters_per_group =
3412 le32_to_cpu(es->s_clusters_per_group);
3413 if (sbi->s_clusters_per_group > blocksize * 8) {
3414 ext4_msg(sb, KERN_ERR,
3415 "#clusters per group too big: %lu",
3416 sbi->s_clusters_per_group);
3419 if (sbi->s_blocks_per_group !=
3420 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3421 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3422 "clusters per group (%lu) inconsistent",
3423 sbi->s_blocks_per_group,
3424 sbi->s_clusters_per_group);
3428 if (clustersize != blocksize) {
3429 ext4_warning(sb, "fragment/cluster size (%d) != "
3430 "block size (%d)", clustersize,
3432 clustersize = blocksize;
3434 if (sbi->s_blocks_per_group > blocksize * 8) {
3435 ext4_msg(sb, KERN_ERR,
3436 "#blocks per group too big: %lu",
3437 sbi->s_blocks_per_group);
3440 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3441 sbi->s_cluster_bits = 0;
3443 sbi->s_cluster_ratio = clustersize / blocksize;
3445 if (sbi->s_inodes_per_group > blocksize * 8) {
3446 ext4_msg(sb, KERN_ERR,
3447 "#inodes per group too big: %lu",
3448 sbi->s_inodes_per_group);
3453 * Test whether we have more sectors than will fit in sector_t,
3454 * and whether the max offset is addressable by the page cache.
3456 err = generic_check_addressable(sb->s_blocksize_bits,
3457 ext4_blocks_count(es));
3459 ext4_msg(sb, KERN_ERR, "filesystem"
3460 " too large to mount safely on this system");
3461 if (sizeof(sector_t) < 8)
3462 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3467 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3470 /* check blocks count against device size */
3471 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3472 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3473 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3474 "exceeds size of device (%llu blocks)",
3475 ext4_blocks_count(es), blocks_count);
3480 * It makes no sense for the first data block to be beyond the end
3481 * of the filesystem.
3483 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3484 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3485 "block %u is beyond end of filesystem (%llu)",
3486 le32_to_cpu(es->s_first_data_block),
3487 ext4_blocks_count(es));
3490 blocks_count = (ext4_blocks_count(es) -
3491 le32_to_cpu(es->s_first_data_block) +
3492 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3493 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3494 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3495 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3496 "(block count %llu, first data block %u, "
3497 "blocks per group %lu)", sbi->s_groups_count,
3498 ext4_blocks_count(es),
3499 le32_to_cpu(es->s_first_data_block),
3500 EXT4_BLOCKS_PER_GROUP(sb));
3503 sbi->s_groups_count = blocks_count;
3504 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3505 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3506 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3507 EXT4_DESC_PER_BLOCK(sb);
3508 sbi->s_group_desc = ext4_kvmalloc(db_count *
3509 sizeof(struct buffer_head *),
3511 if (sbi->s_group_desc == NULL) {
3512 ext4_msg(sb, KERN_ERR, "not enough memory");
3517 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3519 bgl_lock_init(sbi->s_blockgroup_lock);
3521 for (i = 0; i < db_count; i++) {
3522 block = descriptor_loc(sb, logical_sb_block, i);
3523 sbi->s_group_desc[i] = sb_bread(sb, block);
3524 if (!sbi->s_group_desc[i]) {
3525 ext4_msg(sb, KERN_ERR,
3526 "can't read group descriptor %d", i);
3531 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3532 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3535 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3536 if (!ext4_fill_flex_info(sb)) {
3537 ext4_msg(sb, KERN_ERR,
3538 "unable to initialize "
3539 "flex_bg meta info!");
3543 sbi->s_gdb_count = db_count;
3544 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3545 spin_lock_init(&sbi->s_next_gen_lock);
3547 init_timer(&sbi->s_err_report);
3548 sbi->s_err_report.function = print_daily_error_info;
3549 sbi->s_err_report.data = (unsigned long) sb;
3551 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3552 ext4_count_free_clusters(sb));
3554 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3555 ext4_count_free_inodes(sb));
3558 err = percpu_counter_init(&sbi->s_dirs_counter,
3559 ext4_count_dirs(sb));
3562 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3565 ext4_msg(sb, KERN_ERR, "insufficient memory");
3569 sbi->s_stripe = ext4_get_stripe_size(sbi);
3570 sbi->s_max_writeback_mb_bump = 128;
3573 * set up enough so that it can read an inode
3575 if (!test_opt(sb, NOLOAD) &&
3576 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3577 sb->s_op = &ext4_sops;
3579 sb->s_op = &ext4_nojournal_sops;
3580 sb->s_export_op = &ext4_export_ops;
3581 sb->s_xattr = ext4_xattr_handlers;
3583 sb->s_qcop = &ext4_qctl_operations;
3584 sb->dq_op = &ext4_quota_operations;
3586 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3588 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3589 mutex_init(&sbi->s_orphan_lock);
3590 sbi->s_resize_flags = 0;
3594 needs_recovery = (es->s_last_orphan != 0 ||
3595 EXT4_HAS_INCOMPAT_FEATURE(sb,
3596 EXT4_FEATURE_INCOMPAT_RECOVER));
3598 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3599 !(sb->s_flags & MS_RDONLY))
3600 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3604 * The first inode we look at is the journal inode. Don't try
3605 * root first: it may be modified in the journal!
