2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 void *ext4_kvmalloc(size_t size, gfp_t flags)
118 ret = kmalloc(size, flags);
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
124 void *ext4_kvzalloc(size_t size, gfp_t flags)
128 ret = kzalloc(size, flags);
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
134 void ext4_kvfree(void *ptr)
136 if (is_vmalloc_addr(ptr))
143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
159 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
162 return le32_to_cpu(bg->bg_inode_table_lo) |
163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
167 __u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
175 __u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
183 __u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
191 __u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
199 void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
207 void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
215 void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
223 void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
231 void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
239 void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
247 void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
256 /* Just increment the non-pointer handle value */
257 static handle_t *ext4_get_nojournal(void)
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
265 handle = (handle_t *)ref_cnt;
267 current->journal_info = handle;
272 /* Decrement the non-pointer handle value */
273 static void ext4_put_nojournal(handle_t *handle)
275 unsigned long ref_cnt = (unsigned long)handle;
277 BUG_ON(ref_cnt == 0);
280 handle = (handle_t *)ref_cnt;
282 current->journal_info = handle;
286 * Wrappers for jbd2_journal_start/end.
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
306 journal = EXT4_SB(sb)->s_journal;
307 handle = ext4_journal_current_handle();
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
320 return ext4_get_nojournal();
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
330 return jbd2_journal_start(journal, nblocks);
334 * The only special thing we need to do here is to make sure that all
335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
336 * that sync() will call the filesystem's write_super callback if
339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
341 struct super_block *sb;
345 if (!ext4_handle_valid(handle)) {
346 ext4_put_nojournal(handle);
349 sb = handle->h_transaction->t_journal->j_private;
351 rc = jbd2_journal_stop(handle);
356 __ext4_std_error(sb, where, line, err);
360 void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
367 BUG_ON(!ext4_handle_valid(handle));
370 BUFFER_TRACE(bh, "abort");
375 if (is_handle_aborted(handle))
378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
381 jbd2_journal_abort_handle(handle);
384 static void __save_error_info(struct super_block *sb, const char *func,
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
403 * Start the daily error reporting function if it hasn't been
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
411 static void save_error_info(struct super_block *sb, const char *func,
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
426 static int block_device_ejected(struct super_block *sb)
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431 return bdi->dev == NULL;
435 /* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
438 * On ext2, we can store the error state of the filesystem in the
439 * superblock. That is not possible on ext4, because we may have other
440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
445 * We'll just use the jbd2_journal_abort() error code to record an error in
446 * the journal instead. On recovery, the journal will complain about
447 * that error until we've noted it down and cleared it.
450 static void ext4_handle_error(struct super_block *sb)
452 if (sb->s_flags & MS_RDONLY)
455 if (!test_opt(sb, ERRORS_CONT)) {
456 journal_t *journal = EXT4_SB(sb)->s_journal;
458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
460 jbd2_journal_abort(journal, -EIO);
462 if (test_opt(sb, ERRORS_RO)) {
463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
464 sb->s_flags |= MS_RDONLY;
466 if (test_opt(sb, ERRORS_PANIC))
467 panic("EXT4-fs (device %s): panic forced after error\n",
471 void __ext4_error(struct super_block *sb, const char *function,
472 unsigned int line, const char *fmt, ...)
474 struct va_format vaf;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
483 save_error_info(sb, function, line);
485 ext4_handle_error(sb);
488 void ext4_error_inode(struct inode *inode, const char *function,
489 unsigned int line, ext4_fsblk_t block,
490 const char *fmt, ...)
493 struct va_format vaf;
494 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
496 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
497 es->s_last_error_block = cpu_to_le64(block);
498 save_error_info(inode->i_sb, function, line);
502 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
503 inode->i_sb->s_id, function, line, inode->i_ino);
505 printk(KERN_CONT "block %llu: ", block);
506 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
509 ext4_handle_error(inode->i_sb);
512 void ext4_error_file(struct file *file, const char *function,
513 unsigned int line, ext4_fsblk_t block,
514 const char *fmt, ...)
517 struct va_format vaf;
518 struct ext4_super_block *es;
519 struct inode *inode = file->f_dentry->d_inode;
520 char pathname[80], *path;
522 es = EXT4_SB(inode->i_sb)->s_es;
523 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
524 save_error_info(inode->i_sb, function, line);
525 path = d_path(&(file->f_path), pathname, sizeof(pathname));
529 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
530 inode->i_sb->s_id, function, line, inode->i_ino);
532 printk(KERN_CONT "block %llu: ", block);
536 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
539 ext4_handle_error(inode->i_sb);
542 static const char *ext4_decode_error(struct super_block *sb, int errno,
549 errstr = "IO failure";
552 errstr = "Out of memory";
555 if (!sb || (EXT4_SB(sb)->s_journal &&
556 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
557 errstr = "Journal has aborted";
559 errstr = "Readonly filesystem";
562 /* If the caller passed in an extra buffer for unknown
563 * errors, textualise them now. Else we just return
566 /* Check for truncated error codes... */
567 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
576 /* __ext4_std_error decodes expected errors from journaling functions
577 * automatically and invokes the appropriate error response. */
579 void __ext4_std_error(struct super_block *sb, const char *function,
580 unsigned int line, int errno)
585 /* Special case: if the error is EROFS, and we're not already
586 * inside a transaction, then there's really no point in logging
588 if (errno == -EROFS && journal_current_handle() == NULL &&
589 (sb->s_flags & MS_RDONLY))
592 errstr = ext4_decode_error(sb, errno, nbuf);
593 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
594 sb->s_id, function, line, errstr);
595 save_error_info(sb, function, line);
597 ext4_handle_error(sb);
601 * ext4_abort is a much stronger failure handler than ext4_error. The
602 * abort function may be used to deal with unrecoverable failures such
603 * as journal IO errors or ENOMEM at a critical moment in log management.
605 * We unconditionally force the filesystem into an ABORT|READONLY state,
606 * unless the error response on the fs has been set to panic in which
607 * case we take the easy way out and panic immediately.
610 void __ext4_abort(struct super_block *sb, const char *function,
611 unsigned int line, const char *fmt, ...)
615 save_error_info(sb, function, line);
617 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
623 if ((sb->s_flags & MS_RDONLY) == 0) {
624 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
625 sb->s_flags |= MS_RDONLY;
626 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
627 if (EXT4_SB(sb)->s_journal)
628 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
629 save_error_info(sb, function, line);
631 if (test_opt(sb, ERRORS_PANIC))
632 panic("EXT4-fs panic from previous error\n");
635 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
637 struct va_format vaf;
643 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
647 void __ext4_warning(struct super_block *sb, const char *function,
648 unsigned int line, const char *fmt, ...)
650 struct va_format vaf;
656 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
657 sb->s_id, function, line, &vaf);
661 void __ext4_grp_locked_error(const char *function, unsigned int line,
662 struct super_block *sb, ext4_group_t grp,
663 unsigned long ino, ext4_fsblk_t block,
664 const char *fmt, ...)
668 struct va_format vaf;
670 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
672 es->s_last_error_ino = cpu_to_le32(ino);
673 es->s_last_error_block = cpu_to_le64(block);
674 __save_error_info(sb, function, line);
680 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
681 sb->s_id, function, line, grp);
683 printk(KERN_CONT "inode %lu: ", ino);
685 printk(KERN_CONT "block %llu:", (unsigned long long) block);
686 printk(KERN_CONT "%pV\n", &vaf);
689 if (test_opt(sb, ERRORS_CONT)) {
690 ext4_commit_super(sb, 0);
694 ext4_unlock_group(sb, grp);
695 ext4_handle_error(sb);
697 * We only get here in the ERRORS_RO case; relocking the group
698 * may be dangerous, but nothing bad will happen since the
699 * filesystem will have already been marked read/only and the
700 * journal has been aborted. We return 1 as a hint to callers
701 * who might what to use the return value from
702 * ext4_grp_locked_error() to distinguish between the
703 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
704 * aggressively from the ext4 function in question, with a
705 * more appropriate error code.
707 ext4_lock_group(sb, grp);
711 void ext4_update_dynamic_rev(struct super_block *sb)
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
715 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
719 "updating to rev %d because of new feature flag, "
720 "running e2fsck is recommended",
723 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
724 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
725 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
726 /* leave es->s_feature_*compat flags alone */
727 /* es->s_uuid will be set by e2fsck if empty */
730 * The rest of the superblock fields should be zero, and if not it
731 * means they are likely already in use, so leave them alone. We
732 * can leave it up to e2fsck to clean up any inconsistencies there.
737 * Open the external journal device
739 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
741 struct block_device *bdev;
742 char b[BDEVNAME_SIZE];
744 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
750 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
751 __bdevname(dev, b), PTR_ERR(bdev));
756 * Release the journal device
758 static int ext4_blkdev_put(struct block_device *bdev)
760 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
763 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
765 struct block_device *bdev;
768 bdev = sbi->journal_bdev;
770 ret = ext4_blkdev_put(bdev);
771 sbi->journal_bdev = NULL;
776 static inline struct inode *orphan_list_entry(struct list_head *l)
778 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
781 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
785 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
786 le32_to_cpu(sbi->s_es->s_last_orphan));
788 printk(KERN_ERR "sb_info orphan list:\n");
789 list_for_each(l, &sbi->s_orphan) {
790 struct inode *inode = orphan_list_entry(l);
792 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
793 inode->i_sb->s_id, inode->i_ino, inode,
794 inode->i_mode, inode->i_nlink,
799 static void ext4_put_super(struct super_block *sb)
801 struct ext4_sb_info *sbi = EXT4_SB(sb);
802 struct ext4_super_block *es = sbi->s_es;
805 ext4_unregister_li_request(sb);
806 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
808 flush_workqueue(sbi->dio_unwritten_wq);
809 destroy_workqueue(sbi->dio_unwritten_wq);
813 ext4_commit_super(sb, 1);
815 if (sbi->s_journal) {
816 err = jbd2_journal_destroy(sbi->s_journal);
817 sbi->s_journal = NULL;
819 ext4_abort(sb, "Couldn't clean up the journal");
822 del_timer(&sbi->s_err_report);
823 ext4_release_system_zone(sb);
825 ext4_ext_release(sb);
826 ext4_xattr_put_super(sb);
828 if (!(sb->s_flags & MS_RDONLY)) {
829 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
830 es->s_state = cpu_to_le16(sbi->s_mount_state);
831 ext4_commit_super(sb, 1);
834 remove_proc_entry(sb->s_id, ext4_proc_root);
836 kobject_del(&sbi->s_kobj);
838 for (i = 0; i < sbi->s_gdb_count; i++)
839 brelse(sbi->s_group_desc[i]);
840 ext4_kvfree(sbi->s_group_desc);
841 ext4_kvfree(sbi->s_flex_groups);
842 percpu_counter_destroy(&sbi->s_freeclusters_counter);
843 percpu_counter_destroy(&sbi->s_freeinodes_counter);
844 percpu_counter_destroy(&sbi->s_dirs_counter);
845 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
848 for (i = 0; i < MAXQUOTAS; i++)
849 kfree(sbi->s_qf_names[i]);
852 /* Debugging code just in case the in-memory inode orphan list
853 * isn't empty. The on-disk one can be non-empty if we've
854 * detected an error and taken the fs readonly, but the
855 * in-memory list had better be clean by this point. */
856 if (!list_empty(&sbi->s_orphan))
857 dump_orphan_list(sb, sbi);
858 J_ASSERT(list_empty(&sbi->s_orphan));
860 invalidate_bdev(sb->s_bdev);
861 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
863 * Invalidate the journal device's buffers. We don't want them
864 * floating about in memory - the physical journal device may
865 * hotswapped, and it breaks the `ro-after' testing code.
867 sync_blockdev(sbi->journal_bdev);
868 invalidate_bdev(sbi->journal_bdev);
869 ext4_blkdev_remove(sbi);
872 kthread_stop(sbi->s_mmp_tsk);
873 sb->s_fs_info = NULL;
875 * Now that we are completely done shutting down the
876 * superblock, we need to actually destroy the kobject.
