2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 void *ext4_kvmalloc(size_t size, gfp_t flags)
118 ret = kmalloc(size, flags);
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
124 void *ext4_kvzalloc(size_t size, gfp_t flags)
128 ret = kzalloc(size, flags);
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
134 void ext4_kvfree(void *ptr)
136 if (is_vmalloc_addr(ptr))
143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
159 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
162 return le32_to_cpu(bg->bg_inode_table_lo) |
163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
167 __u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
175 __u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
183 __u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
191 __u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
199 void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
207 void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
215 void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
223 void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
231 void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
239 void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
247 void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
256 /* Just increment the non-pointer handle value */
257 static handle_t *ext4_get_nojournal(void)
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
265 handle = (handle_t *)ref_cnt;
267 current->journal_info = handle;
272 /* Decrement the non-pointer handle value */
273 static void ext4_put_nojournal(handle_t *handle)
275 unsigned long ref_cnt = (unsigned long)handle;
277 BUG_ON(ref_cnt == 0);
280 handle = (handle_t *)ref_cnt;
282 current->journal_info = handle;
286 * Wrappers for jbd2_journal_start/end.
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
306 journal = EXT4_SB(sb)->s_journal;
307 handle = ext4_journal_current_handle();
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
320 return ext4_get_nojournal();
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
330 return jbd2_journal_start(journal, nblocks);
334 * The only special thing we need to do here is to make sure that all
335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
336 * that sync() will call the filesystem's write_super callback if
339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
341 struct super_block *sb;
345 if (!ext4_handle_valid(handle)) {
346 ext4_put_nojournal(handle);
349 sb = handle->h_transaction->t_journal->j_private;
351 rc = jbd2_journal_stop(handle);
356 __ext4_std_error(sb, where, line, err);
360 void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
367 BUG_ON(!ext4_handle_valid(handle));
370 BUFFER_TRACE(bh, "abort");
375 if (is_handle_aborted(handle))
378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
381 jbd2_journal_abort_handle(handle);
384 static void __save_error_info(struct super_block *sb, const char *func,
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
403 * Start the daily error reporting function if it hasn't been
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
411 static void save_error_info(struct super_block *sb, const char *func,
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
426 static int block_device_ejected(struct super_block *sb)
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431 return bdi->dev == NULL;
435 /* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
438 * On ext2, we can store the error state of the filesystem in the
439 * superblock. That is not possible on ext4, because we may have other
440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
445 * We'll just use the jbd2_journal_abort() error code to record an error in
446 * the journal instead. On recovery, the journal will complain about
447 * that error until we've noted it down and cleared it.
450 static void ext4_handle_error(struct super_block *sb)
452 if (sb->s_flags & MS_RDONLY)
455 if (!test_opt(sb, ERRORS_CONT)) {
456 journal_t *journal = EXT4_SB(sb)->s_journal;
458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
460 jbd2_journal_abort(journal, -EIO);
462 if (test_opt(sb, ERRORS_RO)) {
463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
464 sb->s_flags |= MS_RDONLY;
466 if (test_opt(sb, ERRORS_PANIC)) {
467 if (EXT4_SB(sb)->s_journal &&
468 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
470 panic("EXT4-fs (device %s): panic forced after error\n",
475 void __ext4_error(struct super_block *sb, const char *function,
476 unsigned int line, const char *fmt, ...)
478 struct va_format vaf;
484 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
485 sb->s_id, function, line, current->comm, &vaf);
487 save_error_info(sb, function, line);
489 ext4_handle_error(sb);
492 void ext4_error_inode(struct inode *inode, const char *function,
493 unsigned int line, ext4_fsblk_t block,
494 const char *fmt, ...)
497 struct va_format vaf;
498 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
500 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
501 es->s_last_error_block = cpu_to_le64(block);
502 save_error_info(inode->i_sb, function, line);
506 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
507 inode->i_sb->s_id, function, line, inode->i_ino);
509 printk(KERN_CONT "block %llu: ", block);
510 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
513 ext4_handle_error(inode->i_sb);
516 void ext4_error_file(struct file *file, const char *function,
517 unsigned int line, ext4_fsblk_t block,
518 const char *fmt, ...)
521 struct va_format vaf;
522 struct ext4_super_block *es;
523 struct inode *inode = file->f_dentry->d_inode;
524 char pathname[80], *path;
526 es = EXT4_SB(inode->i_sb)->s_es;
527 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
528 save_error_info(inode->i_sb, function, line);
529 path = d_path(&(file->f_path), pathname, sizeof(pathname));
533 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
534 inode->i_sb->s_id, function, line, inode->i_ino);
536 printk(KERN_CONT "block %llu: ", block);
540 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
543 ext4_handle_error(inode->i_sb);
546 static const char *ext4_decode_error(struct super_block *sb, int errno,
553 errstr = "IO failure";
556 errstr = "Out of memory";
559 if (!sb || (EXT4_SB(sb)->s_journal &&
560 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
561 errstr = "Journal has aborted";
563 errstr = "Readonly filesystem";
566 /* If the caller passed in an extra buffer for unknown
567 * errors, textualise them now. Else we just return
570 /* Check for truncated error codes... */
571 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
580 /* __ext4_std_error decodes expected errors from journaling functions
581 * automatically and invokes the appropriate error response. */
583 void __ext4_std_error(struct super_block *sb, const char *function,
584 unsigned int line, int errno)
589 /* Special case: if the error is EROFS, and we're not already
590 * inside a transaction, then there's really no point in logging
592 if (errno == -EROFS && journal_current_handle() == NULL &&
593 (sb->s_flags & MS_RDONLY))
596 errstr = ext4_decode_error(sb, errno, nbuf);
597 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
598 sb->s_id, function, line, errstr);
599 save_error_info(sb, function, line);
601 ext4_handle_error(sb);
605 * ext4_abort is a much stronger failure handler than ext4_error. The
606 * abort function may be used to deal with unrecoverable failures such
607 * as journal IO errors or ENOMEM at a critical moment in log management.
609 * We unconditionally force the filesystem into an ABORT|READONLY state,
610 * unless the error response on the fs has been set to panic in which
611 * case we take the easy way out and panic immediately.
614 void __ext4_abort(struct super_block *sb, const char *function,
615 unsigned int line, const char *fmt, ...)
619 save_error_info(sb, function, line);
621 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
627 if ((sb->s_flags & MS_RDONLY) == 0) {
628 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
629 sb->s_flags |= MS_RDONLY;
630 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
631 if (EXT4_SB(sb)->s_journal)
632 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
633 save_error_info(sb, function, line);
635 if (test_opt(sb, ERRORS_PANIC)) {
636 if (EXT4_SB(sb)->s_journal &&
637 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
639 panic("EXT4-fs panic from previous error\n");
643 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
645 struct va_format vaf;
651 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
655 void __ext4_warning(struct super_block *sb, const char *function,
656 unsigned int line, const char *fmt, ...)
658 struct va_format vaf;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
669 void __ext4_grp_locked_error(const char *function, unsigned int line,
670 struct super_block *sb, ext4_group_t grp,
671 unsigned long ino, ext4_fsblk_t block,
672 const char *fmt, ...)
676 struct va_format vaf;
678 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
680 es->s_last_error_ino = cpu_to_le32(ino);
681 es->s_last_error_block = cpu_to_le64(block);
682 __save_error_info(sb, function, line);
688 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
689 sb->s_id, function, line, grp);
691 printk(KERN_CONT "inode %lu: ", ino);
693 printk(KERN_CONT "block %llu:", (unsigned long long) block);
694 printk(KERN_CONT "%pV\n", &vaf);
697 if (test_opt(sb, ERRORS_CONT)) {
698 ext4_commit_super(sb, 0);
702 ext4_unlock_group(sb, grp);
703 ext4_handle_error(sb);
705 * We only get here in the ERRORS_RO case; relocking the group
706 * may be dangerous, but nothing bad will happen since the
707 * filesystem will have already been marked read/only and the
708 * journal has been aborted. We return 1 as a hint to callers
709 * who might what to use the return value from
710 * ext4_grp_locked_error() to distinguish between the
711 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
712 * aggressively from the ext4 function in question, with a
713 * more appropriate error code.
715 ext4_lock_group(sb, grp);
719 void ext4_update_dynamic_rev(struct super_block *sb)
721 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
723 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
727 "updating to rev %d because of new feature flag, "
728 "running e2fsck is recommended",
731 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
732 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
733 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
734 /* leave es->s_feature_*compat flags alone */
735 /* es->s_uuid will be set by e2fsck if empty */
738 * The rest of the superblock fields should be zero, and if not it
739 * means they are likely already in use, so leave them alone. We
740 * can leave it up to e2fsck to clean up any inconsistencies there.
745 * Open the external journal device
747 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
749 struct block_device *bdev;
750 char b[BDEVNAME_SIZE];
752 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
758 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
759 __bdevname(dev, b), PTR_ERR(bdev));
764 * Release the journal device
766 static int ext4_blkdev_put(struct block_device *bdev)
768 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
771 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
773 struct block_device *bdev;
776 bdev = sbi->journal_bdev;
778 ret = ext4_blkdev_put(bdev);
779 sbi->journal_bdev = NULL;
784 static inline struct inode *orphan_list_entry(struct list_head *l)
786 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
789 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
793 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
794 le32_to_cpu(sbi->s_es->s_last_orphan));
796 printk(KERN_ERR "sb_info orphan list:\n");
797 list_for_each(l, &sbi->s_orphan) {
798 struct inode *inode = orphan_list_entry(l);
800 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
801 inode->i_sb->s_id, inode->i_ino, inode,
802 inode->i_mode, inode->i_nlink,
807 static void ext4_put_super(struct super_block *sb)
809 struct ext4_sb_info *sbi = EXT4_SB(sb);
810 struct ext4_super_block *es = sbi->s_es;
813 ext4_unregister_li_request(sb);
814 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
816 flush_workqueue(sbi->dio_unwritten_wq);
817 destroy_workqueue(sbi->dio_unwritten_wq);
821 ext4_commit_super(sb, 1);
823 if (sbi->s_journal) {
824 err = jbd2_journal_destroy(sbi->s_journal);
825 sbi->s_journal = NULL;
827 ext4_abort(sb, "Couldn't clean up the journal");
830 del_timer_sync(&sbi->s_err_report);
831 ext4_release_system_zone(sb);
833 ext4_ext_release(sb);
834 ext4_xattr_put_super(sb);
836 if (!(sb->s_flags & MS_RDONLY)) {
837 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
838 es->s_state = cpu_to_le16(sbi->s_mount_state);
839 ext4_commit_super(sb, 1);
842 remove_proc_entry(sb->s_id, ext4_proc_root);
844 kobject_del(&sbi->s_kobj);
846 for (i = 0; i < sbi->s_gdb_count; i++)
847 brelse(sbi->s_group_desc[i]);
848 ext4_kvfree(sbi->s_group_desc);
849 ext4_kvfree(sbi->s_flex_groups);
850 percpu_counter_destroy(&sbi->s_freeclusters_counter);
851 percpu_counter_destroy(&sbi->s_freeinodes_counter);
852 percpu_counter_destroy(&sbi->s_dirs_counter);
853 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
856 for (i = 0; i < MAXQUOTAS; i++)
857 kfree(sbi->s_qf_names[i]);
860 /* Debugging code just in case the in-memory inode orphan list
861 * isn't empty. The on-disk one can be non-empty if we've
862 * detected an error and taken the fs readonly, but the
863 * in-memory list had better be clean by this point. */
864 if (!list_empty(&sbi->s_orphan))
865 dump_orphan_list(sb, sbi);
866 J_ASSERT(list_empty(&sbi->s_orphan));
868 sync_blockdev(sb->s_bdev);
869 invalidate_bdev(sb->s_bdev);
870 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
872 * Invalidate the journal device's buffers. We don't want them
873 * floating about in memory - the physical journal device may
874 * hotswapped, and it breaks the `ro-after' testing code.
876 sync_blockdev(sbi->journal_bdev);
877 invalidate_bdev(sbi->journal_bdev);
878 ext4_blkdev_remove(sbi);
881 kthread_stop(sbi->s_mmp_tsk);
882 sb->s_fs_info = NULL;
884 * Now that we are completely done shutting down the
885 * superblock, we need to actually destroy the kobject.