3607 if (!test_opt(sb, NOLOAD) &&
3608 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3609 if (ext4_load_journal(sb, es, journal_devnum))
3611 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3612 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3613 ext4_msg(sb, KERN_ERR, "required journal recovery "
3614 "suppressed and not mounted read-only");
3615 goto failed_mount_wq;
3617 clear_opt(sb, DATA_FLAGS);
3618 sbi->s_journal = NULL;
3623 if (ext4_blocks_count(es) > 0xffffffffULL &&
3624 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3625 JBD2_FEATURE_INCOMPAT_64BIT)) {
3626 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3627 goto failed_mount_wq;
3630 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3631 jbd2_journal_set_features(sbi->s_journal,
3632 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3633 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3634 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3635 jbd2_journal_set_features(sbi->s_journal,
3636 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3637 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3638 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3640 jbd2_journal_clear_features(sbi->s_journal,
3641 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3642 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3645 /* We have now updated the journal if required, so we can
3646 * validate the data journaling mode. */
3647 switch (test_opt(sb, DATA_FLAGS)) {
3649 /* No mode set, assume a default based on the journal
3650 * capabilities: ORDERED_DATA if the journal can
3651 * cope, else JOURNAL_DATA
3653 if (jbd2_journal_check_available_features
3654 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3655 set_opt(sb, ORDERED_DATA);
3657 set_opt(sb, JOURNAL_DATA);
3660 case EXT4_MOUNT_ORDERED_DATA:
3661 case EXT4_MOUNT_WRITEBACK_DATA:
3662 if (!jbd2_journal_check_available_features
3663 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3664 ext4_msg(sb, KERN_ERR, "Journal does not support "
3665 "requested data journaling mode");
3666 goto failed_mount_wq;
3671 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3673 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3676 * The journal may have updated the bg summary counts, so we
3677 * need to update the global counters.
3679 percpu_counter_set(&sbi->s_freeclusters_counter,
3680 ext4_count_free_clusters(sb));
3681 percpu_counter_set(&sbi->s_freeinodes_counter,
3682 ext4_count_free_inodes(sb));
3683 percpu_counter_set(&sbi->s_dirs_counter,
3684 ext4_count_dirs(sb));
3685 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3689 * The maximum number of concurrent works can be high and
3690 * concurrency isn't really necessary. Limit it to 1.
3692 EXT4_SB(sb)->dio_unwritten_wq =
3693 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3694 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3695 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3696 goto failed_mount_wq;
3700 * The jbd2_journal_load will have done any necessary log recovery,
3701 * so we can safely mount the rest of the filesystem now.
3704 root = ext4_iget(sb, EXT4_ROOT_INO);
3706 ext4_msg(sb, KERN_ERR, "get root inode failed");
3707 ret = PTR_ERR(root);
3711 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3712 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3716 sb->s_root = d_alloc_root(root);
3719 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3724 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3726 /* determine the minimum size of new large inodes, if present */
3727 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3728 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3729 EXT4_GOOD_OLD_INODE_SIZE;
3730 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3731 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3732 if (sbi->s_want_extra_isize <
3733 le16_to_cpu(es->s_want_extra_isize))
3734 sbi->s_want_extra_isize =
3735 le16_to_cpu(es->s_want_extra_isize);
3736 if (sbi->s_want_extra_isize <
3737 le16_to_cpu(es->s_min_extra_isize))
3738 sbi->s_want_extra_isize =
3739 le16_to_cpu(es->s_min_extra_isize);
3742 /* Check if enough inode space is available */
3743 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3744 sbi->s_inode_size) {
3745 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3746 EXT4_GOOD_OLD_INODE_SIZE;
3747 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3751 err = ext4_setup_system_zone(sb);
3753 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3755 goto failed_mount4a;
3759 err = ext4_mb_init(sb, needs_recovery);
3761 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3766 err = ext4_register_li_request(sb, first_not_zeroed);
3770 sbi->s_kobj.kset = ext4_kset;
3771 init_completion(&sbi->s_kobj_unregister);
3772 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3777 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3778 ext4_orphan_cleanup(sb, es);
3779 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3780 if (needs_recovery) {
3781 ext4_msg(sb, KERN_INFO, "recovery complete");
3782 ext4_mark_recovery_complete(sb, es);
3784 if (EXT4_SB(sb)->s_journal) {
3785 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3786 descr = " journalled data mode";
3787 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3788 descr = " ordered data mode";
3790 descr = " writeback data mode";
3792 descr = "out journal";
3794 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3795 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3796 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3798 if (es->s_error_count)
3799 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3806 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3810 ext4_unregister_li_request(sb);
