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_sync(&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 sync_blockdev(sb->s_bdev);
861 invalidate_bdev(sb->s_bdev);
862 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
864 * Invalidate the journal device's buffers. We don't want them
865 * floating about in memory - the physical journal device may
866 * hotswapped, and it breaks the `ro-after' testing code.
868 sync_blockdev(sbi->journal_bdev);
869 invalidate_bdev(sbi->journal_bdev);
870 ext4_blkdev_remove(sbi);
873 kthread_stop(sbi->s_mmp_tsk);
874 sb->s_fs_info = NULL;
876 * Now that we are completely done shutting down the
877 * superblock, we need to actually destroy the kobject.
880 kobject_put(&sbi->s_kobj);
881 wait_for_completion(&sbi->s_kobj_unregister);
882 kfree(sbi->s_blockgroup_lock);
886 static struct kmem_cache *ext4_inode_cachep;
889 * Called inside transaction, so use GFP_NOFS
891 static struct inode *ext4_alloc_inode(struct super_block *sb)
893 struct ext4_inode_info *ei;
895 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
899 ei->vfs_inode.i_version = 1;
900 ei->vfs_inode.i_data.writeback_index = 0;
901 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
902 INIT_LIST_HEAD(&ei->i_prealloc_list);
903 spin_lock_init(&ei->i_prealloc_lock);
904 ei->i_reserved_data_blocks = 0;
905 ei->i_reserved_meta_blocks = 0;
906 ei->i_allocated_meta_blocks = 0;
907 ei->i_da_metadata_calc_len = 0;
908 ei->i_da_metadata_calc_last_lblock = 0;
909 spin_lock_init(&(ei->i_block_reservation_lock));
911 ei->i_reserved_quota = 0;
914 INIT_LIST_HEAD(&ei->i_completed_io_list);
915 spin_lock_init(&ei->i_completed_io_lock);
916 ei->cur_aio_dio = NULL;
918 ei->i_datasync_tid = 0;
919 atomic_set(&ei->i_ioend_count, 0);
920 atomic_set(&ei->i_aiodio_unwritten, 0);
922 return &ei->vfs_inode;
925 static int ext4_drop_inode(struct inode *inode)
927 int drop = generic_drop_inode(inode);
929 trace_ext4_drop_inode(inode, drop);
933 static void ext4_i_callback(struct rcu_head *head)
935 struct inode *inode = container_of(head, struct inode, i_rcu);
936 INIT_LIST_HEAD(&inode->i_dentry);
937 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
940 static void ext4_destroy_inode(struct inode *inode)
942 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
943 ext4_msg(inode->i_sb, KERN_ERR,
944 "Inode %lu (%p): orphan list check failed!",
945 inode->i_ino, EXT4_I(inode));
946 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
947 EXT4_I(inode), sizeof(struct ext4_inode_info),
951 call_rcu(&inode->i_rcu, ext4_i_callback);
954 static void init_once(void *foo)
956 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
958 INIT_LIST_HEAD(&ei->i_orphan);
959 #ifdef CONFIG_EXT4_FS_XATTR
960 init_rwsem(&ei->xattr_sem);
962 init_rwsem(&ei->i_data_sem);
963 inode_init_once(&ei->vfs_inode);
966 static int init_inodecache(void)
968 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
969 sizeof(struct ext4_inode_info),
970 0, (SLAB_RECLAIM_ACCOUNT|
973 if (ext4_inode_cachep == NULL)
978 static void destroy_inodecache(void)
980 kmem_cache_destroy(ext4_inode_cachep);
983 void ext4_clear_inode(struct inode *inode)
985 invalidate_inode_buffers(inode);
986 end_writeback(inode);
988 ext4_discard_preallocations(inode);
989 if (EXT4_I(inode)->jinode) {
990 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
991 EXT4_I(inode)->jinode);
992 jbd2_free_inode(EXT4_I(inode)->jinode);
993 EXT4_I(inode)->jinode = NULL;
997 static inline void ext4_show_quota_options(struct seq_file *seq,
998 struct super_block *sb)
1000 #if defined(CONFIG_QUOTA)
1001 struct ext4_sb_info *sbi = EXT4_SB(sb);
1003 if (sbi->s_jquota_fmt) {
1006 switch (sbi->s_jquota_fmt) {
1017 seq_printf(seq, ",jqfmt=%s", fmtname);
1020 if (sbi->s_qf_names[USRQUOTA])
1021 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1023 if (sbi->s_qf_names[GRPQUOTA])
1024 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1026 if (test_opt(sb, USRQUOTA))
1027 seq_puts(seq, ",usrquota");
1029 if (test_opt(sb, GRPQUOTA))
1030 seq_puts(seq, ",grpquota");
1036 * - it's set to a non-default value OR
1037 * - if the per-sb default is different from the global default
1039 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1042 unsigned long def_mount_opts;
1043 struct super_block *sb = vfs->mnt_sb;
1044 struct ext4_sb_info *sbi = EXT4_SB(sb);
1045 struct ext4_super_block *es = sbi->s_es;
1047 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1048 def_errors = le16_to_cpu(es->s_errors);
1050 if (sbi->s_sb_block != 1)
1051 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1052 if (test_opt(sb, MINIX_DF))
1053 seq_puts(seq, ",minixdf");
1054 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1055 seq_puts(seq, ",grpid");
1056 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1057 seq_puts(seq, ",nogrpid");
1058 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1059 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1060 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1062 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1063 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1064 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1066 if (test_opt(sb, ERRORS_RO)) {
1067 if (def_errors == EXT4_ERRORS_PANIC ||
1068 def_errors == EXT4_ERRORS_CONTINUE) {
1069 seq_puts(seq, ",errors=remount-ro");
1072 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1073 seq_puts(seq, ",errors=continue");
1074 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1075 seq_puts(seq, ",errors=panic");
1076 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1077 seq_puts(seq, ",nouid32");
1078 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1079 seq_puts(seq, ",debug");
1080 #ifdef CONFIG_EXT4_FS_XATTR
1081 if (test_opt(sb, XATTR_USER))
1082 seq_puts(seq, ",user_xattr");
1083 if (!test_opt(sb, XATTR_USER))
1084 seq_puts(seq, ",nouser_xattr");
1086 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1087 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1088 seq_puts(seq, ",acl");
1089 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1090 seq_puts(seq, ",noacl");
1092 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1093 seq_printf(seq, ",commit=%u",
1094 (unsigned) (sbi->s_commit_interval / HZ));
1096 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1097 seq_printf(seq, ",min_batch_time=%u",
1098 (unsigned) sbi->s_min_batch_time);
1100 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1101 seq_printf(seq, ",max_batch_time=%u",
1102 (unsigned) sbi->s_max_batch_time);
1106 * We're changing the default of barrier mount option, so
1107 * let's always display its mount state so it's clear what its
1110 seq_puts(seq, ",barrier=");
1111 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1112 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1113 seq_puts(seq, ",journal_async_commit");
1114 else if (test_opt(sb, JOURNAL_CHECKSUM))
1115 seq_puts(seq, ",journal_checksum");
1116 if (test_opt(sb, I_VERSION))
1117 seq_puts(seq, ",i_version");
1118 if (!test_opt(sb, DELALLOC) &&
1119 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1120 seq_puts(seq, ",nodelalloc");
1122 if (!test_opt(sb, MBLK_IO_SUBMIT))
1123 seq_puts(seq, ",nomblk_io_submit");
1125 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1127 * journal mode get enabled in different ways
1128 * So just print the value even if we didn't specify it
1130 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1131 seq_puts(seq, ",data=journal");
1132 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1133 seq_puts(seq, ",data=ordered");
1134 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1135 seq_puts(seq, ",data=writeback");
1137 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1138 seq_printf(seq, ",inode_readahead_blks=%u",
1139 sbi->s_inode_readahead_blks);
1141 if (test_opt(sb, DATA_ERR_ABORT))
1142 seq_puts(seq, ",data_err=abort");
1144 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1145 seq_puts(seq, ",noauto_da_alloc");
1147 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1148 seq_puts(seq, ",discard");
1150 if (test_opt(sb, NOLOAD))
1151 seq_puts(seq, ",norecovery");
1153 if (test_opt(sb, DIOREAD_NOLOCK))
1154 seq_puts(seq, ",dioread_nolock");
1156 if (test_opt(sb, BLOCK_VALIDITY) &&
1157 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1158 seq_puts(seq, ",block_validity");
1160 if (!test_opt(sb, INIT_INODE_TABLE))
1161 seq_puts(seq, ",noinit_itable");
1162 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1163 seq_printf(seq, ",init_itable=%u",
1164 (unsigned) sbi->s_li_wait_mult);
1166 ext4_show_quota_options(seq, sb);
1171 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1172 u64 ino, u32 generation)
1174 struct inode *inode;
1176 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1177 return ERR_PTR(-ESTALE);
1178 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1179 return ERR_PTR(-ESTALE);
1181 /* iget isn't really right if the inode is currently unallocated!!
1183 * ext4_read_inode will return a bad_inode if the inode had been
1184 * deleted, so we should be safe.
1186 * Currently we don't know the generation for parent directory, so
1187 * a generation of 0 means "accept any"
1189 inode = ext4_iget_normal(sb, ino);
1191 return ERR_CAST(inode);
1192 if (generation && inode->i_generation != generation) {
1194 return ERR_PTR(-ESTALE);
1200 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1201 int fh_len, int fh_type)
1203 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1204 ext4_nfs_get_inode);
1207 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1208 int fh_len, int fh_type)
1210 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1211 ext4_nfs_get_inode);
1215 * Try to release metadata pages (indirect blocks, directories) which are
1216 * mapped via the block device. Since these pages could have journal heads
1217 * which would prevent try_to_free_buffers() from freeing them, we must use
1218 * jbd2 layer's try_to_free_buffers() function to release them.