888 kobject_put(&sbi->s_kobj);
889 wait_for_completion(&sbi->s_kobj_unregister);
890 kfree(sbi->s_blockgroup_lock);
894 static struct kmem_cache *ext4_inode_cachep;
897 * Called inside transaction, so use GFP_NOFS
899 static struct inode *ext4_alloc_inode(struct super_block *sb)
901 struct ext4_inode_info *ei;
903 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
907 ei->vfs_inode.i_version = 1;
908 ei->vfs_inode.i_data.writeback_index = 0;
909 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
910 INIT_LIST_HEAD(&ei->i_prealloc_list);
911 spin_lock_init(&ei->i_prealloc_lock);
912 ei->i_reserved_data_blocks = 0;
913 ei->i_reserved_meta_blocks = 0;
914 ei->i_allocated_meta_blocks = 0;
915 ei->i_da_metadata_calc_len = 0;
916 ei->i_da_metadata_calc_last_lblock = 0;
917 spin_lock_init(&(ei->i_block_reservation_lock));
919 ei->i_reserved_quota = 0;
922 INIT_LIST_HEAD(&ei->i_completed_io_list);
923 spin_lock_init(&ei->i_completed_io_lock);
924 ei->cur_aio_dio = NULL;
926 ei->i_datasync_tid = 0;
927 atomic_set(&ei->i_ioend_count, 0);
928 atomic_set(&ei->i_aiodio_unwritten, 0);
930 return &ei->vfs_inode;
933 static int ext4_drop_inode(struct inode *inode)
935 int drop = generic_drop_inode(inode);
937 trace_ext4_drop_inode(inode, drop);
941 static void ext4_i_callback(struct rcu_head *head)
943 struct inode *inode = container_of(head, struct inode, i_rcu);
944 INIT_LIST_HEAD(&inode->i_dentry);
945 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
948 static void ext4_destroy_inode(struct inode *inode)
950 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
951 ext4_msg(inode->i_sb, KERN_ERR,
952 "Inode %lu (%p): orphan list check failed!",
953 inode->i_ino, EXT4_I(inode));
954 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
955 EXT4_I(inode), sizeof(struct ext4_inode_info),
959 call_rcu(&inode->i_rcu, ext4_i_callback);
962 static void init_once(void *foo)
964 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
966 INIT_LIST_HEAD(&ei->i_orphan);
967 #ifdef CONFIG_EXT4_FS_XATTR
968 init_rwsem(&ei->xattr_sem);
970 init_rwsem(&ei->i_data_sem);
971 inode_init_once(&ei->vfs_inode);
974 static int init_inodecache(void)
976 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
977 sizeof(struct ext4_inode_info),
978 0, (SLAB_RECLAIM_ACCOUNT|
981 if (ext4_inode_cachep == NULL)
986 static void destroy_inodecache(void)
988 kmem_cache_destroy(ext4_inode_cachep);
991 void ext4_clear_inode(struct inode *inode)
993 invalidate_inode_buffers(inode);
994 end_writeback(inode);
996 ext4_discard_preallocations(inode);
997 if (EXT4_I(inode)->jinode) {
998 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
999 EXT4_I(inode)->jinode);
1000 jbd2_free_inode(EXT4_I(inode)->jinode);
1001 EXT4_I(inode)->jinode = NULL;
1005 static inline void ext4_show_quota_options(struct seq_file *seq,
1006 struct super_block *sb)
1008 #if defined(CONFIG_QUOTA)
1009 struct ext4_sb_info *sbi = EXT4_SB(sb);
1011 if (sbi->s_jquota_fmt) {
1014 switch (sbi->s_jquota_fmt) {
1025 seq_printf(seq, ",jqfmt=%s", fmtname);
1028 if (sbi->s_qf_names[USRQUOTA])
1029 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1031 if (sbi->s_qf_names[GRPQUOTA])
1032 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1034 if (test_opt(sb, USRQUOTA))
1035 seq_puts(seq, ",usrquota");
1037 if (test_opt(sb, GRPQUOTA))
1038 seq_puts(seq, ",grpquota");
1044 * - it's set to a non-default value OR
1045 * - if the per-sb default is different from the global default
1047 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1050 unsigned long def_mount_opts;
1051 struct super_block *sb = vfs->mnt_sb;
1052 struct ext4_sb_info *sbi = EXT4_SB(sb);
1053 struct ext4_super_block *es = sbi->s_es;
1055 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1056 def_errors = le16_to_cpu(es->s_errors);
1058 if (sbi->s_sb_block != 1)
1059 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1060 if (test_opt(sb, MINIX_DF))
1061 seq_puts(seq, ",minixdf");
1062 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1063 seq_puts(seq, ",grpid");
1064 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1065 seq_puts(seq, ",nogrpid");
1066 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1067 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1068 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1070 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1071 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1072 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1074 if (test_opt(sb, ERRORS_RO)) {
1075 if (def_errors == EXT4_ERRORS_PANIC ||
1076 def_errors == EXT4_ERRORS_CONTINUE) {
1077 seq_puts(seq, ",errors=remount-ro");
1080 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1081 seq_puts(seq, ",errors=continue");
1082 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1083 seq_puts(seq, ",errors=panic");
1084 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1085 seq_puts(seq, ",nouid32");
1086 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1087 seq_puts(seq, ",debug");
1088 #ifdef CONFIG_EXT4_FS_XATTR
1089 if (test_opt(sb, XATTR_USER))
1090 seq_puts(seq, ",user_xattr");
1091 if (!test_opt(sb, XATTR_USER))
1092 seq_puts(seq, ",nouser_xattr");
1094 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1095 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1096 seq_puts(seq, ",acl");
1097 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1098 seq_puts(seq, ",noacl");
1100 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1101 seq_printf(seq, ",commit=%u",
1102 (unsigned) (sbi->s_commit_interval / HZ));
1104 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1105 seq_printf(seq, ",min_batch_time=%u",
1106 (unsigned) sbi->s_min_batch_time);
1108 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1109 seq_printf(seq, ",max_batch_time=%u",
1110 (unsigned) sbi->s_max_batch_time);
1114 * We're changing the default of barrier mount option, so
1115 * let's always display its mount state so it's clear what its
1118 seq_puts(seq, ",barrier=");
1119 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1120 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1121 seq_puts(seq, ",journal_async_commit");
1122 else if (test_opt(sb, JOURNAL_CHECKSUM))
1123 seq_puts(seq, ",journal_checksum");
1124 if (test_opt(sb, I_VERSION))
1125 seq_puts(seq, ",i_version");
1126 if (!test_opt(sb, DELALLOC) &&
1127 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1128 seq_puts(seq, ",nodelalloc");
1130 if (!test_opt(sb, MBLK_IO_SUBMIT))
1131 seq_puts(seq, ",nomblk_io_submit");
1133 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1135 * journal mode get enabled in different ways
1136 * So just print the value even if we didn't specify it
1138 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1139 seq_puts(seq, ",data=journal");
1140 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1141 seq_puts(seq, ",data=ordered");
1142 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1143 seq_puts(seq, ",data=writeback");
1145 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1146 seq_printf(seq, ",inode_readahead_blks=%u",
1147 sbi->s_inode_readahead_blks);
1149 if (test_opt(sb, DATA_ERR_ABORT))
1150 seq_puts(seq, ",data_err=abort");
1152 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1153 seq_puts(seq, ",noauto_da_alloc");
1155 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1156 seq_puts(seq, ",discard");
1158 if (test_opt(sb, NOLOAD))
1159 seq_puts(seq, ",norecovery");
1161 if (test_opt(sb, DIOREAD_NOLOCK))
1162 seq_puts(seq, ",dioread_nolock");
1164 if (test_opt(sb, BLOCK_VALIDITY) &&
1165 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1166 seq_puts(seq, ",block_validity");
1168 if (!test_opt(sb, INIT_INODE_TABLE))
1169 seq_puts(seq, ",noinit_itable");
1170 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1171 seq_printf(seq, ",init_itable=%u",
1172 (unsigned) sbi->s_li_wait_mult);
1174 ext4_show_quota_options(seq, sb);
1179 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1180 u64 ino, u32 generation)
1182 struct inode *inode;
1184 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1185 return ERR_PTR(-ESTALE);
1186 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1187 return ERR_PTR(-ESTALE);
1189 /* iget isn't really right if the inode is currently unallocated!!
1191 * ext4_read_inode will return a bad_inode if the inode had been
1192 * deleted, so we should be safe.
1194 * Currently we don't know the generation for parent directory, so
1195 * a generation of 0 means "accept any"
1197 inode = ext4_iget_normal(sb, ino);
1199 return ERR_CAST(inode);
1200 if (generation && inode->i_generation != generation) {
1202 return ERR_PTR(-ESTALE);
1208 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1209 int fh_len, int fh_type)
1211 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1212 ext4_nfs_get_inode);
1215 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1216 int fh_len, int fh_type)
1218 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1219 ext4_nfs_get_inode);
1223 * Try to release metadata pages (indirect blocks, directories) which are
1224 * mapped via the block device. Since these pages could have journal heads
1225 * which would prevent try_to_free_buffers() from freeing them, we must use
1226 * jbd2 layer's try_to_free_buffers() function to release them.
1228 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1231 journal_t *journal = EXT4_SB(sb)->s_journal;
1233 WARN_ON(PageChecked(page));
1234 if (!page_has_buffers(page))
1237 return jbd2_journal_try_to_free_buffers(journal, page,
1238 wait & ~__GFP_WAIT);
1239 return try_to_free_buffers(page);
1243 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1244 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1246 static int ext4_write_dquot(struct dquot *dquot);
1247 static int ext4_acquire_dquot(struct dquot *dquot);
1248 static int ext4_release_dquot(struct dquot *dquot);
1249 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1250 static int ext4_write_info(struct super_block *sb, int type);
1251 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1253 static int ext4_quota_off(struct super_block *sb, int type);
1254 static int ext4_quota_on_mount(struct super_block *sb, int type);
1255 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1256 size_t len, loff_t off);
1257 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1258 const char *data, size_t len, loff_t off);
1260 static const struct dquot_operations ext4_quota_operations = {
1261 .get_reserved_space = ext4_get_reserved_space,
1262 .write_dquot = ext4_write_dquot,
1263 .acquire_dquot = ext4_acquire_dquot,
1264 .release_dquot = ext4_release_dquot,
1265 .mark_dirty = ext4_mark_dquot_dirty,
1266 .write_info = ext4_write_info,
1267 .alloc_dquot = dquot_alloc,
1268 .destroy_dquot = dquot_destroy,
1271 static const struct quotactl_ops ext4_qctl_operations = {
1272 .quota_on = ext4_quota_on,
1273 .quota_off = ext4_quota_off,
1274 .quota_sync = dquot_quota_sync,
1275 .get_info = dquot_get_dqinfo,
1276 .set_info = dquot_set_dqinfo,
1277 .get_dqblk = dquot_get_dqblk,
1278 .set_dqblk = dquot_set_dqblk
1282 static const struct super_operations ext4_sops = {
1283 .alloc_inode = ext4_alloc_inode,
1284 .destroy_inode = ext4_destroy_inode,
1285 .write_inode = ext4_write_inode,
1286 .dirty_inode = ext4_dirty_inode,
1287 .drop_inode = ext4_drop_inode,
1288 .evict_inode = ext4_evict_inode,
1289 .put_super = ext4_put_super,
1290 .sync_fs = ext4_sync_fs,
1291 .freeze_fs = ext4_freeze,
1292 .unfreeze_fs = ext4_unfreeze,
1293 .statfs = ext4_statfs,
1294 .remount_fs = ext4_remount,
1295 .show_options = ext4_show_options,
1297 .quota_read = ext4_quota_read,
1298 .quota_write = ext4_quota_write,
1300 .bdev_try_to_free_page = bdev_try_to_free_page,
1303 static const struct super_operations ext4_nojournal_sops = {
1304 .alloc_inode = ext4_alloc_inode,
1305 .destroy_inode = ext4_destroy_inode,
1306 .write_inode = ext4_write_inode,
1307 .dirty_inode = ext4_dirty_inode,
1308 .drop_inode = ext4_drop_inode,
1309 .evict_inode = ext4_evict_inode,
1310 .write_super = ext4_write_super,
1311 .put_super = ext4_put_super,
1312 .statfs = ext4_statfs,
1313 .remount_fs = ext4_remount,
1314 .show_options = ext4_show_options,
1316 .quota_read = ext4_quota_read,
1317 .quota_write = ext4_quota_write,
1319 .bdev_try_to_free_page = bdev_try_to_free_page,
1322 static const struct export_operations ext4_export_ops = {
1323 .fh_to_dentry = ext4_fh_to_dentry,
1324 .fh_to_parent = ext4_fh_to_parent,
1325 .get_parent = ext4_get_parent,
1329 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1330 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1331 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1332 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1333 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1334 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1335 Opt_journal_update, Opt_journal_dev,
1336 Opt_journal_checksum, Opt_journal_async_commit,
1337 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1338 Opt_data_err_abort, Opt_data_err_ignore,
1339 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1340 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1341 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1342 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1343 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1344 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1345 Opt_inode_readahead_blks, Opt_journal_ioprio,
1346 Opt_dioread_nolock, Opt_dioread_lock,
1347 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1350 static const match_table_t tokens = {
1351 {Opt_bsd_df, "bsddf"},
1352 {Opt_minix_df, "minixdf"},
1353 {Opt_grpid, "grpid"},
1354 {Opt_grpid, "bsdgroups"},
1355 {Opt_nogrpid, "nogrpid"},
1356 {Opt_nogrpid, "sysvgroups"},
1357 {Opt_resgid, "resgid=%u"},
1358 {Opt_resuid, "resuid=%u"},
1360 {Opt_err_cont, "errors=continue"},
1361 {Opt_err_panic, "errors=panic"},
1362 {Opt_err_ro, "errors=remount-ro"},
1363 {Opt_nouid32, "nouid32"},
1364 {Opt_debug, "debug"},
1365 {Opt_oldalloc, "oldalloc"},
1366 {Opt_orlov, "orlov"},
1367 {Opt_user_xattr, "user_xattr"},
1368 {Opt_nouser_xattr, "nouser_xattr"},
1370 {Opt_noacl, "noacl"},
1371 {Opt_noload, "noload"},
1372 {Opt_noload, "norecovery"},
1375 {Opt_commit, "commit=%u"},
1376 {Opt_min_batch_time, "min_batch_time=%u"},
1377 {Opt_max_batch_time, "max_batch_time=%u"},
1378 {Opt_journal_update, "journal=update"},
1379 {Opt_journal_dev, "journal_dev=%u"},
1380 {Opt_journal_checksum, "journal_checksum"},
1381 {Opt_journal_async_commit, "journal_async_commit"},
1382 {Opt_abort, "abort"},
1383 {Opt_data_journal, "data=journal"},
1384 {Opt_data_ordered, "data=ordered"},
1385 {Opt_data_writeback, "data=writeback"},
1386 {Opt_data_err_abort, "data_err=abort"},
1387 {Opt_data_err_ignore, "data_err=ignore"},
1388 {Opt_offusrjquota, "usrjquota="},
1389 {Opt_usrjquota, "usrjquota=%s"},
1390 {Opt_offgrpjquota, "grpjquota="},
1391 {Opt_grpjquota, "grpjquota=%s"},
1392 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1393 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1394 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1395 {Opt_grpquota, "grpquota"},
1396 {Opt_noquota, "noquota"},
1397 {Opt_quota, "quota"},
1398 {Opt_usrquota, "usrquota"},
1399 {Opt_barrier, "barrier=%u"},
1400 {Opt_barrier, "barrier"},
1401 {Opt_nobarrier, "nobarrier"},
1402 {Opt_i_version, "i_version"},
1403 {Opt_stripe, "stripe=%u"},
1404 {Opt_resize, "resize"},
1405 {Opt_delalloc, "delalloc"},
1406 {Opt_nodelalloc, "nodelalloc"},
1407 {Opt_mblk_io_submit, "mblk_io_submit"},
1408 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1409 {Opt_block_validity, "block_validity"},
1410 {Opt_noblock_validity, "noblock_validity"},
1411 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1412 {Opt_journal_ioprio, "journal_ioprio=%u"},
1413 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1414 {Opt_auto_da_alloc, "auto_da_alloc"},
1415 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1416 {Opt_dioread_nolock, "dioread_nolock"},
1417 {Opt_dioread_lock, "dioread_lock"},
1418 {Opt_discard, "discard"},
1419 {Opt_nodiscard, "nodiscard"},
1420 {Opt_init_itable, "init_itable=%u"},
1421 {Opt_init_itable, "init_itable"},
1422 {Opt_noinit_itable, "noinit_itable"},
1426 static ext4_fsblk_t get_sb_block(void **data)
1428 ext4_fsblk_t sb_block;
1429 char *options = (char *) *data;
1431 if (!options || strncmp(options, "sb=", 3) != 0)
1432 return 1; /* Default location */
1435 /* TODO: use simple_strtoll with >32bit ext4 */
1436 sb_block = simple_strtoul(options, &options, 0);
1437 if (*options && *options != ',') {
1438 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1442 if (*options == ',')
1444 *data = (void *) options;
1449 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1450 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1451 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1454 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1456 struct ext4_sb_info *sbi = EXT4_SB(sb);
1459 if (sb_any_quota_loaded(sb) &&
1460 !sbi->s_qf_names[qtype]) {
1461 ext4_msg(sb, KERN_ERR,
1462 "Cannot change journaled "
1463 "quota options when quota turned on");
1466 qname = match_strdup(args);
1468 ext4_msg(sb, KERN_ERR,
1469 "Not enough memory for storing quotafile name");
1472 if (sbi->s_qf_names[qtype] &&
1473 strcmp(sbi->s_qf_names[qtype], qname)) {
1474 ext4_msg(sb, KERN_ERR,
1475 "%s quota file already specified", QTYPE2NAME(qtype));
1479 sbi->s_qf_names[qtype] = qname;
1480 if (strchr(sbi->s_qf_names[qtype], '/')) {
1481 ext4_msg(sb, KERN_ERR,
1482 "quotafile must be on filesystem root");
1483 kfree(sbi->s_qf_names[qtype]);
1484 sbi->s_qf_names[qtype] = NULL;
1491 static int clear_qf_name(struct super_block *sb, int qtype)
1494 struct ext4_sb_info *sbi = EXT4_SB(sb);
1496 if (sb_any_quota_loaded(sb) &&
1497 sbi->s_qf_names[qtype]) {
1498 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1499 " when quota turned on");
1503 * The space will be released later when all options are confirmed
1506 sbi->s_qf_names[qtype] = NULL;
1511 static int parse_options(char *options, struct super_block *sb,
1512 unsigned long *journal_devnum,
1513 unsigned int *journal_ioprio,
1514 ext4_fsblk_t *n_blocks_count, int is_remount)
1516 struct ext4_sb_info *sbi = EXT4_SB(sb);
1518 substring_t args[MAX_OPT_ARGS];
1528 while ((p = strsep(&options, ",")) != NULL) {
1534 * Initialize args struct so we know whether arg was
1535 * found; some options take optional arguments.