3812 ext4_mb_release(sb);
3814 ext4_ext_release(sb);
3815 ext4_release_system_zone(sb);
3820 ext4_msg(sb, KERN_ERR, "mount failed");
3821 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3823 if (sbi->s_journal) {
3824 jbd2_journal_destroy(sbi->s_journal);
3825 sbi->s_journal = NULL;
3828 del_timer(&sbi->s_err_report);
3829 if (sbi->s_flex_groups)
3830 ext4_kvfree(sbi->s_flex_groups);
3831 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3832 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3833 percpu_counter_destroy(&sbi->s_dirs_counter);
3834 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3836 kthread_stop(sbi->s_mmp_tsk);
3838 for (i = 0; i < db_count; i++)
3839 brelse(sbi->s_group_desc[i]);
3840 ext4_kvfree(sbi->s_group_desc);
3843 remove_proc_entry(sb->s_id, ext4_proc_root);
3846 for (i = 0; i < MAXQUOTAS; i++)
3847 kfree(sbi->s_qf_names[i]);
3849 ext4_blkdev_remove(sbi);
3852 sb->s_fs_info = NULL;
3853 kfree(sbi->s_blockgroup_lock);
3861 * Setup any per-fs journal parameters now. We'll do this both on
3862 * initial mount, once the journal has been initialised but before we've
3863 * done any recovery; and again on any subsequent remount.
3865 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3867 struct ext4_sb_info *sbi = EXT4_SB(sb);
3869 journal->j_commit_interval = sbi->s_commit_interval;
3870 journal->j_min_batch_time = sbi->s_min_batch_time;
3871 journal->j_max_batch_time = sbi->s_max_batch_time;
3873 write_lock(&journal->j_state_lock);
3874 if (test_opt(sb, BARRIER))
3875 journal->j_flags |= JBD2_BARRIER;
3877 journal->j_flags &= ~JBD2_BARRIER;
3878 if (test_opt(sb, DATA_ERR_ABORT))
3879 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3881 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3882 write_unlock(&journal->j_state_lock);
3885 static journal_t *ext4_get_journal(struct super_block *sb,
3886 unsigned int journal_inum)
3888 struct inode *journal_inode;
3891 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3893 /* First, test for the existence of a valid inode on disk. Bad
3894 * things happen if we iget() an unused inode, as the subsequent
3895 * iput() will try to delete it. */
3897 journal_inode = ext4_iget(sb, journal_inum);
3898 if (IS_ERR(journal_inode)) {
3899 ext4_msg(sb, KERN_ERR, "no journal found");
3902 if (!journal_inode->i_nlink) {
3903 make_bad_inode(journal_inode);
3904 iput(journal_inode);
3905 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3909 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3910 journal_inode, journal_inode->i_size);
3911 if (!S_ISREG(journal_inode->i_mode)) {
3912 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3913 iput(journal_inode);
3917 journal = jbd2_journal_init_inode(journal_inode);
3919 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3920 iput(journal_inode);
3923 journal->j_private = sb;
3924 ext4_init_journal_params(sb, journal);
3928 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3931 struct buffer_head *bh;
3935 int hblock, blocksize;
3936 ext4_fsblk_t sb_block;
3937 unsigned long offset;
3938 struct ext4_super_block *es;
3939 struct block_device *bdev;
3941 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3943 bdev = ext4_blkdev_get(j_dev, sb);
3947 blocksize = sb->s_blocksize;
3948 hblock = bdev_logical_block_size(bdev);
3949 if (blocksize < hblock) {
3950 ext4_msg(sb, KERN_ERR,
3951 "blocksize too small for journal device");
3955 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3956 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3957 set_blocksize(bdev, blocksize);
3958 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3959 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3960 "external journal");
3964 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3965 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3966 !(le32_to_cpu(es->s_feature_incompat) &
3967 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3968 ext4_msg(sb, KERN_ERR, "external journal has "
3974 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3975 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3980 len = ext4_blocks_count(es);
3981 start = sb_block + 1;
3982 brelse(bh); /* we're done with the superblock */
3984 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3985 start, len, blocksize);
3987 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3990 journal->j_private = sb;
3991 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3992 wait_on_buffer(journal->j_sb_buffer);
3993 if (!buffer_uptodate(journal->j_sb_buffer)) {
3994 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3997 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3998 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3999 "user (unsupported) - %d",
4000 be32_to_cpu(journal->j_superblock->s_nr_users));
4003 EXT4_SB(sb)->journal_bdev = bdev;
4004 ext4_init_journal_params(sb, journal);
4008 jbd2_journal_destroy(journal);
4010 ext4_blkdev_put(bdev);
4014 static int ext4_load_journal(struct super_block *sb,
4015 struct ext4_super_block *es,
4016 unsigned long journal_devnum)
4019 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4022 int really_read_only;
4024 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4026 if (journal_devnum &&
4027 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4028 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4029 "numbers have changed");
4030 journal_dev = new_decode_dev(journal_devnum);
4032 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4034 really_read_only = bdev_read_only(sb->s_bdev);
4037 * Are we loading a blank journal or performing recovery after a
4038 * crash? For recovery, we need to check in advance whether we
4039 * can get read-write access to the device.