1220 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1223 journal_t *journal = EXT4_SB(sb)->s_journal;
1225 WARN_ON(PageChecked(page));
1226 if (!page_has_buffers(page))
1229 return jbd2_journal_try_to_free_buffers(journal, page,
1230 wait & ~__GFP_WAIT);
1231 return try_to_free_buffers(page);
1235 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1236 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1238 static int ext4_write_dquot(struct dquot *dquot);
1239 static int ext4_acquire_dquot(struct dquot *dquot);
1240 static int ext4_release_dquot(struct dquot *dquot);
1241 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1242 static int ext4_write_info(struct super_block *sb, int type);
1243 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1245 static int ext4_quota_off(struct super_block *sb, int type);
1246 static int ext4_quota_on_mount(struct super_block *sb, int type);
1247 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1248 size_t len, loff_t off);
1249 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1250 const char *data, size_t len, loff_t off);
1252 static const struct dquot_operations ext4_quota_operations = {
1253 .get_reserved_space = ext4_get_reserved_space,
1254 .write_dquot = ext4_write_dquot,
1255 .acquire_dquot = ext4_acquire_dquot,
1256 .release_dquot = ext4_release_dquot,
1257 .mark_dirty = ext4_mark_dquot_dirty,
1258 .write_info = ext4_write_info,
1259 .alloc_dquot = dquot_alloc,
1260 .destroy_dquot = dquot_destroy,
1263 static const struct quotactl_ops ext4_qctl_operations = {
1264 .quota_on = ext4_quota_on,
1265 .quota_off = ext4_quota_off,
1266 .quota_sync = dquot_quota_sync,
1267 .get_info = dquot_get_dqinfo,
1268 .set_info = dquot_set_dqinfo,
1269 .get_dqblk = dquot_get_dqblk,
1270 .set_dqblk = dquot_set_dqblk
1274 static const struct super_operations ext4_sops = {
1275 .alloc_inode = ext4_alloc_inode,
1276 .destroy_inode = ext4_destroy_inode,
1277 .write_inode = ext4_write_inode,
1278 .dirty_inode = ext4_dirty_inode,
1279 .drop_inode = ext4_drop_inode,
1280 .evict_inode = ext4_evict_inode,
1281 .put_super = ext4_put_super,
1282 .sync_fs = ext4_sync_fs,
1283 .freeze_fs = ext4_freeze,
1284 .unfreeze_fs = ext4_unfreeze,
1285 .statfs = ext4_statfs,
1286 .remount_fs = ext4_remount,
1287 .show_options = ext4_show_options,
1289 .quota_read = ext4_quota_read,
1290 .quota_write = ext4_quota_write,
1292 .bdev_try_to_free_page = bdev_try_to_free_page,
1295 static const struct super_operations ext4_nojournal_sops = {
1296 .alloc_inode = ext4_alloc_inode,
1297 .destroy_inode = ext4_destroy_inode,
1298 .write_inode = ext4_write_inode,
1299 .dirty_inode = ext4_dirty_inode,
1300 .drop_inode = ext4_drop_inode,
1301 .evict_inode = ext4_evict_inode,
1302 .write_super = ext4_write_super,
1303 .put_super = ext4_put_super,
1304 .statfs = ext4_statfs,
1305 .remount_fs = ext4_remount,
1306 .show_options = ext4_show_options,
1308 .quota_read = ext4_quota_read,
1309 .quota_write = ext4_quota_write,
1311 .bdev_try_to_free_page = bdev_try_to_free_page,
1314 static const struct export_operations ext4_export_ops = {
1315 .fh_to_dentry = ext4_fh_to_dentry,
1316 .fh_to_parent = ext4_fh_to_parent,
1317 .get_parent = ext4_get_parent,
1321 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1322 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1323 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1324 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1325 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1326 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1327 Opt_journal_update, Opt_journal_dev,
1328 Opt_journal_checksum, Opt_journal_async_commit,
1329 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1330 Opt_data_err_abort, Opt_data_err_ignore,
1331 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1332 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1333 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1334 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1335 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1336 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1337 Opt_inode_readahead_blks, Opt_journal_ioprio,
1338 Opt_dioread_nolock, Opt_dioread_lock,
1339 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1342 static const match_table_t tokens = {
1343 {Opt_bsd_df, "bsddf"},
1344 {Opt_minix_df, "minixdf"},
1345 {Opt_grpid, "grpid"},
1346 {Opt_grpid, "bsdgroups"},
1347 {Opt_nogrpid, "nogrpid"},
1348 {Opt_nogrpid, "sysvgroups"},
1349 {Opt_resgid, "resgid=%u"},
1350 {Opt_resuid, "resuid=%u"},
1352 {Opt_err_cont, "errors=continue"},
1353 {Opt_err_panic, "errors=panic"},
1354 {Opt_err_ro, "errors=remount-ro"},
1355 {Opt_nouid32, "nouid32"},
1356 {Opt_debug, "debug"},
1357 {Opt_oldalloc, "oldalloc"},
1358 {Opt_orlov, "orlov"},
1359 {Opt_user_xattr, "user_xattr"},
1360 {Opt_nouser_xattr, "nouser_xattr"},
1362 {Opt_noacl, "noacl"},
1363 {Opt_noload, "noload"},
1364 {Opt_noload, "norecovery"},
1367 {Opt_commit, "commit=%u"},
1368 {Opt_min_batch_time, "min_batch_time=%u"},
1369 {Opt_max_batch_time, "max_batch_time=%u"},
1370 {Opt_journal_update, "journal=update"},
1371 {Opt_journal_dev, "journal_dev=%u"},
1372 {Opt_journal_checksum, "journal_checksum"},
1373 {Opt_journal_async_commit, "journal_async_commit"},
1374 {Opt_abort, "abort"},
1375 {Opt_data_journal, "data=journal"},
1376 {Opt_data_ordered, "data=ordered"},
1377 {Opt_data_writeback, "data=writeback"},
1378 {Opt_data_err_abort, "data_err=abort"},
1379 {Opt_data_err_ignore, "data_err=ignore"},
1380 {Opt_offusrjquota, "usrjquota="},
1381 {Opt_usrjquota, "usrjquota=%s"},
1382 {Opt_offgrpjquota, "grpjquota="},
1383 {Opt_grpjquota, "grpjquota=%s"},
1384 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1385 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1386 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1387 {Opt_grpquota, "grpquota"},
1388 {Opt_noquota, "noquota"},
1389 {Opt_quota, "quota"},
1390 {Opt_usrquota, "usrquota"},
1391 {Opt_barrier, "barrier=%u"},
1392 {Opt_barrier, "barrier"},
1393 {Opt_nobarrier, "nobarrier"},
1394 {Opt_i_version, "i_version"},
1395 {Opt_stripe, "stripe=%u"},
1396 {Opt_resize, "resize"},
1397 {Opt_delalloc, "delalloc"},
1398 {Opt_nodelalloc, "nodelalloc"},
1399 {Opt_mblk_io_submit, "mblk_io_submit"},
1400 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1401 {Opt_block_validity, "block_validity"},
1402 {Opt_noblock_validity, "noblock_validity"},
1403 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1404 {Opt_journal_ioprio, "journal_ioprio=%u"},
1405 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1406 {Opt_auto_da_alloc, "auto_da_alloc"},
1407 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1408 {Opt_dioread_nolock, "dioread_nolock"},
1409 {Opt_dioread_lock, "dioread_lock"},
1410 {Opt_discard, "discard"},
1411 {Opt_nodiscard, "nodiscard"},
1412 {Opt_init_itable, "init_itable=%u"},
1413 {Opt_init_itable, "init_itable"},
1414 {Opt_noinit_itable, "noinit_itable"},
1418 static ext4_fsblk_t get_sb_block(void **data)
1420 ext4_fsblk_t sb_block;
1421 char *options = (char *) *data;
1423 if (!options || strncmp(options, "sb=", 3) != 0)
1424 return 1; /* Default location */
1427 /* TODO: use simple_strtoll with >32bit ext4 */
1428 sb_block = simple_strtoul(options, &options, 0);
1429 if (*options && *options != ',') {
1430 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1434 if (*options == ',')
1436 *data = (void *) options;
1441 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1442 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1443 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1446 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1448 struct ext4_sb_info *sbi = EXT4_SB(sb);
1451 if (sb_any_quota_loaded(sb) &&
1452 !sbi->s_qf_names[qtype]) {
1453 ext4_msg(sb, KERN_ERR,
1454 "Cannot change journaled "
1455 "quota options when quota turned on");
1458 qname = match_strdup(args);
1460 ext4_msg(sb, KERN_ERR,
1461 "Not enough memory for storing quotafile name");
1464 if (sbi->s_qf_names[qtype] &&
1465 strcmp(sbi->s_qf_names[qtype], qname)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "%s quota file already specified", QTYPE2NAME(qtype));
1471 sbi->s_qf_names[qtype] = qname;
1472 if (strchr(sbi->s_qf_names[qtype], '/')) {
1473 ext4_msg(sb, KERN_ERR,
1474 "quotafile must be on filesystem root");
1475 kfree(sbi->s_qf_names[qtype]);
1476 sbi->s_qf_names[qtype] = NULL;
1483 static int clear_qf_name(struct super_block *sb, int qtype)
1486 struct ext4_sb_info *sbi = EXT4_SB(sb);
1488 if (sb_any_quota_loaded(sb) &&
1489 sbi->s_qf_names[qtype]) {
1490 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1491 " when quota turned on");
1495 * The space will be released later when all options are confirmed
1498 sbi->s_qf_names[qtype] = NULL;
1503 static int parse_options(char *options, struct super_block *sb,
1504 unsigned long *journal_devnum,
1505 unsigned int *journal_ioprio,
1506 ext4_fsblk_t *n_blocks_count, int is_remount)
1508 struct ext4_sb_info *sbi = EXT4_SB(sb);
1510 substring_t args[MAX_OPT_ARGS];
1520 while ((p = strsep(&options, ",")) != NULL) {
1526 * Initialize args struct so we know whether arg was
1527 * found; some options take optional arguments.
1529 args[0].to = args[0].from = NULL;
1530 token = match_token(p, tokens, args);
1533 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1534 clear_opt(sb, MINIX_DF);
1537 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1538 set_opt(sb, MINIX_DF);
1542 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1547 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1548 clear_opt(sb, GRPID);
1552 if (match_int(&args[0], &option))
1554 sbi->s_resuid = option;
1557 if (match_int(&args[0], &option))
1559 sbi->s_resgid = option;
1562 /* handled by get_sb_block() instead of here */
1563 /* *sb_block = match_int(&args[0]); */
1566 clear_opt(sb, ERRORS_CONT);
1567 clear_opt(sb, ERRORS_RO);
1568 set_opt(sb, ERRORS_PANIC);
1571 clear_opt(sb, ERRORS_CONT);
1572 clear_opt(sb, ERRORS_PANIC);
1573 set_opt(sb, ERRORS_RO);
1576 clear_opt(sb, ERRORS_RO);
1577 clear_opt(sb, ERRORS_PANIC);
1578 set_opt(sb, ERRORS_CONT);
1581 set_opt(sb, NO_UID32);
1587 ext4_msg(sb, KERN_WARNING,
1588 "Ignoring deprecated oldalloc option");
1591 ext4_msg(sb, KERN_WARNING,
1592 "Ignoring deprecated orlov option");
1594 #ifdef CONFIG_EXT4_FS_XATTR
1595 case Opt_user_xattr:
1596 set_opt(sb, XATTR_USER);
1598 case Opt_nouser_xattr:
1599 clear_opt(sb, XATTR_USER);
1602 case Opt_user_xattr:
1603 case Opt_nouser_xattr:
1604 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1607 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1609 set_opt(sb, POSIX_ACL);
1612 clear_opt(sb, POSIX_ACL);
1617 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1620 case Opt_journal_update:
1622 /* Eventually we will want to be able to create
1623 a journal file here. For now, only allow the
1624 user to specify an existing inode to be the
1627 ext4_msg(sb, KERN_ERR,
1628 "Cannot specify journal on remount");
1631 set_opt(sb, UPDATE_JOURNAL);
1633 case Opt_journal_dev:
1635 ext4_msg(sb, KERN_ERR,
1636 "Cannot specify journal on remount");
1639 if (match_int(&args[0], &option))
1641 *journal_devnum = option;
1643 case Opt_journal_checksum:
1644 set_opt(sb, JOURNAL_CHECKSUM);
1646 case Opt_journal_async_commit:
1647 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1648 set_opt(sb, JOURNAL_CHECKSUM);
1651 set_opt(sb, NOLOAD);
1654 if (match_int(&args[0], &option))
1659 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1660 sbi->s_commit_interval = HZ * option;
1662 case Opt_max_batch_time:
1663 if (match_int(&args[0], &option))
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 "
1937 if (test_opt(sb, DIOREAD_NOLOCK)) {
1939 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1941 if (blocksize < PAGE_CACHE_SIZE) {
1942 ext4_msg(sb, KERN_ERR, "can't mount with "
1943 "dioread_nolock if block size != PAGE_SIZE");
1950 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1953 struct ext4_sb_info *sbi = EXT4_SB(sb);
1956 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1957 ext4_msg(sb, KERN_ERR, "revision level too high, "
1958 "forcing read-only mode");
1963 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1964 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1965 "running e2fsck is recommended");
1966 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1967 ext4_msg(sb, KERN_WARNING,
1968 "warning: mounting fs with errors, "
1969 "running e2fsck is recommended");
1970 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1971 le16_to_cpu(es->s_mnt_count) >=
1972 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1973 ext4_msg(sb, KERN_WARNING,
1974 "warning: maximal mount count reached, "
1975 "running e2fsck is recommended");
1976 else if (le32_to_cpu(es->s_checkinterval) &&
1977 (le32_to_cpu(es->s_lastcheck) +
1978 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1979 ext4_msg(sb, KERN_WARNING,
1980 "warning: checktime reached, "
1981 "running e2fsck is recommended");
1982 if (!sbi->s_journal)
1983 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1984 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1985 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1986 le16_add_cpu(&es->s_mnt_count, 1);
1987 es->s_mtime = cpu_to_le32(get_seconds());
1988 ext4_update_dynamic_rev(sb);