1537 args[0].to = args[0].from = NULL;
1538 token = match_token(p, tokens, args);
1541 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1542 clear_opt(sb, MINIX_DF);
1545 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1546 set_opt(sb, MINIX_DF);
1550 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1555 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1556 clear_opt(sb, GRPID);
1560 if (match_int(&args[0], &option))
1562 sbi->s_resuid = option;
1565 if (match_int(&args[0], &option))
1567 sbi->s_resgid = option;
1570 /* handled by get_sb_block() instead of here */
1571 /* *sb_block = match_int(&args[0]); */
1574 clear_opt(sb, ERRORS_CONT);
1575 clear_opt(sb, ERRORS_RO);
1576 set_opt(sb, ERRORS_PANIC);
1579 clear_opt(sb, ERRORS_CONT);
1580 clear_opt(sb, ERRORS_PANIC);
1581 set_opt(sb, ERRORS_RO);
1584 clear_opt(sb, ERRORS_RO);
1585 clear_opt(sb, ERRORS_PANIC);
1586 set_opt(sb, ERRORS_CONT);
1589 set_opt(sb, NO_UID32);
1595 ext4_msg(sb, KERN_WARNING,
1596 "Ignoring deprecated oldalloc option");
1599 ext4_msg(sb, KERN_WARNING,
1600 "Ignoring deprecated orlov option");
1602 #ifdef CONFIG_EXT4_FS_XATTR
1603 case Opt_user_xattr:
1604 set_opt(sb, XATTR_USER);
1606 case Opt_nouser_xattr:
1607 clear_opt(sb, XATTR_USER);
1610 case Opt_user_xattr:
1611 case Opt_nouser_xattr:
1612 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1615 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1617 set_opt(sb, POSIX_ACL);
1620 clear_opt(sb, POSIX_ACL);
1625 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1628 case Opt_journal_update:
1630 /* Eventually we will want to be able to create
1631 a journal file here. For now, only allow the
1632 user to specify an existing inode to be the
1635 ext4_msg(sb, KERN_ERR,
1636 "Cannot specify journal on remount");
1639 set_opt(sb, UPDATE_JOURNAL);
1641 case Opt_journal_dev:
1643 ext4_msg(sb, KERN_ERR,
1644 "Cannot specify journal on remount");
1647 if (match_int(&args[0], &option))
1649 *journal_devnum = option;
1651 case Opt_journal_checksum:
1652 set_opt(sb, JOURNAL_CHECKSUM);
1654 case Opt_journal_async_commit:
1655 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1656 set_opt(sb, JOURNAL_CHECKSUM);
1659 set_opt(sb, NOLOAD);
1662 if (match_int(&args[0], &option))
1667 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1668 sbi->s_commit_interval = HZ * option;
1670 case Opt_max_batch_time:
1671 if (match_int(&args[0], &option))
1675 sbi->s_max_batch_time = option;
1677 case Opt_min_batch_time:
1678 if (match_int(&args[0], &option))
1682 sbi->s_min_batch_time = option;
1684 case Opt_data_journal:
1685 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1687 case Opt_data_ordered:
1688 data_opt = EXT4_MOUNT_ORDERED_DATA;
1690 case Opt_data_writeback:
1691 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1694 if (!sbi->s_journal)
1695 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1696 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1697 ext4_msg(sb, KERN_ERR,
1698 "Cannot change data mode on remount");
1702 clear_opt(sb, DATA_FLAGS);
1703 sbi->s_mount_opt |= data_opt;
1706 case Opt_data_err_abort:
1707 set_opt(sb, DATA_ERR_ABORT);
1709 case Opt_data_err_ignore:
1710 clear_opt(sb, DATA_ERR_ABORT);
1714 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1718 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1721 case Opt_offusrjquota:
1722 if (!clear_qf_name(sb, USRQUOTA))
1725 case Opt_offgrpjquota:
1726 if (!clear_qf_name(sb, GRPQUOTA))
1730 case Opt_jqfmt_vfsold:
1731 qfmt = QFMT_VFS_OLD;
1733 case Opt_jqfmt_vfsv0:
1736 case Opt_jqfmt_vfsv1:
1739 if (sb_any_quota_loaded(sb) &&
1740 sbi->s_jquota_fmt != qfmt) {
1741 ext4_msg(sb, KERN_ERR, "Cannot change "
1742 "journaled quota options when "
1746 sbi->s_jquota_fmt = qfmt;
1751 set_opt(sb, USRQUOTA);
1755 set_opt(sb, GRPQUOTA);
1758 if (sb_any_quota_loaded(sb)) {
1759 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1760 "options when quota turned on");
1763 clear_opt(sb, QUOTA);
1764 clear_opt(sb, USRQUOTA);
1765 clear_opt(sb, GRPQUOTA);
1771 ext4_msg(sb, KERN_ERR,
1772 "quota options not supported");
1776 case Opt_offusrjquota:
1777 case Opt_offgrpjquota:
1778 case Opt_jqfmt_vfsold:
1779 case Opt_jqfmt_vfsv0:
1780 case Opt_jqfmt_vfsv1:
1781 ext4_msg(sb, KERN_ERR,
1782 "journaled quota options not supported");
1788 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1791 clear_opt(sb, BARRIER);
1795 if (match_int(&args[0], &option))
1798 option = 1; /* No argument, default to 1 */
1800 set_opt(sb, BARRIER);
1802 clear_opt(sb, BARRIER);
1808 ext4_msg(sb, KERN_ERR,
1809 "resize option only available "
1813 if (match_int(&args[0], &option) != 0)
1815 *n_blocks_count = option;
1818 ext4_msg(sb, KERN_WARNING,
1819 "Ignoring deprecated nobh option");
1822 ext4_msg(sb, KERN_WARNING,
1823 "Ignoring deprecated bh option");
1826 set_opt(sb, I_VERSION);
1827 sb->s_flags |= MS_I_VERSION;
1829 case Opt_nodelalloc:
1830 clear_opt(sb, DELALLOC);
1831 clear_opt2(sb, EXPLICIT_DELALLOC);
1833 case Opt_mblk_io_submit:
1834 set_opt(sb, MBLK_IO_SUBMIT);
1836 case Opt_nomblk_io_submit:
1837 clear_opt(sb, MBLK_IO_SUBMIT);
1840 if (match_int(&args[0], &option))
1844 sbi->s_stripe = option;
1847 set_opt(sb, DELALLOC);
1848 set_opt2(sb, EXPLICIT_DELALLOC);
1850 case Opt_block_validity:
1851 set_opt(sb, BLOCK_VALIDITY);
1853 case Opt_noblock_validity:
1854 clear_opt(sb, BLOCK_VALIDITY);
1856 case Opt_inode_readahead_blks:
1857 if (match_int(&args[0], &option))
1859 if (option < 0 || option > (1 << 30))
1861 if (option && !is_power_of_2(option)) {
1862 ext4_msg(sb, KERN_ERR,
1863 "EXT4-fs: inode_readahead_blks"
1864 " must be a power of 2");
1867 sbi->s_inode_readahead_blks = option;
1869 case Opt_journal_ioprio:
1870 if (match_int(&args[0], &option))
1872 if (option < 0 || option > 7)
1874 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1877 case Opt_noauto_da_alloc:
1878 set_opt(sb, NO_AUTO_DA_ALLOC);
1880 case Opt_auto_da_alloc:
1882 if (match_int(&args[0], &option))
1885 option = 1; /* No argument, default to 1 */
1887 clear_opt(sb, NO_AUTO_DA_ALLOC);
1889 set_opt(sb,NO_AUTO_DA_ALLOC);
1892 set_opt(sb, DISCARD);
1895 clear_opt(sb, DISCARD);
1897 case Opt_dioread_nolock:
1898 set_opt(sb, DIOREAD_NOLOCK);
1900 case Opt_dioread_lock:
1901 clear_opt(sb, DIOREAD_NOLOCK);
1903 case Opt_init_itable:
1904 set_opt(sb, INIT_INODE_TABLE);
1906 if (match_int(&args[0], &option))
1909 option = EXT4_DEF_LI_WAIT_MULT;
1912 sbi->s_li_wait_mult = option;
1914 case Opt_noinit_itable:
1915 clear_opt(sb, INIT_INODE_TABLE);
1918 ext4_msg(sb, KERN_ERR,
1919 "Unrecognized mount option \"%s\" "
1920 "or missing value", p);
1925 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1926 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1927 clear_opt(sb, USRQUOTA);
1929 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1930 clear_opt(sb, GRPQUOTA);
1932 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1933 ext4_msg(sb, KERN_ERR, "old and new quota "
1938 if (!sbi->s_jquota_fmt) {
1939 ext4_msg(sb, KERN_ERR, "journaled quota format "
1945 if (test_opt(sb, DIOREAD_NOLOCK)) {
1947 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1949 if (blocksize < PAGE_CACHE_SIZE) {
1950 ext4_msg(sb, KERN_ERR, "can't mount with "
1951 "dioread_nolock if block size != PAGE_SIZE");
1958 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1961 struct ext4_sb_info *sbi = EXT4_SB(sb);
1964 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1965 ext4_msg(sb, KERN_ERR, "revision level too high, "
1966 "forcing read-only mode");
1971 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1972 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1973 "running e2fsck is recommended");
1974 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1975 ext4_msg(sb, KERN_WARNING,
1976 "warning: mounting fs with errors, "
1977 "running e2fsck is recommended");
1978 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1979 le16_to_cpu(es->s_mnt_count) >=
1980 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1981 ext4_msg(sb, KERN_WARNING,
1982 "warning: maximal mount count reached, "
1983 "running e2fsck is recommended");
1984 else if (le32_to_cpu(es->s_checkinterval) &&
1985 (le32_to_cpu(es->s_lastcheck) +
1986 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1987 ext4_msg(sb, KERN_WARNING,
1988 "warning: checktime reached, "
1989 "running e2fsck is recommended");
1990 if (!sbi->s_journal)
1991 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1992 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1993 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1994 le16_add_cpu(&es->s_mnt_count, 1);
1995 es->s_mtime = cpu_to_le32(get_seconds());
1996 ext4_update_dynamic_rev(sb);
1998 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2000 ext4_commit_super(sb, 1);
2002 if (test_opt(sb, DEBUG))
2003 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2004 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2006 sbi->s_groups_count,
2007 EXT4_BLOCKS_PER_GROUP(sb),
2008 EXT4_INODES_PER_GROUP(sb),
2009 sbi->s_mount_opt, sbi->s_mount_opt2);
2011 cleancache_init_fs(sb);
2015 static int ext4_fill_flex_info(struct super_block *sb)
2017 struct ext4_sb_info *sbi = EXT4_SB(sb);
2018 struct ext4_group_desc *gdp = NULL;
2019 ext4_group_t flex_group_count;
2020 ext4_group_t flex_group;
2021 unsigned int groups_per_flex = 0;
2025 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2026 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2027 sbi->s_log_groups_per_flex = 0;
2030 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2032 /* We allocate both existing and potentially added groups */
2033 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2034 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2035 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2036 size = flex_group_count * sizeof(struct flex_groups);
2037 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2038 if (sbi->s_flex_groups == NULL) {
2039 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2044 for (i = 0; i < sbi->s_groups_count; i++) {
2045 gdp = ext4_get_group_desc(sb, i, NULL);
2047 flex_group = ext4_flex_group(sbi, i);
2048 atomic_add(ext4_free_inodes_count(sb, gdp),
2049 &sbi->s_flex_groups[flex_group].free_inodes);
2050 atomic64_add(ext4_free_group_clusters(sb, gdp),
2051 &sbi->s_flex_groups[flex_group].free_clusters);
2052 atomic_add(ext4_used_dirs_count(sb, gdp),
2053 &sbi->s_flex_groups[flex_group].used_dirs);
2061 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2062 struct ext4_group_desc *gdp)
2066 if (sbi->s_es->s_feature_ro_compat &
2067 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2068 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2069 __le32 le_group = cpu_to_le32(block_group);
2071 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2072 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2073 crc = crc16(crc, (__u8 *)gdp, offset);
2074 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2075 /* for checksum of struct ext4_group_desc do the rest...*/
2076 if ((sbi->s_es->s_feature_incompat &
2077 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2078 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2079 crc = crc16(crc, (__u8 *)gdp + offset,
2080 le16_to_cpu(sbi->s_es->s_desc_size) -
2084 return cpu_to_le16(crc);
2087 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2088 struct ext4_group_desc *gdp)
2090 if ((sbi->s_es->s_feature_ro_compat &
2091 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2092 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2098 /* Called at mount-time, super-block is locked */
2099 static int ext4_check_descriptors(struct super_block *sb,
2100 ext4_fsblk_t sb_block,
2101 ext4_group_t *first_not_zeroed)
2103 struct ext4_sb_info *sbi = EXT4_SB(sb);
2104 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2105 ext4_fsblk_t last_block;
2106 ext4_fsblk_t block_bitmap;
2107 ext4_fsblk_t inode_bitmap;
2108 ext4_fsblk_t inode_table;
2109 int flexbg_flag = 0;
2110 ext4_group_t i, grp = sbi->s_groups_count;
2112 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2115 ext4_debug("Checking group descriptors");
2117 for (i = 0; i < sbi->s_groups_count; i++) {
2118 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2120 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2121 last_block = ext4_blocks_count(sbi->s_es) - 1;
2123 last_block = first_block +
2124 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2126 if ((grp == sbi->s_groups_count) &&
2127 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2130 block_bitmap = ext4_block_bitmap(sb, gdp);
2131 if (block_bitmap == sb_block) {
2132 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2133 "Block bitmap for group %u overlaps "
2136 if (block_bitmap < first_block || block_bitmap > last_block) {
2137 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2138 "Block bitmap for group %u not in group "
2139 "(block %llu)!", i, block_bitmap);
2142 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2143 if (inode_bitmap == sb_block) {
2144 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2145 "Inode bitmap for group %u overlaps "
2148 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2149 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2150 "Inode bitmap for group %u not in group "
2151 "(block %llu)!", i, inode_bitmap);
2154 inode_table = ext4_inode_table(sb, gdp);
2155 if (inode_table == sb_block) {
2156 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2157 "Inode table for group %u overlaps "
2160 if (inode_table < first_block ||
2161 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2162 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2163 "Inode table for group %u not in group "
2164 "(block %llu)!", i, inode_table);
2167 ext4_lock_group(sb, i);
2168 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2169 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2170 "Checksum for group %u failed (%u!=%u)",
2171 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2172 gdp)), le16_to_cpu(gdp->bg_checksum));
2173 if (!(sb->s_flags & MS_RDONLY)) {
2174 ext4_unlock_group(sb, i);
2178 ext4_unlock_group(sb, i);
2180 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2182 if (NULL != first_not_zeroed)
2183 *first_not_zeroed = grp;
2185 ext4_free_blocks_count_set(sbi->s_es,
2186 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2187 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2191 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2192 * the superblock) which were deleted from all directories, but held open by
2193 * a process at the time of a crash. We walk the list and try to delete these
2194 * inodes at recovery time (only with a read-write filesystem).