4041 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4042 if (sb->s_flags & MS_RDONLY) {
4043 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4044 "required on readonly filesystem");
4045 if (really_read_only) {
4046 ext4_msg(sb, KERN_ERR, "write access "
4047 "unavailable, cannot proceed");
4050 ext4_msg(sb, KERN_INFO, "write access will "
4051 "be enabled during recovery");
4055 if (journal_inum && journal_dev) {
4056 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4057 "and inode journals!");
4062 if (!(journal = ext4_get_journal(sb, journal_inum)))
4065 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4069 if (!(journal->j_flags & JBD2_BARRIER))
4070 ext4_msg(sb, KERN_INFO, "barriers disabled");
4072 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4073 err = jbd2_journal_wipe(journal, !really_read_only);
4075 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4077 memcpy(save, ((char *) es) +
4078 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4079 err = jbd2_journal_load(journal);
4081 memcpy(((char *) es) + EXT4_S_ERR_START,
4082 save, EXT4_S_ERR_LEN);
4087 ext4_msg(sb, KERN_ERR, "error loading journal");
4088 jbd2_journal_destroy(journal);
4092 EXT4_SB(sb)->s_journal = journal;
4093 ext4_clear_journal_err(sb, es);
4095 if (!really_read_only && journal_devnum &&
4096 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4097 es->s_journal_dev = cpu_to_le32(journal_devnum);
4099 /* Make sure we flush the recovery flag to disk. */
4100 ext4_commit_super(sb, 1);
4106 static int ext4_commit_super(struct super_block *sb, int sync)
4108 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4109 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4112 if (!sbh || block_device_ejected(sb))
4114 if (buffer_write_io_error(sbh)) {
4116 * Oh, dear. A previous attempt to write the
4117 * superblock failed. This could happen because the
4118 * USB device was yanked out. Or it could happen to
4119 * be a transient write error and maybe the block will
4120 * be remapped. Nothing we can do but to retry the
4121 * write and hope for the best.
4123 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4124 "superblock detected");
4125 clear_buffer_write_io_error(sbh);
4126 set_buffer_uptodate(sbh);
4129 * If the file system is mounted read-only, don't update the
4130 * superblock write time. This avoids updating the superblock
4131 * write time when we are mounting the root file system
4132 * read/only but we need to replay the journal; at that point,
4133 * for people who are east of GMT and who make their clock
4134 * tick in localtime for Windows bug-for-bug compatibility,
4135 * the clock is set in the future, and this will cause e2fsck
4136 * to complain and force a full file system check.
4138 if (!(sb->s_flags & MS_RDONLY))
4139 es->s_wtime = cpu_to_le32(get_seconds());
4140 if (sb->s_bdev->bd_part)
4141 es->s_kbytes_written =
4142 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4143 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4144 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4146 es->s_kbytes_written =
4147 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4148 ext4_free_blocks_count_set(es,
4149 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4150 &EXT4_SB(sb)->s_freeclusters_counter)));
4151 es->s_free_inodes_count =
4152 cpu_to_le32(percpu_counter_sum_positive(
4153 &EXT4_SB(sb)->s_freeinodes_counter));
4155 BUFFER_TRACE(sbh, "marking dirty");
4156 mark_buffer_dirty(sbh);
4158 error = sync_dirty_buffer(sbh);
4162 error = buffer_write_io_error(sbh);
4164 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4166 clear_buffer_write_io_error(sbh);
4167 set_buffer_uptodate(sbh);
4174 * Have we just finished recovery? If so, and if we are mounting (or
4175 * remounting) the filesystem readonly, then we will end up with a
4176 * consistent fs on disk. Record that fact.
4178 static void ext4_mark_recovery_complete(struct super_block *sb,
4179 struct ext4_super_block *es)
4181 journal_t *journal = EXT4_SB(sb)->s_journal;
4183 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4184 BUG_ON(journal != NULL);
4187 jbd2_journal_lock_updates(journal);
4188 if (jbd2_journal_flush(journal) < 0)
4191 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4192 sb->s_flags & MS_RDONLY) {
4193 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4194 ext4_commit_super(sb, 1);
4198 jbd2_journal_unlock_updates(journal);
4202 * If we are mounting (or read-write remounting) a filesystem whose journal
4203 * has recorded an error from a previous lifetime, move that error to the
4204 * main filesystem now.