879 kobject_put(&sbi->s_kobj);
880 wait_for_completion(&sbi->s_kobj_unregister);
881 kfree(sbi->s_blockgroup_lock);
885 static struct kmem_cache *ext4_inode_cachep;
888 * Called inside transaction, so use GFP_NOFS
890 static struct inode *ext4_alloc_inode(struct super_block *sb)
892 struct ext4_inode_info *ei;
894 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
898 ei->vfs_inode.i_version = 1;
899 ei->vfs_inode.i_data.writeback_index = 0;
900 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
901 INIT_LIST_HEAD(&ei->i_prealloc_list);
902 spin_lock_init(&ei->i_prealloc_lock);
903 ei->i_reserved_data_blocks = 0;
904 ei->i_reserved_meta_blocks = 0;
905 ei->i_allocated_meta_blocks = 0;
906 ei->i_da_metadata_calc_len = 0;
907 spin_lock_init(&(ei->i_block_reservation_lock));
909 ei->i_reserved_quota = 0;
912 INIT_LIST_HEAD(&ei->i_completed_io_list);
913 spin_lock_init(&ei->i_completed_io_lock);
914 ei->cur_aio_dio = NULL;
916 ei->i_datasync_tid = 0;
917 atomic_set(&ei->i_ioend_count, 0);
918 atomic_set(&ei->i_aiodio_unwritten, 0);
920 return &ei->vfs_inode;
923 static int ext4_drop_inode(struct inode *inode)
925 int drop = generic_drop_inode(inode);
927 trace_ext4_drop_inode(inode, drop);
931 static void ext4_i_callback(struct rcu_head *head)
933 struct inode *inode = container_of(head, struct inode, i_rcu);
934 INIT_LIST_HEAD(&inode->i_dentry);
935 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
938 static void ext4_destroy_inode(struct inode *inode)
940 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
941 ext4_msg(inode->i_sb, KERN_ERR,
942 "Inode %lu (%p): orphan list check failed!",
943 inode->i_ino, EXT4_I(inode));
944 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
945 EXT4_I(inode), sizeof(struct ext4_inode_info),
949 call_rcu(&inode->i_rcu, ext4_i_callback);
952 static void init_once(void *foo)
954 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
956 INIT_LIST_HEAD(&ei->i_orphan);
957 #ifdef CONFIG_EXT4_FS_XATTR
958 init_rwsem(&ei->xattr_sem);
960 init_rwsem(&ei->i_data_sem);
961 inode_init_once(&ei->vfs_inode);
964 static int init_inodecache(void)
966 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
967 sizeof(struct ext4_inode_info),
968 0, (SLAB_RECLAIM_ACCOUNT|
971 if (ext4_inode_cachep == NULL)
976 static void destroy_inodecache(void)
978 kmem_cache_destroy(ext4_inode_cachep);
981 void ext4_clear_inode(struct inode *inode)
983 invalidate_inode_buffers(inode);
984 end_writeback(inode);
986 ext4_discard_preallocations(inode);
987 if (EXT4_I(inode)->jinode) {
988 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
989 EXT4_I(inode)->jinode);
990 jbd2_free_inode(EXT4_I(inode)->jinode);
991 EXT4_I(inode)->jinode = NULL;
995 static inline void ext4_show_quota_options(struct seq_file *seq,
996 struct super_block *sb)
998 #if defined(CONFIG_QUOTA)
999 struct ext4_sb_info *sbi = EXT4_SB(sb);
1001 if (sbi->s_jquota_fmt) {
1004 switch (sbi->s_jquota_fmt) {
1015 seq_printf(seq, ",jqfmt=%s", fmtname);
1018 if (sbi->s_qf_names[USRQUOTA])
1019 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1021 if (sbi->s_qf_names[GRPQUOTA])
1022 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1024 if (test_opt(sb, USRQUOTA))
1025 seq_puts(seq, ",usrquota");
1027 if (test_opt(sb, GRPQUOTA))
1028 seq_puts(seq, ",grpquota");
1034 * - it's set to a non-default value OR
1035 * - if the per-sb default is different from the global default
1037 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1040 unsigned long def_mount_opts;
1041 struct super_block *sb = vfs->mnt_sb;
1042 struct ext4_sb_info *sbi = EXT4_SB(sb);
1043 struct ext4_super_block *es = sbi->s_es;
1045 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1046 def_errors = le16_to_cpu(es->s_errors);
1048 if (sbi->s_sb_block != 1)
1049 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1050 if (test_opt(sb, MINIX_DF))
1051 seq_puts(seq, ",minixdf");
1052 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1053 seq_puts(seq, ",grpid");
1054 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1055 seq_puts(seq, ",nogrpid");
1056 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1057 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1058 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1060 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1061 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1062 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1064 if (test_opt(sb, ERRORS_RO)) {
1065 if (def_errors == EXT4_ERRORS_PANIC ||
1066 def_errors == EXT4_ERRORS_CONTINUE) {
1067 seq_puts(seq, ",errors=remount-ro");
1070 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1071 seq_puts(seq, ",errors=continue");
1072 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1073 seq_puts(seq, ",errors=panic");
1074 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1075 seq_puts(seq, ",nouid32");
1076 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1077 seq_puts(seq, ",debug");
1078 #ifdef CONFIG_EXT4_FS_XATTR
1079 if (test_opt(sb, XATTR_USER))
1080 seq_puts(seq, ",user_xattr");
1081 if (!test_opt(sb, XATTR_USER))
1082 seq_puts(seq, ",nouser_xattr");
1084 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1085 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1086 seq_puts(seq, ",acl");
1087 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1088 seq_puts(seq, ",noacl");
1090 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1091 seq_printf(seq, ",commit=%u",
1092 (unsigned) (sbi->s_commit_interval / HZ));
1094 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1095 seq_printf(seq, ",min_batch_time=%u",
1096 (unsigned) sbi->s_min_batch_time);
1098 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1099 seq_printf(seq, ",max_batch_time=%u",
1100 (unsigned) sbi->s_max_batch_time);
1104 * We're changing the default of barrier mount option, so
1105 * let's always display its mount state so it's clear what its
1108 seq_puts(seq, ",barrier=");
1109 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1110 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1111 seq_puts(seq, ",journal_async_commit");
1112 else if (test_opt(sb, JOURNAL_CHECKSUM))
1113 seq_puts(seq, ",journal_checksum");
1114 if (test_opt(sb, I_VERSION))
1115 seq_puts(seq, ",i_version");
1116 if (!test_opt(sb, DELALLOC) &&
1117 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1118 seq_puts(seq, ",nodelalloc");
1120 if (!test_opt(sb, MBLK_IO_SUBMIT))
1121 seq_puts(seq, ",nomblk_io_submit");
1123 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1125 * journal mode get enabled in different ways
1126 * So just print the value even if we didn't specify it
1128 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1129 seq_puts(seq, ",data=journal");
1130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1131 seq_puts(seq, ",data=ordered");
1132 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1133 seq_puts(seq, ",data=writeback");
1135 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1136 seq_printf(seq, ",inode_readahead_blks=%u",
1137 sbi->s_inode_readahead_blks);
1139 if (test_opt(sb, DATA_ERR_ABORT))
1140 seq_puts(seq, ",data_err=abort");
1142 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1143 seq_puts(seq, ",noauto_da_alloc");
1145 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1146 seq_puts(seq, ",discard");
1148 if (test_opt(sb, NOLOAD))
1149 seq_puts(seq, ",norecovery");
1151 if (test_opt(sb, DIOREAD_NOLOCK))
1152 seq_puts(seq, ",dioread_nolock");
1154 if (test_opt(sb, BLOCK_VALIDITY) &&
1155 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1156 seq_puts(seq, ",block_validity");
1158 if (!test_opt(sb, INIT_INODE_TABLE))
1159 seq_puts(seq, ",noinit_itable");
1160 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1161 seq_printf(seq, ",init_itable=%u",
1162 (unsigned) sbi->s_li_wait_mult);
1164 ext4_show_quota_options(seq, sb);
1169 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1170 u64 ino, u32 generation)
1172 struct inode *inode;
1174 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1175 return ERR_PTR(-ESTALE);
1176 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1177 return ERR_PTR(-ESTALE);
1179 /* iget isn't really right if the inode is currently unallocated!!
1181 * ext4_read_inode will return a bad_inode if the inode had been
1182 * deleted, so we should be safe.
1184 * Currently we don't know the generation for parent directory, so
1185 * a generation of 0 means "accept any"
1187 inode = ext4_iget(sb, ino);
1189 return ERR_CAST(inode);
1190 if (generation && inode->i_generation != generation) {
1192 return ERR_PTR(-ESTALE);
1198 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1199 int fh_len, int fh_type)
1201 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1202 ext4_nfs_get_inode);
1205 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1206 int fh_len, int fh_type)
1208 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1209 ext4_nfs_get_inode);
1213 * Try to release metadata pages (indirect blocks, directories) which are
1214 * mapped via the block device. Since these pages could have journal heads
1215 * which would prevent try_to_free_buffers() from freeing them, we must use
1216 * jbd2 layer's try_to_free_buffers() function to release them.
1218 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1221 journal_t *journal = EXT4_SB(sb)->s_journal;
1223 WARN_ON(PageChecked(page));
1224 if (!page_has_buffers(page))
1227 return jbd2_journal_try_to_free_buffers(journal, page,
1228 wait & ~__GFP_WAIT);
1229 return try_to_free_buffers(page);
1233 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1234 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1236 static int ext4_write_dquot(struct dquot *dquot);
1237 static int ext4_acquire_dquot(struct dquot *dquot);
1238 static int ext4_release_dquot(struct dquot *dquot);
1239 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1240 static int ext4_write_info(struct super_block *sb, int type);
1241 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1243 static int ext4_quota_off(struct super_block *sb, int type);
1244 static int ext4_quota_on_mount(struct super_block *sb, int type);
1245 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1246 size_t len, loff_t off);
1247 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1248 const char *data, size_t len, loff_t off);
1250 static const struct dquot_operations ext4_quota_operations = {
1251 .get_reserved_space = ext4_get_reserved_space,
1252 .write_dquot = ext4_write_dquot,
1253 .acquire_dquot = ext4_acquire_dquot,
1254 .release_dquot = ext4_release_dquot,
1255 .mark_dirty = ext4_mark_dquot_dirty,
1256 .write_info = ext4_write_info,
1257 .alloc_dquot = dquot_alloc,
1258 .destroy_dquot = dquot_destroy,
1261 static const struct quotactl_ops ext4_qctl_operations = {
1262 .quota_on = ext4_quota_on,
1263 .quota_off = ext4_quota_off,
1264 .quota_sync = dquot_quota_sync,
1265 .get_info = dquot_get_dqinfo,
1266 .set_info = dquot_set_dqinfo,
1267 .get_dqblk = dquot_get_dqblk,
1268 .set_dqblk = dquot_set_dqblk
1272 static const struct super_operations ext4_sops = {
1273 .alloc_inode = ext4_alloc_inode,
1274 .destroy_inode = ext4_destroy_inode,
1275 .write_inode = ext4_write_inode,
1276 .dirty_inode = ext4_dirty_inode,
1277 .drop_inode = ext4_drop_inode,
1278 .evict_inode = ext4_evict_inode,
1279 .put_super = ext4_put_super,
1280 .sync_fs = ext4_sync_fs,
1281 .freeze_fs = ext4_freeze,
1282 .unfreeze_fs = ext4_unfreeze,
1283 .statfs = ext4_statfs,
1284 .remount_fs = ext4_remount,
1285 .show_options = ext4_show_options,
1287 .quota_read = ext4_quota_read,
1288 .quota_write = ext4_quota_write,
1290 .bdev_try_to_free_page = bdev_try_to_free_page,
1293 static const struct super_operations ext4_nojournal_sops = {
1294 .alloc_inode = ext4_alloc_inode,
1295 .destroy_inode = ext4_destroy_inode,
1296 .write_inode = ext4_write_inode,
1297 .dirty_inode = ext4_dirty_inode,
1298 .drop_inode = ext4_drop_inode,
1299 .evict_inode = ext4_evict_inode,
1300 .write_super = ext4_write_super,
1301 .put_super = ext4_put_super,
1302 .statfs = ext4_statfs,
1303 .remount_fs = ext4_remount,
1304 .show_options = ext4_show_options,
1306 .quota_read = ext4_quota_read,
1307 .quota_write = ext4_quota_write,
1309 .bdev_try_to_free_page = bdev_try_to_free_page,
1312 static const struct export_operations ext4_export_ops = {
1313 .fh_to_dentry = ext4_fh_to_dentry,
1314 .fh_to_parent = ext4_fh_to_parent,
1315 .get_parent = ext4_get_parent,
1319 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1320 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1321 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1322 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1323 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1324 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1325 Opt_journal_update, Opt_journal_dev,
1326 Opt_journal_checksum, Opt_journal_async_commit,
1327 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1328 Opt_data_err_abort, Opt_data_err_ignore,
1329 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1330 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1331 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1332 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1333 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1334 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1335 Opt_inode_readahead_blks, Opt_journal_ioprio,
1336 Opt_dioread_nolock, Opt_dioread_lock,
1337 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1340 static const match_table_t tokens = {
1341 {Opt_bsd_df, "bsddf"},
1342 {Opt_minix_df, "minixdf"},
1343 {Opt_grpid, "grpid"},
1344 {Opt_grpid, "bsdgroups"},
1345 {Opt_nogrpid, "nogrpid"},
1346 {Opt_nogrpid, "sysvgroups"},
1347 {Opt_resgid, "resgid=%u"},
1348 {Opt_resuid, "resuid=%u"},
1350 {Opt_err_cont, "errors=continue"},
1351 {Opt_err_panic, "errors=panic"},
1352 {Opt_err_ro, "errors=remount-ro"},
1353 {Opt_nouid32, "nouid32"},
1354 {Opt_debug, "debug"},
1355 {Opt_oldalloc, "oldalloc"},
1356 {Opt_orlov, "orlov"},
1357 {Opt_user_xattr, "user_xattr"},
1358 {Opt_nouser_xattr, "nouser_xattr"},
1360 {Opt_noacl, "noacl"},
1361 {Opt_noload, "noload"},
1362 {Opt_noload, "norecovery"},
1365 {Opt_commit, "commit=%u"},
1366 {Opt_min_batch_time, "min_batch_time=%u"},
1367 {Opt_max_batch_time, "max_batch_time=%u"},
1368 {Opt_journal_update, "journal=update"},
1369 {Opt_journal_dev, "journal_dev=%u"},
1370 {Opt_journal_checksum, "journal_checksum"},
1371 {Opt_journal_async_commit, "journal_async_commit"},
1372 {Opt_abort, "abort"},
1373 {Opt_data_journal, "data=journal"},
1374 {Opt_data_ordered, "data=ordered"},
1375 {Opt_data_writeback, "data=writeback"},
1376 {Opt_data_err_abort, "data_err=abort"},
1377 {Opt_data_err_ignore, "data_err=ignore"},
1378 {Opt_offusrjquota, "usrjquota="},
1379 {Opt_usrjquota, "usrjquota=%s"},
1380 {Opt_offgrpjquota, "grpjquota="},
1381 {Opt_grpjquota, "grpjquota=%s"},
1382 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1383 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1384 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1385 {Opt_grpquota, "grpquota"},
1386 {Opt_noquota, "noquota"},
1387 {Opt_quota, "quota"},
1388 {Opt_usrquota, "usrquota"},
1389 {Opt_barrier, "barrier=%u"},
1390 {Opt_barrier, "barrier"},
1391 {Opt_nobarrier, "nobarrier"},
1392 {Opt_i_version, "i_version"},
1393 {Opt_stripe, "stripe=%u"},
1394 {Opt_resize, "resize"},
1395 {Opt_delalloc, "delalloc"},
1396 {Opt_nodelalloc, "nodelalloc"},
1397 {Opt_mblk_io_submit, "mblk_io_submit"},
1398 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1399 {Opt_block_validity, "block_validity"},
1400 {Opt_noblock_validity, "noblock_validity"},
1401 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1402 {Opt_journal_ioprio, "journal_ioprio=%u"},
1403 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1404 {Opt_auto_da_alloc, "auto_da_alloc"},
1405 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1406 {Opt_dioread_nolock, "dioread_nolock"},
1407 {Opt_dioread_lock, "dioread_lock"},
1408 {Opt_discard, "discard"},
1409 {Opt_nodiscard, "nodiscard"},
1410 {Opt_init_itable, "init_itable=%u"},
1411 {Opt_init_itable, "init_itable"},
1412 {Opt_noinit_itable, "noinit_itable"},
1416 static ext4_fsblk_t get_sb_block(void **data)
1418 ext4_fsblk_t sb_block;
1419 char *options = (char *) *data;
1421 if (!options || strncmp(options, "sb=", 3) != 0)
1422 return 1; /* Default location */
1425 /* TODO: use simple_strtoll with >32bit ext4 */
1426 sb_block = simple_strtoul(options, &options, 0);
1427 if (*options && *options != ',') {
1428 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1432 if (*options == ',')
1434 *data = (void *) options;
1439 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1440 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1441 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1444 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1449 if (sb_any_quota_loaded(sb) &&
1450 !sbi->s_qf_names[qtype]) {
1451 ext4_msg(sb, KERN_ERR,
1452 "Cannot change journaled "
1453 "quota options when quota turned on");
1456 qname = match_strdup(args);
1458 ext4_msg(sb, KERN_ERR,
1459 "Not enough memory for storing quotafile name");
1462 if (sbi->s_qf_names[qtype] &&
1463 strcmp(sbi->s_qf_names[qtype], qname)) {
1464 ext4_msg(sb, KERN_ERR,
1465 "%s quota file already specified", QTYPE2NAME(qtype));
1469 sbi->s_qf_names[qtype] = qname;
1470 if (strchr(sbi->s_qf_names[qtype], '/')) {
1471 ext4_msg(sb, KERN_ERR,
1472 "quotafile must be on filesystem root");
1473 kfree(sbi->s_qf_names[qtype]);
1474 sbi->s_qf_names[qtype] = NULL;
1481 static int clear_qf_name(struct super_block *sb, int qtype)
1484 struct ext4_sb_info *sbi = EXT4_SB(sb);
1486 if (sb_any_quota_loaded(sb) &&
1487 sbi->s_qf_names[qtype]) {
1488 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1489 " when quota turned on");
1493 * The space will be released later when all options are confirmed
1496 sbi->s_qf_names[qtype] = NULL;
1501 static int parse_options(char *options, struct super_block *sb,
1502 unsigned long *journal_devnum,
1503 unsigned int *journal_ioprio,
1504 ext4_fsblk_t *n_blocks_count, int is_remount)
1506 struct ext4_sb_info *sbi = EXT4_SB(sb);
1508 substring_t args[MAX_OPT_ARGS];
1518 while ((p = strsep(&options, ",")) != NULL) {
1524 * Initialize args struct so we know whether arg was
1525 * found; some options take optional arguments.