2196 * In order to keep the orphan inode chain consistent during traversal (in
2197 * case of crash during recovery), we link each inode into the superblock
2198 * orphan list_head and handle it the same way as an inode deletion during
2199 * normal operation (which journals the operations for us).
2201 * We only do an iget() and an iput() on each inode, which is very safe if we
2202 * accidentally point at an in-use or already deleted inode. The worst that
2203 * can happen in this case is that we get a "bit already cleared" message from
2204 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2205 * e2fsck was run on this filesystem, and it must have already done the orphan
2206 * inode cleanup for us, so we can safely abort without any further action.
2208 static void ext4_orphan_cleanup(struct super_block *sb,
2209 struct ext4_super_block *es)
2211 unsigned int s_flags = sb->s_flags;
2212 int nr_orphans = 0, nr_truncates = 0;
2216 if (!es->s_last_orphan) {
2217 jbd_debug(4, "no orphan inodes to clean up\n");
2221 if (bdev_read_only(sb->s_bdev)) {
2222 ext4_msg(sb, KERN_ERR, "write access "
2223 "unavailable, skipping orphan cleanup");
2227 /* Check if feature set would not allow a r/w mount */
2228 if (!ext4_feature_set_ok(sb, 0)) {
2229 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2230 "unknown ROCOMPAT features");
2234 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2235 if (es->s_last_orphan)
2236 jbd_debug(1, "Errors on filesystem, "
2237 "clearing orphan list.\n");
2238 es->s_last_orphan = 0;
2239 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2243 if (s_flags & MS_RDONLY) {
2244 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2245 sb->s_flags &= ~MS_RDONLY;
2248 /* Needed for iput() to work correctly and not trash data */
2249 sb->s_flags |= MS_ACTIVE;
2250 /* Turn on quotas so that they are updated correctly */
2251 for (i = 0; i < MAXQUOTAS; i++) {
2252 if (EXT4_SB(sb)->s_qf_names[i]) {
2253 int ret = ext4_quota_on_mount(sb, i);
2255 ext4_msg(sb, KERN_ERR,
2256 "Cannot turn on journaled "
2257 "quota: error %d", ret);
2262 while (es->s_last_orphan) {
2263 struct inode *inode;
2266 * We may have encountered an error during cleanup; if
2267 * so, skip the rest.
2269 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2270 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2271 es->s_last_orphan = 0;
2275 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2276 if (IS_ERR(inode)) {
2277 es->s_last_orphan = 0;
2281 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2282 dquot_initialize(inode);
2283 if (inode->i_nlink) {
2284 ext4_msg(sb, KERN_DEBUG,
2285 "%s: truncating inode %lu to %lld bytes",
2286 __func__, inode->i_ino, inode->i_size);
2287 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2288 inode->i_ino, inode->i_size);
2289 mutex_lock(&inode->i_mutex);
2290 ext4_truncate(inode);
2291 mutex_unlock(&inode->i_mutex);
2294 ext4_msg(sb, KERN_DEBUG,
2295 "%s: deleting unreferenced inode %lu",
2296 __func__, inode->i_ino);
2297 jbd_debug(2, "deleting unreferenced inode %lu\n",
2301 iput(inode); /* The delete magic happens here! */
2304 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2307 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2308 PLURAL(nr_orphans));
2310 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2311 PLURAL(nr_truncates));
2313 /* Turn quotas off */
2314 for (i = 0; i < MAXQUOTAS; i++) {
2315 if (sb_dqopt(sb)->files[i])
2316 dquot_quota_off(sb, i);
2319 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2323 * Maximal extent format file size.
2324 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2325 * extent format containers, within a sector_t, and within i_blocks
2326 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2327 * so that won't be a limiting factor.
2329 * However there is other limiting factor. We do store extents in the form
2330 * of starting block and length, hence the resulting length of the extent
2331 * covering maximum file size must fit into on-disk format containers as
2332 * well. Given that length is always by 1 unit bigger than max unit (because
2333 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2335 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2337 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2340 loff_t upper_limit = MAX_LFS_FILESIZE;
2342 /* small i_blocks in vfs inode? */
2343 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2345 * CONFIG_LBDAF is not enabled implies the inode
2346 * i_block represent total blocks in 512 bytes
2347 * 32 == size of vfs inode i_blocks * 8
2349 upper_limit = (1LL << 32) - 1;
2351 /* total blocks in file system block size */
2352 upper_limit >>= (blkbits - 9);
2353 upper_limit <<= blkbits;
2357 * 32-bit extent-start container, ee_block. We lower the maxbytes
2358 * by one fs block, so ee_len can cover the extent of maximum file
2361 res = (1LL << 32) - 1;
2364 /* Sanity check against vm- & vfs- imposed limits */
2365 if (res > upper_limit)
2372 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2373 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2374 * We need to be 1 filesystem block less than the 2^48 sector limit.
2376 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2378 loff_t res = EXT4_NDIR_BLOCKS;
2381 /* This is calculated to be the largest file size for a dense, block
2382 * mapped file such that the file's total number of 512-byte sectors,
2383 * including data and all indirect blocks, does not exceed (2^48 - 1).
2385 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2386 * number of 512-byte sectors of the file.
2389 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2391 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2392 * the inode i_block field represents total file blocks in
2393 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2395 upper_limit = (1LL << 32) - 1;
2397 /* total blocks in file system block size */
2398 upper_limit >>= (bits - 9);
2402 * We use 48 bit ext4_inode i_blocks
2403 * With EXT4_HUGE_FILE_FL set the i_blocks
2404 * represent total number of blocks in
2405 * file system block size
2407 upper_limit = (1LL << 48) - 1;
2411 /* indirect blocks */
2413 /* double indirect blocks */
2414 meta_blocks += 1 + (1LL << (bits-2));
2415 /* tripple indirect blocks */
2416 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2418 upper_limit -= meta_blocks;
2419 upper_limit <<= bits;
2421 res += 1LL << (bits-2);
2422 res += 1LL << (2*(bits-2));
2423 res += 1LL << (3*(bits-2));
2425 if (res > upper_limit)
2428 if (res > MAX_LFS_FILESIZE)
2429 res = MAX_LFS_FILESIZE;
2434 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2435 ext4_fsblk_t logical_sb_block, int nr)
2437 struct ext4_sb_info *sbi = EXT4_SB(sb);
2438 ext4_group_t bg, first_meta_bg;
2441 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2443 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2445 return logical_sb_block + nr + 1;
2446 bg = sbi->s_desc_per_block * nr;
2447 if (ext4_bg_has_super(sb, bg))
2450 return (has_super + ext4_group_first_block_no(sb, bg));
2454 * ext4_get_stripe_size: Get the stripe size.
2455 * @sbi: In memory super block info
2457 * If we have specified it via mount option, then
2458 * use the mount option value. If the value specified at mount time is
2459 * greater than the blocks per group use the super block value.
2460 * If the super block value is greater than blocks per group return 0.
2461 * Allocator needs it be less than blocks per group.
2464 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2466 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2467 unsigned long stripe_width =
2468 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2471 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2472 ret = sbi->s_stripe;
2473 else if (stripe_width <= sbi->s_blocks_per_group)
2475 else if (stride <= sbi->s_blocks_per_group)
2481 * If the stripe width is 1, this makes no sense and
2482 * we set it to 0 to turn off stripe handling code.
2493 struct attribute attr;
2494 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2495 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2496 const char *, size_t);
2500 static int parse_strtoul(const char *buf,
2501 unsigned long max, unsigned long *value)
2505 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2506 endp = skip_spaces(endp);
2507 if (*endp || *value > max)
2513 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi,
2517 return snprintf(buf, PAGE_SIZE, "%llu\n",
2519 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2522 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2525 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2527 if (!sb->s_bdev->bd_part)
2528 return snprintf(buf, PAGE_SIZE, "0\n");
2529 return snprintf(buf, PAGE_SIZE, "%lu\n",
2530 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2531 sbi->s_sectors_written_start) >> 1);
2534 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2537 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2539 if (!sb->s_bdev->bd_part)
2540 return snprintf(buf, PAGE_SIZE, "0\n");
2541 return snprintf(buf, PAGE_SIZE, "%llu\n",
2542 (unsigned long long)(sbi->s_kbytes_written +
2543 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2544 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2547 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2548 struct ext4_sb_info *sbi,
2549 const char *buf, size_t count)
2553 if (parse_strtoul(buf, 0x40000000, &t))
2556 if (t && !is_power_of_2(t))
2559 sbi->s_inode_readahead_blks = t;
2563 static ssize_t sbi_ui_show(struct ext4_attr *a,
2564 struct ext4_sb_info *sbi, char *buf)
2566 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2568 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2571 static ssize_t sbi_ui_store(struct ext4_attr *a,
2572 struct ext4_sb_info *sbi,
2573 const char *buf, size_t count)
2575 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2578 if (parse_strtoul(buf, 0xffffffff, &t))
2584 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2585 static struct ext4_attr ext4_attr_##_name = { \
2586 .attr = {.name = __stringify(_name), .mode = _mode }, \
2589 .offset = offsetof(struct ext4_sb_info, _elname), \
2591 #define EXT4_ATTR(name, mode, show, store) \
2592 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2594 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2595 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2596 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2597 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2598 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2599 #define ATTR_LIST(name) &ext4_attr_##name.attr
2601 EXT4_RO_ATTR(delayed_allocation_blocks);
2602 EXT4_RO_ATTR(session_write_kbytes);
2603 EXT4_RO_ATTR(lifetime_write_kbytes);
2604 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2605 inode_readahead_blks_store, s_inode_readahead_blks);
2606 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2607 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2608 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2609 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2610 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2611 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2612 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2613 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2615 static struct attribute *ext4_attrs[] = {
2616 ATTR_LIST(delayed_allocation_blocks),
2617 ATTR_LIST(session_write_kbytes),
2618 ATTR_LIST(lifetime_write_kbytes),
2619 ATTR_LIST(inode_readahead_blks),
2620 ATTR_LIST(inode_goal),
2621 ATTR_LIST(mb_stats),
2622 ATTR_LIST(mb_max_to_scan),
2623 ATTR_LIST(mb_min_to_scan),
2624 ATTR_LIST(mb_order2_req),
2625 ATTR_LIST(mb_stream_req),
2626 ATTR_LIST(mb_group_prealloc),
2627 ATTR_LIST(max_writeback_mb_bump),
2631 /* Features this copy of ext4 supports */
2632 EXT4_INFO_ATTR(lazy_itable_init);
2633 EXT4_INFO_ATTR(batched_discard);
2635 static struct attribute *ext4_feat_attrs[] = {
2636 ATTR_LIST(lazy_itable_init),
2637 ATTR_LIST(batched_discard),
2641 static ssize_t ext4_attr_show(struct kobject *kobj,
2642 struct attribute *attr, char *buf)
2644 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2646 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2648 return a->show ? a->show(a, sbi, buf) : 0;
2651 static ssize_t ext4_attr_store(struct kobject *kobj,
2652 struct attribute *attr,
2653 const char *buf, size_t len)
2655 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2657 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2659 return a->store ? a->store(a, sbi, buf, len) : 0;
2662 static void ext4_sb_release(struct kobject *kobj)
2664 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2666 complete(&sbi->s_kobj_unregister);
2669 static const struct sysfs_ops ext4_attr_ops = {
2670 .show = ext4_attr_show,
2671 .store = ext4_attr_store,
2674 static struct kobj_type ext4_ktype = {
2675 .default_attrs = ext4_attrs,
2676 .sysfs_ops = &ext4_attr_ops,
2677 .release = ext4_sb_release,
2680 static void ext4_feat_release(struct kobject *kobj)
2682 complete(&ext4_feat->f_kobj_unregister);
2685 static struct kobj_type ext4_feat_ktype = {
2686 .default_attrs = ext4_feat_attrs,
2687 .sysfs_ops = &ext4_attr_ops,
2688 .release = ext4_feat_release,
2692 * Check whether this filesystem can be mounted based on
2693 * the features present and the RDONLY/RDWR mount requested.