1990 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1992 ext4_commit_super(sb, 1);
1994 if (test_opt(sb, DEBUG))
1995 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1996 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1998 sbi->s_groups_count,
1999 EXT4_BLOCKS_PER_GROUP(sb),
2000 EXT4_INODES_PER_GROUP(sb),
2001 sbi->s_mount_opt, sbi->s_mount_opt2);
2003 cleancache_init_fs(sb);
2007 static int ext4_fill_flex_info(struct super_block *sb)
2009 struct ext4_sb_info *sbi = EXT4_SB(sb);
2010 struct ext4_group_desc *gdp = NULL;
2011 ext4_group_t flex_group_count;
2012 ext4_group_t flex_group;
2013 unsigned int groups_per_flex = 0;
2017 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2018 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2019 sbi->s_log_groups_per_flex = 0;
2022 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2024 /* We allocate both existing and potentially added groups */
2025 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2026 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2027 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2028 size = flex_group_count * sizeof(struct flex_groups);
2029 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2030 if (sbi->s_flex_groups == NULL) {
2031 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2036 for (i = 0; i < sbi->s_groups_count; i++) {
2037 gdp = ext4_get_group_desc(sb, i, NULL);
2039 flex_group = ext4_flex_group(sbi, i);
2040 atomic_add(ext4_free_inodes_count(sb, gdp),
2041 &sbi->s_flex_groups[flex_group].free_inodes);
2042 atomic64_add(ext4_free_group_clusters(sb, gdp),
2043 &sbi->s_flex_groups[flex_group].free_clusters);
2044 atomic_add(ext4_used_dirs_count(sb, gdp),
2045 &sbi->s_flex_groups[flex_group].used_dirs);
2053 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2054 struct ext4_group_desc *gdp)
2058 if (sbi->s_es->s_feature_ro_compat &
2059 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2060 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2061 __le32 le_group = cpu_to_le32(block_group);
2063 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2064 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2065 crc = crc16(crc, (__u8 *)gdp, offset);
2066 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2067 /* for checksum of struct ext4_group_desc do the rest...*/
2068 if ((sbi->s_es->s_feature_incompat &
2069 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2070 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2071 crc = crc16(crc, (__u8 *)gdp + offset,
2072 le16_to_cpu(sbi->s_es->s_desc_size) -
2076 return cpu_to_le16(crc);
2079 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2080 struct ext4_group_desc *gdp)
2082 if ((sbi->s_es->s_feature_ro_compat &
2083 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2084 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2090 /* Called at mount-time, super-block is locked */
2091 static int ext4_check_descriptors(struct super_block *sb,
2092 ext4_group_t *first_not_zeroed)
2094 struct ext4_sb_info *sbi = EXT4_SB(sb);
2095 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2096 ext4_fsblk_t last_block;
2097 ext4_fsblk_t block_bitmap;
2098 ext4_fsblk_t inode_bitmap;
2099 ext4_fsblk_t inode_table;
2100 int flexbg_flag = 0;
2101 ext4_group_t i, grp = sbi->s_groups_count;
2103 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2106 ext4_debug("Checking group descriptors");
2108 for (i = 0; i < sbi->s_groups_count; i++) {
2109 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2111 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2112 last_block = ext4_blocks_count(sbi->s_es) - 1;
2114 last_block = first_block +
2115 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2117 if ((grp == sbi->s_groups_count) &&
2118 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2121 block_bitmap = ext4_block_bitmap(sb, gdp);
2122 if (block_bitmap < first_block || block_bitmap > last_block) {
2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 "Block bitmap for group %u not in group "
2125 "(block %llu)!", i, block_bitmap);
2128 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2129 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2130 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 "Inode bitmap for group %u not in group "
2132 "(block %llu)!", i, inode_bitmap);
2135 inode_table = ext4_inode_table(sb, gdp);
2136 if (inode_table < first_block ||
2137 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2138 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2139 "Inode table for group %u not in group "
2140 "(block %llu)!", i, inode_table);
2143 ext4_lock_group(sb, i);
2144 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2145 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2146 "Checksum for group %u failed (%u!=%u)",
2147 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2148 gdp)), le16_to_cpu(gdp->bg_checksum));
2149 if (!(sb->s_flags & MS_RDONLY)) {
2150 ext4_unlock_group(sb, i);
2154 ext4_unlock_group(sb, i);
2156 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2158 if (NULL != first_not_zeroed)
2159 *first_not_zeroed = grp;
2161 ext4_free_blocks_count_set(sbi->s_es,
2162 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2163 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2167 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2168 * the superblock) which were deleted from all directories, but held open by
2169 * a process at the time of a crash. We walk the list and try to delete these
2170 * inodes at recovery time (only with a read-write filesystem).
2172 * In order to keep the orphan inode chain consistent during traversal (in
2173 * case of crash during recovery), we link each inode into the superblock
2174 * orphan list_head and handle it the same way as an inode deletion during
2175 * normal operation (which journals the operations for us).
2177 * We only do an iget() and an iput() on each inode, which is very safe if we
2178 * accidentally point at an in-use or already deleted inode. The worst that
2179 * can happen in this case is that we get a "bit already cleared" message from
2180 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2181 * e2fsck was run on this filesystem, and it must have already done the orphan
2182 * inode cleanup for us, so we can safely abort without any further action.
2184 static void ext4_orphan_cleanup(struct super_block *sb,
2185 struct ext4_super_block *es)
2187 unsigned int s_flags = sb->s_flags;
2188 int nr_orphans = 0, nr_truncates = 0;
2192 if (!es->s_last_orphan) {
2193 jbd_debug(4, "no orphan inodes to clean up\n");
2197 if (bdev_read_only(sb->s_bdev)) {
2198 ext4_msg(sb, KERN_ERR, "write access "
2199 "unavailable, skipping orphan cleanup");
2203 /* Check if feature set would not allow a r/w mount */
2204 if (!ext4_feature_set_ok(sb, 0)) {
2205 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2206 "unknown ROCOMPAT features");
2210 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2211 if (es->s_last_orphan)
2212 jbd_debug(1, "Errors on filesystem, "
2213 "clearing orphan list.\n");
2214 es->s_last_orphan = 0;
2215 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2219 if (s_flags & MS_RDONLY) {
2220 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2221 sb->s_flags &= ~MS_RDONLY;
2224 /* Needed for iput() to work correctly and not trash data */
2225 sb->s_flags |= MS_ACTIVE;
2226 /* Turn on quotas so that they are updated correctly */
2227 for (i = 0; i < MAXQUOTAS; i++) {
2228 if (EXT4_SB(sb)->s_qf_names[i]) {
2229 int ret = ext4_quota_on_mount(sb, i);
2231 ext4_msg(sb, KERN_ERR,
2232 "Cannot turn on journaled "
2233 "quota: error %d", ret);
2238 while (es->s_last_orphan) {
2239 struct inode *inode;
2241 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2242 if (IS_ERR(inode)) {
2243 es->s_last_orphan = 0;
2247 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2248 dquot_initialize(inode);
2249 if (inode->i_nlink) {
2250 ext4_msg(sb, KERN_DEBUG,
2251 "%s: truncating inode %lu to %lld bytes",
2252 __func__, inode->i_ino, inode->i_size);
2253 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2254 inode->i_ino, inode->i_size);
2255 mutex_lock(&inode->i_mutex);
2256 ext4_truncate(inode);
2257 mutex_unlock(&inode->i_mutex);
2260 ext4_msg(sb, KERN_DEBUG,
2261 "%s: deleting unreferenced inode %lu",
2262 __func__, inode->i_ino);
2263 jbd_debug(2, "deleting unreferenced inode %lu\n",
2267 iput(inode); /* The delete magic happens here! */
2270 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2273 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2274 PLURAL(nr_orphans));
2276 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2277 PLURAL(nr_truncates));
2279 /* Turn quotas off */
2280 for (i = 0; i < MAXQUOTAS; i++) {
2281 if (sb_dqopt(sb)->files[i])
2282 dquot_quota_off(sb, i);
2285 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2289 * Maximal extent format file size.
2290 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2291 * extent format containers, within a sector_t, and within i_blocks
2292 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2293 * so that won't be a limiting factor.
2295 * However there is other limiting factor. We do store extents in the form
2296 * of starting block and length, hence the resulting length of the extent
2297 * covering maximum file size must fit into on-disk format containers as
2298 * well. Given that length is always by 1 unit bigger than max unit (because
2299 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2301 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2303 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2306 loff_t upper_limit = MAX_LFS_FILESIZE;
2308 /* small i_blocks in vfs inode? */
2309 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2311 * CONFIG_LBDAF is not enabled implies the inode
2312 * i_block represent total blocks in 512 bytes
2313 * 32 == size of vfs inode i_blocks * 8
2315 upper_limit = (1LL << 32) - 1;
2317 /* total blocks in file system block size */
2318 upper_limit >>= (blkbits - 9);
2319 upper_limit <<= blkbits;
2323 * 32-bit extent-start container, ee_block. We lower the maxbytes
2324 * by one fs block, so ee_len can cover the extent of maximum file
2327 res = (1LL << 32) - 1;
2330 /* Sanity check against vm- & vfs- imposed limits */
2331 if (res > upper_limit)
2338 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2339 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2340 * We need to be 1 filesystem block less than the 2^48 sector limit.
2342 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2344 loff_t res = EXT4_NDIR_BLOCKS;
2347 /* This is calculated to be the largest file size for a dense, block
2348 * mapped file such that the file's total number of 512-byte sectors,
2349 * including data and all indirect blocks, does not exceed (2^48 - 1).
2351 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2352 * number of 512-byte sectors of the file.
2355 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2357 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2358 * the inode i_block field represents total file blocks in
2359 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2361 upper_limit = (1LL << 32) - 1;
2363 /* total blocks in file system block size */
2364 upper_limit >>= (bits - 9);
2368 * We use 48 bit ext4_inode i_blocks
2369 * With EXT4_HUGE_FILE_FL set the i_blocks
2370 * represent total number of blocks in
2371 * file system block size
2373 upper_limit = (1LL << 48) - 1;
2377 /* indirect blocks */
2379 /* double indirect blocks */
2380 meta_blocks += 1 + (1LL << (bits-2));
2381 /* tripple indirect blocks */
2382 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2384 upper_limit -= meta_blocks;
2385 upper_limit <<= bits;
2387 res += 1LL << (bits-2);
2388 res += 1LL << (2*(bits-2));
2389 res += 1LL << (3*(bits-2));
2391 if (res > upper_limit)
2394 if (res > MAX_LFS_FILESIZE)
2395 res = MAX_LFS_FILESIZE;
2400 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2401 ext4_fsblk_t logical_sb_block, int nr)
2403 struct ext4_sb_info *sbi = EXT4_SB(sb);
2404 ext4_group_t bg, first_meta_bg;
2407 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2409 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2411 return logical_sb_block + nr + 1;
2412 bg = sbi->s_desc_per_block * nr;
2413 if (ext4_bg_has_super(sb, bg))
2416 return (has_super + ext4_group_first_block_no(sb, bg));
2420 * ext4_get_stripe_size: Get the stripe size.
2421 * @sbi: In memory super block info
2423 * If we have specified it via mount option, then
2424 * use the mount option value. If the value specified at mount time is
2425 * greater than the blocks per group use the super block value.
2426 * If the super block value is greater than blocks per group return 0.
2427 * Allocator needs it be less than blocks per group.
2430 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2432 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2433 unsigned long stripe_width =
2434 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2437 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2438 ret = sbi->s_stripe;
2439 else if (stripe_width <= sbi->s_blocks_per_group)
2441 else if (stride <= sbi->s_blocks_per_group)
2447 * If the stripe width is 1, this makes no sense and
2448 * we set it to 0 to turn off stripe handling code.