4206 static void ext4_clear_journal_err(struct super_block *sb,
4207 struct ext4_super_block *es)
4213 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4215 journal = EXT4_SB(sb)->s_journal;
4218 * Now check for any error status which may have been recorded in the
4219 * journal by a prior ext4_error() or ext4_abort()
4222 j_errno = jbd2_journal_errno(journal);
4226 errstr = ext4_decode_error(sb, j_errno, nbuf);
4227 ext4_warning(sb, "Filesystem error recorded "
4228 "from previous mount: %s", errstr);
4229 ext4_warning(sb, "Marking fs in need of filesystem check.");
4231 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4232 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4233 ext4_commit_super(sb, 1);
4235 jbd2_journal_clear_err(journal);
4240 * Force the running and committing transactions to commit,
4241 * and wait on the commit.
4243 int ext4_force_commit(struct super_block *sb)
4248 if (sb->s_flags & MS_RDONLY)
4251 journal = EXT4_SB(sb)->s_journal;
4253 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4254 ret = ext4_journal_force_commit(journal);
4260 static void ext4_write_super(struct super_block *sb)
4263 ext4_commit_super(sb, 1);
4267 static int ext4_sync_fs(struct super_block *sb, int wait)
4271 struct ext4_sb_info *sbi = EXT4_SB(sb);
4273 trace_ext4_sync_fs(sb, wait);
4274 flush_workqueue(sbi->dio_unwritten_wq);
4275 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4277 jbd2_log_wait_commit(sbi->s_journal, target);
4283 * LVM calls this function before a (read-only) snapshot is created. This
4284 * gives us a chance to flush the journal completely and mark the fs clean.
4286 * Note that only this function cannot bring a filesystem to be in a clean
4287 * state independently, because ext4 prevents a new handle from being started
4288 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4291 static int ext4_freeze(struct super_block *sb)
4296 if (sb->s_flags & MS_RDONLY)
4299 journal = EXT4_SB(sb)->s_journal;
4301 /* Now we set up the journal barrier. */
4302 jbd2_journal_lock_updates(journal);
4305 * Don't clear the needs_recovery flag if we failed to flush
4308 error = jbd2_journal_flush(journal);
4312 /* Journal blocked and flushed, clear needs_recovery flag. */
4313 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4314 error = ext4_commit_super(sb, 1);
4316 /* we rely on s_frozen to stop further updates */
4317 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4322 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4323 * flag here, even though the filesystem is not technically dirty yet.
4325 static int ext4_unfreeze(struct super_block *sb)
4327 if (sb->s_flags & MS_RDONLY)
4331 /* Reset the needs_recovery flag before the fs is unlocked. */
4332 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4333 ext4_commit_super(sb, 1);
4339 * Structure to save mount options for ext4_remount's benefit
4341 struct ext4_mount_options {
4342 unsigned long s_mount_opt;
4343 unsigned long s_mount_opt2;
4346 unsigned long s_commit_interval;
4347 u32 s_min_batch_time, s_max_batch_time;
4350 char *s_qf_names[MAXQUOTAS];
4354 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4356 struct ext4_super_block *es;
4357 struct ext4_sb_info *sbi = EXT4_SB(sb);
4358 unsigned long old_sb_flags;
4359 struct ext4_mount_options old_opts;
4360 int enable_quota = 0;
4362 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4367 char *orig_data = kstrdup(data, GFP_KERNEL);
4369 /* Store the original options */
4371 old_sb_flags = sb->s_flags;
4372 old_opts.s_mount_opt = sbi->s_mount_opt;
4373 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4374 old_opts.s_resuid = sbi->s_resuid;
4375 old_opts.s_resgid = sbi->s_resgid;
4376 old_opts.s_commit_interval = sbi->s_commit_interval;
4377 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4378 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4380 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4381 for (i = 0; i < MAXQUOTAS; i++)
4382 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4384 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4385 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4388 * Allow the "check" option to be passed as a remount option.
4390 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4395 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4396 ext4_abort(sb, "Abort forced by user");
4398 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4399 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4403 if (sbi->s_journal) {
4404 ext4_init_journal_params(sb, sbi->s_journal);
4405 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4408 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4409 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4414 if (*flags & MS_RDONLY) {
4415 err = dquot_suspend(sb, -1);
4420 * First of all, the unconditional stuff we have to do
4421 * to disable replay of the journal when we next remount
4423 sb->s_flags |= MS_RDONLY;
4426 * OK, test if we are remounting a valid rw partition
4427 * readonly, and if so set the rdonly flag and then
4428 * mark the partition as valid again.
4430 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4431 (sbi->s_mount_state & EXT4_VALID_FS))
4432 es->s_state = cpu_to_le16(sbi->s_mount_state);
4435 ext4_mark_recovery_complete(sb, es);
4437 /* Make sure we can mount this feature set readwrite */
4438 if (!ext4_feature_set_ok(sb, 0)) {
4443 * Make sure the group descriptor checksums
4444 * are sane. If they aren't, refuse to remount r/w.