2694 * Returns 1 if this filesystem can be mounted as requested,
2695 * 0 if it cannot be.
2697 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2700 ext4_msg(sb, KERN_ERR,
2701 "Couldn't mount because of "
2702 "unsupported optional features (%x)",
2703 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2704 ~EXT4_FEATURE_INCOMPAT_SUPP));
2711 /* Check that feature set is OK for a read-write mount */
2712 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2713 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2714 "unsupported optional features (%x)",
2715 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2716 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2720 * Large file size enabled file system can only be mounted
2721 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2723 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2724 if (sizeof(blkcnt_t) < sizeof(u64)) {
2725 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2726 "cannot be mounted RDWR without "
2731 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2732 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2733 ext4_msg(sb, KERN_ERR,
2734 "Can't support bigalloc feature without "
2735 "extents feature\n");
2742 * This function is called once a day if we have errors logged
2743 * on the file system
2745 static void print_daily_error_info(unsigned long arg)
2747 struct super_block *sb = (struct super_block *) arg;
2748 struct ext4_sb_info *sbi;
2749 struct ext4_super_block *es;
2754 if (es->s_error_count)
2755 /* fsck newer than v1.41.13 is needed to clean this condition. */
2756 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2757 le32_to_cpu(es->s_error_count));
2758 if (es->s_first_error_time) {
2759 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2760 sb->s_id, le32_to_cpu(es->s_first_error_time),
2761 (int) sizeof(es->s_first_error_func),
2762 es->s_first_error_func,
2763 le32_to_cpu(es->s_first_error_line));
2764 if (es->s_first_error_ino)
2765 printk(": inode %u",
2766 le32_to_cpu(es->s_first_error_ino));
2767 if (es->s_first_error_block)
2768 printk(": block %llu", (unsigned long long)
2769 le64_to_cpu(es->s_first_error_block));
2772 if (es->s_last_error_time) {
2773 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2774 sb->s_id, le32_to_cpu(es->s_last_error_time),
2775 (int) sizeof(es->s_last_error_func),
2776 es->s_last_error_func,
2777 le32_to_cpu(es->s_last_error_line));
2778 if (es->s_last_error_ino)
2779 printk(": inode %u",
2780 le32_to_cpu(es->s_last_error_ino));
2781 if (es->s_last_error_block)
2782 printk(": block %llu", (unsigned long long)
2783 le64_to_cpu(es->s_last_error_block));
2786 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2789 /* Find next suitable group and run ext4_init_inode_table */
2790 static int ext4_run_li_request(struct ext4_li_request *elr)
2792 struct ext4_group_desc *gdp = NULL;
2793 ext4_group_t group, ngroups;
2794 struct super_block *sb;
2795 unsigned long timeout = 0;
2799 ngroups = EXT4_SB(sb)->s_groups_count;
2801 for (group = elr->lr_next_group; group < ngroups; group++) {
2802 gdp = ext4_get_group_desc(sb, group, NULL);
2808 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2812 if (group == ngroups)
2817 ret = ext4_init_inode_table(sb, group,
2818 elr->lr_timeout ? 0 : 1);
2819 if (elr->lr_timeout == 0) {
2820 timeout = (jiffies - timeout) *
2821 elr->lr_sbi->s_li_wait_mult;
2822 elr->lr_timeout = timeout;
2824 elr->lr_next_sched = jiffies + elr->lr_timeout;
2825 elr->lr_next_group = group + 1;
2832 * Remove lr_request from the list_request and free the
2833 * request structure. Should be called with li_list_mtx held
2835 static void ext4_remove_li_request(struct ext4_li_request *elr)
2837 struct ext4_sb_info *sbi;
2844 list_del(&elr->lr_request);
2845 sbi->s_li_request = NULL;
2849 static void ext4_unregister_li_request(struct super_block *sb)
2851 mutex_lock(&ext4_li_mtx);
2852 if (!ext4_li_info) {
2853 mutex_unlock(&ext4_li_mtx);
2857 mutex_lock(&ext4_li_info->li_list_mtx);
2858 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2859 mutex_unlock(&ext4_li_info->li_list_mtx);
2860 mutex_unlock(&ext4_li_mtx);
2863 static struct task_struct *ext4_lazyinit_task;
2866 * This is the function where ext4lazyinit thread lives. It walks
2867 * through the request list searching for next scheduled filesystem.
2868 * When such a fs is found, run the lazy initialization request
2869 * (ext4_rn_li_request) and keep track of the time spend in this
2870 * function. Based on that time we compute next schedule time of
2871 * the request. When walking through the list is complete, compute
2872 * next waking time and put itself into sleep.
2874 static int ext4_lazyinit_thread(void *arg)
2876 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2877 struct list_head *pos, *n;
2878 struct ext4_li_request *elr;
2879 unsigned long next_wakeup, cur;
2881 BUG_ON(NULL == eli);
2885 next_wakeup = MAX_JIFFY_OFFSET;
2887 mutex_lock(&eli->li_list_mtx);
2888 if (list_empty(&eli->li_request_list)) {
2889 mutex_unlock(&eli->li_list_mtx);
2893 list_for_each_safe(pos, n, &eli->li_request_list) {
2894 elr = list_entry(pos, struct ext4_li_request,
2897 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2898 if (ext4_run_li_request(elr) != 0) {
2899 /* error, remove the lazy_init job */
2900 ext4_remove_li_request(elr);
2905 if (time_before(elr->lr_next_sched, next_wakeup))
2906 next_wakeup = elr->lr_next_sched;
2908 mutex_unlock(&eli->li_list_mtx);
2910 if (freezing(current))
2914 if ((time_after_eq(cur, next_wakeup)) ||
2915 (MAX_JIFFY_OFFSET == next_wakeup)) {
2920 schedule_timeout_interruptible(next_wakeup - cur);
2922 if (kthread_should_stop()) {
2923 ext4_clear_request_list();
2930 * It looks like the request list is empty, but we need
2931 * to check it under the li_list_mtx lock, to prevent any
2932 * additions into it, and of course we should lock ext4_li_mtx
2933 * to atomically free the list and ext4_li_info, because at
2934 * this point another ext4 filesystem could be registering
2937 mutex_lock(&ext4_li_mtx);
2938 mutex_lock(&eli->li_list_mtx);
2939 if (!list_empty(&eli->li_request_list)) {
2940 mutex_unlock(&eli->li_list_mtx);
2941 mutex_unlock(&ext4_li_mtx);
2944 mutex_unlock(&eli->li_list_mtx);
2945 kfree(ext4_li_info);
2946 ext4_li_info = NULL;
2947 mutex_unlock(&ext4_li_mtx);
2952 static void ext4_clear_request_list(void)
2954 struct list_head *pos, *n;
2955 struct ext4_li_request *elr;
2957 mutex_lock(&ext4_li_info->li_list_mtx);
2958 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2959 elr = list_entry(pos, struct ext4_li_request,
2961 ext4_remove_li_request(elr);
2963 mutex_unlock(&ext4_li_info->li_list_mtx);
2966 static int ext4_run_lazyinit_thread(void)
2968 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2969 ext4_li_info, "ext4lazyinit");
2970 if (IS_ERR(ext4_lazyinit_task)) {
2971 int err = PTR_ERR(ext4_lazyinit_task);
2972 ext4_clear_request_list();
2973 kfree(ext4_li_info);
2974 ext4_li_info = NULL;
2975 printk(KERN_CRIT "EXT4: error %d creating inode table "
2976 "initialization thread\n",
2980 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2985 * Check whether it make sense to run itable init. thread or not.
2986 * If there is at least one uninitialized inode table, return
2987 * corresponding group number, else the loop goes through all
2988 * groups and return total number of groups.
2990 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2992 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2993 struct ext4_group_desc *gdp = NULL;
2995 for (group = 0; group < ngroups; group++) {
2996 gdp = ext4_get_group_desc(sb, group, NULL);
3000 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3007 static int ext4_li_info_new(void)
3009 struct ext4_lazy_init *eli = NULL;
3011 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3015 INIT_LIST_HEAD(&eli->li_request_list);
3016 mutex_init(&eli->li_list_mtx);
3018 eli->li_state |= EXT4_LAZYINIT_QUIT;
3025 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3028 struct ext4_sb_info *sbi = EXT4_SB(sb);
3029 struct ext4_li_request *elr;
3032 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3038 elr->lr_next_group = start;
3041 * Randomize first schedule time of the request to
3042 * spread the inode table initialization requests
3045 get_random_bytes(&rnd, sizeof(rnd));
3046 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3047 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3052 static int ext4_register_li_request(struct super_block *sb,
3053 ext4_group_t first_not_zeroed)
3055 struct ext4_sb_info *sbi = EXT4_SB(sb);
3056 struct ext4_li_request *elr;
3057 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3060 if (sbi->s_li_request != NULL) {
3062 * Reset timeout so it can be computed again, because
3063 * s_li_wait_mult might have changed.
3065 sbi->s_li_request->lr_timeout = 0;
3069 if (first_not_zeroed == ngroups ||
3070 (sb->s_flags & MS_RDONLY) ||
3071 !test_opt(sb, INIT_INODE_TABLE))
3074 elr = ext4_li_request_new(sb, first_not_zeroed);
3078 mutex_lock(&ext4_li_mtx);
3080 if (NULL == ext4_li_info) {
3081 ret = ext4_li_info_new();
3086 mutex_lock(&ext4_li_info->li_list_mtx);
3087 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3088 mutex_unlock(&ext4_li_info->li_list_mtx);
3090 sbi->s_li_request = elr;
3092 * set elr to NULL here since it has been inserted to
3093 * the request_list and the removal and free of it is
3094 * handled by ext4_clear_request_list from now on.
3098 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3099 ret = ext4_run_lazyinit_thread();
3104 mutex_unlock(&ext4_li_mtx);
3111 * We do not need to lock anything since this is called on
3114 static void ext4_destroy_lazyinit_thread(void)
3117 * If thread exited earlier
3118 * there's nothing to be done.
3120 if (!ext4_li_info || !ext4_lazyinit_task)
3123 kthread_stop(ext4_lazyinit_task);
3127 * Note: calculating the overhead so we can be compatible with
3128 * historical BSD practice is quite difficult in the face of
3129 * clusters/bigalloc. This is because multiple metadata blocks from
3130 * different block group can end up in the same allocation cluster.
3131 * Calculating the exact overhead in the face of clustered allocation
3132 * requires either O(all block bitmaps) in memory or O(number of block
3133 * groups**2) in time. We will still calculate the superblock for
3134 * older file systems --- and if we come across with a bigalloc file
3135 * system with zero in s_overhead_clusters the estimate will be close to
3136 * correct especially for very large cluster sizes --- but for newer
3137 * file systems, it's better to calculate this figure once at mkfs
3138 * time, and store it in the superblock. If the superblock value is
3139 * present (even for non-bigalloc file systems), we will use it.
3141 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3144 struct ext4_sb_info *sbi = EXT4_SB(sb);
3145 struct ext4_group_desc *gdp;
3146 ext4_fsblk_t first_block, last_block, b;
3147 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3148 int s, j, count = 0;
3150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3151 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3152 sbi->s_itb_per_group + 2);
3154 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3155 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3156 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3157 for (i = 0; i < ngroups; i++) {
3158 gdp = ext4_get_group_desc(sb, i, NULL);
3159 b = ext4_block_bitmap(sb, gdp);
3160 if (b >= first_block && b <= last_block) {
3161 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3164 b = ext4_inode_bitmap(sb, gdp);
3165 if (b >= first_block && b <= last_block) {
3166 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3169 b = ext4_inode_table(sb, gdp);
3170 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3171 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3172 int c = EXT4_B2C(sbi, b - first_block);
3173 ext4_set_bit(c, buf);
3179 if (ext4_bg_has_super(sb, grp)) {
3180 ext4_set_bit(s++, buf);
3183 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3184 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3190 return EXT4_CLUSTERS_PER_GROUP(sb) -
3191 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3195 * Compute the overhead and stash it in sbi->s_overhead
3197 int ext4_calculate_overhead(struct super_block *sb)
3199 struct ext4_sb_info *sbi = EXT4_SB(sb);
3200 struct ext4_super_block *es = sbi->s_es;
3201 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3202 ext4_fsblk_t overhead = 0;
3203 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3205 memset(buf, 0, PAGE_SIZE);
3210 * Compute the overhead (FS structures). This is constant
3211 * for a given filesystem unless the number of block groups
3212 * changes so we cache the previous value until it does.
3216 * All of the blocks before first_data_block are overhead
3218 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3221 * Add the overhead found in each block group
3223 for (i = 0; i < ngroups; i++) {
3226 blks = count_overhead(sb, i, buf);
3229 memset(buf, 0, PAGE_SIZE);
3232 sbi->s_overhead = overhead;
3234 free_page((unsigned long) buf);
3238 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3240 char *orig_data = kstrdup(data, GFP_KERNEL);
3241 struct buffer_head *bh;
3242 struct ext4_super_block *es = NULL;
3243 struct ext4_sb_info *sbi;
3245 ext4_fsblk_t sb_block = get_sb_block(&data);
3246 ext4_fsblk_t logical_sb_block;
3247 unsigned long offset = 0;
3248 unsigned long journal_devnum = 0;
3249 unsigned long def_mount_opts;
3254 int blocksize, clustersize;
3255 unsigned int db_count;
3257 int needs_recovery, has_huge_files, has_bigalloc;
3260 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3261 ext4_group_t first_not_zeroed;
3263 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3267 sbi->s_blockgroup_lock =
3268 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3269 if (!sbi->s_blockgroup_lock) {
3273 sb->s_fs_info = sbi;
3274 sbi->s_mount_opt = 0;
3275 sbi->s_resuid = EXT4_DEF_RESUID;
3276 sbi->s_resgid = EXT4_DEF_RESGID;
3277 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3278 sbi->s_sb_block = sb_block;
3279 if (sb->s_bdev->bd_part)
3280 sbi->s_sectors_written_start =
3281 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3283 /* Cleanup superblock name */
3284 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3288 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3290 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3295 * The ext4 superblock will not be buffer aligned for other than 1kB
3296 * block sizes. We need to calculate the offset from buffer start.