2459 struct attribute attr;
2460 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2461 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2462 const char *, size_t);
2466 static int parse_strtoul(const char *buf,
2467 unsigned long max, unsigned long *value)
2471 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2472 endp = skip_spaces(endp);
2473 if (*endp || *value > max)
2479 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2483 return snprintf(buf, PAGE_SIZE, "%llu\n",
2485 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2488 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2489 struct ext4_sb_info *sbi, char *buf)
2491 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2493 if (!sb->s_bdev->bd_part)
2494 return snprintf(buf, PAGE_SIZE, "0\n");
2495 return snprintf(buf, PAGE_SIZE, "%lu\n",
2496 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2497 sbi->s_sectors_written_start) >> 1);
2500 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2501 struct ext4_sb_info *sbi, char *buf)
2503 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2505 if (!sb->s_bdev->bd_part)
2506 return snprintf(buf, PAGE_SIZE, "0\n");
2507 return snprintf(buf, PAGE_SIZE, "%llu\n",
2508 (unsigned long long)(sbi->s_kbytes_written +
2509 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2510 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2513 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi,
2515 const char *buf, size_t count)
2519 if (parse_strtoul(buf, 0x40000000, &t))
2522 if (t && !is_power_of_2(t))
2525 sbi->s_inode_readahead_blks = t;
2529 static ssize_t sbi_ui_show(struct ext4_attr *a,
2530 struct ext4_sb_info *sbi, char *buf)
2532 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2534 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2537 static ssize_t sbi_ui_store(struct ext4_attr *a,
2538 struct ext4_sb_info *sbi,
2539 const char *buf, size_t count)
2541 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2544 if (parse_strtoul(buf, 0xffffffff, &t))
2550 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2551 static struct ext4_attr ext4_attr_##_name = { \
2552 .attr = {.name = __stringify(_name), .mode = _mode }, \
2555 .offset = offsetof(struct ext4_sb_info, _elname), \
2557 #define EXT4_ATTR(name, mode, show, store) \
2558 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2560 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2561 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2562 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2563 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2564 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2565 #define ATTR_LIST(name) &ext4_attr_##name.attr
2567 EXT4_RO_ATTR(delayed_allocation_blocks);
2568 EXT4_RO_ATTR(session_write_kbytes);
2569 EXT4_RO_ATTR(lifetime_write_kbytes);
2570 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2571 inode_readahead_blks_store, s_inode_readahead_blks);
2572 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2573 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2574 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2575 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2576 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2577 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2578 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2579 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2581 static struct attribute *ext4_attrs[] = {
2582 ATTR_LIST(delayed_allocation_blocks),
2583 ATTR_LIST(session_write_kbytes),
2584 ATTR_LIST(lifetime_write_kbytes),
2585 ATTR_LIST(inode_readahead_blks),
2586 ATTR_LIST(inode_goal),
2587 ATTR_LIST(mb_stats),
2588 ATTR_LIST(mb_max_to_scan),
2589 ATTR_LIST(mb_min_to_scan),
2590 ATTR_LIST(mb_order2_req),
2591 ATTR_LIST(mb_stream_req),
2592 ATTR_LIST(mb_group_prealloc),
2593 ATTR_LIST(max_writeback_mb_bump),
2597 /* Features this copy of ext4 supports */
2598 EXT4_INFO_ATTR(lazy_itable_init);
2599 EXT4_INFO_ATTR(batched_discard);
2601 static struct attribute *ext4_feat_attrs[] = {
2602 ATTR_LIST(lazy_itable_init),
2603 ATTR_LIST(batched_discard),
2607 static ssize_t ext4_attr_show(struct kobject *kobj,
2608 struct attribute *attr, char *buf)
2610 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2612 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2614 return a->show ? a->show(a, sbi, buf) : 0;
2617 static ssize_t ext4_attr_store(struct kobject *kobj,
2618 struct attribute *attr,
2619 const char *buf, size_t len)
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2623 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2625 return a->store ? a->store(a, sbi, buf, len) : 0;
2628 static void ext4_sb_release(struct kobject *kobj)
2630 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2632 complete(&sbi->s_kobj_unregister);
2635 static const struct sysfs_ops ext4_attr_ops = {
2636 .show = ext4_attr_show,
2637 .store = ext4_attr_store,
2640 static struct kobj_type ext4_ktype = {
2641 .default_attrs = ext4_attrs,
2642 .sysfs_ops = &ext4_attr_ops,
2643 .release = ext4_sb_release,
2646 static void ext4_feat_release(struct kobject *kobj)
2648 complete(&ext4_feat->f_kobj_unregister);
2651 static struct kobj_type ext4_feat_ktype = {
2652 .default_attrs = ext4_feat_attrs,
2653 .sysfs_ops = &ext4_attr_ops,
2654 .release = ext4_feat_release,
2658 * Check whether this filesystem can be mounted based on
2659 * the features present and the RDONLY/RDWR mount requested.
2660 * Returns 1 if this filesystem can be mounted as requested,
2661 * 0 if it cannot be.
2663 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2665 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2666 ext4_msg(sb, KERN_ERR,
2667 "Couldn't mount because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2670 ~EXT4_FEATURE_INCOMPAT_SUPP));
2677 /* Check that feature set is OK for a read-write mount */
2678 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2679 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2680 "unsupported optional features (%x)",
2681 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2682 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2686 * Large file size enabled file system can only be mounted
2687 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2689 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2690 if (sizeof(blkcnt_t) < sizeof(u64)) {
2691 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2692 "cannot be mounted RDWR without "
2697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2698 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2699 ext4_msg(sb, KERN_ERR,
2700 "Can't support bigalloc feature without "
2701 "extents feature\n");
2708 * This function is called once a day if we have errors logged
2709 * on the file system
2711 static void print_daily_error_info(unsigned long arg)
2713 struct super_block *sb = (struct super_block *) arg;
2714 struct ext4_sb_info *sbi;
2715 struct ext4_super_block *es;
2720 if (es->s_error_count)
2721 /* fsck newer than v1.41.13 is needed to clean this condition. */
2722 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2723 le32_to_cpu(es->s_error_count));
2724 if (es->s_first_error_time) {
2725 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2726 sb->s_id, le32_to_cpu(es->s_first_error_time),
2727 (int) sizeof(es->s_first_error_func),
2728 es->s_first_error_func,
2729 le32_to_cpu(es->s_first_error_line));
2730 if (es->s_first_error_ino)
2731 printk(": inode %u",
2732 le32_to_cpu(es->s_first_error_ino));
2733 if (es->s_first_error_block)
2734 printk(": block %llu", (unsigned long long)
2735 le64_to_cpu(es->s_first_error_block));
2738 if (es->s_last_error_time) {
2739 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2740 sb->s_id, le32_to_cpu(es->s_last_error_time),
2741 (int) sizeof(es->s_last_error_func),
2742 es->s_last_error_func,
2743 le32_to_cpu(es->s_last_error_line));
2744 if (es->s_last_error_ino)
2745 printk(": inode %u",
2746 le32_to_cpu(es->s_last_error_ino));
2747 if (es->s_last_error_block)
2748 printk(": block %llu", (unsigned long long)
2749 le64_to_cpu(es->s_last_error_block));
2752 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2755 /* Find next suitable group and run ext4_init_inode_table */
2756 static int ext4_run_li_request(struct ext4_li_request *elr)
2758 struct ext4_group_desc *gdp = NULL;
2759 ext4_group_t group, ngroups;
2760 struct super_block *sb;
2761 unsigned long timeout = 0;
2765 ngroups = EXT4_SB(sb)->s_groups_count;
2767 for (group = elr->lr_next_group; group < ngroups; group++) {
2768 gdp = ext4_get_group_desc(sb, group, NULL);
2774 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2778 if (group == ngroups)
2783 ret = ext4_init_inode_table(sb, group,
2784 elr->lr_timeout ? 0 : 1);
2785 if (elr->lr_timeout == 0) {
2786 timeout = (jiffies - timeout) *
2787 elr->lr_sbi->s_li_wait_mult;
2788 elr->lr_timeout = timeout;
2790 elr->lr_next_sched = jiffies + elr->lr_timeout;
2791 elr->lr_next_group = group + 1;
2798 * Remove lr_request from the list_request and free the
2799 * request structure. Should be called with li_list_mtx held
2801 static void ext4_remove_li_request(struct ext4_li_request *elr)
2803 struct ext4_sb_info *sbi;
2810 list_del(&elr->lr_request);
2811 sbi->s_li_request = NULL;
2815 static void ext4_unregister_li_request(struct super_block *sb)
2817 mutex_lock(&ext4_li_mtx);
2818 if (!ext4_li_info) {
2819 mutex_unlock(&ext4_li_mtx);
2823 mutex_lock(&ext4_li_info->li_list_mtx);
2824 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2825 mutex_unlock(&ext4_li_info->li_list_mtx);
2826 mutex_unlock(&ext4_li_mtx);
2829 static struct task_struct *ext4_lazyinit_task;
2832 * This is the function where ext4lazyinit thread lives. It walks
2833 * through the request list searching for next scheduled filesystem.
2834 * When such a fs is found, run the lazy initialization request
2835 * (ext4_rn_li_request) and keep track of the time spend in this
2836 * function. Based on that time we compute next schedule time of
2837 * the request. When walking through the list is complete, compute
2838 * next waking time and put itself into sleep.
2840 static int ext4_lazyinit_thread(void *arg)
2842 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2843 struct list_head *pos, *n;
2844 struct ext4_li_request *elr;
2845 unsigned long next_wakeup, cur;
2847 BUG_ON(NULL == eli);
2851 next_wakeup = MAX_JIFFY_OFFSET;
2853 mutex_lock(&eli->li_list_mtx);
2854 if (list_empty(&eli->li_request_list)) {
2855 mutex_unlock(&eli->li_list_mtx);
2859 list_for_each_safe(pos, n, &eli->li_request_list) {
2860 elr = list_entry(pos, struct ext4_li_request,
2863 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2864 if (ext4_run_li_request(elr) != 0) {
2865 /* error, remove the lazy_init job */
2866 ext4_remove_li_request(elr);
2871 if (time_before(elr->lr_next_sched, next_wakeup))
2872 next_wakeup = elr->lr_next_sched;
2874 mutex_unlock(&eli->li_list_mtx);
2876 if (freezing(current))
2880 if ((time_after_eq(cur, next_wakeup)) ||
2881 (MAX_JIFFY_OFFSET == next_wakeup)) {
2886 schedule_timeout_interruptible(next_wakeup - cur);
2888 if (kthread_should_stop()) {
2889 ext4_clear_request_list();
2896 * It looks like the request list is empty, but we need
2897 * to check it under the li_list_mtx lock, to prevent any
2898 * additions into it, and of course we should lock ext4_li_mtx
2899 * to atomically free the list and ext4_li_info, because at
2900 * this point another ext4 filesystem could be registering
2903 mutex_lock(&ext4_li_mtx);
2904 mutex_lock(&eli->li_list_mtx);
2905 if (!list_empty(&eli->li_request_list)) {
2906 mutex_unlock(&eli->li_list_mtx);
2907 mutex_unlock(&ext4_li_mtx);
2910 mutex_unlock(&eli->li_list_mtx);
2911 kfree(ext4_li_info);
2912 ext4_li_info = NULL;
2913 mutex_unlock(&ext4_li_mtx);
2918 static void ext4_clear_request_list(void)
2920 struct list_head *pos, *n;
2921 struct ext4_li_request *elr;
2923 mutex_lock(&ext4_li_info->li_list_mtx);
2924 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2925 elr = list_entry(pos, struct ext4_li_request,
2927 ext4_remove_li_request(elr);
2929 mutex_unlock(&ext4_li_info->li_list_mtx);
2932 static int ext4_run_lazyinit_thread(void)
2934 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2935 ext4_li_info, "ext4lazyinit");
2936 if (IS_ERR(ext4_lazyinit_task)) {
2937 int err = PTR_ERR(ext4_lazyinit_task);
2938 ext4_clear_request_list();
2939 kfree(ext4_li_info);
2940 ext4_li_info = NULL;
2941 printk(KERN_CRIT "EXT4: error %d creating inode table "
2942 "initialization thread\n",
2946 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2951 * Check whether it make sense to run itable init. thread or not.
2952 * If there is at least one uninitialized inode table, return
2953 * corresponding group number, else the loop goes through all
2954 * groups and return total number of groups.
2956 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2958 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2959 struct ext4_group_desc *gdp = NULL;
2961 for (group = 0; group < ngroups; group++) {
2962 gdp = ext4_get_group_desc(sb, group, NULL);
2966 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2973 static int ext4_li_info_new(void)
2975 struct ext4_lazy_init *eli = NULL;
2977 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2981 INIT_LIST_HEAD(&eli->li_request_list);
2982 mutex_init(&eli->li_list_mtx);
2984 eli->li_state |= EXT4_LAZYINIT_QUIT;
2991 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2994 struct ext4_sb_info *sbi = EXT4_SB(sb);
2995 struct ext4_li_request *elr;
2998 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3004 elr->lr_next_group = start;
3007 * Randomize first schedule time of the request to
3008 * spread the inode table initialization requests
3011 get_random_bytes(&rnd, sizeof(rnd));
3012 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3013 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3018 static int ext4_register_li_request(struct super_block *sb,
3019 ext4_group_t first_not_zeroed)
3021 struct ext4_sb_info *sbi = EXT4_SB(sb);
3022 struct ext4_li_request *elr;
3023 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3026 if (sbi->s_li_request != NULL) {
3028 * Reset timeout so it can be computed again, because
3029 * s_li_wait_mult might have changed.