4446 for (g = 0; g < sbi->s_groups_count; g++) {
4447 struct ext4_group_desc *gdp =
4448 ext4_get_group_desc(sb, g, NULL);
4450 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4451 ext4_msg(sb, KERN_ERR,
4452 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4453 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4454 le16_to_cpu(gdp->bg_checksum));
4461 * If we have an unprocessed orphan list hanging
4462 * around from a previously readonly bdev mount,
4463 * require a full umount/remount for now.
4465 if (es->s_last_orphan) {
4466 ext4_msg(sb, KERN_WARNING, "Couldn't "
4467 "remount RDWR because of unprocessed "
4468 "orphan inode list. Please "
4469 "umount/remount instead");
4475 * Mounting a RDONLY partition read-write, so reread
4476 * and store the current valid flag. (It may have
4477 * been changed by e2fsck since we originally mounted
4481 ext4_clear_journal_err(sb, es);
4482 sbi->s_mount_state = le16_to_cpu(es->s_state);
4483 if (!ext4_setup_super(sb, es, 0))
4484 sb->s_flags &= ~MS_RDONLY;
4485 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4486 EXT4_FEATURE_INCOMPAT_MMP))
4487 if (ext4_multi_mount_protect(sb,
4488 le64_to_cpu(es->s_mmp_block))) {
4497 * Reinitialize lazy itable initialization thread based on
4500 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4501 ext4_unregister_li_request(sb);
4503 ext4_group_t first_not_zeroed;
4504 first_not_zeroed = ext4_has_uninit_itable(sb);
4505 ext4_register_li_request(sb, first_not_zeroed);
4508 ext4_setup_system_zone(sb);
4509 if (sbi->s_journal == NULL)
4510 ext4_commit_super(sb, 1);
4513 /* Release old quota file names */
4514 for (i = 0; i < MAXQUOTAS; i++)
4515 if (old_opts.s_qf_names[i] &&
4516 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4517 kfree(old_opts.s_qf_names[i]);
4521 dquot_resume(sb, -1);
4523 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4528 sb->s_flags = old_sb_flags;
4529 sbi->s_mount_opt = old_opts.s_mount_opt;
4530 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4531 sbi->s_resuid = old_opts.s_resuid;
4532 sbi->s_resgid = old_opts.s_resgid;
4533 sbi->s_commit_interval = old_opts.s_commit_interval;
4534 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4535 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4537 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4538 for (i = 0; i < MAXQUOTAS; i++) {
4539 if (sbi->s_qf_names[i] &&
4540 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4541 kfree(sbi->s_qf_names[i]);
4542 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4551 * Note: calculating the overhead so we can be compatible with
4552 * historical BSD practice is quite difficult in the face of
4553 * clusters/bigalloc. This is because multiple metadata blocks from
4554 * different block group can end up in the same allocation cluster.
4555 * Calculating the exact overhead in the face of clustered allocation
4556 * requires either O(all block bitmaps) in memory or O(number of block
4557 * groups**2) in time. We will still calculate the superblock for
4558 * older file systems --- and if we come across with a bigalloc file
4559 * system with zero in s_overhead_clusters the estimate will be close to
4560 * correct especially for very large cluster sizes --- but for newer
4561 * file systems, it's better to calculate this figure once at mkfs
4562 * time, and store it in the superblock. If the superblock value is
4563 * present (even for non-bigalloc file systems), we will use it.
4565 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4567 struct super_block *sb = dentry->d_sb;
4568 struct ext4_sb_info *sbi = EXT4_SB(sb);
4569 struct ext4_super_block *es = sbi->s_es;
4570 struct ext4_group_desc *gdp;
4574 if (test_opt(sb, MINIX_DF)) {
4575 sbi->s_overhead_last = 0;
4576 } else if (es->s_overhead_clusters) {
4577 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4578 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4579 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4580 ext4_fsblk_t overhead = 0;
4583 * Compute the overhead (FS structures). This is constant
4584 * for a given filesystem unless the number of block groups
4585 * changes so we cache the previous value until it does.