3298 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3299 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3300 offset = do_div(logical_sb_block, blocksize);
3302 logical_sb_block = sb_block;
3305 if (!(bh = sb_bread(sb, logical_sb_block))) {
3306 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3310 * Note: s_es must be initialized as soon as possible because
3311 * some ext4 macro-instructions depend on its value
3313 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3315 sb->s_magic = le16_to_cpu(es->s_magic);
3316 if (sb->s_magic != EXT4_SUPER_MAGIC)
3318 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3320 /* Set defaults before we parse the mount options */
3321 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3322 set_opt(sb, INIT_INODE_TABLE);
3323 if (def_mount_opts & EXT4_DEFM_DEBUG)
3325 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3326 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3330 if (def_mount_opts & EXT4_DEFM_UID16)
3331 set_opt(sb, NO_UID32);
3332 /* xattr user namespace & acls are now defaulted on */
3333 #ifdef CONFIG_EXT4_FS_XATTR
3334 set_opt(sb, XATTR_USER);
3336 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3337 set_opt(sb, POSIX_ACL);
3339 set_opt(sb, MBLK_IO_SUBMIT);
3340 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3341 set_opt(sb, JOURNAL_DATA);
3342 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3343 set_opt(sb, ORDERED_DATA);
3344 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3345 set_opt(sb, WRITEBACK_DATA);
3347 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3348 set_opt(sb, ERRORS_PANIC);
3349 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3350 set_opt(sb, ERRORS_CONT);
3352 set_opt(sb, ERRORS_RO);
3353 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3354 set_opt(sb, BLOCK_VALIDITY);
3355 if (def_mount_opts & EXT4_DEFM_DISCARD)
3356 set_opt(sb, DISCARD);
3358 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3359 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3360 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3361 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3362 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3364 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3365 set_opt(sb, BARRIER);
3368 * enable delayed allocation by default
3369 * Use -o nodelalloc to turn it off
3371 if (!IS_EXT3_SB(sb) &&
3372 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3373 set_opt(sb, DELALLOC);
3376 * set default s_li_wait_mult for lazyinit, for the case there is
3377 * no mount option specified.
3379 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3381 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3382 &journal_devnum, &journal_ioprio, NULL, 0)) {
3383 ext4_msg(sb, KERN_WARNING,
3384 "failed to parse options in superblock: %s",
3385 sbi->s_es->s_mount_opts);
3387 if (!parse_options((char *) data, sb, &journal_devnum,
3388 &journal_ioprio, NULL, 0))
3391 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3392 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3393 "with data=journal disables delayed "
3394 "allocation and O_DIRECT support!\n");
3395 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3396 ext4_msg(sb, KERN_ERR, "can't mount with "
3397 "both data=journal and delalloc");
3400 if (test_opt(sb, DIOREAD_NOLOCK)) {
3401 ext4_msg(sb, KERN_ERR, "can't mount with "
3402 "both data=journal and dioread_nolock");
3405 if (test_opt(sb, DELALLOC))
3406 clear_opt(sb, DELALLOC);
3409 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3410 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3412 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3413 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3414 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3415 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3416 ext4_msg(sb, KERN_WARNING,
3417 "feature flags set on rev 0 fs, "
3418 "running e2fsck is recommended");
3420 if (IS_EXT2_SB(sb)) {
3421 if (ext2_feature_set_ok(sb))
3422 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3423 "using the ext4 subsystem");
3425 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3426 "to feature incompatibilities");
3431 if (IS_EXT3_SB(sb)) {
3432 if (ext3_feature_set_ok(sb))
3433 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3434 "using the ext4 subsystem");
3436 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3437 "to feature incompatibilities");
3443 * Check feature flags regardless of the revision level, since we
3444 * previously didn't change the revision level when setting the flags,
3445 * so there is a chance incompat flags are set on a rev 0 filesystem.
3447 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3450 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3451 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3452 blocksize > EXT4_MAX_BLOCK_SIZE) {
3453 ext4_msg(sb, KERN_ERR,
3454 "Unsupported filesystem blocksize %d", blocksize);
3458 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3459 ext4_msg(sb, KERN_ERR,
3460 "Number of reserved GDT blocks insanely large: %d",
3461 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3465 if (sb->s_blocksize != blocksize) {
3466 /* Validate the filesystem blocksize */
3467 if (!sb_set_blocksize(sb, blocksize)) {
3468 ext4_msg(sb, KERN_ERR, "bad block size %d",
3474 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3475 offset = do_div(logical_sb_block, blocksize);
3476 bh = sb_bread(sb, logical_sb_block);
3478 ext4_msg(sb, KERN_ERR,
3479 "Can't read superblock on 2nd try");
3482 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3484 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3485 ext4_msg(sb, KERN_ERR,
3486 "Magic mismatch, very weird!");
3491 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3492 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3493 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3495 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3497 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3498 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3499 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3501 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3502 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3503 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3504 (!is_power_of_2(sbi->s_inode_size)) ||
3505 (sbi->s_inode_size > blocksize)) {
3506 ext4_msg(sb, KERN_ERR,
3507 "unsupported inode size: %d",
3511 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3512 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3515 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3516 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3517 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3518 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3519 !is_power_of_2(sbi->s_desc_size)) {
3520 ext4_msg(sb, KERN_ERR,
3521 "unsupported descriptor size %lu",
3526 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3528 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3529 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3530 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3533 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3534 if (sbi->s_inodes_per_block == 0)
3536 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3537 sbi->s_inodes_per_block;
3538 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3540 sbi->s_mount_state = le16_to_cpu(es->s_state);
3541 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3542 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3544 for (i = 0; i < 4; i++)
3545 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3546 sbi->s_def_hash_version = es->s_def_hash_version;
3547 i = le32_to_cpu(es->s_flags);
3548 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3549 sbi->s_hash_unsigned = 3;
3550 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3551 #ifdef __CHAR_UNSIGNED__
3552 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3553 sbi->s_hash_unsigned = 3;
3555 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3560 /* Handle clustersize */
3561 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3562 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3563 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3565 if (clustersize < blocksize) {
3566 ext4_msg(sb, KERN_ERR,
3567 "cluster size (%d) smaller than "
3568 "block size (%d)", clustersize, blocksize);
3571 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3572 le32_to_cpu(es->s_log_block_size);
3573 sbi->s_clusters_per_group =
3574 le32_to_cpu(es->s_clusters_per_group);
3575 if (sbi->s_clusters_per_group > blocksize * 8) {
3576 ext4_msg(sb, KERN_ERR,
3577 "#clusters per group too big: %lu",
3578 sbi->s_clusters_per_group);
3581 if (sbi->s_blocks_per_group !=
3582 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3583 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3584 "clusters per group (%lu) inconsistent",
3585 sbi->s_blocks_per_group,
3586 sbi->s_clusters_per_group);
3590 if (clustersize != blocksize) {
3591 ext4_warning(sb, "fragment/cluster size (%d) != "
3592 "block size (%d)", clustersize,
3594 clustersize = blocksize;
3596 if (sbi->s_blocks_per_group > blocksize * 8) {
3597 ext4_msg(sb, KERN_ERR,
3598 "#blocks per group too big: %lu",
3599 sbi->s_blocks_per_group);
3602 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3603 sbi->s_cluster_bits = 0;
3605 sbi->s_cluster_ratio = clustersize / blocksize;
3607 if (sbi->s_inodes_per_group > blocksize * 8) {
3608 ext4_msg(sb, KERN_ERR,
3609 "#inodes per group too big: %lu",
3610 sbi->s_inodes_per_group);
3615 * Test whether we have more sectors than will fit in sector_t,
3616 * and whether the max offset is addressable by the page cache.
3618 err = generic_check_addressable(sb->s_blocksize_bits,
3619 ext4_blocks_count(es));
3621 ext4_msg(sb, KERN_ERR, "filesystem"
3622 " too large to mount safely on this system");
3623 if (sizeof(sector_t) < 8)
3624 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3629 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3632 /* check blocks count against device size */
3633 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3634 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3635 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3636 "exceeds size of device (%llu blocks)",
3637 ext4_blocks_count(es), blocks_count);
3642 * It makes no sense for the first data block to be beyond the end
3643 * of the filesystem.
3645 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3646 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3647 "block %u is beyond end of filesystem (%llu)",
3648 le32_to_cpu(es->s_first_data_block),
3649 ext4_blocks_count(es));
3652 blocks_count = (ext4_blocks_count(es) -
3653 le32_to_cpu(es->s_first_data_block) +
3654 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3655 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3656 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3657 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3658 "(block count %llu, first data block %u, "
3659 "blocks per group %lu)", sbi->s_groups_count,
3660 ext4_blocks_count(es),
3661 le32_to_cpu(es->s_first_data_block),
3662 EXT4_BLOCKS_PER_GROUP(sb));
3665 sbi->s_groups_count = blocks_count;
3666 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3667 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3668 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3669 EXT4_DESC_PER_BLOCK(sb);
3670 sbi->s_group_desc = ext4_kvmalloc(db_count *
3671 sizeof(struct buffer_head *),
3673 if (sbi->s_group_desc == NULL) {
3674 ext4_msg(sb, KERN_ERR, "not enough memory");
3679 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3681 bgl_lock_init(sbi->s_blockgroup_lock);
3683 for (i = 0; i < db_count; i++) {
3684 block = descriptor_loc(sb, logical_sb_block, i);
3685 sbi->s_group_desc[i] = sb_bread(sb, block);
3686 if (!sbi->s_group_desc[i]) {
3687 ext4_msg(sb, KERN_ERR,
3688 "can't read group descriptor %d", i);
3693 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3694 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3697 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3698 if (!ext4_fill_flex_info(sb)) {
3699 ext4_msg(sb, KERN_ERR,
3700 "unable to initialize "
3701 "flex_bg meta info!");
3705 sbi->s_gdb_count = db_count;
3706 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3707 spin_lock_init(&sbi->s_next_gen_lock);
3709 init_timer(&sbi->s_err_report);
3710 sbi->s_err_report.function = print_daily_error_info;
3711 sbi->s_err_report.data = (unsigned long) sb;
3713 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3714 ext4_count_free_clusters(sb));
3716 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3717 ext4_count_free_inodes(sb));
3720 err = percpu_counter_init(&sbi->s_dirs_counter,
3721 ext4_count_dirs(sb));
3724 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3727 ext4_msg(sb, KERN_ERR, "insufficient memory");
3731 sbi->s_stripe = ext4_get_stripe_size(sbi);
3732 sbi->s_max_writeback_mb_bump = 128;
3735 * set up enough so that it can read an inode
3737 if (!test_opt(sb, NOLOAD) &&
3738 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3739 sb->s_op = &ext4_sops;
3741 sb->s_op = &ext4_nojournal_sops;
3742 sb->s_export_op = &ext4_export_ops;
3743 sb->s_xattr = ext4_xattr_handlers;
3745 sb->s_qcop = &ext4_qctl_operations;
3746 sb->dq_op = &ext4_quota_operations;
3748 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3750 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3751 mutex_init(&sbi->s_orphan_lock);
3752 sbi->s_resize_flags = 0;
3756 needs_recovery = (es->s_last_orphan != 0 ||
3757 EXT4_HAS_INCOMPAT_FEATURE(sb,
3758 EXT4_FEATURE_INCOMPAT_RECOVER));
3760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3761 !(sb->s_flags & MS_RDONLY))
3762 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3766 * The first inode we look at is the journal inode. Don't try
3767 * root first: it may be modified in the journal!
3769 if (!test_opt(sb, NOLOAD) &&
3770 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3771 if (ext4_load_journal(sb, es, journal_devnum))
3773 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3774 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3775 ext4_msg(sb, KERN_ERR, "required journal recovery "
3776 "suppressed and not mounted read-only");
3777 goto failed_mount_wq;
3779 clear_opt(sb, DATA_FLAGS);
3780 sbi->s_journal = NULL;
3785 if (ext4_blocks_count(es) > 0xffffffffULL &&
3786 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3787 JBD2_FEATURE_INCOMPAT_64BIT)) {
3788 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3789 goto failed_mount_wq;
3792 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3793 jbd2_journal_set_features(sbi->s_journal,
3794 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3795 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3796 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3797 jbd2_journal_set_features(sbi->s_journal,
3798 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3799 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3800 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3802 jbd2_journal_clear_features(sbi->s_journal,
3803 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3804 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3807 /* We have now updated the journal if required, so we can
3808 * validate the data journaling mode. */
3809 switch (test_opt(sb, DATA_FLAGS)) {
3811 /* No mode set, assume a default based on the journal
3812 * capabilities: ORDERED_DATA if the journal can
3813 * cope, else JOURNAL_DATA
3815 if (jbd2_journal_check_available_features
3816 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3817 set_opt(sb, ORDERED_DATA);
3819 set_opt(sb, JOURNAL_DATA);
3822 case EXT4_MOUNT_ORDERED_DATA:
3823 case EXT4_MOUNT_WRITEBACK_DATA:
3824 if (!jbd2_journal_check_available_features
3825 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3826 ext4_msg(sb, KERN_ERR, "Journal does not support "
3827 "requested data journaling mode");
3828 goto failed_mount_wq;
3833 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3836 * The journal may have updated the bg summary counts, so we
3837 * need to update the global counters.
3839 percpu_counter_set(&sbi->s_freeclusters_counter,
3840 ext4_count_free_clusters(sb));
3841 percpu_counter_set(&sbi->s_freeinodes_counter,
3842 ext4_count_free_inodes(sb));
3843 percpu_counter_set(&sbi->s_dirs_counter,
3844 ext4_count_dirs(sb));
3845 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3849 * Get the # of file system overhead blocks from the
3850 * superblock if present.
3852 if (es->s_overhead_clusters)
3853 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3855 ret = ext4_calculate_overhead(sb);
3857 goto failed_mount_wq;
3861 * The maximum number of concurrent works can be high and
3862 * concurrency isn't really necessary. Limit it to 1.
3864 EXT4_SB(sb)->dio_unwritten_wq =
3865 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3866 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3867 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3868 goto failed_mount_wq;
3872 * The jbd2_journal_load will have done any necessary log recovery,
3873 * so we can safely mount the rest of the filesystem now.
