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_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static void ext4_write_super(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
94 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #define IS_EXT2_SB(sb) (0)
100 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101 static struct file_system_type ext3_fs_type = {
102 .owner = THIS_MODULE,
105 .kill_sb = kill_block_super,
106 .fs_flags = FS_REQUIRES_DEV,
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #define IS_EXT3_SB(sb) (0)
113 void *ext4_kvmalloc(size_t size, gfp_t flags)
117 ret = kmalloc(size, flags);
119 ret = __vmalloc(size, flags, PAGE_KERNEL);
123 void *ext4_kvzalloc(size_t size, gfp_t flags)
127 ret = kzalloc(size, flags);
129 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
133 void ext4_kvfree(void *ptr)
135 if (is_vmalloc_addr(ptr))
142 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
143 struct ext4_group_desc *bg)
145 return le32_to_cpu(bg->bg_block_bitmap_lo) |
146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
147 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
150 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
151 struct ext4_group_desc *bg)
153 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
154 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
155 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
158 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
159 struct ext4_group_desc *bg)
161 return le32_to_cpu(bg->bg_inode_table_lo) |
162 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
163 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
166 __u32 ext4_free_group_clusters(struct super_block *sb,
167 struct ext4_group_desc *bg)
169 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
170 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
171 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
174 __u32 ext4_free_inodes_count(struct super_block *sb,
175 struct ext4_group_desc *bg)
177 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
178 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
179 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
182 __u32 ext4_used_dirs_count(struct super_block *sb,
183 struct ext4_group_desc *bg)
185 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
186 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
187 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
190 __u32 ext4_itable_unused_count(struct super_block *sb,
191 struct ext4_group_desc *bg)
193 return le16_to_cpu(bg->bg_itable_unused_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
198 void ext4_block_bitmap_set(struct super_block *sb,
199 struct ext4_group_desc *bg, ext4_fsblk_t blk)
201 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
206 void ext4_inode_bitmap_set(struct super_block *sb,
207 struct ext4_group_desc *bg, ext4_fsblk_t blk)
209 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
210 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
211 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
214 void ext4_inode_table_set(struct super_block *sb,
215 struct ext4_group_desc *bg, ext4_fsblk_t blk)
217 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
218 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
219 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
222 void ext4_free_group_clusters_set(struct super_block *sb,
223 struct ext4_group_desc *bg, __u32 count)
225 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
226 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
227 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
230 void ext4_free_inodes_set(struct super_block *sb,
231 struct ext4_group_desc *bg, __u32 count)
233 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
234 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
235 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
238 void ext4_used_dirs_set(struct super_block *sb,
239 struct ext4_group_desc *bg, __u32 count)
241 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
242 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
243 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
246 void ext4_itable_unused_set(struct super_block *sb,
247 struct ext4_group_desc *bg, __u32 count)
249 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
255 /* Just increment the non-pointer handle value */
256 static handle_t *ext4_get_nojournal(void)
258 handle_t *handle = current->journal_info;
259 unsigned long ref_cnt = (unsigned long)handle;
261 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
264 handle = (handle_t *)ref_cnt;
266 current->journal_info = handle;
271 /* Decrement the non-pointer handle value */
272 static void ext4_put_nojournal(handle_t *handle)
274 unsigned long ref_cnt = (unsigned long)handle;
276 BUG_ON(ref_cnt == 0);
279 handle = (handle_t *)ref_cnt;
281 current->journal_info = handle;
285 * Wrappers for jbd2_journal_start/end.
287 * The only special thing we need to do here is to make sure that all
288 * journal_end calls result in the superblock being marked dirty, so
289 * that sync() will call the filesystem's write_super callback if
292 * To avoid j_barrier hold in userspace when a user calls freeze(),
293 * ext4 prevents a new handle from being started by s_frozen, which
294 * is in an upper layer.
296 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
301 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
302 if (sb->s_flags & MS_RDONLY)
303 return ERR_PTR(-EROFS);
305 journal = EXT4_SB(sb)->s_journal;
306 handle = ext4_journal_current_handle();
309 * If a handle has been started, it should be allowed to
310 * finish, otherwise deadlock could happen between freeze
311 * and others(e.g. truncate) due to the restart of the
312 * journal handle if the filesystem is forzen and active
313 * handles are not stopped.
316 vfs_check_frozen(sb, SB_FREEZE_TRANS);
319 return ext4_get_nojournal();
321 * Special case here: if the journal has aborted behind our
322 * backs (eg. EIO in the commit thread), then we still need to
323 * take the FS itself readonly cleanly.