4589 * All of the blocks before first_data_block are
4592 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4595 * Add the overhead found in each block group
4597 for (i = 0; i < ngroups; i++) {
4598 gdp = ext4_get_group_desc(sb, i, NULL);
4599 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4602 sbi->s_overhead_last = overhead;
4604 sbi->s_blocks_last = ext4_blocks_count(es);
4607 buf->f_type = EXT4_SUPER_MAGIC;
4608 buf->f_bsize = sb->s_blocksize;
4609 buf->f_blocks = (ext4_blocks_count(es) -
4610 EXT4_C2B(sbi, sbi->s_overhead_last));
4611 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4612 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4613 /* prevent underflow in case that few free space is available */
4614 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4615 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4616 if (buf->f_bfree < ext4_r_blocks_count(es))
4618 buf->f_files = le32_to_cpu(es->s_inodes_count);
4619 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4620 buf->f_namelen = EXT4_NAME_LEN;
4621 fsid = le64_to_cpup((void *)es->s_uuid) ^
4622 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4623 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4624 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4629 /* Helper function for writing quotas on sync - we need to start transaction
4630 * before quota file is locked for write. Otherwise the are possible deadlocks:
4631 * Process 1 Process 2
4632 * ext4_create() quota_sync()
4633 * jbd2_journal_start() write_dquot()
4634 * dquot_initialize() down(dqio_mutex)
4635 * down(dqio_mutex) jbd2_journal_start()
4641 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4643 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4646 static int ext4_write_dquot(struct dquot *dquot)
4650 struct inode *inode;
4652 inode = dquot_to_inode(dquot);
4653 handle = ext4_journal_start(inode,
4654 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4656 return PTR_ERR(handle);
4657 ret = dquot_commit(dquot);
4658 err = ext4_journal_stop(handle);
4664 static int ext4_acquire_dquot(struct dquot *dquot)
4669 handle = ext4_journal_start(dquot_to_inode(dquot),
4670 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4672 return PTR_ERR(handle);
4673 ret = dquot_acquire(dquot);
4674 err = ext4_journal_stop(handle);
4680 static int ext4_release_dquot(struct dquot *dquot)
4685 handle = ext4_journal_start(dquot_to_inode(dquot),
4686 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4687 if (IS_ERR(handle)) {
4688 /* Release dquot anyway to avoid endless cycle in dqput() */
4689 dquot_release(dquot);
4690 return PTR_ERR(handle);
4692 ret = dquot_release(dquot);
4693 err = ext4_journal_stop(handle);
4699 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4701 /* Are we journaling quotas? */
4702 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4703 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4704 dquot_mark_dquot_dirty(dquot);
4705 return ext4_write_dquot(dquot);
4707 return dquot_mark_dquot_dirty(dquot);
4711 static int ext4_write_info(struct super_block *sb, int type)
4716 /* Data block + inode block */
4717 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4719 return PTR_ERR(handle);
4720 ret = dquot_commit_info(sb, type);
4721 err = ext4_journal_stop(handle);
4728 * Turn on quotas during mount time - we need to find
4729 * the quota file and such...
4731 static int ext4_quota_on_mount(struct super_block *sb, int type)
4733 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4734 EXT4_SB(sb)->s_jquota_fmt, type);
4738 * Standard function to be called on quota_on
4740 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4745 if (!test_opt(sb, QUOTA))
4748 /* Quotafile not on the same filesystem? */
4749 if (path->dentry->d_sb != sb)
4751 /* Journaling quota? */
4752 if (EXT4_SB(sb)->s_qf_names[type]) {
4753 /* Quotafile not in fs root? */
4754 if (path->dentry->d_parent != sb->s_root)
4755 ext4_msg(sb, KERN_WARNING,
4756 "Quota file not on filesystem root. "
4757 "Journaled quota will not work");
4761 * When we journal data on quota file, we have to flush journal to see
4762 * all updates to the file when we bypass pagecache...
4764 if (EXT4_SB(sb)->s_journal &&
4765 ext4_should_journal_data(path->dentry->d_inode)) {
4767 * We don't need to lock updates but journal_flush() could
4768 * otherwise be livelocked...
4770 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4771 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4772 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4777 return dquot_quota_on(sb, type, format_id, path);
4780 static int ext4_quota_off(struct super_block *sb, int type)
4782 struct inode *inode = sb_dqopt(sb)->files[type];
4785 /* Force all delayed allocation blocks to be allocated.
4786 * Caller already holds s_umount sem */
4787 if (test_opt(sb, DELALLOC))
4788 sync_filesystem(sb);
4793 /* Update modification times of quota files when userspace can
4794 * start looking at them */
4795 handle = ext4_journal_start(inode, 1);
4798 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4799 ext4_mark_inode_dirty(handle, inode);
4800 ext4_journal_stop(handle);
4803 return dquot_quota_off(sb, type);
4806 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4807 * acquiring the locks... As quota files are never truncated and quota code
4808 * itself serializes the operations (and no one else should touch the files)
4809 * we don't have to be afraid of races */
4810 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4811 size_t len, loff_t off)
4813 struct inode *inode = sb_dqopt(sb)->files[type];
4814 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4816 int offset = off & (sb->s_blocksize - 1);
4819 struct buffer_head *bh;
4820 loff_t i_size = i_size_read(inode);
4824 if (off+len > i_size)
4827 while (toread > 0) {
4828 tocopy = sb->s_blocksize - offset < toread ?