1527 args[0].to = args[0].from = NULL;
1528 token = match_token(p, tokens, args);
1531 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1532 clear_opt(sb, MINIX_DF);
1535 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1536 set_opt(sb, MINIX_DF);
1540 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1545 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1546 clear_opt(sb, GRPID);
1550 if (match_int(&args[0], &option))
1552 sbi->s_resuid = option;
1555 if (match_int(&args[0], &option))
1557 sbi->s_resgid = option;
1560 /* handled by get_sb_block() instead of here */
1561 /* *sb_block = match_int(&args[0]); */
1564 clear_opt(sb, ERRORS_CONT);
1565 clear_opt(sb, ERRORS_RO);
1566 set_opt(sb, ERRORS_PANIC);
1569 clear_opt(sb, ERRORS_CONT);
1570 clear_opt(sb, ERRORS_PANIC);
1571 set_opt(sb, ERRORS_RO);
1574 clear_opt(sb, ERRORS_RO);
1575 clear_opt(sb, ERRORS_PANIC);
1576 set_opt(sb, ERRORS_CONT);
1579 set_opt(sb, NO_UID32);
1585 ext4_msg(sb, KERN_WARNING,
1586 "Ignoring deprecated oldalloc option");
1589 ext4_msg(sb, KERN_WARNING,
1590 "Ignoring deprecated orlov option");
1592 #ifdef CONFIG_EXT4_FS_XATTR
1593 case Opt_user_xattr:
1594 set_opt(sb, XATTR_USER);
1596 case Opt_nouser_xattr:
1597 clear_opt(sb, XATTR_USER);
1600 case Opt_user_xattr:
1601 case Opt_nouser_xattr:
1602 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1605 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1607 set_opt(sb, POSIX_ACL);
1610 clear_opt(sb, POSIX_ACL);
1615 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1618 case Opt_journal_update:
1620 /* Eventually we will want to be able to create
1621 a journal file here. For now, only allow the
1622 user to specify an existing inode to be the
1625 ext4_msg(sb, KERN_ERR,
1626 "Cannot specify journal on remount");
1629 set_opt(sb, UPDATE_JOURNAL);
1631 case Opt_journal_dev:
1633 ext4_msg(sb, KERN_ERR,
1634 "Cannot specify journal on remount");
1637 if (match_int(&args[0], &option))
1639 *journal_devnum = option;
1641 case Opt_journal_checksum:
1642 set_opt(sb, JOURNAL_CHECKSUM);
1644 case Opt_journal_async_commit:
1645 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1646 set_opt(sb, JOURNAL_CHECKSUM);
1649 set_opt(sb, NOLOAD);
1652 if (match_int(&args[0], &option))
1657 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1658 sbi->s_commit_interval = HZ * option;
1660 case Opt_max_batch_time:
1661 if (match_int(&args[0], &option))
1666 option = EXT4_DEF_MAX_BATCH_TIME;
1667 sbi->s_max_batch_time = option;
1669 case Opt_min_batch_time:
1670 if (match_int(&args[0], &option))
1674 sbi->s_min_batch_time = option;
1676 case Opt_data_journal:
1677 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1679 case Opt_data_ordered:
1680 data_opt = EXT4_MOUNT_ORDERED_DATA;
1682 case Opt_data_writeback:
1683 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1686 if (!sbi->s_journal)
1687 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1688 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1689 ext4_msg(sb, KERN_ERR,
1690 "Cannot change data mode on remount");
1694 clear_opt(sb, DATA_FLAGS);
1695 sbi->s_mount_opt |= data_opt;
1698 case Opt_data_err_abort:
1699 set_opt(sb, DATA_ERR_ABORT);
1701 case Opt_data_err_ignore:
1702 clear_opt(sb, DATA_ERR_ABORT);
1706 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1710 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1713 case Opt_offusrjquota:
1714 if (!clear_qf_name(sb, USRQUOTA))
1717 case Opt_offgrpjquota:
1718 if (!clear_qf_name(sb, GRPQUOTA))
1722 case Opt_jqfmt_vfsold:
1723 qfmt = QFMT_VFS_OLD;
1725 case Opt_jqfmt_vfsv0:
1728 case Opt_jqfmt_vfsv1:
1731 if (sb_any_quota_loaded(sb) &&
1732 sbi->s_jquota_fmt != qfmt) {
1733 ext4_msg(sb, KERN_ERR, "Cannot change "
1734 "journaled quota options when "
1738 sbi->s_jquota_fmt = qfmt;
1743 set_opt(sb, USRQUOTA);
1747 set_opt(sb, GRPQUOTA);
1750 if (sb_any_quota_loaded(sb)) {
1751 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1752 "options when quota turned on");
1755 clear_opt(sb, QUOTA);
1756 clear_opt(sb, USRQUOTA);
1757 clear_opt(sb, GRPQUOTA);
1763 ext4_msg(sb, KERN_ERR,
1764 "quota options not supported");
1768 case Opt_offusrjquota:
1769 case Opt_offgrpjquota:
1770 case Opt_jqfmt_vfsold:
1771 case Opt_jqfmt_vfsv0:
1772 case Opt_jqfmt_vfsv1:
1773 ext4_msg(sb, KERN_ERR,
1774 "journaled quota options not supported");
1780 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1783 clear_opt(sb, BARRIER);
1787 if (match_int(&args[0], &option))
1790 option = 1; /* No argument, default to 1 */
1792 set_opt(sb, BARRIER);
1794 clear_opt(sb, BARRIER);
1800 ext4_msg(sb, KERN_ERR,
1801 "resize option only available "
1805 if (match_int(&args[0], &option) != 0)
1807 *n_blocks_count = option;
1810 ext4_msg(sb, KERN_WARNING,
1811 "Ignoring deprecated nobh option");
1814 ext4_msg(sb, KERN_WARNING,
1815 "Ignoring deprecated bh option");
1818 set_opt(sb, I_VERSION);
1819 sb->s_flags |= MS_I_VERSION;
1821 case Opt_nodelalloc:
1822 clear_opt(sb, DELALLOC);
1823 clear_opt2(sb, EXPLICIT_DELALLOC);
1825 case Opt_mblk_io_submit:
1826 set_opt(sb, MBLK_IO_SUBMIT);
1828 case Opt_nomblk_io_submit:
1829 clear_opt(sb, MBLK_IO_SUBMIT);
1832 if (match_int(&args[0], &option))
1836 sbi->s_stripe = option;
1839 set_opt(sb, DELALLOC);
1840 set_opt2(sb, EXPLICIT_DELALLOC);
1842 case Opt_block_validity:
1843 set_opt(sb, BLOCK_VALIDITY);
1845 case Opt_noblock_validity:
1846 clear_opt(sb, BLOCK_VALIDITY);
1848 case Opt_inode_readahead_blks:
1849 if (match_int(&args[0], &option))
1851 if (option < 0 || option > (1 << 30))
1853 if (option && !is_power_of_2(option)) {
1854 ext4_msg(sb, KERN_ERR,
1855 "EXT4-fs: inode_readahead_blks"
1856 " must be a power of 2");
1859 sbi->s_inode_readahead_blks = option;
1861 case Opt_journal_ioprio:
1862 if (match_int(&args[0], &option))
1864 if (option < 0 || option > 7)
1866 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1869 case Opt_noauto_da_alloc:
1870 set_opt(sb, NO_AUTO_DA_ALLOC);
1872 case Opt_auto_da_alloc:
1874 if (match_int(&args[0], &option))
1877 option = 1; /* No argument, default to 1 */
1879 clear_opt(sb, NO_AUTO_DA_ALLOC);
1881 set_opt(sb,NO_AUTO_DA_ALLOC);
1884 set_opt(sb, DISCARD);
1887 clear_opt(sb, DISCARD);
1889 case Opt_dioread_nolock:
1890 set_opt(sb, DIOREAD_NOLOCK);
1892 case Opt_dioread_lock:
1893 clear_opt(sb, DIOREAD_NOLOCK);
1895 case Opt_init_itable:
1896 set_opt(sb, INIT_INODE_TABLE);
1898 if (match_int(&args[0], &option))
1901 option = EXT4_DEF_LI_WAIT_MULT;
1904 sbi->s_li_wait_mult = option;
1906 case Opt_noinit_itable:
1907 clear_opt(sb, INIT_INODE_TABLE);
1910 ext4_msg(sb, KERN_ERR,
1911 "Unrecognized mount option \"%s\" "
1912 "or missing value", p);
1917 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1918 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1919 clear_opt(sb, USRQUOTA);
1921 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1922 clear_opt(sb, GRPQUOTA);
1924 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1925 ext4_msg(sb, KERN_ERR, "old and new quota "
1930 if (!sbi->s_jquota_fmt) {
1931 ext4_msg(sb, KERN_ERR, "journaled quota format "
1936 if (sbi->s_jquota_fmt) {
1937 ext4_msg(sb, KERN_ERR, "journaled quota format "
1938 "specified with no journaling "
1947 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1953 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1954 ext4_msg(sb, KERN_ERR, "revision level too high, "
1955 "forcing read-only mode");
1960 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1961 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1962 "running e2fsck is recommended");
1963 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1964 ext4_msg(sb, KERN_WARNING,
1965 "warning: mounting fs with errors, "
1966 "running e2fsck is recommended");
1967 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1968 le16_to_cpu(es->s_mnt_count) >=
1969 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1970 ext4_msg(sb, KERN_WARNING,
1971 "warning: maximal mount count reached, "
1972 "running e2fsck is recommended");
1973 else if (le32_to_cpu(es->s_checkinterval) &&
1974 (le32_to_cpu(es->s_lastcheck) +
1975 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1976 ext4_msg(sb, KERN_WARNING,
1977 "warning: checktime reached, "
1978 "running e2fsck is recommended");
1979 if (!sbi->s_journal)
1980 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1981 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1982 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1983 le16_add_cpu(&es->s_mnt_count, 1);
1984 es->s_mtime = cpu_to_le32(get_seconds());
1985 ext4_update_dynamic_rev(sb);
1987 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1989 ext4_commit_super(sb, 1);
1991 if (test_opt(sb, DEBUG))
1992 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1993 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1995 sbi->s_groups_count,
1996 EXT4_BLOCKS_PER_GROUP(sb),
1997 EXT4_INODES_PER_GROUP(sb),
1998 sbi->s_mount_opt, sbi->s_mount_opt2);
2000 cleancache_init_fs(sb);
2004 static int ext4_fill_flex_info(struct super_block *sb)
2006 struct ext4_sb_info *sbi = EXT4_SB(sb);
2007 struct ext4_group_desc *gdp = NULL;
2008 ext4_group_t flex_group_count;
2009 ext4_group_t flex_group;
2010 unsigned int groups_per_flex = 0;
2014 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2015 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2016 sbi->s_log_groups_per_flex = 0;
2019 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2021 /* We allocate both existing and potentially added groups */
2022 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2023 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2024 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2025 size = flex_group_count * sizeof(struct flex_groups);
2026 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2027 if (sbi->s_flex_groups == NULL) {
2028 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2033 for (i = 0; i < sbi->s_groups_count; i++) {
2034 gdp = ext4_get_group_desc(sb, i, NULL);
2036 flex_group = ext4_flex_group(sbi, i);
2037 atomic_add(ext4_free_inodes_count(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_inodes);
2039 atomic_add(ext4_free_group_clusters(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].free_clusters);
2041 atomic_add(ext4_used_dirs_count(sb, gdp),
2042 &sbi->s_flex_groups[flex_group].used_dirs);
2050 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2055 if (sbi->s_es->s_feature_ro_compat &
2056 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2057 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2058 __le32 le_group = cpu_to_le32(block_group);
2060 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2061 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2062 crc = crc16(crc, (__u8 *)gdp, offset);
2063 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2064 /* for checksum of struct ext4_group_desc do the rest...*/
2065 if ((sbi->s_es->s_feature_incompat &
2066 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2067 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2068 crc = crc16(crc, (__u8 *)gdp + offset,
2069 le16_to_cpu(sbi->s_es->s_desc_size) -
2073 return cpu_to_le16(crc);
2076 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2077 struct ext4_group_desc *gdp)
2079 if ((sbi->s_es->s_feature_ro_compat &
2080 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2081 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2087 /* Called at mount-time, super-block is locked */
2088 static int ext4_check_descriptors(struct super_block *sb,
2089 ext4_group_t *first_not_zeroed)
2091 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2093 ext4_fsblk_t last_block;
2094 ext4_fsblk_t block_bitmap;
2095 ext4_fsblk_t inode_bitmap;
2096 ext4_fsblk_t inode_table;
2097 int flexbg_flag = 0;
2098 ext4_group_t i, grp = sbi->s_groups_count;
2100 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2103 ext4_debug("Checking group descriptors");
2105 for (i = 0; i < sbi->s_groups_count; i++) {
2106 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2108 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2109 last_block = ext4_blocks_count(sbi->s_es) - 1;
2111 last_block = first_block +
2112 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2114 if ((grp == sbi->s_groups_count) &&
2115 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2118 block_bitmap = ext4_block_bitmap(sb, gdp);
2119 if (block_bitmap < first_block || block_bitmap > last_block) {
2120 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2121 "Block bitmap for group %u not in group "
2122 "(block %llu)!", i, block_bitmap);
2125 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2126 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2127 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2128 "Inode bitmap for group %u not in group "
2129 "(block %llu)!", i, inode_bitmap);
2132 inode_table = ext4_inode_table(sb, gdp);
2133 if (inode_table < first_block ||
2134 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2135 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2136 "Inode table for group %u not in group "
2137 "(block %llu)!", i, inode_table);
2140 ext4_lock_group(sb, i);
2141 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2142 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2143 "Checksum for group %u failed (%u!=%u)",
2144 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2145 gdp)), le16_to_cpu(gdp->bg_checksum));
2146 if (!(sb->s_flags & MS_RDONLY)) {
2147 ext4_unlock_group(sb, i);
2151 ext4_unlock_group(sb, i);
2153 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2155 if (NULL != first_not_zeroed)
2156 *first_not_zeroed = grp;
2158 ext4_free_blocks_count_set(sbi->s_es,
2159 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2160 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2164 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2165 * the superblock) which were deleted from all directories, but held open by
2166 * a process at the time of a crash. We walk the list and try to delete these
2167 * inodes at recovery time (only with a read-write filesystem).