3876 root = ext4_iget(sb, EXT4_ROOT_INO);
3878 ext4_msg(sb, KERN_ERR, "get root inode failed");
3879 ret = PTR_ERR(root);
3883 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3884 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3887 sb->s_root = d_alloc_root(root);
3889 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3894 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3895 sb->s_flags |= MS_RDONLY;
3897 /* determine the minimum size of new large inodes, if present */
3898 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3899 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3900 EXT4_GOOD_OLD_INODE_SIZE;
3901 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3902 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3903 if (sbi->s_want_extra_isize <
3904 le16_to_cpu(es->s_want_extra_isize))
3905 sbi->s_want_extra_isize =
3906 le16_to_cpu(es->s_want_extra_isize);
3907 if (sbi->s_want_extra_isize <
3908 le16_to_cpu(es->s_min_extra_isize))
3909 sbi->s_want_extra_isize =
3910 le16_to_cpu(es->s_min_extra_isize);
3913 /* Check if enough inode space is available */
3914 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3915 sbi->s_inode_size) {
3916 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3917 EXT4_GOOD_OLD_INODE_SIZE;
3918 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3922 err = ext4_setup_system_zone(sb);
3924 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3930 err = ext4_mb_init(sb, needs_recovery);
3932 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3937 err = ext4_register_li_request(sb, first_not_zeroed);
3941 sbi->s_kobj.kset = ext4_kset;
3942 init_completion(&sbi->s_kobj_unregister);
3943 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3948 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3949 ext4_orphan_cleanup(sb, es);
3950 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3951 if (needs_recovery) {
3952 ext4_msg(sb, KERN_INFO, "recovery complete");
3953 ext4_mark_recovery_complete(sb, es);
3955 if (EXT4_SB(sb)->s_journal) {
3956 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3957 descr = " journalled data mode";
3958 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3959 descr = " ordered data mode";
3961 descr = " writeback data mode";
3963 descr = "out journal";
3965 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3966 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3967 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3969 if (es->s_error_count)
3970 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3977 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3981 ext4_unregister_li_request(sb);
3983 ext4_ext_release(sb);
3985 ext4_mb_release(sb);
3986 ext4_release_system_zone(sb);
3990 ext4_msg(sb, KERN_ERR, "mount failed");
3991 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3993 if (sbi->s_journal) {
3994 jbd2_journal_destroy(sbi->s_journal);
3995 sbi->s_journal = NULL;
3998 del_timer_sync(&sbi->s_err_report);
3999 if (sbi->s_flex_groups)
4000 ext4_kvfree(sbi->s_flex_groups);
4001 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4002 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4003 percpu_counter_destroy(&sbi->s_dirs_counter);
4004 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4006 kthread_stop(sbi->s_mmp_tsk);
4008 for (i = 0; i < db_count; i++)
4009 brelse(sbi->s_group_desc[i]);
4010 ext4_kvfree(sbi->s_group_desc);
4013 remove_proc_entry(sb->s_id, ext4_proc_root);
4016 for (i = 0; i < MAXQUOTAS; i++)
4017 kfree(sbi->s_qf_names[i]);
4019 ext4_blkdev_remove(sbi);
4022 sb->s_fs_info = NULL;
4023 kfree(sbi->s_blockgroup_lock);
4031 * Setup any per-fs journal parameters now. We'll do this both on
4032 * initial mount, once the journal has been initialised but before we've
4033 * done any recovery; and again on any subsequent remount.
4035 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4037 struct ext4_sb_info *sbi = EXT4_SB(sb);
4039 journal->j_commit_interval = sbi->s_commit_interval;
4040 journal->j_min_batch_time = sbi->s_min_batch_time;
4041 journal->j_max_batch_time = sbi->s_max_batch_time;
4043 write_lock(&journal->j_state_lock);
4044 if (test_opt(sb, BARRIER))
4045 journal->j_flags |= JBD2_BARRIER;
4047 journal->j_flags &= ~JBD2_BARRIER;
4048 if (test_opt(sb, DATA_ERR_ABORT))
4049 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4051 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4052 write_unlock(&journal->j_state_lock);
4055 static journal_t *ext4_get_journal(struct super_block *sb,
4056 unsigned int journal_inum)
4058 struct inode *journal_inode;
4061 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4063 /* First, test for the existence of a valid inode on disk. Bad
4064 * things happen if we iget() an unused inode, as the subsequent
4065 * iput() will try to delete it. */
4067 journal_inode = ext4_iget(sb, journal_inum);
4068 if (IS_ERR(journal_inode)) {
4069 ext4_msg(sb, KERN_ERR, "no journal found");
4072 if (!journal_inode->i_nlink) {
4073 make_bad_inode(journal_inode);
4074 iput(journal_inode);
4075 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4079 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4080 journal_inode, journal_inode->i_size);
4081 if (!S_ISREG(journal_inode->i_mode)) {
4082 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4083 iput(journal_inode);
4087 journal = jbd2_journal_init_inode(journal_inode);
4089 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4090 iput(journal_inode);
4093 journal->j_private = sb;
4094 ext4_init_journal_params(sb, journal);
4098 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4101 struct buffer_head *bh;
4105 int hblock, blocksize;
4106 ext4_fsblk_t sb_block;
4107 unsigned long offset;
4108 struct ext4_super_block *es;
4109 struct block_device *bdev;
4111 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4113 bdev = ext4_blkdev_get(j_dev, sb);
4117 blocksize = sb->s_blocksize;
4118 hblock = bdev_logical_block_size(bdev);
4119 if (blocksize < hblock) {
4120 ext4_msg(sb, KERN_ERR,
4121 "blocksize too small for journal device");
4125 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4126 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4127 set_blocksize(bdev, blocksize);
4128 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4129 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4130 "external journal");
4134 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4135 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4136 !(le32_to_cpu(es->s_feature_incompat) &
4137 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4138 ext4_msg(sb, KERN_ERR, "external journal has "
4144 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4145 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4150 len = ext4_blocks_count(es);
4151 start = sb_block + 1;
4152 brelse(bh); /* we're done with the superblock */
4154 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4155 start, len, blocksize);
4157 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4160 journal->j_private = sb;
4161 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4162 wait_on_buffer(journal->j_sb_buffer);
4163 if (!buffer_uptodate(journal->j_sb_buffer)) {
4164 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4167 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4168 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4169 "user (unsupported) - %d",
4170 be32_to_cpu(journal->j_superblock->s_nr_users));
4173 EXT4_SB(sb)->journal_bdev = bdev;
4174 ext4_init_journal_params(sb, journal);
4178 jbd2_journal_destroy(journal);
4180 ext4_blkdev_put(bdev);
4184 static int ext4_load_journal(struct super_block *sb,
4185 struct ext4_super_block *es,
4186 unsigned long journal_devnum)
4189 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4192 int really_read_only;
4194 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4196 if (journal_devnum &&
4197 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4198 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4199 "numbers have changed");
4200 journal_dev = new_decode_dev(journal_devnum);
4202 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4204 really_read_only = bdev_read_only(sb->s_bdev);
4207 * Are we loading a blank journal or performing recovery after a
4208 * crash? For recovery, we need to check in advance whether we
4209 * can get read-write access to the device.
4211 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4212 if (sb->s_flags & MS_RDONLY) {
4213 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4214 "required on readonly filesystem");
4215 if (really_read_only) {
4216 ext4_msg(sb, KERN_ERR, "write access "
4217 "unavailable, cannot proceed");
4220 ext4_msg(sb, KERN_INFO, "write access will "
4221 "be enabled during recovery");
4225 if (journal_inum && journal_dev) {
4226 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4227 "and inode journals!");
4232 if (!(journal = ext4_get_journal(sb, journal_inum)))
4235 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4239 if (!(journal->j_flags & JBD2_BARRIER))
4240 ext4_msg(sb, KERN_INFO, "barriers disabled");
4242 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4243 err = jbd2_journal_update_format(journal);
4245 ext4_msg(sb, KERN_ERR, "error updating journal");
4246 jbd2_journal_destroy(journal);
4251 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4252 err = jbd2_journal_wipe(journal, !really_read_only);
4254 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4256 memcpy(save, ((char *) es) +
4257 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4258 err = jbd2_journal_load(journal);
4260 memcpy(((char *) es) + EXT4_S_ERR_START,
4261 save, EXT4_S_ERR_LEN);
4266 ext4_msg(sb, KERN_ERR, "error loading journal");
4267 jbd2_journal_destroy(journal);
4271 EXT4_SB(sb)->s_journal = journal;
4272 ext4_clear_journal_err(sb, es);
4274 if (!really_read_only && journal_devnum &&
4275 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4276 es->s_journal_dev = cpu_to_le32(journal_devnum);
4278 /* Make sure we flush the recovery flag to disk. */
4279 ext4_commit_super(sb, 1);
4285 static int ext4_commit_super(struct super_block *sb, int sync)
4287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4288 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4291 if (!sbh || block_device_ejected(sb))
4293 if (buffer_write_io_error(sbh)) {
4295 * Oh, dear. A previous attempt to write the
4296 * superblock failed. This could happen because the
4297 * USB device was yanked out. Or it could happen to
4298 * be a transient write error and maybe the block will
4299 * be remapped. Nothing we can do but to retry the
4300 * write and hope for the best.
4302 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4303 "superblock detected");
4304 clear_buffer_write_io_error(sbh);
4305 set_buffer_uptodate(sbh);
4308 * If the file system is mounted read-only, don't update the
4309 * superblock write time. This avoids updating the superblock
4310 * write time when we are mounting the root file system
4311 * read/only but we need to replay the journal; at that point,
4312 * for people who are east of GMT and who make their clock
4313 * tick in localtime for Windows bug-for-bug compatibility,
4314 * the clock is set in the future, and this will cause e2fsck
4315 * to complain and force a full file system check.
4317 if (!(sb->s_flags & MS_RDONLY))
4318 es->s_wtime = cpu_to_le32(get_seconds());
4319 if (sb->s_bdev->bd_part)
4320 es->s_kbytes_written =
4321 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4322 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4323 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4325 es->s_kbytes_written =
4326 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4327 ext4_free_blocks_count_set(es,
4328 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4329 &EXT4_SB(sb)->s_freeclusters_counter)));
4330 es->s_free_inodes_count =
4331 cpu_to_le32(percpu_counter_sum_positive(
4332 &EXT4_SB(sb)->s_freeinodes_counter));
4334 BUFFER_TRACE(sbh, "marking dirty");
4335 mark_buffer_dirty(sbh);
4337 error = sync_dirty_buffer(sbh);
4341 error = buffer_write_io_error(sbh);
4343 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4345 clear_buffer_write_io_error(sbh);
4346 set_buffer_uptodate(sbh);
4353 * Have we just finished recovery? If so, and if we are mounting (or
4354 * remounting) the filesystem readonly, then we will end up with a
4355 * consistent fs on disk. Record that fact.
4357 static void ext4_mark_recovery_complete(struct super_block *sb,
4358 struct ext4_super_block *es)
4360 journal_t *journal = EXT4_SB(sb)->s_journal;
4362 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4363 BUG_ON(journal != NULL);
4366 jbd2_journal_lock_updates(journal);
4367 if (jbd2_journal_flush(journal) < 0)
4370 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4371 sb->s_flags & MS_RDONLY) {
4372 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4373 ext4_commit_super(sb, 1);
4377 jbd2_journal_unlock_updates(journal);
4381 * If we are mounting (or read-write remounting) a filesystem whose journal
4382 * has recorded an error from a previous lifetime, move that error to the
4383 * main filesystem now.
4385 static void ext4_clear_journal_err(struct super_block *sb,
4386 struct ext4_super_block *es)
4392 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4394 journal = EXT4_SB(sb)->s_journal;
4397 * Now check for any error status which may have been recorded in the
4398 * journal by a prior ext4_error() or ext4_abort()
4401 j_errno = jbd2_journal_errno(journal);
4405 errstr = ext4_decode_error(sb, j_errno, nbuf);
4406 ext4_warning(sb, "Filesystem error recorded "
4407 "from previous mount: %s", errstr);
4408 ext4_warning(sb, "Marking fs in need of filesystem check.");
4410 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4411 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4412 ext4_commit_super(sb, 1);
4414 jbd2_journal_clear_err(journal);
4419 * Force the running and committing transactions to commit,
4420 * and wait on the commit.
4422 int ext4_force_commit(struct super_block *sb)
4427 if (sb->s_flags & MS_RDONLY)
4430 journal = EXT4_SB(sb)->s_journal;
4432 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4433 ret = ext4_journal_force_commit(journal);
4439 static void ext4_write_super(struct super_block *sb)
4442 ext4_commit_super(sb, 1);
4446 static int ext4_sync_fs(struct super_block *sb, int wait)
4450 struct ext4_sb_info *sbi = EXT4_SB(sb);
4452 trace_ext4_sync_fs(sb, wait);
4453 flush_workqueue(sbi->dio_unwritten_wq);
4454 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4456 jbd2_log_wait_commit(sbi->s_journal, target);
4462 * LVM calls this function before a (read-only) snapshot is created. This
4463 * gives us a chance to flush the journal completely and mark the fs clean.
4465 * Note that only this function cannot bring a filesystem to be in a clean
4466 * state independently, because ext4 prevents a new handle from being started
4467 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4470 static int ext4_freeze(struct super_block *sb)
4475 if (sb->s_flags & MS_RDONLY)
4478 journal = EXT4_SB(sb)->s_journal;
4480 /* Now we set up the journal barrier. */
4481 jbd2_journal_lock_updates(journal);
4484 * Don't clear the needs_recovery flag if we failed to flush
4487 error = jbd2_journal_flush(journal);
4491 /* Journal blocked and flushed, clear needs_recovery flag. */
4492 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4493 error = ext4_commit_super(sb, 1);
4495 /* we rely on s_frozen to stop further updates */
4496 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4501 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4502 * flag here, even though the filesystem is not technically dirty yet.
4504 static int ext4_unfreeze(struct super_block *sb)
4506 if (sb->s_flags & MS_RDONLY)
4510 /* Reset the needs_recovery flag before the fs is unlocked. */
4511 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4512 ext4_commit_super(sb, 1);
4518 * Structure to save mount options for ext4_remount's benefit
4520 struct ext4_mount_options {
4521 unsigned long s_mount_opt;
4522 unsigned long s_mount_opt2;
4525 unsigned long s_commit_interval;
4526 u32 s_min_batch_time, s_max_batch_time;
4529 char *s_qf_names[MAXQUOTAS];
4533 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4535 struct ext4_super_block *es;
4536 struct ext4_sb_info *sbi = EXT4_SB(sb);
4537 ext4_fsblk_t n_blocks_count = 0;
4538 unsigned long old_sb_flags;
4539 struct ext4_mount_options old_opts;
4540 int enable_quota = 0;
4542 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4547 char *orig_data = kstrdup(data, GFP_KERNEL);
4549 /* Store the original options */
4551 old_sb_flags = sb->s_flags;
4552 old_opts.s_mount_opt = sbi->s_mount_opt;
4553 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4554 old_opts.s_resuid = sbi->s_resuid;
4555 old_opts.s_resgid = sbi->s_resgid;
4556 old_opts.s_commit_interval = sbi->s_commit_interval;
4557 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4558 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4560 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4561 for (i = 0; i < MAXQUOTAS; i++)
4562 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4564 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4565 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4568 * Allow the "check" option to be passed as a remount option.