3031 sbi->s_li_request->lr_timeout = 0;
3035 if (first_not_zeroed == ngroups ||
3036 (sb->s_flags & MS_RDONLY) ||
3037 !test_opt(sb, INIT_INODE_TABLE))
3040 elr = ext4_li_request_new(sb, first_not_zeroed);
3044 mutex_lock(&ext4_li_mtx);
3046 if (NULL == ext4_li_info) {
3047 ret = ext4_li_info_new();
3052 mutex_lock(&ext4_li_info->li_list_mtx);
3053 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3054 mutex_unlock(&ext4_li_info->li_list_mtx);
3056 sbi->s_li_request = elr;
3058 * set elr to NULL here since it has been inserted to
3059 * the request_list and the removal and free of it is
3060 * handled by ext4_clear_request_list from now on.
3064 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3065 ret = ext4_run_lazyinit_thread();
3070 mutex_unlock(&ext4_li_mtx);
3077 * We do not need to lock anything since this is called on
3080 static void ext4_destroy_lazyinit_thread(void)
3083 * If thread exited earlier
3084 * there's nothing to be done.
3086 if (!ext4_li_info || !ext4_lazyinit_task)
3089 kthread_stop(ext4_lazyinit_task);
3093 * Note: calculating the overhead so we can be compatible with
3094 * historical BSD practice is quite difficult in the face of
3095 * clusters/bigalloc. This is because multiple metadata blocks from
3096 * different block group can end up in the same allocation cluster.
3097 * Calculating the exact overhead in the face of clustered allocation
3098 * requires either O(all block bitmaps) in memory or O(number of block
3099 * groups**2) in time. We will still calculate the superblock for
3100 * older file systems --- and if we come across with a bigalloc file
3101 * system with zero in s_overhead_clusters the estimate will be close to
3102 * correct especially for very large cluster sizes --- but for newer
3103 * file systems, it's better to calculate this figure once at mkfs
3104 * time, and store it in the superblock. If the superblock value is
3105 * present (even for non-bigalloc file systems), we will use it.
3107 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3110 struct ext4_sb_info *sbi = EXT4_SB(sb);
3111 struct ext4_group_desc *gdp;
3112 ext4_fsblk_t first_block, last_block, b;
3113 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3114 int s, j, count = 0;
3116 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3117 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3118 sbi->s_itb_per_group + 2);
3120 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3121 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3122 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3123 for (i = 0; i < ngroups; i++) {
3124 gdp = ext4_get_group_desc(sb, i, NULL);
3125 b = ext4_block_bitmap(sb, gdp);
3126 if (b >= first_block && b <= last_block) {
3127 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3130 b = ext4_inode_bitmap(sb, gdp);
3131 if (b >= first_block && b <= last_block) {
3132 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3135 b = ext4_inode_table(sb, gdp);
3136 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3137 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3138 int c = EXT4_B2C(sbi, b - first_block);
3139 ext4_set_bit(c, buf);
3145 if (ext4_bg_has_super(sb, grp)) {
3146 ext4_set_bit(s++, buf);
3149 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3150 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3156 return EXT4_CLUSTERS_PER_GROUP(sb) -
3157 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3161 * Compute the overhead and stash it in sbi->s_overhead
3163 int ext4_calculate_overhead(struct super_block *sb)
3165 struct ext4_sb_info *sbi = EXT4_SB(sb);
3166 struct ext4_super_block *es = sbi->s_es;
3167 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3168 ext4_fsblk_t overhead = 0;
3169 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3171 memset(buf, 0, PAGE_SIZE);
3176 * Compute the overhead (FS structures). This is constant
3177 * for a given filesystem unless the number of block groups
3178 * changes so we cache the previous value until it does.
3182 * All of the blocks before first_data_block are overhead
3184 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3187 * Add the overhead found in each block group
3189 for (i = 0; i < ngroups; i++) {
3192 blks = count_overhead(sb, i, buf);
3195 memset(buf, 0, PAGE_SIZE);
3198 sbi->s_overhead = overhead;
3200 free_page((unsigned long) buf);
3204 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3206 char *orig_data = kstrdup(data, GFP_KERNEL);
3207 struct buffer_head *bh;
3208 struct ext4_super_block *es = NULL;
3209 struct ext4_sb_info *sbi;
3211 ext4_fsblk_t sb_block = get_sb_block(&data);
3212 ext4_fsblk_t logical_sb_block;
3213 unsigned long offset = 0;
3214 unsigned long journal_devnum = 0;
3215 unsigned long def_mount_opts;
3220 int blocksize, clustersize;
3221 unsigned int db_count;
3223 int needs_recovery, has_huge_files, has_bigalloc;
3226 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3227 ext4_group_t first_not_zeroed;
3229 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3233 sbi->s_blockgroup_lock =
3234 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3235 if (!sbi->s_blockgroup_lock) {
3239 sb->s_fs_info = sbi;
3240 sbi->s_mount_opt = 0;
3241 sbi->s_resuid = EXT4_DEF_RESUID;
3242 sbi->s_resgid = EXT4_DEF_RESGID;
3243 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3244 sbi->s_sb_block = sb_block;
3245 if (sb->s_bdev->bd_part)
3246 sbi->s_sectors_written_start =
3247 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3249 /* Cleanup superblock name */
3250 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3254 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3256 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3261 * The ext4 superblock will not be buffer aligned for other than 1kB
3262 * block sizes. We need to calculate the offset from buffer start.
3264 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3265 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3266 offset = do_div(logical_sb_block, blocksize);
3268 logical_sb_block = sb_block;
3271 if (!(bh = sb_bread(sb, logical_sb_block))) {
3272 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3276 * Note: s_es must be initialized as soon as possible because
3277 * some ext4 macro-instructions depend on its value
3279 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3281 sb->s_magic = le16_to_cpu(es->s_magic);
3282 if (sb->s_magic != EXT4_SUPER_MAGIC)
3284 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3286 /* Set defaults before we parse the mount options */
3287 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3288 set_opt(sb, INIT_INODE_TABLE);
3289 if (def_mount_opts & EXT4_DEFM_DEBUG)
3291 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3292 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3296 if (def_mount_opts & EXT4_DEFM_UID16)
3297 set_opt(sb, NO_UID32);
3298 /* xattr user namespace & acls are now defaulted on */
3299 #ifdef CONFIG_EXT4_FS_XATTR
3300 set_opt(sb, XATTR_USER);
3302 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3303 set_opt(sb, POSIX_ACL);
3305 set_opt(sb, MBLK_IO_SUBMIT);
3306 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3307 set_opt(sb, JOURNAL_DATA);
3308 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3309 set_opt(sb, ORDERED_DATA);
3310 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3311 set_opt(sb, WRITEBACK_DATA);
3313 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3314 set_opt(sb, ERRORS_PANIC);
3315 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3316 set_opt(sb, ERRORS_CONT);
3318 set_opt(sb, ERRORS_RO);
3319 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3320 set_opt(sb, BLOCK_VALIDITY);
3321 if (def_mount_opts & EXT4_DEFM_DISCARD)
3322 set_opt(sb, DISCARD);
3324 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3325 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3326 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3327 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3328 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3330 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3331 set_opt(sb, BARRIER);
3334 * enable delayed allocation by default
3335 * Use -o nodelalloc to turn it off
3337 if (!IS_EXT3_SB(sb) &&
3338 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3339 set_opt(sb, DELALLOC);
3342 * set default s_li_wait_mult for lazyinit, for the case there is
3343 * no mount option specified.
3345 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3347 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3348 &journal_devnum, &journal_ioprio, NULL, 0)) {
3349 ext4_msg(sb, KERN_WARNING,
3350 "failed to parse options in superblock: %s",
3351 sbi->s_es->s_mount_opts);
3353 if (!parse_options((char *) data, sb, &journal_devnum,
3354 &journal_ioprio, NULL, 0))
3357 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3358 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3359 "with data=journal disables delayed "
3360 "allocation and O_DIRECT support!\n");
3361 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3362 ext4_msg(sb, KERN_ERR, "can't mount with "
3363 "both data=journal and delalloc");
3366 if (test_opt(sb, DIOREAD_NOLOCK)) {
3367 ext4_msg(sb, KERN_ERR, "can't mount with "
3368 "both data=journal and dioread_nolock");
3371 if (test_opt(sb, DELALLOC))
3372 clear_opt(sb, DELALLOC);
3375 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3376 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3378 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3379 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3380 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3381 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3382 ext4_msg(sb, KERN_WARNING,
3383 "feature flags set on rev 0 fs, "
3384 "running e2fsck is recommended");
3386 if (IS_EXT2_SB(sb)) {
3387 if (ext2_feature_set_ok(sb))
3388 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3389 "using the ext4 subsystem");
3391 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3392 "to feature incompatibilities");
3397 if (IS_EXT3_SB(sb)) {
3398 if (ext3_feature_set_ok(sb))
3399 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3400 "using the ext4 subsystem");
3402 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3403 "to feature incompatibilities");
3409 * Check feature flags regardless of the revision level, since we
3410 * previously didn't change the revision level when setting the flags,
3411 * so there is a chance incompat flags are set on a rev 0 filesystem.
3413 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3416 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3417 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3418 blocksize > EXT4_MAX_BLOCK_SIZE) {
3419 ext4_msg(sb, KERN_ERR,
3420 "Unsupported filesystem blocksize %d", blocksize);
3424 if (sb->s_blocksize != blocksize) {
3425 /* Validate the filesystem blocksize */
3426 if (!sb_set_blocksize(sb, blocksize)) {
3427 ext4_msg(sb, KERN_ERR, "bad block size %d",
3433 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3434 offset = do_div(logical_sb_block, blocksize);
3435 bh = sb_bread(sb, logical_sb_block);
3437 ext4_msg(sb, KERN_ERR,
3438 "Can't read superblock on 2nd try");
3441 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3443 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3444 ext4_msg(sb, KERN_ERR,
3445 "Magic mismatch, very weird!");
3450 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3451 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3452 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3454 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3456 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3457 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3458 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3460 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3461 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3462 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3463 (!is_power_of_2(sbi->s_inode_size)) ||
3464 (sbi->s_inode_size > blocksize)) {
3465 ext4_msg(sb, KERN_ERR,
3466 "unsupported inode size: %d",
3470 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3471 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3474 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3475 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3476 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3477 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3478 !is_power_of_2(sbi->s_desc_size)) {
3479 ext4_msg(sb, KERN_ERR,
3480 "unsupported descriptor size %lu",
3485 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3487 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3488 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3489 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3492 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3493 if (sbi->s_inodes_per_block == 0)
3495 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3496 sbi->s_inodes_per_block;
3497 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3499 sbi->s_mount_state = le16_to_cpu(es->s_state);
3500 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3501 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3503 for (i = 0; i < 4; i++)
3504 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3505 sbi->s_def_hash_version = es->s_def_hash_version;
3506 i = le32_to_cpu(es->s_flags);
3507 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3508 sbi->s_hash_unsigned = 3;
3509 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3510 #ifdef __CHAR_UNSIGNED__
3511 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3512 sbi->s_hash_unsigned = 3;
3514 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3519 /* Handle clustersize */
3520 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3521 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3522 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3524 if (clustersize < blocksize) {
3525 ext4_msg(sb, KERN_ERR,
3526 "cluster size (%d) smaller than "
3527 "block size (%d)", clustersize, blocksize);
3530 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3531 le32_to_cpu(es->s_log_block_size);
3532 sbi->s_clusters_per_group =
3533 le32_to_cpu(es->s_clusters_per_group);
3534 if (sbi->s_clusters_per_group > blocksize * 8) {
3535 ext4_msg(sb, KERN_ERR,
3536 "#clusters per group too big: %lu",
3537 sbi->s_clusters_per_group);
3540 if (sbi->s_blocks_per_group !=
3541 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3542 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3543 "clusters per group (%lu) inconsistent",
3544 sbi->s_blocks_per_group,
3545 sbi->s_clusters_per_group);
3549 if (clustersize != blocksize) {
3550 ext4_warning(sb, "fragment/cluster size (%d) != "
3551 "block size (%d)", clustersize,
3553 clustersize = blocksize;
3555 if (sbi->s_blocks_per_group > blocksize * 8) {
3556 ext4_msg(sb, KERN_ERR,
3557 "#blocks per group too big: %lu",
3558 sbi->s_blocks_per_group);
3561 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3562 sbi->s_cluster_bits = 0;
3564 sbi->s_cluster_ratio = clustersize / blocksize;
3566 if (sbi->s_inodes_per_group > blocksize * 8) {
3567 ext4_msg(sb, KERN_ERR,
3568 "#inodes per group too big: %lu",
3569 sbi->s_inodes_per_group);
3574 * Test whether we have more sectors than will fit in sector_t,
3575 * and whether the max offset is addressable by the page cache.