2169 * In order to keep the orphan inode chain consistent during traversal (in
2170 * case of crash during recovery), we link each inode into the superblock
2171 * orphan list_head and handle it the same way as an inode deletion during
2172 * normal operation (which journals the operations for us).
2174 * We only do an iget() and an iput() on each inode, which is very safe if we
2175 * accidentally point at an in-use or already deleted inode. The worst that
2176 * can happen in this case is that we get a "bit already cleared" message from
2177 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2178 * e2fsck was run on this filesystem, and it must have already done the orphan
2179 * inode cleanup for us, so we can safely abort without any further action.
2181 static void ext4_orphan_cleanup(struct super_block *sb,
2182 struct ext4_super_block *es)
2184 unsigned int s_flags = sb->s_flags;
2185 int nr_orphans = 0, nr_truncates = 0;
2189 if (!es->s_last_orphan) {
2190 jbd_debug(4, "no orphan inodes to clean up\n");
2194 if (bdev_read_only(sb->s_bdev)) {
2195 ext4_msg(sb, KERN_ERR, "write access "
2196 "unavailable, skipping orphan cleanup");
2200 /* Check if feature set would not allow a r/w mount */
2201 if (!ext4_feature_set_ok(sb, 0)) {
2202 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2203 "unknown ROCOMPAT features");
2207 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2208 if (es->s_last_orphan)
2209 jbd_debug(1, "Errors on filesystem, "
2210 "clearing orphan list.\n");
2211 es->s_last_orphan = 0;
2212 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2216 if (s_flags & MS_RDONLY) {
2217 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2218 sb->s_flags &= ~MS_RDONLY;
2221 /* Needed for iput() to work correctly and not trash data */
2222 sb->s_flags |= MS_ACTIVE;
2223 /* Turn on quotas so that they are updated correctly */
2224 for (i = 0; i < MAXQUOTAS; i++) {
2225 if (EXT4_SB(sb)->s_qf_names[i]) {
2226 int ret = ext4_quota_on_mount(sb, i);
2228 ext4_msg(sb, KERN_ERR,
2229 "Cannot turn on journaled "
2230 "quota: error %d", ret);
2235 while (es->s_last_orphan) {
2236 struct inode *inode;
2238 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2239 if (IS_ERR(inode)) {
2240 es->s_last_orphan = 0;
2244 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2245 dquot_initialize(inode);
2246 if (inode->i_nlink) {
2247 ext4_msg(sb, KERN_DEBUG,
2248 "%s: truncating inode %lu to %lld bytes",
2249 __func__, inode->i_ino, inode->i_size);
2250 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2251 inode->i_ino, inode->i_size);
2252 ext4_truncate(inode);
2255 ext4_msg(sb, KERN_DEBUG,
2256 "%s: deleting unreferenced inode %lu",
2257 __func__, inode->i_ino);
2258 jbd_debug(2, "deleting unreferenced inode %lu\n",
2262 iput(inode); /* The delete magic happens here! */
2265 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2268 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2269 PLURAL(nr_orphans));
2271 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2272 PLURAL(nr_truncates));
2274 /* Turn quotas off */
2275 for (i = 0; i < MAXQUOTAS; i++) {
2276 if (sb_dqopt(sb)->files[i])
2277 dquot_quota_off(sb, i);
2280 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2284 * Maximal extent format file size.
2285 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2286 * extent format containers, within a sector_t, and within i_blocks
2287 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2288 * so that won't be a limiting factor.
2290 * However there is other limiting factor. We do store extents in the form
2291 * of starting block and length, hence the resulting length of the extent
2292 * covering maximum file size must fit into on-disk format containers as
2293 * well. Given that length is always by 1 unit bigger than max unit (because
2294 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2296 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2298 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2301 loff_t upper_limit = MAX_LFS_FILESIZE;
2303 /* small i_blocks in vfs inode? */
2304 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2306 * CONFIG_LBDAF is not enabled implies the inode
2307 * i_block represent total blocks in 512 bytes
2308 * 32 == size of vfs inode i_blocks * 8
2310 upper_limit = (1LL << 32) - 1;
2312 /* total blocks in file system block size */
2313 upper_limit >>= (blkbits - 9);
2314 upper_limit <<= blkbits;
2318 * 32-bit extent-start container, ee_block. We lower the maxbytes
2319 * by one fs block, so ee_len can cover the extent of maximum file
2322 res = (1LL << 32) - 1;
2325 /* Sanity check against vm- & vfs- imposed limits */
2326 if (res > upper_limit)
2333 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2334 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2335 * We need to be 1 filesystem block less than the 2^48 sector limit.
2337 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2339 loff_t res = EXT4_NDIR_BLOCKS;
2342 /* This is calculated to be the largest file size for a dense, block
2343 * mapped file such that the file's total number of 512-byte sectors,
2344 * including data and all indirect blocks, does not exceed (2^48 - 1).
2346 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2347 * number of 512-byte sectors of the file.
2350 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2352 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2353 * the inode i_block field represents total file blocks in
2354 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2356 upper_limit = (1LL << 32) - 1;
2358 /* total blocks in file system block size */
2359 upper_limit >>= (bits - 9);
2363 * We use 48 bit ext4_inode i_blocks
2364 * With EXT4_HUGE_FILE_FL set the i_blocks
2365 * represent total number of blocks in
2366 * file system block size
2368 upper_limit = (1LL << 48) - 1;
2372 /* indirect blocks */
2374 /* double indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2));
2376 /* tripple indirect blocks */
2377 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2379 upper_limit -= meta_blocks;
2380 upper_limit <<= bits;
2382 res += 1LL << (bits-2);
2383 res += 1LL << (2*(bits-2));
2384 res += 1LL << (3*(bits-2));
2386 if (res > upper_limit)
2389 if (res > MAX_LFS_FILESIZE)
2390 res = MAX_LFS_FILESIZE;
2395 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2396 ext4_fsblk_t logical_sb_block, int nr)
2398 struct ext4_sb_info *sbi = EXT4_SB(sb);
2399 ext4_group_t bg, first_meta_bg;
2402 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2404 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2406 return logical_sb_block + nr + 1;
2407 bg = sbi->s_desc_per_block * nr;
2408 if (ext4_bg_has_super(sb, bg))
2411 return (has_super + ext4_group_first_block_no(sb, bg));
2415 * ext4_get_stripe_size: Get the stripe size.
2416 * @sbi: In memory super block info
2418 * If we have specified it via mount option, then
2419 * use the mount option value. If the value specified at mount time is
2420 * greater than the blocks per group use the super block value.
2421 * If the super block value is greater than blocks per group return 0.
2422 * Allocator needs it be less than blocks per group.
2425 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2427 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2428 unsigned long stripe_width =
2429 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2432 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2433 ret = sbi->s_stripe;
2434 else if (stripe_width <= sbi->s_blocks_per_group)
2436 else if (stride <= sbi->s_blocks_per_group)
2442 * If the stripe width is 1, this makes no sense and
2443 * we set it to 0 to turn off stripe handling code.
2454 struct attribute attr;
2455 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2456 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2457 const char *, size_t);
2461 static int parse_strtoul(const char *buf,
2462 unsigned long max, unsigned long *value)
2466 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2467 endp = skip_spaces(endp);
2468 if (*endp || *value > max)
2474 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi,
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2480 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2483 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2484 struct ext4_sb_info *sbi, char *buf)
2486 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2488 if (!sb->s_bdev->bd_part)
2489 return snprintf(buf, PAGE_SIZE, "0\n");
2490 return snprintf(buf, PAGE_SIZE, "%lu\n",
2491 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2492 sbi->s_sectors_written_start) >> 1);
2495 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2498 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2500 if (!sb->s_bdev->bd_part)
2501 return snprintf(buf, PAGE_SIZE, "0\n");
2502 return snprintf(buf, PAGE_SIZE, "%llu\n",
2503 (unsigned long long)(sbi->s_kbytes_written +
2504 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2505 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2508 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2509 struct ext4_sb_info *sbi,
2510 const char *buf, size_t count)
2514 if (parse_strtoul(buf, 0x40000000, &t))
2517 if (t && !is_power_of_2(t))
2520 sbi->s_inode_readahead_blks = t;
2524 static ssize_t sbi_ui_show(struct ext4_attr *a,
2525 struct ext4_sb_info *sbi, char *buf)
2527 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2529 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2532 static ssize_t sbi_ui_store(struct ext4_attr *a,
2533 struct ext4_sb_info *sbi,
2534 const char *buf, size_t count)
2536 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2539 if (parse_strtoul(buf, 0xffffffff, &t))
2545 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2546 static struct ext4_attr ext4_attr_##_name = { \
2547 .attr = {.name = __stringify(_name), .mode = _mode }, \
2550 .offset = offsetof(struct ext4_sb_info, _elname), \
2552 #define EXT4_ATTR(name, mode, show, store) \
2553 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2555 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2556 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2557 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2558 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2559 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2560 #define ATTR_LIST(name) &ext4_attr_##name.attr
2562 EXT4_RO_ATTR(delayed_allocation_blocks);
2563 EXT4_RO_ATTR(session_write_kbytes);
2564 EXT4_RO_ATTR(lifetime_write_kbytes);
2565 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2566 inode_readahead_blks_store, s_inode_readahead_blks);
2567 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2568 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2569 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2570 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2571 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2572 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2573 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2574 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2576 static struct attribute *ext4_attrs[] = {
2577 ATTR_LIST(delayed_allocation_blocks),
2578 ATTR_LIST(session_write_kbytes),
2579 ATTR_LIST(lifetime_write_kbytes),
2580 ATTR_LIST(inode_readahead_blks),
2581 ATTR_LIST(inode_goal),
2582 ATTR_LIST(mb_stats),
2583 ATTR_LIST(mb_max_to_scan),
2584 ATTR_LIST(mb_min_to_scan),
2585 ATTR_LIST(mb_order2_req),
2586 ATTR_LIST(mb_stream_req),
2587 ATTR_LIST(mb_group_prealloc),
2588 ATTR_LIST(max_writeback_mb_bump),
2592 /* Features this copy of ext4 supports */
2593 EXT4_INFO_ATTR(lazy_itable_init);
2594 EXT4_INFO_ATTR(batched_discard);
2596 static struct attribute *ext4_feat_attrs[] = {
2597 ATTR_LIST(lazy_itable_init),
2598 ATTR_LIST(batched_discard),
2602 static ssize_t ext4_attr_show(struct kobject *kobj,
2603 struct attribute *attr, char *buf)
2605 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2607 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2609 return a->show ? a->show(a, sbi, buf) : 0;
2612 static ssize_t ext4_attr_store(struct kobject *kobj,
2613 struct attribute *attr,
2614 const char *buf, size_t len)
2616 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2618 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2620 return a->store ? a->store(a, sbi, buf, len) : 0;
2623 static void ext4_sb_release(struct kobject *kobj)
2625 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2627 complete(&sbi->s_kobj_unregister);
2630 static const struct sysfs_ops ext4_attr_ops = {
2631 .show = ext4_attr_show,
2632 .store = ext4_attr_store,
2635 static struct kobj_type ext4_ktype = {
2636 .default_attrs = ext4_attrs,
2637 .sysfs_ops = &ext4_attr_ops,
2638 .release = ext4_sb_release,
2641 static void ext4_feat_release(struct kobject *kobj)
2643 complete(&ext4_feat->f_kobj_unregister);
2646 static struct kobj_type ext4_feat_ktype = {
2647 .default_attrs = ext4_feat_attrs,
2648 .sysfs_ops = &ext4_attr_ops,
2649 .release = ext4_feat_release,
2653 * Check whether this filesystem can be mounted based on
2654 * the features present and the RDONLY/RDWR mount requested.
2655 * Returns 1 if this filesystem can be mounted as requested,
2656 * 0 if it cannot be.