4570 if (!parse_options(data, sb, NULL, &journal_ioprio,
4571 &n_blocks_count, 1)) {
4576 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4577 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4578 ext4_msg(sb, KERN_ERR, "can't mount with "
4579 "both data=journal and delalloc");
4583 if (test_opt(sb, DIOREAD_NOLOCK)) {
4584 ext4_msg(sb, KERN_ERR, "can't mount with "
4585 "both data=journal and dioread_nolock");
4591 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4592 ext4_abort(sb, "Abort forced by user");
4594 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4595 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4599 if (sbi->s_journal) {
4600 ext4_init_journal_params(sb, sbi->s_journal);
4601 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4604 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4605 n_blocks_count > ext4_blocks_count(es)) {
4606 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4611 if (*flags & MS_RDONLY) {
4612 err = dquot_suspend(sb, -1);
4617 * First of all, the unconditional stuff we have to do
4618 * to disable replay of the journal when we next remount
4620 sb->s_flags |= MS_RDONLY;
4623 * OK, test if we are remounting a valid rw partition
4624 * readonly, and if so set the rdonly flag and then
4625 * mark the partition as valid again.
4627 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4628 (sbi->s_mount_state & EXT4_VALID_FS))
4629 es->s_state = cpu_to_le16(sbi->s_mount_state);
4632 ext4_mark_recovery_complete(sb, es);
4634 /* Make sure we can mount this feature set readwrite */
4635 if (!ext4_feature_set_ok(sb, 0)) {
4640 * Make sure the group descriptor checksums
4641 * are sane. If they aren't, refuse to remount r/w.
4643 for (g = 0; g < sbi->s_groups_count; g++) {
4644 struct ext4_group_desc *gdp =
4645 ext4_get_group_desc(sb, g, NULL);
4647 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4648 ext4_msg(sb, KERN_ERR,
4649 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4650 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4651 le16_to_cpu(gdp->bg_checksum));
4658 * If we have an unprocessed orphan list hanging
4659 * around from a previously readonly bdev mount,
4660 * require a full umount/remount for now.
4662 if (es->s_last_orphan) {
4663 ext4_msg(sb, KERN_WARNING, "Couldn't "
4664 "remount RDWR because of unprocessed "
4665 "orphan inode list. Please "
4666 "umount/remount instead");
4672 * Mounting a RDONLY partition read-write, so reread
4673 * and store the current valid flag. (It may have
4674 * been changed by e2fsck since we originally mounted
4678 ext4_clear_journal_err(sb, es);
4679 sbi->s_mount_state = le16_to_cpu(es->s_state);
4680 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4682 if (!ext4_setup_super(sb, es, 0))
4683 sb->s_flags &= ~MS_RDONLY;
4684 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4685 EXT4_FEATURE_INCOMPAT_MMP))
4686 if (ext4_multi_mount_protect(sb,
4687 le64_to_cpu(es->s_mmp_block))) {
4696 * Reinitialize lazy itable initialization thread based on
4699 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4700 ext4_unregister_li_request(sb);
4702 ext4_group_t first_not_zeroed;
4703 first_not_zeroed = ext4_has_uninit_itable(sb);
4704 ext4_register_li_request(sb, first_not_zeroed);
4707 ext4_setup_system_zone(sb);
4708 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4709 ext4_commit_super(sb, 1);
4712 /* Release old quota file names */
4713 for (i = 0; i < MAXQUOTAS; i++)
4714 if (old_opts.s_qf_names[i] &&
4715 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4716 kfree(old_opts.s_qf_names[i]);
4720 dquot_resume(sb, -1);
4722 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4727 sb->s_flags = old_sb_flags;
4728 sbi->s_mount_opt = old_opts.s_mount_opt;
4729 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4730 sbi->s_resuid = old_opts.s_resuid;
4731 sbi->s_resgid = old_opts.s_resgid;
4732 sbi->s_commit_interval = old_opts.s_commit_interval;
4733 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4734 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4736 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4737 for (i = 0; i < MAXQUOTAS; i++) {
4738 if (sbi->s_qf_names[i] &&
4739 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4740 kfree(sbi->s_qf_names[i]);
4741 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4749 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4751 struct super_block *sb = dentry->d_sb;
4752 struct ext4_sb_info *sbi = EXT4_SB(sb);
4753 struct ext4_super_block *es = sbi->s_es;
4754 ext4_fsblk_t overhead = 0;
4758 if (!test_opt(sb, MINIX_DF))
4759 overhead = sbi->s_overhead;
4761 buf->f_type = EXT4_SUPER_MAGIC;
4762 buf->f_bsize = sb->s_blocksize;
4763 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4764 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4765 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4766 /* prevent underflow in case that few free space is available */
4767 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4768 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4769 if (buf->f_bfree < ext4_r_blocks_count(es))
4771 buf->f_files = le32_to_cpu(es->s_inodes_count);
4772 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4773 buf->f_namelen = EXT4_NAME_LEN;
4774 fsid = le64_to_cpup((void *)es->s_uuid) ^
4775 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4776 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4777 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4782 /* Helper function for writing quotas on sync - we need to start transaction
4783 * before quota file is locked for write. Otherwise the are possible deadlocks:
4784 * Process 1 Process 2
4785 * ext4_create() quota_sync()
4786 * jbd2_journal_start() write_dquot()
4787 * dquot_initialize() down(dqio_mutex)
4788 * down(dqio_mutex) jbd2_journal_start()
4794 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4796 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4799 static int ext4_write_dquot(struct dquot *dquot)
4803 struct inode *inode;
4805 inode = dquot_to_inode(dquot);
4806 handle = ext4_journal_start(inode,
4807 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4809 return PTR_ERR(handle);
4810 ret = dquot_commit(dquot);
4811 err = ext4_journal_stop(handle);
4817 static int ext4_acquire_dquot(struct dquot *dquot)
4822 handle = ext4_journal_start(dquot_to_inode(dquot),
4823 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4825 return PTR_ERR(handle);
4826 ret = dquot_acquire(dquot);
4827 err = ext4_journal_stop(handle);
4833 static int ext4_release_dquot(struct dquot *dquot)
4838 handle = ext4_journal_start(dquot_to_inode(dquot),
4839 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4840 if (IS_ERR(handle)) {
4841 /* Release dquot anyway to avoid endless cycle in dqput() */
4842 dquot_release(dquot);
4843 return PTR_ERR(handle);
4845 ret = dquot_release(dquot);
4846 err = ext4_journal_stop(handle);
4852 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4854 /* Are we journaling quotas? */
4855 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4856 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4857 dquot_mark_dquot_dirty(dquot);
4858 return ext4_write_dquot(dquot);
4860 return dquot_mark_dquot_dirty(dquot);
4864 static int ext4_write_info(struct super_block *sb, int type)
4869 /* Data block + inode block */
4870 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4872 return PTR_ERR(handle);
4873 ret = dquot_commit_info(sb, type);
4874 err = ext4_journal_stop(handle);
4881 * Turn on quotas during mount time - we need to find
4882 * the quota file and such...
4884 static int ext4_quota_on_mount(struct super_block *sb, int type)
4886 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4887 EXT4_SB(sb)->s_jquota_fmt, type);
4891 * Standard function to be called on quota_on
4893 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4898 if (!test_opt(sb, QUOTA))
4901 /* Quotafile not on the same filesystem? */
4902 if (path->mnt->mnt_sb != sb)
4904 /* Journaling quota? */
4905 if (EXT4_SB(sb)->s_qf_names[type]) {
4906 /* Quotafile not in fs root? */
4907 if (path->dentry->d_parent != sb->s_root)
4908 ext4_msg(sb, KERN_WARNING,
4909 "Quota file not on filesystem root. "
4910 "Journaled quota will not work");
4914 * When we journal data on quota file, we have to flush journal to see
4915 * all updates to the file when we bypass pagecache...
4917 if (EXT4_SB(sb)->s_journal &&
4918 ext4_should_journal_data(path->dentry->d_inode)) {
4920 * We don't need to lock updates but journal_flush() could
4921 * otherwise be livelocked...
4923 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4924 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4925 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4930 return dquot_quota_on(sb, type, format_id, path);
4933 static int ext4_quota_off(struct super_block *sb, int type)
4935 struct inode *inode = sb_dqopt(sb)->files[type];
4938 /* Force all delayed allocation blocks to be allocated.
4939 * Caller already holds s_umount sem */
4940 if (test_opt(sb, DELALLOC))
4941 sync_filesystem(sb);
4946 /* Update modification times of quota files when userspace can
4947 * start looking at them */
4948 handle = ext4_journal_start(inode, 1);
4951 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4952 ext4_mark_inode_dirty(handle, inode);
4953 ext4_journal_stop(handle);
4956 return dquot_quota_off(sb, type);
4959 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4960 * acquiring the locks... As quota files are never truncated and quota code
4961 * itself serializes the operations (and no one else should touch the files)
4962 * we don't have to be afraid of races */
4963 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4964 size_t len, loff_t off)
4966 struct inode *inode = sb_dqopt(sb)->files[type];
4967 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4969 int offset = off & (sb->s_blocksize - 1);
4972 struct buffer_head *bh;
4973 loff_t i_size = i_size_read(inode);
4977 if (off+len > i_size)
4980 while (toread > 0) {
4981 tocopy = sb->s_blocksize - offset < toread ?
4982 sb->s_blocksize - offset : toread;
4983 bh = ext4_bread(NULL, inode, blk, 0, &err);
4986 if (!bh) /* A hole? */
4987 memset(data, 0, tocopy);
4989 memcpy(data, bh->b_data+offset, tocopy);
4999 /* Write to quotafile (we know the transaction is already started and has
5000 * enough credits) */
5001 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5002 const char *data, size_t len, loff_t off)
5004 struct inode *inode = sb_dqopt(sb)->files[type];
5005 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5007 int offset = off & (sb->s_blocksize - 1);
5008 struct buffer_head *bh;
5009 handle_t *handle = journal_current_handle();
5011 if (EXT4_SB(sb)->s_journal && !handle) {
5012 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5013 " cancelled because transaction is not started",
5014 (unsigned long long)off, (unsigned long long)len);
5018 * Since we account only one data block in transaction credits,
5019 * then it is impossible to cross a block boundary.
5021 if (sb->s_blocksize - offset < len) {
5022 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5023 " cancelled because not block aligned",
5024 (unsigned long long)off, (unsigned long long)len);
5028 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
5029 bh = ext4_bread(handle, inode, blk, 1, &err);
5032 err = ext4_journal_get_write_access(handle, bh);
5038 memcpy(bh->b_data+offset, data, len);
5039 flush_dcache_page(bh->b_page);
5041 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5045 mutex_unlock(&inode->i_mutex);
5048 if (inode->i_size < off + len) {
5049 i_size_write(inode, off + len);
5050 EXT4_I(inode)->i_disksize = inode->i_size;
5051 ext4_mark_inode_dirty(handle, inode);
5053 mutex_unlock(&inode->i_mutex);
5059 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5060 const char *dev_name, void *data)
5062 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5065 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5066 static inline void register_as_ext2(void)
5068 int err = register_filesystem(&ext2_fs_type);
5071 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5074 static inline void unregister_as_ext2(void)
5076 unregister_filesystem(&ext2_fs_type);
5079 static inline int ext2_feature_set_ok(struct super_block *sb)
5081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5083 if (sb->s_flags & MS_RDONLY)
5085 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5089 MODULE_ALIAS("ext2");
5091 static inline void register_as_ext2(void) { }
5092 static inline void unregister_as_ext2(void) { }
5093 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5096 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5097 static inline void register_as_ext3(void)
5099 int err = register_filesystem(&ext3_fs_type);
5102 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5105 static inline void unregister_as_ext3(void)
5107 unregister_filesystem(&ext3_fs_type);
5110 static inline int ext3_feature_set_ok(struct super_block *sb)
5112 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5114 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5116 if (sb->s_flags & MS_RDONLY)
5118 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5122 MODULE_ALIAS("ext3");
5124 static inline void register_as_ext3(void) { }
5125 static inline void unregister_as_ext3(void) { }
5126 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5129 static struct file_system_type ext4_fs_type = {
5130 .owner = THIS_MODULE,
5132 .mount = ext4_mount,
5133 .kill_sb = kill_block_super,
5134 .fs_flags = FS_REQUIRES_DEV,
5137 static int __init ext4_init_feat_adverts(void)
5139 struct ext4_features *ef;
5142 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5146 ef->f_kobj.kset = ext4_kset;
5147 init_completion(&ef->f_kobj_unregister);
5148 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5161 static void ext4_exit_feat_adverts(void)
5163 kobject_put(&ext4_feat->f_kobj);
5164 wait_for_completion(&ext4_feat->f_kobj_unregister);
5168 /* Shared across all ext4 file systems */
5169 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5170 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5172 static int __init ext4_init_fs(void)
5176 ext4_check_flag_values();
5178 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5179 mutex_init(&ext4__aio_mutex[i]);
5180 init_waitqueue_head(&ext4__ioend_wq[i]);
5183 err = ext4_init_pageio();
5186 err = ext4_init_system_zone();
5189 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5192 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5194 err = ext4_init_feat_adverts();
5198 err = ext4_init_mballoc();
5202 err = ext4_init_xattr();
5205 err = init_inodecache();
5210 err = register_filesystem(&ext4_fs_type);
5214 ext4_li_info = NULL;
5215 mutex_init(&ext4_li_mtx);
5218 unregister_as_ext2();
5219 unregister_as_ext3();
5220 destroy_inodecache();
5224 ext4_exit_mballoc();
5226 ext4_exit_feat_adverts();
5229 remove_proc_entry("fs/ext4", NULL);
5230 kset_unregister(ext4_kset);
5232 ext4_exit_system_zone();
5238 static void __exit ext4_exit_fs(void)
5240 ext4_destroy_lazyinit_thread();
5241 unregister_as_ext2();
5242 unregister_as_ext3();
5243 unregister_filesystem(&ext4_fs_type);
5244 destroy_inodecache();
5246 ext4_exit_mballoc();
5247 ext4_exit_feat_adverts();
5248 remove_proc_entry("fs/ext4", NULL);
5249 kset_unregister(ext4_kset);
5250 ext4_exit_system_zone();
5254 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5255 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5256 MODULE_LICENSE("GPL");
5257 module_init(ext4_init_fs)
5258 module_exit(ext4_exit_fs)