3577 err = generic_check_addressable(sb->s_blocksize_bits,
3578 ext4_blocks_count(es));
3580 ext4_msg(sb, KERN_ERR, "filesystem"
3581 " too large to mount safely on this system");
3582 if (sizeof(sector_t) < 8)
3583 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3588 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3591 /* check blocks count against device size */
3592 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3593 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3594 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3595 "exceeds size of device (%llu blocks)",
3596 ext4_blocks_count(es), blocks_count);
3601 * It makes no sense for the first data block to be beyond the end
3602 * of the filesystem.
3604 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3605 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3606 "block %u is beyond end of filesystem (%llu)",
3607 le32_to_cpu(es->s_first_data_block),
3608 ext4_blocks_count(es));
3611 blocks_count = (ext4_blocks_count(es) -
3612 le32_to_cpu(es->s_first_data_block) +
3613 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3614 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3615 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3616 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3617 "(block count %llu, first data block %u, "
3618 "blocks per group %lu)", sbi->s_groups_count,
3619 ext4_blocks_count(es),
3620 le32_to_cpu(es->s_first_data_block),
3621 EXT4_BLOCKS_PER_GROUP(sb));
3624 sbi->s_groups_count = blocks_count;
3625 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3626 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3627 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3628 EXT4_DESC_PER_BLOCK(sb);
3629 sbi->s_group_desc = ext4_kvmalloc(db_count *
3630 sizeof(struct buffer_head *),
3632 if (sbi->s_group_desc == NULL) {
3633 ext4_msg(sb, KERN_ERR, "not enough memory");
3638 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3640 bgl_lock_init(sbi->s_blockgroup_lock);
3642 for (i = 0; i < db_count; i++) {
3643 block = descriptor_loc(sb, logical_sb_block, i);
3644 sbi->s_group_desc[i] = sb_bread(sb, block);
3645 if (!sbi->s_group_desc[i]) {
3646 ext4_msg(sb, KERN_ERR,
3647 "can't read group descriptor %d", i);
3652 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3653 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3657 if (!ext4_fill_flex_info(sb)) {
3658 ext4_msg(sb, KERN_ERR,
3659 "unable to initialize "
3660 "flex_bg meta info!");
3664 sbi->s_gdb_count = db_count;
3665 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3666 spin_lock_init(&sbi->s_next_gen_lock);
3668 init_timer(&sbi->s_err_report);
3669 sbi->s_err_report.function = print_daily_error_info;
3670 sbi->s_err_report.data = (unsigned long) sb;
3672 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3673 ext4_count_free_clusters(sb));
3675 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3676 ext4_count_free_inodes(sb));
3679 err = percpu_counter_init(&sbi->s_dirs_counter,
3680 ext4_count_dirs(sb));
3683 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3686 ext4_msg(sb, KERN_ERR, "insufficient memory");
3690 sbi->s_stripe = ext4_get_stripe_size(sbi);
3691 sbi->s_max_writeback_mb_bump = 128;
3694 * set up enough so that it can read an inode
3696 if (!test_opt(sb, NOLOAD) &&
3697 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3698 sb->s_op = &ext4_sops;
3700 sb->s_op = &ext4_nojournal_sops;
3701 sb->s_export_op = &ext4_export_ops;
3702 sb->s_xattr = ext4_xattr_handlers;
3704 sb->s_qcop = &ext4_qctl_operations;
3705 sb->dq_op = &ext4_quota_operations;
3707 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3709 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3710 mutex_init(&sbi->s_orphan_lock);
3711 sbi->s_resize_flags = 0;
3715 needs_recovery = (es->s_last_orphan != 0 ||
3716 EXT4_HAS_INCOMPAT_FEATURE(sb,
3717 EXT4_FEATURE_INCOMPAT_RECOVER));
3719 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3720 !(sb->s_flags & MS_RDONLY))
3721 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3725 * The first inode we look at is the journal inode. Don't try
3726 * root first: it may be modified in the journal!
3728 if (!test_opt(sb, NOLOAD) &&
3729 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3730 if (ext4_load_journal(sb, es, journal_devnum))
3732 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3733 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3734 ext4_msg(sb, KERN_ERR, "required journal recovery "
3735 "suppressed and not mounted read-only");
3736 goto failed_mount_wq;
3738 clear_opt(sb, DATA_FLAGS);
3739 sbi->s_journal = NULL;
3744 if (ext4_blocks_count(es) > 0xffffffffULL &&
3745 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3746 JBD2_FEATURE_INCOMPAT_64BIT)) {
3747 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3748 goto failed_mount_wq;
3751 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3752 jbd2_journal_set_features(sbi->s_journal,
3753 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3754 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3755 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3756 jbd2_journal_set_features(sbi->s_journal,
3757 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3758 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3759 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3761 jbd2_journal_clear_features(sbi->s_journal,
3762 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3763 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3766 /* We have now updated the journal if required, so we can
3767 * validate the data journaling mode. */
3768 switch (test_opt(sb, DATA_FLAGS)) {
3770 /* No mode set, assume a default based on the journal
3771 * capabilities: ORDERED_DATA if the journal can
3772 * cope, else JOURNAL_DATA
3774 if (jbd2_journal_check_available_features
3775 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3776 set_opt(sb, ORDERED_DATA);
3778 set_opt(sb, JOURNAL_DATA);
3781 case EXT4_MOUNT_ORDERED_DATA:
3782 case EXT4_MOUNT_WRITEBACK_DATA:
3783 if (!jbd2_journal_check_available_features
3784 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3785 ext4_msg(sb, KERN_ERR, "Journal does not support "
3786 "requested data journaling mode");
3787 goto failed_mount_wq;
3792 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3795 * The journal may have updated the bg summary counts, so we
3796 * need to update the global counters.
3798 percpu_counter_set(&sbi->s_freeclusters_counter,
3799 ext4_count_free_clusters(sb));
3800 percpu_counter_set(&sbi->s_freeinodes_counter,
3801 ext4_count_free_inodes(sb));
3802 percpu_counter_set(&sbi->s_dirs_counter,
3803 ext4_count_dirs(sb));
3804 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3808 * Get the # of file system overhead blocks from the
3809 * superblock if present.
3811 if (es->s_overhead_clusters)
3812 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3814 ret = ext4_calculate_overhead(sb);
3816 goto failed_mount_wq;
3820 * The maximum number of concurrent works can be high and
3821 * concurrency isn't really necessary. Limit it to 1.
3823 EXT4_SB(sb)->dio_unwritten_wq =
3824 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3825 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3826 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3827 goto failed_mount_wq;
3831 * The jbd2_journal_load will have done any necessary log recovery,
3832 * so we can safely mount the rest of the filesystem now.
3835 root = ext4_iget(sb, EXT4_ROOT_INO);
3837 ext4_msg(sb, KERN_ERR, "get root inode failed");
3838 ret = PTR_ERR(root);
3842 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3843 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3846 sb->s_root = d_alloc_root(root);
3848 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3853 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3854 sb->s_flags |= MS_RDONLY;
3856 /* determine the minimum size of new large inodes, if present */
3857 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3858 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3859 EXT4_GOOD_OLD_INODE_SIZE;
3860 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3861 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3862 if (sbi->s_want_extra_isize <
3863 le16_to_cpu(es->s_want_extra_isize))
3864 sbi->s_want_extra_isize =
3865 le16_to_cpu(es->s_want_extra_isize);
3866 if (sbi->s_want_extra_isize <
3867 le16_to_cpu(es->s_min_extra_isize))
3868 sbi->s_want_extra_isize =
3869 le16_to_cpu(es->s_min_extra_isize);
3872 /* Check if enough inode space is available */
3873 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3874 sbi->s_inode_size) {
3875 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3876 EXT4_GOOD_OLD_INODE_SIZE;
3877 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3881 err = ext4_setup_system_zone(sb);
3883 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3889 err = ext4_mb_init(sb, needs_recovery);
3891 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3896 err = ext4_register_li_request(sb, first_not_zeroed);
3900 sbi->s_kobj.kset = ext4_kset;
3901 init_completion(&sbi->s_kobj_unregister);
3902 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3907 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3908 ext4_orphan_cleanup(sb, es);
3909 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3910 if (needs_recovery) {
3911 ext4_msg(sb, KERN_INFO, "recovery complete");
3912 ext4_mark_recovery_complete(sb, es);
3914 if (EXT4_SB(sb)->s_journal) {
3915 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3916 descr = " journalled data mode";
3917 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3918 descr = " ordered data mode";
3920 descr = " writeback data mode";
3922 descr = "out journal";
3924 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3925 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3926 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3928 if (es->s_error_count)
3929 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3936 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3940 ext4_unregister_li_request(sb);
3942 ext4_ext_release(sb);
3944 ext4_mb_release(sb);
3945 ext4_release_system_zone(sb);
3949 ext4_msg(sb, KERN_ERR, "mount failed");
3950 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3952 if (sbi->s_journal) {
3953 jbd2_journal_destroy(sbi->s_journal);
3954 sbi->s_journal = NULL;
3957 del_timer_sync(&sbi->s_err_report);
3958 if (sbi->s_flex_groups)
3959 ext4_kvfree(sbi->s_flex_groups);
3960 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3961 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3962 percpu_counter_destroy(&sbi->s_dirs_counter);
3963 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3965 kthread_stop(sbi->s_mmp_tsk);
3967 for (i = 0; i < db_count; i++)
3968 brelse(sbi->s_group_desc[i]);
3969 ext4_kvfree(sbi->s_group_desc);
3972 remove_proc_entry(sb->s_id, ext4_proc_root);
3975 for (i = 0; i < MAXQUOTAS; i++)
3976 kfree(sbi->s_qf_names[i]);
3978 ext4_blkdev_remove(sbi);
3981 sb->s_fs_info = NULL;
3982 kfree(sbi->s_blockgroup_lock);
3990 * Setup any per-fs journal parameters now. We'll do this both on
3991 * initial mount, once the journal has been initialised but before we've
3992 * done any recovery; and again on any subsequent remount.