2658 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2660 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2661 ext4_msg(sb, KERN_ERR,
2662 "Couldn't mount because of "
2663 "unsupported optional features (%x)",
2664 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2665 ~EXT4_FEATURE_INCOMPAT_SUPP));
2672 /* Check that feature set is OK for a read-write mount */
2673 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2674 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2675 "unsupported optional features (%x)",
2676 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2677 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2681 * Large file size enabled file system can only be mounted
2682 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2684 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2685 if (sizeof(blkcnt_t) < sizeof(u64)) {
2686 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2687 "cannot be mounted RDWR without "
2692 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2693 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2694 ext4_msg(sb, KERN_ERR,
2695 "Can't support bigalloc feature without "
2696 "extents feature\n");
2703 * This function is called once a day if we have errors logged
2704 * on the file system
2706 static void print_daily_error_info(unsigned long arg)
2708 struct super_block *sb = (struct super_block *) arg;
2709 struct ext4_sb_info *sbi;
2710 struct ext4_super_block *es;
2715 if (es->s_error_count)
2716 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2717 le32_to_cpu(es->s_error_count));
2718 if (es->s_first_error_time) {
2719 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2720 sb->s_id, le32_to_cpu(es->s_first_error_time),
2721 (int) sizeof(es->s_first_error_func),
2722 es->s_first_error_func,
2723 le32_to_cpu(es->s_first_error_line));
2724 if (es->s_first_error_ino)
2725 printk(": inode %u",
2726 le32_to_cpu(es->s_first_error_ino));
2727 if (es->s_first_error_block)
2728 printk(": block %llu", (unsigned long long)
2729 le64_to_cpu(es->s_first_error_block));
2732 if (es->s_last_error_time) {
2733 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2734 sb->s_id, le32_to_cpu(es->s_last_error_time),
2735 (int) sizeof(es->s_last_error_func),
2736 es->s_last_error_func,
2737 le32_to_cpu(es->s_last_error_line));
2738 if (es->s_last_error_ino)
2739 printk(": inode %u",
2740 le32_to_cpu(es->s_last_error_ino));
2741 if (es->s_last_error_block)
2742 printk(": block %llu", (unsigned long long)
2743 le64_to_cpu(es->s_last_error_block));
2746 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2749 /* Find next suitable group and run ext4_init_inode_table */
2750 static int ext4_run_li_request(struct ext4_li_request *elr)
2752 struct ext4_group_desc *gdp = NULL;
2753 ext4_group_t group, ngroups;
2754 struct super_block *sb;
2755 unsigned long timeout = 0;
2759 ngroups = EXT4_SB(sb)->s_groups_count;
2761 for (group = elr->lr_next_group; group < ngroups; group++) {
2762 gdp = ext4_get_group_desc(sb, group, NULL);
2768 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2772 if (group == ngroups)
2777 ret = ext4_init_inode_table(sb, group,
2778 elr->lr_timeout ? 0 : 1);
2779 if (elr->lr_timeout == 0) {
2780 timeout = (jiffies - timeout) *
2781 elr->lr_sbi->s_li_wait_mult;
2782 elr->lr_timeout = timeout;
2784 elr->lr_next_sched = jiffies + elr->lr_timeout;
2785 elr->lr_next_group = group + 1;
2792 * Remove lr_request from the list_request and free the
2793 * request structure. Should be called with li_list_mtx held
2795 static void ext4_remove_li_request(struct ext4_li_request *elr)
2797 struct ext4_sb_info *sbi;
2804 list_del(&elr->lr_request);
2805 sbi->s_li_request = NULL;
2809 static void ext4_unregister_li_request(struct super_block *sb)
2811 mutex_lock(&ext4_li_mtx);
2812 if (!ext4_li_info) {
2813 mutex_unlock(&ext4_li_mtx);
2817 mutex_lock(&ext4_li_info->li_list_mtx);
2818 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2819 mutex_unlock(&ext4_li_info->li_list_mtx);
2820 mutex_unlock(&ext4_li_mtx);
2823 static struct task_struct *ext4_lazyinit_task;
2826 * This is the function where ext4lazyinit thread lives. It walks
2827 * through the request list searching for next scheduled filesystem.
2828 * When such a fs is found, run the lazy initialization request
2829 * (ext4_rn_li_request) and keep track of the time spend in this
2830 * function. Based on that time we compute next schedule time of
2831 * the request. When walking through the list is complete, compute
2832 * next waking time and put itself into sleep.
2834 static int ext4_lazyinit_thread(void *arg)
2836 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2837 struct list_head *pos, *n;
2838 struct ext4_li_request *elr;
2839 unsigned long next_wakeup, cur;
2841 BUG_ON(NULL == eli);
2845 next_wakeup = MAX_JIFFY_OFFSET;
2847 mutex_lock(&eli->li_list_mtx);
2848 if (list_empty(&eli->li_request_list)) {
2849 mutex_unlock(&eli->li_list_mtx);
2853 list_for_each_safe(pos, n, &eli->li_request_list) {
2854 elr = list_entry(pos, struct ext4_li_request,
2857 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2858 if (ext4_run_li_request(elr) != 0) {
2859 /* error, remove the lazy_init job */
2860 ext4_remove_li_request(elr);
2865 if (time_before(elr->lr_next_sched, next_wakeup))
2866 next_wakeup = elr->lr_next_sched;
2868 mutex_unlock(&eli->li_list_mtx);
2870 if (freezing(current))
2874 if ((time_after_eq(cur, next_wakeup)) ||
2875 (MAX_JIFFY_OFFSET == next_wakeup)) {
2880 schedule_timeout_interruptible(next_wakeup - cur);
2882 if (kthread_should_stop()) {
2883 ext4_clear_request_list();
2890 * It looks like the request list is empty, but we need
2891 * to check it under the li_list_mtx lock, to prevent any
2892 * additions into it, and of course we should lock ext4_li_mtx
2893 * to atomically free the list and ext4_li_info, because at
2894 * this point another ext4 filesystem could be registering
2897 mutex_lock(&ext4_li_mtx);
2898 mutex_lock(&eli->li_list_mtx);
2899 if (!list_empty(&eli->li_request_list)) {
2900 mutex_unlock(&eli->li_list_mtx);
2901 mutex_unlock(&ext4_li_mtx);
2904 mutex_unlock(&eli->li_list_mtx);
2905 kfree(ext4_li_info);
2906 ext4_li_info = NULL;
2907 mutex_unlock(&ext4_li_mtx);
2912 static void ext4_clear_request_list(void)
2914 struct list_head *pos, *n;
2915 struct ext4_li_request *elr;
2917 mutex_lock(&ext4_li_info->li_list_mtx);
2918 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2919 elr = list_entry(pos, struct ext4_li_request,
2921 ext4_remove_li_request(elr);
2923 mutex_unlock(&ext4_li_info->li_list_mtx);
2926 static int ext4_run_lazyinit_thread(void)
2928 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2929 ext4_li_info, "ext4lazyinit");
2930 if (IS_ERR(ext4_lazyinit_task)) {
2931 int err = PTR_ERR(ext4_lazyinit_task);
2932 ext4_clear_request_list();
2933 kfree(ext4_li_info);
2934 ext4_li_info = NULL;
2935 printk(KERN_CRIT "EXT4: error %d creating inode table "
2936 "initialization thread\n",
2940 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2945 * Check whether it make sense to run itable init. thread or not.
2946 * If there is at least one uninitialized inode table, return
2947 * corresponding group number, else the loop goes through all
2948 * groups and return total number of groups.
2950 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2952 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2953 struct ext4_group_desc *gdp = NULL;
2955 for (group = 0; group < ngroups; group++) {
2956 gdp = ext4_get_group_desc(sb, group, NULL);
2960 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2967 static int ext4_li_info_new(void)
2969 struct ext4_lazy_init *eli = NULL;
2971 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2975 INIT_LIST_HEAD(&eli->li_request_list);
2976 mutex_init(&eli->li_list_mtx);
2978 eli->li_state |= EXT4_LAZYINIT_QUIT;
2985 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2988 struct ext4_sb_info *sbi = EXT4_SB(sb);
2989 struct ext4_li_request *elr;
2992 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2998 elr->lr_next_group = start;
3001 * Randomize first schedule time of the request to
3002 * spread the inode table initialization requests
3005 get_random_bytes(&rnd, sizeof(rnd));
3006 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3007 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3012 static int ext4_register_li_request(struct super_block *sb,
3013 ext4_group_t first_not_zeroed)
3015 struct ext4_sb_info *sbi = EXT4_SB(sb);
3016 struct ext4_li_request *elr;
3017 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3020 if (sbi->s_li_request != NULL) {
3022 * Reset timeout so it can be computed again, because
3023 * s_li_wait_mult might have changed.
3025 sbi->s_li_request->lr_timeout = 0;
3029 if (first_not_zeroed == ngroups ||
3030 (sb->s_flags & MS_RDONLY) ||
3031 !test_opt(sb, INIT_INODE_TABLE))
3034 elr = ext4_li_request_new(sb, first_not_zeroed);
3038 mutex_lock(&ext4_li_mtx);
3040 if (NULL == ext4_li_info) {
3041 ret = ext4_li_info_new();
3046 mutex_lock(&ext4_li_info->li_list_mtx);
3047 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3048 mutex_unlock(&ext4_li_info->li_list_mtx);
3050 sbi->s_li_request = elr;
3052 * set elr to NULL here since it has been inserted to
3053 * the request_list and the removal and free of it is
3054 * handled by ext4_clear_request_list from now on.
3058 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3059 ret = ext4_run_lazyinit_thread();
3064 mutex_unlock(&ext4_li_mtx);
3071 * We do not need to lock anything since this is called on
3074 static void ext4_destroy_lazyinit_thread(void)
3077 * If thread exited earlier
3078 * there's nothing to be done.
3080 if (!ext4_li_info || !ext4_lazyinit_task)
3083 kthread_stop(ext4_lazyinit_task);
3087 * Note: calculating the overhead so we can be compatible with
3088 * historical BSD practice is quite difficult in the face of
3089 * clusters/bigalloc. This is because multiple metadata blocks from
3090 * different block group can end up in the same allocation cluster.
3091 * Calculating the exact overhead in the face of clustered allocation
3092 * requires either O(all block bitmaps) in memory or O(number of block
3093 * groups**2) in time. We will still calculate the superblock for
3094 * older file systems --- and if we come across with a bigalloc file
3095 * system with zero in s_overhead_clusters the estimate will be close to
3096 * correct especially for very large cluster sizes --- but for newer
3097 * file systems, it's better to calculate this figure once at mkfs
3098 * time, and store it in the superblock. If the superblock value is
3099 * present (even for non-bigalloc file systems), we will use it.
3101 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3105 struct ext4_group_desc *gdp;
3106 ext4_fsblk_t first_block, last_block, b;
3107 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3108 int s, j, count = 0;
3110 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3111 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3112 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3113 for (i = 0; i < ngroups; i++) {
3114 gdp = ext4_get_group_desc(sb, i, NULL);
3115 b = ext4_block_bitmap(sb, gdp);
3116 if (b >= first_block && b <= last_block) {
3117 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3120 b = ext4_inode_bitmap(sb, gdp);
3121 if (b >= first_block && b <= last_block) {
3122 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3125 b = ext4_inode_table(sb, gdp);
3126 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3127 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3128 int c = EXT4_B2C(sbi, b - first_block);
3129 ext4_set_bit(c, buf);
3135 if (ext4_bg_has_super(sb, grp)) {
3136 ext4_set_bit(s++, buf);
3139 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3140 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3146 return EXT4_CLUSTERS_PER_GROUP(sb) -
3147 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3151 * Compute the overhead and stash it in sbi->s_overhead
3153 int ext4_calculate_overhead(struct super_block *sb)
3155 struct ext4_sb_info *sbi = EXT4_SB(sb);
3156 struct ext4_super_block *es = sbi->s_es;
3157 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3158 ext4_fsblk_t overhead = 0;
3159 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3161 memset(buf, 0, PAGE_SIZE);
3166 * Compute the overhead (FS structures). This is constant
3167 * for a given filesystem unless the number of block groups
3168 * changes so we cache the previous value until it does.
3172 * All of the blocks before first_data_block are overhead
3174 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3177 * Add the overhead found in each block group
3179 for (i = 0; i < ngroups; i++) {
3182 blks = count_overhead(sb, i, buf);
3185 memset(buf, 0, PAGE_SIZE);
3188 sbi->s_overhead = overhead;
3190 free_page((unsigned long) buf);
3194 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3196 char *orig_data = kstrdup(data, GFP_KERNEL);
3197 struct buffer_head *bh;
3198 struct ext4_super_block *es = NULL;
3199 struct ext4_sb_info *sbi;
3201 ext4_fsblk_t sb_block = get_sb_block(&data);
3202 ext4_fsblk_t logical_sb_block;
3203 unsigned long offset = 0;
3204 unsigned long journal_devnum = 0;
3205 unsigned long def_mount_opts;
3210 int blocksize, clustersize;
3211 unsigned int db_count;
3213 int needs_recovery, has_huge_files, has_bigalloc;
3216 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3217 ext4_group_t first_not_zeroed;
3219 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3223 sbi->s_blockgroup_lock =
3224 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3225 if (!sbi->s_blockgroup_lock) {
3229 sb->s_fs_info = sbi;
3230 sbi->s_mount_opt = 0;
3231 sbi->s_resuid = EXT4_DEF_RESUID;
3232 sbi->s_resgid = EXT4_DEF_RESGID;
3233 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3234 sbi->s_sb_block = sb_block;
3235 if (sb->s_bdev->bd_part)
3236 sbi->s_sectors_written_start =
3237 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3239 /* Cleanup superblock name */
3240 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3244 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3246 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3251 * The ext4 superblock will not be buffer aligned for other than 1kB
3252 * block sizes. We need to calculate the offset from buffer start.