4829 sb->s_blocksize - offset : toread;
4830 bh = ext4_bread(NULL, inode, blk, 0, &err);
4833 if (!bh) /* A hole? */
4834 memset(data, 0, tocopy);
4836 memcpy(data, bh->b_data+offset, tocopy);
4846 /* Write to quotafile (we know the transaction is already started and has
4847 * enough credits) */
4848 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4849 const char *data, size_t len, loff_t off)
4851 struct inode *inode = sb_dqopt(sb)->files[type];
4852 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4854 int offset = off & (sb->s_blocksize - 1);
4855 struct buffer_head *bh;
4856 handle_t *handle = journal_current_handle();
4858 if (EXT4_SB(sb)->s_journal && !handle) {
4859 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4860 " cancelled because transaction is not started",
4861 (unsigned long long)off, (unsigned long long)len);
4865 * Since we account only one data block in transaction credits,
4866 * then it is impossible to cross a block boundary.
4868 if (sb->s_blocksize - offset < len) {
4869 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4870 " cancelled because not block aligned",
4871 (unsigned long long)off, (unsigned long long)len);
4875 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4876 bh = ext4_bread(handle, inode, blk, 1, &err);
4879 err = ext4_journal_get_write_access(handle, bh);
4885 memcpy(bh->b_data+offset, data, len);
4886 flush_dcache_page(bh->b_page);
4888 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4892 mutex_unlock(&inode->i_mutex);
4895 if (inode->i_size < off + len) {
4896 i_size_write(inode, off + len);
4897 EXT4_I(inode)->i_disksize = inode->i_size;
4898 ext4_mark_inode_dirty(handle, inode);
4900 mutex_unlock(&inode->i_mutex);
4906 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4907 const char *dev_name, void *data)
4909 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4912 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4913 static inline void register_as_ext2(void)
4915 int err = register_filesystem(&ext2_fs_type);
4918 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4921 static inline void unregister_as_ext2(void)
4923 unregister_filesystem(&ext2_fs_type);
4926 static inline int ext2_feature_set_ok(struct super_block *sb)
4928 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4930 if (sb->s_flags & MS_RDONLY)
4932 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4936 MODULE_ALIAS("ext2");
4938 static inline void register_as_ext2(void) { }
4939 static inline void unregister_as_ext2(void) { }
4940 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4943 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4944 static inline void register_as_ext3(void)
4946 int err = register_filesystem(&ext3_fs_type);
4949 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4952 static inline void unregister_as_ext3(void)
4954 unregister_filesystem(&ext3_fs_type);
4957 static inline int ext3_feature_set_ok(struct super_block *sb)
4959 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4961 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4963 if (sb->s_flags & MS_RDONLY)
4965 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4969 MODULE_ALIAS("ext3");
4971 static inline void register_as_ext3(void) { }
4972 static inline void unregister_as_ext3(void) { }
4973 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4976 static struct file_system_type ext4_fs_type = {
4977 .owner = THIS_MODULE,
4979 .mount = ext4_mount,
4980 .kill_sb = kill_block_super,
4981 .fs_flags = FS_REQUIRES_DEV,
4984 static int __init ext4_init_feat_adverts(void)
4986 struct ext4_features *ef;
4989 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4993 ef->f_kobj.kset = ext4_kset;
4994 init_completion(&ef->f_kobj_unregister);
4995 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5008 static void ext4_exit_feat_adverts(void)
5010 kobject_put(&ext4_feat->f_kobj);
5011 wait_for_completion(&ext4_feat->f_kobj_unregister);
5015 /* Shared across all ext4 file systems */
5016 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5017 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5019 static int __init ext4_init_fs(void)
5023 ext4_check_flag_values();
5025 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5026 mutex_init(&ext4__aio_mutex[i]);
5027 init_waitqueue_head(&ext4__ioend_wq[i]);
5030 err = ext4_init_pageio();
5033 err = ext4_init_system_zone();
5036 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5039 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5041 err = ext4_init_feat_adverts();
5045 err = ext4_init_mballoc();
5049 err = ext4_init_xattr();
5052 err = init_inodecache();
5057 err = register_filesystem(&ext4_fs_type);
5061 ext4_li_info = NULL;
5062 mutex_init(&ext4_li_mtx);
5065 unregister_as_ext2();
5066 unregister_as_ext3();
5067 destroy_inodecache();
5071 ext4_exit_mballoc();
5073 ext4_exit_feat_adverts();
5076 remove_proc_entry("fs/ext4", NULL);
5077 kset_unregister(ext4_kset);
5079 ext4_exit_system_zone();
5085 static void __exit ext4_exit_fs(void)
5087 ext4_destroy_lazyinit_thread();
5088 unregister_as_ext2();
5089 unregister_as_ext3();
5090 unregister_filesystem(&ext4_fs_type);
5091 destroy_inodecache();
5093 ext4_exit_mballoc();
5094 ext4_exit_feat_adverts();
5095 remove_proc_entry("fs/ext4", NULL);
5096 kset_unregister(ext4_kset);
5097 ext4_exit_system_zone();
5101 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5102 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5103 MODULE_LICENSE("GPL");
5104 module_init(ext4_init_fs)
5105 module_exit(ext4_exit_fs)