3994 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3996 struct ext4_sb_info *sbi = EXT4_SB(sb);
3998 journal->j_commit_interval = sbi->s_commit_interval;
3999 journal->j_min_batch_time = sbi->s_min_batch_time;
4000 journal->j_max_batch_time = sbi->s_max_batch_time;
4002 write_lock(&journal->j_state_lock);
4003 if (test_opt(sb, BARRIER))
4004 journal->j_flags |= JBD2_BARRIER;
4006 journal->j_flags &= ~JBD2_BARRIER;
4007 if (test_opt(sb, DATA_ERR_ABORT))
4008 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4010 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4011 write_unlock(&journal->j_state_lock);
4014 static journal_t *ext4_get_journal(struct super_block *sb,
4015 unsigned int journal_inum)
4017 struct inode *journal_inode;
4020 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4022 /* First, test for the existence of a valid inode on disk. Bad
4023 * things happen if we iget() an unused inode, as the subsequent
4024 * iput() will try to delete it. */
4026 journal_inode = ext4_iget(sb, journal_inum);
4027 if (IS_ERR(journal_inode)) {
4028 ext4_msg(sb, KERN_ERR, "no journal found");
4031 if (!journal_inode->i_nlink) {
4032 make_bad_inode(journal_inode);
4033 iput(journal_inode);
4034 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4038 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4039 journal_inode, journal_inode->i_size);
4040 if (!S_ISREG(journal_inode->i_mode)) {
4041 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4042 iput(journal_inode);
4046 journal = jbd2_journal_init_inode(journal_inode);
4048 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4049 iput(journal_inode);
4052 journal->j_private = sb;
4053 ext4_init_journal_params(sb, journal);
4057 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4060 struct buffer_head *bh;
4064 int hblock, blocksize;
4065 ext4_fsblk_t sb_block;
4066 unsigned long offset;
4067 struct ext4_super_block *es;
4068 struct block_device *bdev;
4070 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4072 bdev = ext4_blkdev_get(j_dev, sb);
4076 blocksize = sb->s_blocksize;
4077 hblock = bdev_logical_block_size(bdev);
4078 if (blocksize < hblock) {
4079 ext4_msg(sb, KERN_ERR,
4080 "blocksize too small for journal device");
4084 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4085 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4086 set_blocksize(bdev, blocksize);
4087 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4088 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4089 "external journal");
4093 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4094 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4095 !(le32_to_cpu(es->s_feature_incompat) &
4096 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4097 ext4_msg(sb, KERN_ERR, "external journal has "
4103 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4104 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4109 len = ext4_blocks_count(es);
4110 start = sb_block + 1;
4111 brelse(bh); /* we're done with the superblock */
4113 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4114 start, len, blocksize);
4116 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4119 journal->j_private = sb;
4120 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4121 wait_on_buffer(journal->j_sb_buffer);
4122 if (!buffer_uptodate(journal->j_sb_buffer)) {
4123 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4126 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4127 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4128 "user (unsupported) - %d",
4129 be32_to_cpu(journal->j_superblock->s_nr_users));
4132 EXT4_SB(sb)->journal_bdev = bdev;
4133 ext4_init_journal_params(sb, journal);
4137 jbd2_journal_destroy(journal);
4139 ext4_blkdev_put(bdev);
4143 static int ext4_load_journal(struct super_block *sb,
4144 struct ext4_super_block *es,
4145 unsigned long journal_devnum)
4148 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4151 int really_read_only;
4153 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4155 if (journal_devnum &&
4156 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4157 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4158 "numbers have changed");
4159 journal_dev = new_decode_dev(journal_devnum);
4161 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4163 really_read_only = bdev_read_only(sb->s_bdev);
4166 * Are we loading a blank journal or performing recovery after a
4167 * crash? For recovery, we need to check in advance whether we
4168 * can get read-write access to the device.
4170 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4171 if (sb->s_flags & MS_RDONLY) {
4172 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4173 "required on readonly filesystem");
4174 if (really_read_only) {
4175 ext4_msg(sb, KERN_ERR, "write access "
4176 "unavailable, cannot proceed");
4179 ext4_msg(sb, KERN_INFO, "write access will "
4180 "be enabled during recovery");
4184 if (journal_inum && journal_dev) {
4185 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4186 "and inode journals!");
4191 if (!(journal = ext4_get_journal(sb, journal_inum)))
4194 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4198 if (!(journal->j_flags & JBD2_BARRIER))
4199 ext4_msg(sb, KERN_INFO, "barriers disabled");
4201 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4202 err = jbd2_journal_update_format(journal);
4204 ext4_msg(sb, KERN_ERR, "error updating journal");
4205 jbd2_journal_destroy(journal);
4210 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4211 err = jbd2_journal_wipe(journal, !really_read_only);
4213 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4215 memcpy(save, ((char *) es) +
4216 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4217 err = jbd2_journal_load(journal);
4219 memcpy(((char *) es) + EXT4_S_ERR_START,
4220 save, EXT4_S_ERR_LEN);
4225 ext4_msg(sb, KERN_ERR, "error loading journal");
4226 jbd2_journal_destroy(journal);
4230 EXT4_SB(sb)->s_journal = journal;
4231 ext4_clear_journal_err(sb, es);
4233 if (!really_read_only && journal_devnum &&
4234 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4235 es->s_journal_dev = cpu_to_le32(journal_devnum);
4237 /* Make sure we flush the recovery flag to disk. */
4238 ext4_commit_super(sb, 1);
4244 static int ext4_commit_super(struct super_block *sb, int sync)
4246 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4247 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4250 if (!sbh || block_device_ejected(sb))
4252 if (buffer_write_io_error(sbh)) {
4254 * Oh, dear. A previous attempt to write the
4255 * superblock failed. This could happen because the
4256 * USB device was yanked out. Or it could happen to
4257 * be a transient write error and maybe the block will
4258 * be remapped. Nothing we can do but to retry the
4259 * write and hope for the best.
4261 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4262 "superblock detected");
4263 clear_buffer_write_io_error(sbh);
4264 set_buffer_uptodate(sbh);
4267 * If the file system is mounted read-only, don't update the
4268 * superblock write time. This avoids updating the superblock
4269 * write time when we are mounting the root file system
4270 * read/only but we need to replay the journal; at that point,
4271 * for people who are east of GMT and who make their clock
4272 * tick in localtime for Windows bug-for-bug compatibility,
4273 * the clock is set in the future, and this will cause e2fsck
4274 * to complain and force a full file system check.
4276 if (!(sb->s_flags & MS_RDONLY))
4277 es->s_wtime = cpu_to_le32(get_seconds());
4278 if (sb->s_bdev->bd_part)
4279 es->s_kbytes_written =
4280 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4281 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4282 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4284 es->s_kbytes_written =
4285 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4286 ext4_free_blocks_count_set(es,
4287 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4288 &EXT4_SB(sb)->s_freeclusters_counter)));
4289 es->s_free_inodes_count =
4290 cpu_to_le32(percpu_counter_sum_positive(
4291 &EXT4_SB(sb)->s_freeinodes_counter));
4293 BUFFER_TRACE(sbh, "marking dirty");
4294 mark_buffer_dirty(sbh);
4296 error = sync_dirty_buffer(sbh);
4300 error = buffer_write_io_error(sbh);
4302 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4304 clear_buffer_write_io_error(sbh);
4305 set_buffer_uptodate(sbh);
4312 * Have we just finished recovery? If so, and if we are mounting (or
4313 * remounting) the filesystem readonly, then we will end up with a
4314 * consistent fs on disk. Record that fact.
4316 static void ext4_mark_recovery_complete(struct super_block *sb,
4317 struct ext4_super_block *es)
4319 journal_t *journal = EXT4_SB(sb)->s_journal;
4321 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4322 BUG_ON(journal != NULL);
4325 jbd2_journal_lock_updates(journal);
4326 if (jbd2_journal_flush(journal) < 0)
4329 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4330 sb->s_flags & MS_RDONLY) {
4331 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4332 ext4_commit_super(sb, 1);
4336 jbd2_journal_unlock_updates(journal);
4340 * If we are mounting (or read-write remounting) a filesystem whose journal
4341 * has recorded an error from a previous lifetime, move that error to the
4342 * main filesystem now.
4344 static void ext4_clear_journal_err(struct super_block *sb,
4345 struct ext4_super_block *es)
4351 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4353 journal = EXT4_SB(sb)->s_journal;
4356 * Now check for any error status which may have been recorded in the
4357 * journal by a prior ext4_error() or ext4_abort()
4360 j_errno = jbd2_journal_errno(journal);
4364 errstr = ext4_decode_error(sb, j_errno, nbuf);
4365 ext4_warning(sb, "Filesystem error recorded "
4366 "from previous mount: %s", errstr);
4367 ext4_warning(sb, "Marking fs in need of filesystem check.");
4369 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4370 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4371 ext4_commit_super(sb, 1);
4373 jbd2_journal_clear_err(journal);
4378 * Force the running and committing transactions to commit,
4379 * and wait on the commit.
4381 int ext4_force_commit(struct super_block *sb)
4386 if (sb->s_flags & MS_RDONLY)
4389 journal = EXT4_SB(sb)->s_journal;
4391 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4392 ret = ext4_journal_force_commit(journal);
4398 static void ext4_write_super(struct super_block *sb)
4401 ext4_commit_super(sb, 1);
4405 static int ext4_sync_fs(struct super_block *sb, int wait)
4409 struct ext4_sb_info *sbi = EXT4_SB(sb);
4411 trace_ext4_sync_fs(sb, wait);
4412 flush_workqueue(sbi->dio_unwritten_wq);
4413 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4415 jbd2_log_wait_commit(sbi->s_journal, target);
4421 * LVM calls this function before a (read-only) snapshot is created. This
4422 * gives us a chance to flush the journal completely and mark the fs clean.
4424 * Note that only this function cannot bring a filesystem to be in a clean
4425 * state independently, because ext4 prevents a new handle from being started
4426 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4429 static int ext4_freeze(struct super_block *sb)
4434 if (sb->s_flags & MS_RDONLY)
4437 journal = EXT4_SB(sb)->s_journal;
4439 /* Now we set up the journal barrier. */
4440 jbd2_journal_lock_updates(journal);
4443 * Don't clear the needs_recovery flag if we failed to flush
4446 error = jbd2_journal_flush(journal);
4450 /* Journal blocked and flushed, clear needs_recovery flag. */
4451 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4452 error = ext4_commit_super(sb, 1);
4454 /* we rely on s_frozen to stop further updates */
4455 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4460 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4461 * flag here, even though the filesystem is not technically dirty yet.
4463 static int ext4_unfreeze(struct super_block *sb)
4465 if (sb->s_flags & MS_RDONLY)
4469 /* Reset the needs_recovery flag before the fs is unlocked. */
4470 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4471 ext4_commit_super(sb, 1);
4477 * Structure to save mount options for ext4_remount's benefit
4479 struct ext4_mount_options {
4480 unsigned long s_mount_opt;
4481 unsigned long s_mount_opt2;
4484 unsigned long s_commit_interval;
4485 u32 s_min_batch_time, s_max_batch_time;
4488 char *s_qf_names[MAXQUOTAS];
4492 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4494 struct ext4_super_block *es;
4495 struct ext4_sb_info *sbi = EXT4_SB(sb);
4496 ext4_fsblk_t n_blocks_count = 0;
4497 unsigned long old_sb_flags;
4498 struct ext4_mount_options old_opts;
4499 int enable_quota = 0;
4501 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4506 char *orig_data = kstrdup(data, GFP_KERNEL);
4508 /* Store the original options */
4510 old_sb_flags = sb->s_flags;
4511 old_opts.s_mount_opt = sbi->s_mount_opt;
4512 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4513 old_opts.s_resuid = sbi->s_resuid;
4514 old_opts.s_resgid = sbi->s_resgid;
4515 old_opts.s_commit_interval = sbi->s_commit_interval;
4516 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4517 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4519 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4520 for (i = 0; i < MAXQUOTAS; i++)
4521 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4523 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4524 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4527 * Allow the "check" option to be passed as a remount option.
4529 if (!parse_options(data, sb, NULL, &journal_ioprio,
4530 &n_blocks_count, 1)) {
4535 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4536 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4537 ext4_msg(sb, KERN_ERR, "can't mount with "
4538 "both data=journal and delalloc");
4542 if (test_opt(sb, DIOREAD_NOLOCK)) {
4543 ext4_msg(sb, KERN_ERR, "can't mount with "
4544 "both data=journal and dioread_nolock");
4550 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4551 ext4_abort(sb, "Abort forced by user");
4553 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4554 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4558 if (sbi->s_journal) {
4559 ext4_init_journal_params(sb, sbi->s_journal);
4560 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4563 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4564 n_blocks_count > ext4_blocks_count(es)) {
4565 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4570 if (*flags & MS_RDONLY) {
4571 err = dquot_suspend(sb, -1);
4576 * First of all, the unconditional stuff we have to do
4577 * to disable replay of the journal when we next remount
4579 sb->s_flags |= MS_RDONLY;
4582 * OK, test if we are remounting a valid rw partition
4583 * readonly, and if so set the rdonly flag and then
4584 * mark the partition as valid again.
4586 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4587 (sbi->s_mount_state & EXT4_VALID_FS))
4588 es->s_state = cpu_to_le16(sbi->s_mount_state);
4591 ext4_mark_recovery_complete(sb, es);
4593 /* Make sure we can mount this feature set readwrite */
4594 if (!ext4_feature_set_ok(sb, 0)) {
4599 * Make sure the group descriptor checksums
4600 * are sane. If they aren't, refuse to remount r/w.
4602 for (g = 0; g < sbi->s_groups_count; g++) {
4603 struct ext4_group_desc *gdp =
4604 ext4_get_group_desc(sb, g, NULL);
4606 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4607 ext4_msg(sb, KERN_ERR,
4608 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4609 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4610 le16_to_cpu(gdp->bg_checksum));
4617 * If we have an unprocessed orphan list hanging
4618 * around from a previously readonly bdev mount,
4619 * require a full umount/remount for now.