3254 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3255 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3256 offset = do_div(logical_sb_block, blocksize);
3258 logical_sb_block = sb_block;
3261 if (!(bh = sb_bread(sb, logical_sb_block))) {
3262 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3266 * Note: s_es must be initialized as soon as possible because
3267 * some ext4 macro-instructions depend on its value
3269 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3271 sb->s_magic = le16_to_cpu(es->s_magic);
3272 if (sb->s_magic != EXT4_SUPER_MAGIC)
3274 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3276 /* Set defaults before we parse the mount options */
3277 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3278 set_opt(sb, INIT_INODE_TABLE);
3279 if (def_mount_opts & EXT4_DEFM_DEBUG)
3281 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3282 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3286 if (def_mount_opts & EXT4_DEFM_UID16)
3287 set_opt(sb, NO_UID32);
3288 /* xattr user namespace & acls are now defaulted on */
3289 #ifdef CONFIG_EXT4_FS_XATTR
3290 set_opt(sb, XATTR_USER);
3292 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3293 set_opt(sb, POSIX_ACL);
3295 set_opt(sb, MBLK_IO_SUBMIT);
3296 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3297 set_opt(sb, JOURNAL_DATA);
3298 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3299 set_opt(sb, ORDERED_DATA);
3300 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3301 set_opt(sb, WRITEBACK_DATA);
3303 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3304 set_opt(sb, ERRORS_PANIC);
3305 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3306 set_opt(sb, ERRORS_CONT);
3308 set_opt(sb, ERRORS_RO);
3309 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3310 set_opt(sb, BLOCK_VALIDITY);
3311 if (def_mount_opts & EXT4_DEFM_DISCARD)
3312 set_opt(sb, DISCARD);
3314 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3315 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3316 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3317 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3318 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3320 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3321 set_opt(sb, BARRIER);
3324 * enable delayed allocation by default
3325 * Use -o nodelalloc to turn it off
3327 if (!IS_EXT3_SB(sb) &&
3328 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3329 set_opt(sb, DELALLOC);
3332 * set default s_li_wait_mult for lazyinit, for the case there is
3333 * no mount option specified.
3335 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3337 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3338 &journal_devnum, &journal_ioprio, NULL, 0)) {
3339 ext4_msg(sb, KERN_WARNING,
3340 "failed to parse options in superblock: %s",
3341 sbi->s_es->s_mount_opts);
3343 if (!parse_options((char *) data, sb, &journal_devnum,
3344 &journal_ioprio, NULL, 0))
3347 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3348 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3349 "with data=journal disables delayed "
3350 "allocation and O_DIRECT support!\n");
3351 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3352 ext4_msg(sb, KERN_ERR, "can't mount with "
3353 "both data=journal and delalloc");
3356 if (test_opt(sb, DIOREAD_NOLOCK)) {
3357 ext4_msg(sb, KERN_ERR, "can't mount with "
3358 "both data=journal and delalloc");
3361 if (test_opt(sb, DELALLOC))
3362 clear_opt(sb, DELALLOC);
3365 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3366 if (test_opt(sb, DIOREAD_NOLOCK)) {
3367 if (blocksize < PAGE_SIZE) {
3368 ext4_msg(sb, KERN_ERR, "can't mount with "
3369 "dioread_nolock if block size != PAGE_SIZE");
3374 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3375 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3377 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3378 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3379 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3380 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3381 ext4_msg(sb, KERN_WARNING,
3382 "feature flags set on rev 0 fs, "
3383 "running e2fsck is recommended");
3385 if (IS_EXT2_SB(sb)) {
3386 if (ext2_feature_set_ok(sb))
3387 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3388 "using the ext4 subsystem");
3390 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3391 "to feature incompatibilities");
3396 if (IS_EXT3_SB(sb)) {
3397 if (ext3_feature_set_ok(sb))
3398 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3399 "using the ext4 subsystem");
3401 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3402 "to feature incompatibilities");
3408 * Check feature flags regardless of the revision level, since we
3409 * previously didn't change the revision level when setting the flags,
3410 * so there is a chance incompat flags are set on a rev 0 filesystem.
3412 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3415 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3416 blocksize > EXT4_MAX_BLOCK_SIZE) {
3417 ext4_msg(sb, KERN_ERR,
3418 "Unsupported filesystem blocksize %d", blocksize);
3422 if (sb->s_blocksize != blocksize) {
3423 /* Validate the filesystem blocksize */
3424 if (!sb_set_blocksize(sb, blocksize)) {
3425 ext4_msg(sb, KERN_ERR, "bad block size %d",
3431 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3432 offset = do_div(logical_sb_block, blocksize);
3433 bh = sb_bread(sb, logical_sb_block);
3435 ext4_msg(sb, KERN_ERR,
3436 "Can't read superblock on 2nd try");
3439 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3441 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3442 ext4_msg(sb, KERN_ERR,
3443 "Magic mismatch, very weird!");
3448 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3449 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3450 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3452 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3454 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3455 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3456 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3458 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3459 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3460 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3461 (!is_power_of_2(sbi->s_inode_size)) ||
3462 (sbi->s_inode_size > blocksize)) {
3463 ext4_msg(sb, KERN_ERR,
3464 "unsupported inode size: %d",
3468 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3469 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3472 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3473 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3474 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3475 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3476 !is_power_of_2(sbi->s_desc_size)) {
3477 ext4_msg(sb, KERN_ERR,
3478 "unsupported descriptor size %lu",
3483 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3485 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3486 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3487 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3490 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3491 if (sbi->s_inodes_per_block == 0)
3493 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3494 sbi->s_inodes_per_block;
3495 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3497 sbi->s_mount_state = le16_to_cpu(es->s_state);
3498 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3499 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3501 for (i = 0; i < 4; i++)
3502 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3503 sbi->s_def_hash_version = es->s_def_hash_version;
3504 i = le32_to_cpu(es->s_flags);
3505 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3506 sbi->s_hash_unsigned = 3;
3507 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3508 #ifdef __CHAR_UNSIGNED__
3509 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3510 sbi->s_hash_unsigned = 3;
3512 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3517 /* Handle clustersize */
3518 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3519 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3520 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3522 if (clustersize < blocksize) {
3523 ext4_msg(sb, KERN_ERR,
3524 "cluster size (%d) smaller than "
3525 "block size (%d)", clustersize, blocksize);
3528 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3529 le32_to_cpu(es->s_log_block_size);
3530 sbi->s_clusters_per_group =
3531 le32_to_cpu(es->s_clusters_per_group);
3532 if (sbi->s_clusters_per_group > blocksize * 8) {
3533 ext4_msg(sb, KERN_ERR,
3534 "#clusters per group too big: %lu",
3535 sbi->s_clusters_per_group);
3538 if (sbi->s_blocks_per_group !=
3539 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3540 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3541 "clusters per group (%lu) inconsistent",
3542 sbi->s_blocks_per_group,
3543 sbi->s_clusters_per_group);
3547 if (clustersize != blocksize) {
3548 ext4_warning(sb, "fragment/cluster size (%d) != "
3549 "block size (%d)", clustersize,
3551 clustersize = blocksize;
3553 if (sbi->s_blocks_per_group > blocksize * 8) {
3554 ext4_msg(sb, KERN_ERR,
3555 "#blocks per group too big: %lu",
3556 sbi->s_blocks_per_group);
3559 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3560 sbi->s_cluster_bits = 0;
3562 sbi->s_cluster_ratio = clustersize / blocksize;
3564 if (sbi->s_inodes_per_group > blocksize * 8) {
3565 ext4_msg(sb, KERN_ERR,
3566 "#inodes per group too big: %lu",
3567 sbi->s_inodes_per_group);
3572 * Test whether we have more sectors than will fit in sector_t,
3573 * and whether the max offset is addressable by the page cache.
3575 err = generic_check_addressable(sb->s_blocksize_bits,
3576 ext4_blocks_count(es));
3578 ext4_msg(sb, KERN_ERR, "filesystem"
3579 " too large to mount safely on this system");
3580 if (sizeof(sector_t) < 8)
3581 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3586 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3589 /* check blocks count against device size */
3590 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3591 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3592 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3593 "exceeds size of device (%llu blocks)",
3594 ext4_blocks_count(es), blocks_count);
3599 * It makes no sense for the first data block to be beyond the end
3600 * of the filesystem.
3602 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3603 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3604 "block %u is beyond end of filesystem (%llu)",
3605 le32_to_cpu(es->s_first_data_block),
3606 ext4_blocks_count(es));
3609 blocks_count = (ext4_blocks_count(es) -
3610 le32_to_cpu(es->s_first_data_block) +
3611 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3612 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3613 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3614 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3615 "(block count %llu, first data block %u, "
3616 "blocks per group %lu)", sbi->s_groups_count,
3617 ext4_blocks_count(es),
3618 le32_to_cpu(es->s_first_data_block),
3619 EXT4_BLOCKS_PER_GROUP(sb));
3622 sbi->s_groups_count = blocks_count;
3623 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3624 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3625 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3626 EXT4_DESC_PER_BLOCK(sb);
3627 sbi->s_group_desc = ext4_kvmalloc(db_count *
3628 sizeof(struct buffer_head *),
3630 if (sbi->s_group_desc == NULL) {
3631 ext4_msg(sb, KERN_ERR, "not enough memory");
3636 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3638 bgl_lock_init(sbi->s_blockgroup_lock);
3640 for (i = 0; i < db_count; i++) {
3641 block = descriptor_loc(sb, logical_sb_block, i);
3642 sbi->s_group_desc[i] = sb_bread(sb, block);
3643 if (!sbi->s_group_desc[i]) {
3644 ext4_msg(sb, KERN_ERR,
3645 "can't read group descriptor %d", i);
3650 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3651 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3654 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3655 if (!ext4_fill_flex_info(sb)) {
3656 ext4_msg(sb, KERN_ERR,
3657 "unable to initialize "
3658 "flex_bg meta info!");
3662 sbi->s_gdb_count = db_count;
3663 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3664 spin_lock_init(&sbi->s_next_gen_lock);
3666 init_timer(&sbi->s_err_report);
3667 sbi->s_err_report.function = print_daily_error_info;
3668 sbi->s_err_report.data = (unsigned long) sb;
3670 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3671 ext4_count_free_clusters(sb));
3673 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3674 ext4_count_free_inodes(sb));
3677 err = percpu_counter_init(&sbi->s_dirs_counter,
3678 ext4_count_dirs(sb));
3681 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3684 ext4_msg(sb, KERN_ERR, "insufficient memory");
3688 sbi->s_stripe = ext4_get_stripe_size(sbi);
3689 sbi->s_max_writeback_mb_bump = 128;
3692 * set up enough so that it can read an inode
3694 if (!test_opt(sb, NOLOAD) &&
3695 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3696 sb->s_op = &ext4_sops;
3698 sb->s_op = &ext4_nojournal_sops;
3699 sb->s_export_op = &ext4_export_ops;
3700 sb->s_xattr = ext4_xattr_handlers;
3702 sb->s_qcop = &ext4_qctl_operations;
3703 sb->dq_op = &ext4_quota_operations;
3705 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3707 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3708 mutex_init(&sbi->s_orphan_lock);
3709 sbi->s_resize_flags = 0;
3713 needs_recovery = (es->s_last_orphan != 0 ||
3714 EXT4_HAS_INCOMPAT_FEATURE(sb,
3715 EXT4_FEATURE_INCOMPAT_RECOVER));
3717 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3718 !(sb->s_flags & MS_RDONLY))
3719 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3723 * The first inode we look at is the journal inode. Don't try
3724 * root first: it may be modified in the journal!
3726 if (!test_opt(sb, NOLOAD) &&
3727 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3728 if (ext4_load_journal(sb, es, journal_devnum))
3730 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3731 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3732 ext4_msg(sb, KERN_ERR, "required journal recovery "
3733 "suppressed and not mounted read-only");
3734 goto failed_mount_wq;
3736 clear_opt(sb, DATA_FLAGS);
3737 sbi->s_journal = NULL;
3742 if (ext4_blocks_count(es) > 0xffffffffULL &&
3743 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3744 JBD2_FEATURE_INCOMPAT_64BIT)) {
3745 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3746 goto failed_mount_wq;
3749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3750 jbd2_journal_set_features(sbi->s_journal,
3751 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3752 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3753 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3754 jbd2_journal_set_features(sbi->s_journal,
3755 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3756 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3757 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3759 jbd2_journal_clear_features(sbi->s_journal,
3760 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3761 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3764 /* We have now updated the journal if required, so we can
3765 * validate the data journaling mode. */
3766 switch (test_opt(sb, DATA_FLAGS)) {
3768 /* No mode set, assume a default based on the journal
3769 * capabilities: ORDERED_DATA if the journal can
3770 * cope, else JOURNAL_DATA
3772 if (jbd2_journal_check_available_features
3773 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3774 set_opt(sb, ORDERED_DATA);
3776 set_opt(sb, JOURNAL_DATA);
3779 case EXT4_MOUNT_ORDERED_DATA:
3780 case EXT4_MOUNT_WRITEBACK_DATA:
3781 if (!jbd2_journal_check_available_features
3782 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3783 ext4_msg(sb, KERN_ERR, "Journal does not support "
3784 "requested data journaling mode");
3785 goto failed_mount_wq;
3790 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3793 * The journal may have updated the bg summary counts, so we
3794 * need to update the global counters.
3796 percpu_counter_set(&sbi->s_freeclusters_counter,
3797 ext4_count_free_clusters(sb));
3798 percpu_counter_set(&sbi->s_freeinodes_counter,
3799 ext4_count_free_inodes(sb));
3800 percpu_counter_set(&sbi->s_dirs_counter,
3801 ext4_count_dirs(sb));
3802 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3806 * Get the # of file system overhead blocks from the
3807 * superblock if present.
3809 if (es->s_overhead_clusters)
3810 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3812 ret = ext4_calculate_overhead(sb);
3814 goto failed_mount_wq;
3818 * The maximum number of concurrent works can be high and
3819 * concurrency isn't really necessary. Limit it to 1.
3821 EXT4_SB(sb)->dio_unwritten_wq =
3822 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3823 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3824 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3825 goto failed_mount_wq;
3829 * The jbd2_journal_load will have done any necessary log recovery,
3830 * so we can safely mount the rest of the filesystem now.
3833 root = ext4_iget(sb, EXT4_ROOT_INO);
3835 ext4_msg(sb, KERN_ERR, "get root inode failed");
3836 ret = PTR_ERR(root);
3840 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3841 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3844 sb->s_root = d_alloc_root(root);
3846 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3851 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3852 sb->s_flags |= MS_RDONLY;
3854 /* determine the minimum size of new large inodes, if present */
3855 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3856 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3857 EXT4_GOOD_OLD_INODE_SIZE;
3858 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3859 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3860 if (sbi->s_want_extra_isize <
3861 le16_to_cpu(es->s_want_extra_isize))
3862 sbi->s_want_extra_isize =
3863 le16_to_cpu(es->s_want_extra_isize);
3864 if (sbi->s_want_extra_isize <
3865 le16_to_cpu(es->s_min_extra_isize))
3866 sbi->s_want_extra_isize =
3867 le16_to_cpu(es->s_min_extra_isize);
3870 /* Check if enough inode space is available */
3871 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3872 sbi->s_inode_size) {
3873 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3874 EXT4_GOOD_OLD_INODE_SIZE;
3875 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3879 err = ext4_setup_system_zone(sb);
3881 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3887 err = ext4_mb_init(sb, needs_recovery);
3889 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3894 err = ext4_register_li_request(sb, first_not_zeroed);
3898 sbi->s_kobj.kset = ext4_kset;
3899 init_completion(&sbi->s_kobj_unregister);
3900 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3905 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3906 ext4_orphan_cleanup(sb, es);
3907 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3908 if (needs_recovery) {
3909 ext4_msg(sb, KERN_INFO, "recovery complete");
3910 ext4_mark_recovery_complete(sb, es);
3912 if (EXT4_SB(sb)->s_journal) {
3913 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3914 descr = " journalled data mode";
3915 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3916 descr = " ordered data mode";
3918 descr = " writeback data mode";
3920 descr = "out journal";
3922 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3923 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3924 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3926 if (es->s_error_count)
3927 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3934 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3938 ext4_unregister_li_request(sb);
3940 ext4_ext_release(sb);
3942 ext4_mb_release(sb);
3943 ext4_release_system_zone(sb);
3947 ext4_msg(sb, KERN_ERR, "mount failed");
3948 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3950 if (sbi->s_journal) {
3951 jbd2_journal_destroy(sbi->s_journal);
3952 sbi->s_journal = NULL;
3955 del_timer(&sbi->s_err_report);
3956 if (sbi->s_flex_groups)
3957 ext4_kvfree(sbi->s_flex_groups);
3958 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3959 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3960 percpu_counter_destroy(&sbi->s_dirs_counter);
3961 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3963 kthread_stop(sbi->s_mmp_tsk);
3965 for (i = 0; i < db_count; i++)
3966 brelse(sbi->s_group_desc[i]);
3967 ext4_kvfree(sbi->s_group_desc);
3970 remove_proc_entry(sb->s_id, ext4_proc_root);
3973 for (i = 0; i < MAXQUOTAS; i++)
3974 kfree(sbi->s_qf_names[i]);
3976 ext4_blkdev_remove(sbi);
3979 sb->s_fs_info = NULL;
3980 kfree(sbi->s_blockgroup_lock);
3988 * Setup any per-fs journal parameters now. We'll do this both on
3989 * initial mount, once the journal has been initialised but before we've
3990 * done any recovery; and again on any subsequent remount.
3992 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3994 struct ext4_sb_info *sbi = EXT4_SB(sb);
3996 journal->j_commit_interval = sbi->s_commit_interval;
3997 journal->j_min_batch_time = sbi->s_min_batch_time;
3998 journal->j_max_batch_time = sbi->s_max_batch_time;
4000 write_lock(&journal->j_state_lock);
4001 if (test_opt(sb, BARRIER))
4002 journal->j_flags |= JBD2_BARRIER;
4004 journal->j_flags &= ~JBD2_BARRIER;
4005 if (test_opt(sb, DATA_ERR_ABORT))
4006 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4008 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4009 write_unlock(&journal->j_state_lock);
4012 static journal_t *ext4_get_journal(struct super_block *sb,
4013 unsigned int journal_inum)
4015 struct inode *journal_inode;
4018 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4020 /* First, test for the existence of a valid inode on disk. Bad
4021 * things happen if we iget() an unused inode, as the subsequent
4022 * iput() will try to delete it. */
4024 journal_inode = ext4_iget(sb, journal_inum);
4025 if (IS_ERR(journal_inode)) {
4026 ext4_msg(sb, KERN_ERR, "no journal found");
4029 if (!journal_inode->i_nlink) {
4030 make_bad_inode(journal_inode);
4031 iput(journal_inode);
4032 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4036 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4037 journal_inode, journal_inode->i_size);
4038 if (!S_ISREG(journal_inode->i_mode)) {
4039 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4040 iput(journal_inode);
4044 journal = jbd2_journal_init_inode(journal_inode);
4046 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4047 iput(journal_inode);
4050 journal->j_private = sb;
4051 ext4_init_journal_params(sb, journal);
4055 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4058 struct buffer_head *bh;
4062 int hblock, blocksize;
4063 ext4_fsblk_t sb_block;
4064 unsigned long offset;
4065 struct ext4_super_block *es;
4066 struct block_device *bdev;
4068 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4070 bdev = ext4_blkdev_get(j_dev, sb);
4074 blocksize = sb->s_blocksize;
4075 hblock = bdev_logical_block_size(bdev);
4076 if (blocksize < hblock) {
4077 ext4_msg(sb, KERN_ERR,
4078 "blocksize too small for journal device");
4082 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4083 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4084 set_blocksize(bdev, blocksize);
4085 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4086 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4087 "external journal");
4091 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4092 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4093 !(le32_to_cpu(es->s_feature_incompat) &
4094 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4095 ext4_msg(sb, KERN_ERR, "external journal has "
4101 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4102 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4107 len = ext4_blocks_count(es);
4108 start = sb_block + 1;
4109 brelse(bh); /* we're done with the superblock */
4111 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4112 start, len, blocksize);
4114 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4117 journal->j_private = sb;
4118 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4119 wait_on_buffer(journal->j_sb_buffer);
4120 if (!buffer_uptodate(journal->j_sb_buffer)) {
4121 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4124 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4125 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4126 "user (unsupported) - %d",
4127 be32_to_cpu(journal->j_superblock->s_nr_users));
4130 EXT4_SB(sb)->journal_bdev = bdev;
4131 ext4_init_journal_params(sb, journal);
4135 jbd2_journal_destroy(journal);
4137 ext4_blkdev_put(bdev);
4141 static int ext4_load_journal(struct super_block *sb,
4142 struct ext4_super_block *es,
4143 unsigned long journal_devnum)
4146 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4149 int really_read_only;
4151 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4153 if (journal_devnum &&
4154 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4155 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4156 "numbers have changed");
4157 journal_dev = new_decode_dev(journal_devnum);
4159 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4161 really_read_only = bdev_read_only(sb->s_bdev);
4164 * Are we loading a blank journal or performing recovery after a
4165 * crash? For recovery, we need to check in advance whether we
4166 * can get read-write access to the device.
4168 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4169 if (sb->s_flags & MS_RDONLY) {
4170 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4171 "required on readonly filesystem");
4172 if (really_read_only) {
4173 ext4_msg(sb, KERN_ERR, "write access "
4174 "unavailable, cannot proceed");
4177 ext4_msg(sb, KERN_INFO, "write access will "
4178 "be enabled during recovery");
4182 if (journal_inum && journal_dev) {
4183 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4184 "and inode journals!");
4189 if (!(journal = ext4_get_journal(sb, journal_inum)))
4192 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4196 if (!(journal->j_flags & JBD2_BARRIER))
4197 ext4_msg(sb, KERN_INFO, "barriers disabled");
4199 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4200 err = jbd2_journal_update_format(journal);
4202 ext4_msg(sb, KERN_ERR, "error updating journal");
4203 jbd2_journal_destroy(journal);
4208 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4209 err = jbd2_journal_wipe(journal, !really_read_only);
4211 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4213 memcpy(save, ((char *) es) +
4214 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4215 err = jbd2_journal_load(journal);
4217 memcpy(((char *) es) + EXT4_S_ERR_START,
4218 save, EXT4_S_ERR_LEN);
4223 ext4_msg(sb, KERN_ERR, "error loading journal");
4224 jbd2_journal_destroy(journal);
4228 EXT4_SB(sb)->s_journal = journal;
4229 ext4_clear_journal_err(sb, es);
4231 if (!really_read_only && journal_devnum &&
4232 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4233 es->s_journal_dev = cpu_to_le32(journal_devnum);
4235 /* Make sure we flush the recovery flag to disk. */
4236 ext4_commit_super(sb, 1);
4242 static int ext4_commit_super(struct super_block *sb, int sync)
4244 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4245 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4248 if (!sbh || block_device_ejected(sb))
4250 if (buffer_write_io_error(sbh)) {
4252 * Oh, dear. A previous attempt to write the
4253 * superblock failed. This could happen because the
4254 * USB device was yanked out. Or it could happen to
4255 * be a transient write error and maybe the block will
4256 * be remapped. Nothing we can do but to retry the
4257 * write and hope for the best.
4259 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4260 "superblock detected");
4261 clear_buffer_write_io_error(sbh);
4262 set_buffer_uptodate(sbh);
4265 * If the file system is mounted read-only, don't update the
4266 * superblock write time. This avoids updating the superblock
4267 * write time when we are mounting the root file system
4268 * read/only but we need to replay the journal; at that point,
4269 * for people who are east of GMT and who make their clock
4270 * tick in localtime for Windows bug-for-bug compatibility,
4271 * the clock is set in the future, and this will cause e2fsck
4272 * to complain and force a full file system check.
4274 if (!(sb->s_flags & MS_RDONLY))
4275 es->s_wtime = cpu_to_le32(get_seconds());
4276 if (sb->s_bdev->bd_part)
4277 es->s_kbytes_written =
4278 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4279 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4280 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4282 es->s_kbytes_written =
4283 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4284 ext4_free_blocks_count_set(es,
4285 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4286 &EXT4_SB(sb)->s_freeclusters_counter)));
4287 es->s_free_inodes_count =
4288 cpu_to_le32(percpu_counter_sum_positive(
4289 &EXT4_SB(sb)->s_freeinodes_counter));
4291 BUFFER_TRACE(sbh, "marking dirty");
4292 mark_buffer_dirty(sbh);
4294 error = sync_dirty_buffer(sbh);
4298 error = buffer_write_io_error(sbh);
4300 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4302 clear_buffer_write_io_error(sbh);
4303 set_buffer_uptodate(sbh);
4310 * Have we just finished recovery? If so, and if we are mounting (or
4311 * remounting) the filesystem readonly, then we will end up with a
4312 * consistent fs on disk. Record that fact.
4314 static void ext4_mark_recovery_complete(struct super_block *sb,
4315 struct ext4_super_block *es)
4317 journal_t *journal = EXT4_SB(sb)->s_journal;
4319 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4320 BUG_ON(journal != NULL);
4323 jbd2_journal_lock_updates(journal);
4324 if (jbd2_journal_flush(journal) < 0)
4327 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4328 sb->s_flags & MS_RDONLY) {
4329 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4330 ext4_commit_super(sb, 1);
4334 jbd2_journal_unlock_updates(journal);
4338 * If we are mounting (or read-write remounting) a filesystem whose journal
4339 * has recorded an error from a previous lifetime, move that error to the
4340 * main filesystem now.
4342 static void ext4_clear_journal_err(struct super_block *sb,
4343 struct ext4_super_block *es)
4349 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4351 journal = EXT4_SB(sb)->s_journal;
4354 * Now check for any error status which may have been recorded in the
4355 * journal by a prior ext4_error() or ext4_abort()
4358 j_errno = jbd2_journal_errno(journal);
4362 errstr = ext4_decode_error(sb, j_errno, nbuf);
4363 ext4_warning(sb, "Filesystem error recorded "
4364 "from previous mount: %s", errstr);
4365 ext4_warning(sb, "Marking fs in need of filesystem check.");
4367 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4368 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4369 ext4_commit_super(sb, 1);
4371 jbd2_journal_clear_err(journal);
4376 * Force the running and committing transactions to commit,
4377 * and wait on the commit.
4379 int ext4_force_commit(struct super_block *sb)
4384 if (sb->s_flags & MS_RDONLY)
4387 journal = EXT4_SB(sb)->s_journal;
4389 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4390 ret = ext4_journal_force_commit(journal);
4396 static void ext4_write_super(struct super_block *sb)
4399 ext4_commit_super(sb, 1);
4403 static int ext4_sync_fs(struct super_block *sb, int wait)
4407 struct ext4_sb_info *sbi = EXT4_SB(sb);
4409 trace_ext4_sync_fs(sb, wait);
4410 flush_workqueue(sbi->dio_unwritten_wq);
4411 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4413 jbd2_log_wait_commit(sbi->s_journal, target);
4419 * LVM calls this function before a (read-only) snapshot is created. This
4420 * gives us a chance to flush the journal completely and mark the fs clean.
4422 * Note that only this function cannot bring a filesystem to be in a clean
4423 * state independently, because ext4 prevents a new handle from being started
4424 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4427 static int ext4_freeze(struct super_block *sb)
4432 if (sb->s_flags & MS_RDONLY)
4435 journal = EXT4_SB(sb)->s_journal;
4437 /* Now we set up the journal barrier. */
4438 jbd2_journal_lock_updates(journal);
4441 * Don't clear the needs_recovery flag if we failed to flush
4444 error = jbd2_journal_flush(journal);
4448 /* Journal blocked and flushed, clear needs_recovery flag. */
4449 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4450 error = ext4_commit_super(sb, 1);
4452 /* we rely on s_frozen to stop further updates */
4453 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4458 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4459 * flag here, even though the filesystem is not technically dirty yet.
4461 static int ext4_unfreeze(struct super_block *sb)
4463 if (sb->s_flags & MS_RDONLY)
4467 /* Reset the needs_recovery flag before the fs is unlocked. */
4468 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4469 ext4_commit_super(sb, 1);
4475 * Structure to save mount options for ext4_remount's benefit
4477 struct ext4_mount_options {
4478 unsigned long s_mount_opt;
4479 unsigned long s_mount_opt2;
4482 unsigned long s_commit_interval;
4483 u32 s_min_batch_time, s_max_batch_time;
4486 char *s_qf_names[MAXQUOTAS];
4490 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4492 struct ext4_super_block *es;
4493 struct ext4_sb_info *sbi = EXT4_SB(sb);
4494 ext4_fsblk_t n_blocks_count = 0;
4495 unsigned long old_sb_flags;
4496 struct ext4_mount_options old_opts;
4497 int enable_quota = 0;
4499 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4504 char *orig_data = kstrdup(data, GFP_KERNEL);
4506 /* Store the original options */
4508 old_sb_flags = sb->s_flags;
4509 old_opts.s_mount_opt = sbi->s_mount_opt;
4510 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4511 old_opts.s_resuid = sbi->s_resuid;
4512 old_opts.s_resgid = sbi->s_resgid;
4513 old_opts.s_commit_interval = sbi->s_commit_interval;
4514 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4515 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4517 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4518 for (i = 0; i < MAXQUOTAS; i++)
4519 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4521 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4522 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4525 * Allow the "check" option to be passed as a remount option.