4621 if (es->s_last_orphan) {
4622 ext4_msg(sb, KERN_WARNING, "Couldn't "
4623 "remount RDWR because of unprocessed "
4624 "orphan inode list. Please "
4625 "umount/remount instead");
4631 * Mounting a RDONLY partition read-write, so reread
4632 * and store the current valid flag. (It may have
4633 * been changed by e2fsck since we originally mounted
4637 ext4_clear_journal_err(sb, es);
4638 sbi->s_mount_state = le16_to_cpu(es->s_state);
4639 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4641 if (!ext4_setup_super(sb, es, 0))
4642 sb->s_flags &= ~MS_RDONLY;
4643 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4644 EXT4_FEATURE_INCOMPAT_MMP))
4645 if (ext4_multi_mount_protect(sb,
4646 le64_to_cpu(es->s_mmp_block))) {
4655 * Reinitialize lazy itable initialization thread based on
4658 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4659 ext4_unregister_li_request(sb);
4661 ext4_group_t first_not_zeroed;
4662 first_not_zeroed = ext4_has_uninit_itable(sb);
4663 ext4_register_li_request(sb, first_not_zeroed);
4666 ext4_setup_system_zone(sb);
4667 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4668 ext4_commit_super(sb, 1);
4671 /* Release old quota file names */
4672 for (i = 0; i < MAXQUOTAS; i++)
4673 if (old_opts.s_qf_names[i] &&
4674 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4675 kfree(old_opts.s_qf_names[i]);
4679 dquot_resume(sb, -1);
4681 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4686 sb->s_flags = old_sb_flags;
4687 sbi->s_mount_opt = old_opts.s_mount_opt;
4688 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4689 sbi->s_resuid = old_opts.s_resuid;
4690 sbi->s_resgid = old_opts.s_resgid;
4691 sbi->s_commit_interval = old_opts.s_commit_interval;
4692 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4693 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4695 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4696 for (i = 0; i < MAXQUOTAS; i++) {
4697 if (sbi->s_qf_names[i] &&
4698 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4699 kfree(sbi->s_qf_names[i]);
4700 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4708 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4710 struct super_block *sb = dentry->d_sb;
4711 struct ext4_sb_info *sbi = EXT4_SB(sb);
4712 struct ext4_super_block *es = sbi->s_es;
4713 ext4_fsblk_t overhead = 0;
4717 if (!test_opt(sb, MINIX_DF))
4718 overhead = sbi->s_overhead;
4720 buf->f_type = EXT4_SUPER_MAGIC;
4721 buf->f_bsize = sb->s_blocksize;
4722 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4723 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4724 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4725 /* prevent underflow in case that few free space is available */
4726 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4727 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4728 if (buf->f_bfree < ext4_r_blocks_count(es))
4730 buf->f_files = le32_to_cpu(es->s_inodes_count);
4731 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4732 buf->f_namelen = EXT4_NAME_LEN;
4733 fsid = le64_to_cpup((void *)es->s_uuid) ^
4734 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4735 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4736 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4741 /* Helper function for writing quotas on sync - we need to start transaction
4742 * before quota file is locked for write. Otherwise the are possible deadlocks:
4743 * Process 1 Process 2
4744 * ext4_create() quota_sync()
4745 * jbd2_journal_start() write_dquot()
4746 * dquot_initialize() down(dqio_mutex)
4747 * down(dqio_mutex) jbd2_journal_start()
4753 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4755 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4758 static int ext4_write_dquot(struct dquot *dquot)
4762 struct inode *inode;
4764 inode = dquot_to_inode(dquot);
4765 handle = ext4_journal_start(inode,
4766 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4768 return PTR_ERR(handle);
4769 ret = dquot_commit(dquot);
4770 err = ext4_journal_stop(handle);
4776 static int ext4_acquire_dquot(struct dquot *dquot)
4781 handle = ext4_journal_start(dquot_to_inode(dquot),
4782 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4784 return PTR_ERR(handle);
4785 ret = dquot_acquire(dquot);
4786 err = ext4_journal_stop(handle);
4792 static int ext4_release_dquot(struct dquot *dquot)
4797 handle = ext4_journal_start(dquot_to_inode(dquot),
4798 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4799 if (IS_ERR(handle)) {
4800 /* Release dquot anyway to avoid endless cycle in dqput() */
4801 dquot_release(dquot);
4802 return PTR_ERR(handle);
4804 ret = dquot_release(dquot);
4805 err = ext4_journal_stop(handle);
4811 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4813 /* Are we journaling quotas? */
4814 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4815 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4816 dquot_mark_dquot_dirty(dquot);
4817 return ext4_write_dquot(dquot);
4819 return dquot_mark_dquot_dirty(dquot);
4823 static int ext4_write_info(struct super_block *sb, int type)
4828 /* Data block + inode block */
4829 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4831 return PTR_ERR(handle);
4832 ret = dquot_commit_info(sb, type);
4833 err = ext4_journal_stop(handle);
4840 * Turn on quotas during mount time - we need to find
4841 * the quota file and such...
4843 static int ext4_quota_on_mount(struct super_block *sb, int type)
4845 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4846 EXT4_SB(sb)->s_jquota_fmt, type);
4850 * Standard function to be called on quota_on
4852 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4857 if (!test_opt(sb, QUOTA))
4860 /* Quotafile not on the same filesystem? */
4861 if (path->mnt->mnt_sb != sb)
4863 /* Journaling quota? */
4864 if (EXT4_SB(sb)->s_qf_names[type]) {
4865 /* Quotafile not in fs root? */
4866 if (path->dentry->d_parent != sb->s_root)
4867 ext4_msg(sb, KERN_WARNING,
4868 "Quota file not on filesystem root. "
4869 "Journaled quota will not work");
4873 * When we journal data on quota file, we have to flush journal to see
4874 * all updates to the file when we bypass pagecache...
4876 if (EXT4_SB(sb)->s_journal &&
4877 ext4_should_journal_data(path->dentry->d_inode)) {
4879 * We don't need to lock updates but journal_flush() could
4880 * otherwise be livelocked...
4882 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4883 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4884 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4889 return dquot_quota_on(sb, type, format_id, path);
4892 static int ext4_quota_off(struct super_block *sb, int type)
4894 struct inode *inode = sb_dqopt(sb)->files[type];
4897 /* Force all delayed allocation blocks to be allocated.
4898 * Caller already holds s_umount sem */
4899 if (test_opt(sb, DELALLOC))
4900 sync_filesystem(sb);
4905 /* Update modification times of quota files when userspace can
4906 * start looking at them */
4907 handle = ext4_journal_start(inode, 1);
4910 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4911 ext4_mark_inode_dirty(handle, inode);
4912 ext4_journal_stop(handle);
4915 return dquot_quota_off(sb, type);
4918 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4919 * acquiring the locks... As quota files are never truncated and quota code
4920 * itself serializes the operations (and no one else should touch the files)
4921 * we don't have to be afraid of races */
4922 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4923 size_t len, loff_t off)
4925 struct inode *inode = sb_dqopt(sb)->files[type];
4926 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4928 int offset = off & (sb->s_blocksize - 1);
4931 struct buffer_head *bh;
4932 loff_t i_size = i_size_read(inode);
4936 if (off+len > i_size)
4939 while (toread > 0) {
4940 tocopy = sb->s_blocksize - offset < toread ?
4941 sb->s_blocksize - offset : toread;
4942 bh = ext4_bread(NULL, inode, blk, 0, &err);
4945 if (!bh) /* A hole? */
4946 memset(data, 0, tocopy);
4948 memcpy(data, bh->b_data+offset, tocopy);
4958 /* Write to quotafile (we know the transaction is already started and has
4959 * enough credits) */
4960 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4961 const char *data, size_t len, loff_t off)
4963 struct inode *inode = sb_dqopt(sb)->files[type];
4964 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4966 int offset = off & (sb->s_blocksize - 1);
4967 struct buffer_head *bh;
4968 handle_t *handle = journal_current_handle();
4970 if (EXT4_SB(sb)->s_journal && !handle) {
4971 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4972 " cancelled because transaction is not started",
4973 (unsigned long long)off, (unsigned long long)len);
4977 * Since we account only one data block in transaction credits,
4978 * then it is impossible to cross a block boundary.
4980 if (sb->s_blocksize - offset < len) {
4981 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4982 " cancelled because not block aligned",
4983 (unsigned long long)off, (unsigned long long)len);
4987 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4988 bh = ext4_bread(handle, inode, blk, 1, &err);
4991 err = ext4_journal_get_write_access(handle, bh);
4997 memcpy(bh->b_data+offset, data, len);
4998 flush_dcache_page(bh->b_page);
5000 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5004 mutex_unlock(&inode->i_mutex);
5007 if (inode->i_size < off + len) {
5008 i_size_write(inode, off + len);
5009 EXT4_I(inode)->i_disksize = inode->i_size;
5010 ext4_mark_inode_dirty(handle, inode);
5012 mutex_unlock(&inode->i_mutex);
5018 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5019 const char *dev_name, void *data)
5021 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5024 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5025 static inline void register_as_ext2(void)
5027 int err = register_filesystem(&ext2_fs_type);
5030 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5033 static inline void unregister_as_ext2(void)
5035 unregister_filesystem(&ext2_fs_type);
5038 static inline int ext2_feature_set_ok(struct super_block *sb)
5040 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5042 if (sb->s_flags & MS_RDONLY)
5044 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5048 MODULE_ALIAS("ext2");
5050 static inline void register_as_ext2(void) { }
5051 static inline void unregister_as_ext2(void) { }
5052 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5055 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5056 static inline void register_as_ext3(void)
5058 int err = register_filesystem(&ext3_fs_type);
5061 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5064 static inline void unregister_as_ext3(void)
5066 unregister_filesystem(&ext3_fs_type);
5069 static inline int ext3_feature_set_ok(struct super_block *sb)
5071 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5073 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5075 if (sb->s_flags & MS_RDONLY)
5077 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5081 MODULE_ALIAS("ext3");
5083 static inline void register_as_ext3(void) { }
5084 static inline void unregister_as_ext3(void) { }
5085 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5088 static struct file_system_type ext4_fs_type = {
5089 .owner = THIS_MODULE,
5091 .mount = ext4_mount,
5092 .kill_sb = kill_block_super,
5093 .fs_flags = FS_REQUIRES_DEV,
5096 static int __init ext4_init_feat_adverts(void)
5098 struct ext4_features *ef;
5101 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5105 ef->f_kobj.kset = ext4_kset;
5106 init_completion(&ef->f_kobj_unregister);
5107 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5120 static void ext4_exit_feat_adverts(void)
5122 kobject_put(&ext4_feat->f_kobj);
5123 wait_for_completion(&ext4_feat->f_kobj_unregister);
5127 /* Shared across all ext4 file systems */
5128 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5129 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5131 static int __init ext4_init_fs(void)
5135 ext4_check_flag_values();
5137 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5138 mutex_init(&ext4__aio_mutex[i]);
5139 init_waitqueue_head(&ext4__ioend_wq[i]);
5142 err = ext4_init_pageio();
5145 err = ext4_init_system_zone();
5148 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5151 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5153 err = ext4_init_feat_adverts();
5157 err = ext4_init_mballoc();
5161 err = ext4_init_xattr();
5164 err = init_inodecache();
5169 err = register_filesystem(&ext4_fs_type);
5173 ext4_li_info = NULL;
5174 mutex_init(&ext4_li_mtx);
5177 unregister_as_ext2();
5178 unregister_as_ext3();
5179 destroy_inodecache();
5183 ext4_exit_mballoc();
5185 ext4_exit_feat_adverts();
5188 remove_proc_entry("fs/ext4", NULL);
5189 kset_unregister(ext4_kset);
5191 ext4_exit_system_zone();
5197 static void __exit ext4_exit_fs(void)
5199 ext4_destroy_lazyinit_thread();
5200 unregister_as_ext2();
5201 unregister_as_ext3();
5202 unregister_filesystem(&ext4_fs_type);
5203 destroy_inodecache();
5205 ext4_exit_mballoc();
5206 ext4_exit_feat_adverts();
5207 remove_proc_entry("fs/ext4", NULL);
5208 kset_unregister(ext4_kset);
5209 ext4_exit_system_zone();
5213 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5214 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5215 MODULE_LICENSE("GPL");
5216 module_init(ext4_init_fs)
5217 module_exit(ext4_exit_fs)