4527 if (!parse_options(data, sb, NULL, &journal_ioprio,
4528 &n_blocks_count, 1)) {
4533 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4534 ext4_abort(sb, "Abort forced by user");
4536 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4537 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4541 if (sbi->s_journal) {
4542 ext4_init_journal_params(sb, sbi->s_journal);
4543 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4546 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4547 n_blocks_count > ext4_blocks_count(es)) {
4548 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4553 if (*flags & MS_RDONLY) {
4554 err = dquot_suspend(sb, -1);
4559 * First of all, the unconditional stuff we have to do
4560 * to disable replay of the journal when we next remount
4562 sb->s_flags |= MS_RDONLY;
4565 * OK, test if we are remounting a valid rw partition
4566 * readonly, and if so set the rdonly flag and then
4567 * mark the partition as valid again.
4569 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4570 (sbi->s_mount_state & EXT4_VALID_FS))
4571 es->s_state = cpu_to_le16(sbi->s_mount_state);
4574 ext4_mark_recovery_complete(sb, es);
4576 /* Make sure we can mount this feature set readwrite */
4577 if (!ext4_feature_set_ok(sb, 0)) {
4582 * Make sure the group descriptor checksums
4583 * are sane. If they aren't, refuse to remount r/w.
4585 for (g = 0; g < sbi->s_groups_count; g++) {
4586 struct ext4_group_desc *gdp =
4587 ext4_get_group_desc(sb, g, NULL);
4589 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4590 ext4_msg(sb, KERN_ERR,
4591 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4592 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4593 le16_to_cpu(gdp->bg_checksum));
4600 * If we have an unprocessed orphan list hanging
4601 * around from a previously readonly bdev mount,
4602 * require a full umount/remount for now.
4604 if (es->s_last_orphan) {
4605 ext4_msg(sb, KERN_WARNING, "Couldn't "
4606 "remount RDWR because of unprocessed "
4607 "orphan inode list. Please "
4608 "umount/remount instead");
4614 * Mounting a RDONLY partition read-write, so reread
4615 * and store the current valid flag. (It may have
4616 * been changed by e2fsck since we originally mounted
4620 ext4_clear_journal_err(sb, es);
4621 sbi->s_mount_state = le16_to_cpu(es->s_state);
4622 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4624 if (!ext4_setup_super(sb, es, 0))
4625 sb->s_flags &= ~MS_RDONLY;
4626 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4627 EXT4_FEATURE_INCOMPAT_MMP))
4628 if (ext4_multi_mount_protect(sb,
4629 le64_to_cpu(es->s_mmp_block))) {
4638 * Reinitialize lazy itable initialization thread based on
4641 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4642 ext4_unregister_li_request(sb);
4644 ext4_group_t first_not_zeroed;
4645 first_not_zeroed = ext4_has_uninit_itable(sb);
4646 ext4_register_li_request(sb, first_not_zeroed);
4649 ext4_setup_system_zone(sb);
4650 if (sbi->s_journal == NULL)
4651 ext4_commit_super(sb, 1);
4654 /* Release old quota file names */
4655 for (i = 0; i < MAXQUOTAS; i++)
4656 if (old_opts.s_qf_names[i] &&
4657 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4658 kfree(old_opts.s_qf_names[i]);
4662 dquot_resume(sb, -1);
4664 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4669 sb->s_flags = old_sb_flags;
4670 sbi->s_mount_opt = old_opts.s_mount_opt;
4671 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4672 sbi->s_resuid = old_opts.s_resuid;
4673 sbi->s_resgid = old_opts.s_resgid;
4674 sbi->s_commit_interval = old_opts.s_commit_interval;
4675 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4676 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4678 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4679 for (i = 0; i < MAXQUOTAS; i++) {
4680 if (sbi->s_qf_names[i] &&
4681 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4682 kfree(sbi->s_qf_names[i]);
4683 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4691 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4693 struct super_block *sb = dentry->d_sb;
4694 struct ext4_sb_info *sbi = EXT4_SB(sb);
4695 struct ext4_super_block *es = sbi->s_es;
4696 ext4_fsblk_t overhead = 0;
4700 if (!test_opt(sb, MINIX_DF))
4701 overhead = sbi->s_overhead;
4703 buf->f_type = EXT4_SUPER_MAGIC;
4704 buf->f_bsize = sb->s_blocksize;
4705 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4706 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4707 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4708 /* prevent underflow in case that few free space is available */
4709 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4710 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4711 if (buf->f_bfree < ext4_r_blocks_count(es))
4713 buf->f_files = le32_to_cpu(es->s_inodes_count);
4714 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4715 buf->f_namelen = EXT4_NAME_LEN;
4716 fsid = le64_to_cpup((void *)es->s_uuid) ^
4717 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4718 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4719 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4724 /* Helper function for writing quotas on sync - we need to start transaction
4725 * before quota file is locked for write. Otherwise the are possible deadlocks:
4726 * Process 1 Process 2
4727 * ext4_create() quota_sync()
4728 * jbd2_journal_start() write_dquot()
4729 * dquot_initialize() down(dqio_mutex)
4730 * down(dqio_mutex) jbd2_journal_start()
4736 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4738 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4741 static int ext4_write_dquot(struct dquot *dquot)
4745 struct inode *inode;
4747 inode = dquot_to_inode(dquot);
4748 handle = ext4_journal_start(inode,
4749 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4751 return PTR_ERR(handle);
4752 ret = dquot_commit(dquot);
4753 err = ext4_journal_stop(handle);
4759 static int ext4_acquire_dquot(struct dquot *dquot)
4764 handle = ext4_journal_start(dquot_to_inode(dquot),
4765 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4767 return PTR_ERR(handle);
4768 ret = dquot_acquire(dquot);
4769 err = ext4_journal_stop(handle);
4775 static int ext4_release_dquot(struct dquot *dquot)
4780 handle = ext4_journal_start(dquot_to_inode(dquot),
4781 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4782 if (IS_ERR(handle)) {
4783 /* Release dquot anyway to avoid endless cycle in dqput() */
4784 dquot_release(dquot);
4785 return PTR_ERR(handle);
4787 ret = dquot_release(dquot);
4788 err = ext4_journal_stop(handle);
4794 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4796 /* Are we journaling quotas? */
4797 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4798 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4799 dquot_mark_dquot_dirty(dquot);
4800 return ext4_write_dquot(dquot);
4802 return dquot_mark_dquot_dirty(dquot);
4806 static int ext4_write_info(struct super_block *sb, int type)
4811 /* Data block + inode block */
4812 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4814 return PTR_ERR(handle);
4815 ret = dquot_commit_info(sb, type);
4816 err = ext4_journal_stop(handle);
4823 * Turn on quotas during mount time - we need to find
4824 * the quota file and such...
4826 static int ext4_quota_on_mount(struct super_block *sb, int type)
4828 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4829 EXT4_SB(sb)->s_jquota_fmt, type);
4833 * Standard function to be called on quota_on
4835 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4840 if (!test_opt(sb, QUOTA))
4843 /* Quotafile not on the same filesystem? */
4844 if (path->mnt->mnt_sb != sb)
4846 /* Journaling quota? */
4847 if (EXT4_SB(sb)->s_qf_names[type]) {
4848 /* Quotafile not in fs root? */
4849 if (path->dentry->d_parent != sb->s_root)
4850 ext4_msg(sb, KERN_WARNING,
4851 "Quota file not on filesystem root. "
4852 "Journaled quota will not work");
4856 * When we journal data on quota file, we have to flush journal to see
4857 * all updates to the file when we bypass pagecache...
4859 if (EXT4_SB(sb)->s_journal &&
4860 ext4_should_journal_data(path->dentry->d_inode)) {
4862 * We don't need to lock updates but journal_flush() could
4863 * otherwise be livelocked...
4865 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4866 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4867 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4872 return dquot_quota_on(sb, type, format_id, path);
4875 static int ext4_quota_off(struct super_block *sb, int type)
4877 struct inode *inode = sb_dqopt(sb)->files[type];
4880 /* Force all delayed allocation blocks to be allocated.
4881 * Caller already holds s_umount sem */
4882 if (test_opt(sb, DELALLOC))
4883 sync_filesystem(sb);
4888 /* Update modification times of quota files when userspace can
4889 * start looking at them */
4890 handle = ext4_journal_start(inode, 1);
4893 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4894 ext4_mark_inode_dirty(handle, inode);
4895 ext4_journal_stop(handle);
4898 return dquot_quota_off(sb, type);
4901 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4902 * acquiring the locks... As quota files are never truncated and quota code
4903 * itself serializes the operations (and no one else should touch the files)
4904 * we don't have to be afraid of races */
4905 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4906 size_t len, loff_t off)
4908 struct inode *inode = sb_dqopt(sb)->files[type];
4909 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4911 int offset = off & (sb->s_blocksize - 1);
4914 struct buffer_head *bh;
4915 loff_t i_size = i_size_read(inode);
4919 if (off+len > i_size)
4922 while (toread > 0) {
4923 tocopy = sb->s_blocksize - offset < toread ?
4924 sb->s_blocksize - offset : toread;
4925 bh = ext4_bread(NULL, inode, blk, 0, &err);
4928 if (!bh) /* A hole? */
4929 memset(data, 0, tocopy);
4931 memcpy(data, bh->b_data+offset, tocopy);
4941 /* Write to quotafile (we know the transaction is already started and has
4942 * enough credits) */
4943 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4944 const char *data, size_t len, loff_t off)
4946 struct inode *inode = sb_dqopt(sb)->files[type];
4947 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4949 int offset = off & (sb->s_blocksize - 1);
4950 struct buffer_head *bh;
4951 handle_t *handle = journal_current_handle();
4953 if (EXT4_SB(sb)->s_journal && !handle) {
4954 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4955 " cancelled because transaction is not started",
4956 (unsigned long long)off, (unsigned long long)len);
4960 * Since we account only one data block in transaction credits,
4961 * then it is impossible to cross a block boundary.
4963 if (sb->s_blocksize - offset < len) {
4964 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4965 " cancelled because not block aligned",
4966 (unsigned long long)off, (unsigned long long)len);
4970 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4971 bh = ext4_bread(handle, inode, blk, 1, &err);
4974 err = ext4_journal_get_write_access(handle, bh);
4980 memcpy(bh->b_data+offset, data, len);
4981 flush_dcache_page(bh->b_page);
4983 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4987 mutex_unlock(&inode->i_mutex);
4990 if (inode->i_size < off + len) {
4991 i_size_write(inode, off + len);
4992 EXT4_I(inode)->i_disksize = inode->i_size;
4993 ext4_mark_inode_dirty(handle, inode);
4995 mutex_unlock(&inode->i_mutex);
5001 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5002 const char *dev_name, void *data)
5004 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5007 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5008 static inline void register_as_ext2(void)
5010 int err = register_filesystem(&ext2_fs_type);
5013 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5016 static inline void unregister_as_ext2(void)
5018 unregister_filesystem(&ext2_fs_type);
5021 static inline int ext2_feature_set_ok(struct super_block *sb)
5023 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5025 if (sb->s_flags & MS_RDONLY)
5027 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5031 MODULE_ALIAS("ext2");
5033 static inline void register_as_ext2(void) { }
5034 static inline void unregister_as_ext2(void) { }
5035 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5038 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5039 static inline void register_as_ext3(void)
5041 int err = register_filesystem(&ext3_fs_type);
5044 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5047 static inline void unregister_as_ext3(void)
5049 unregister_filesystem(&ext3_fs_type);
5052 static inline int ext3_feature_set_ok(struct super_block *sb)
5054 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5056 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5058 if (sb->s_flags & MS_RDONLY)
5060 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5064 MODULE_ALIAS("ext3");
5066 static inline void register_as_ext3(void) { }
5067 static inline void unregister_as_ext3(void) { }
5068 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5071 static struct file_system_type ext4_fs_type = {
5072 .owner = THIS_MODULE,
5074 .mount = ext4_mount,
5075 .kill_sb = kill_block_super,
5076 .fs_flags = FS_REQUIRES_DEV,
5079 static int __init ext4_init_feat_adverts(void)
5081 struct ext4_features *ef;
5084 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5088 ef->f_kobj.kset = ext4_kset;
5089 init_completion(&ef->f_kobj_unregister);
5090 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5103 static void ext4_exit_feat_adverts(void)
5105 kobject_put(&ext4_feat->f_kobj);
5106 wait_for_completion(&ext4_feat->f_kobj_unregister);
5110 /* Shared across all ext4 file systems */
5111 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5112 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5114 static int __init ext4_init_fs(void)
5118 ext4_check_flag_values();
5120 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5121 mutex_init(&ext4__aio_mutex[i]);
5122 init_waitqueue_head(&ext4__ioend_wq[i]);
5125 err = ext4_init_pageio();
5128 err = ext4_init_system_zone();
5131 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5134 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5136 err = ext4_init_feat_adverts();
5140 err = ext4_init_mballoc();
5144 err = ext4_init_xattr();
5147 err = init_inodecache();
5152 err = register_filesystem(&ext4_fs_type);
5156 ext4_li_info = NULL;
5157 mutex_init(&ext4_li_mtx);
5160 unregister_as_ext2();
5161 unregister_as_ext3();
5162 destroy_inodecache();
5166 ext4_exit_mballoc();
5168 ext4_exit_feat_adverts();
5171 remove_proc_entry("fs/ext4", NULL);
5172 kset_unregister(ext4_kset);
5174 ext4_exit_system_zone();
5180 static void __exit ext4_exit_fs(void)
5182 ext4_destroy_lazyinit_thread();
5183 unregister_as_ext2();
5184 unregister_as_ext3();
5185 unregister_filesystem(&ext4_fs_type);
5186 destroy_inodecache();
5188 ext4_exit_mballoc();
5189 ext4_exit_feat_adverts();
5190 remove_proc_entry("fs/ext4", NULL);
5191 kset_unregister(ext4_kset);
5192 ext4_exit_system_zone();
5196 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5197 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5198 MODULE_LICENSE("GPL");
5199 module_init(ext4_init_fs)
5200 module_exit(ext4_exit_fs)