325 if (is_journal_aborted(journal)) {
326 ext4_abort(sb, "Detected aborted journal");
327 return ERR_PTR(-EROFS);
329 return jbd2_journal_start(journal, nblocks);
333 * The only special thing we need to do here is to make sure that all
334 * jbd2_journal_stop calls result in the superblock being marked dirty, so
335 * that sync() will call the filesystem's write_super callback if
338 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
340 struct super_block *sb;
344 if (!ext4_handle_valid(handle)) {
345 ext4_put_nojournal(handle);
348 sb = handle->h_transaction->t_journal->j_private;
350 rc = jbd2_journal_stop(handle);
355 __ext4_std_error(sb, where, line, err);
359 void ext4_journal_abort_handle(const char *caller, unsigned int line,
360 const char *err_fn, struct buffer_head *bh,
361 handle_t *handle, int err)
364 const char *errstr = ext4_decode_error(NULL, err, nbuf);
366 BUG_ON(!ext4_handle_valid(handle));
369 BUFFER_TRACE(bh, "abort");
374 if (is_handle_aborted(handle))
377 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
378 caller, line, errstr, err_fn);
380 jbd2_journal_abort_handle(handle);
383 static void __save_error_info(struct super_block *sb, const char *func,
386 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
388 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
389 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
390 es->s_last_error_time = cpu_to_le32(get_seconds());
391 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
392 es->s_last_error_line = cpu_to_le32(line);
393 if (!es->s_first_error_time) {
394 es->s_first_error_time = es->s_last_error_time;
395 strncpy(es->s_first_error_func, func,
396 sizeof(es->s_first_error_func));
397 es->s_first_error_line = cpu_to_le32(line);
398 es->s_first_error_ino = es->s_last_error_ino;
399 es->s_first_error_block = es->s_last_error_block;
402 * Start the daily error reporting function if it hasn't been
405 if (!es->s_error_count)
406 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
407 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
410 static void save_error_info(struct super_block *sb, const char *func,
413 __save_error_info(sb, func, line);
414 ext4_commit_super(sb, 1);
418 * The del_gendisk() function uninitializes the disk-specific data
419 * structures, including the bdi structure, without telling anyone
420 * else. Once this happens, any attempt to call mark_buffer_dirty()
421 * (for example, by ext4_commit_super), will cause a kernel OOPS.
422 * This is a kludge to prevent these oops until we can put in a proper
423 * hook in del_gendisk() to inform the VFS and file system layers.
425 static int block_device_ejected(struct super_block *sb)
427 struct inode *bd_inode = sb->s_bdev->bd_inode;
428 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430 return bdi->dev == NULL;
434 /* Deal with the reporting of failure conditions on a filesystem such as
435 * inconsistencies detected or read IO failures.
437 * On ext2, we can store the error state of the filesystem in the
438 * superblock. That is not possible on ext4, because we may have other
439 * write ordering constraints on the superblock which prevent us from
440 * writing it out straight away; and given that the journal is about to
441 * be aborted, we can't rely on the current, or future, transactions to
442 * write out the superblock safely.
444 * We'll just use the jbd2_journal_abort() error code to record an error in
445 * the journal instead. On recovery, the journal will complain about
446 * that error until we've noted it down and cleared it.
449 static void ext4_handle_error(struct super_block *sb)
451 if (sb->s_flags & MS_RDONLY)
454 if (!test_opt(sb, ERRORS_CONT)) {
455 journal_t *journal = EXT4_SB(sb)->s_journal;
457 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
459 jbd2_journal_abort(journal, -EIO);
461 if (test_opt(sb, ERRORS_RO)) {
462 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
463 sb->s_flags |= MS_RDONLY;
465 if (test_opt(sb, ERRORS_PANIC))
466 panic("EXT4-fs (device %s): panic forced after error\n",
470 void __ext4_error(struct super_block *sb, const char *function,
471 unsigned int line, const char *fmt, ...)
473 struct va_format vaf;
479 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
480 sb->s_id, function, line, current->comm, &vaf);
483 ext4_handle_error(sb);
486 void ext4_error_inode(struct inode *inode, const char *function,
487 unsigned int line, ext4_fsblk_t block,
488 const char *fmt, ...)
491 struct va_format vaf;
492 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
494 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
495 es->s_last_error_block = cpu_to_le64(block);
496 save_error_info(inode->i_sb, function, line);
500 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
501 inode->i_sb->s_id, function, line, inode->i_ino);
503 printk(KERN_CONT "block %llu: ", block);
504 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
507 ext4_handle_error(inode->i_sb);
510 void ext4_error_file(struct file *file, const char *function,
511 unsigned int line, ext4_fsblk_t block,
512 const char *fmt, ...)
515 struct va_format vaf;
516 struct ext4_super_block *es;
517 struct inode *inode = file->f_dentry->d_inode;
518 char pathname[80], *path;
520 es = EXT4_SB(inode->i_sb)->s_es;
521 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
522 save_error_info(inode->i_sb, function, line);
523 path = d_path(&(file->f_path), pathname, sizeof(pathname));
527 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
528 inode->i_sb->s_id, function, line, inode->i_ino);
530 printk(KERN_CONT "block %llu: ", block);
534 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
537 ext4_handle_error(inode->i_sb);
540 static const char *ext4_decode_error(struct super_block *sb, int errno,
547 errstr = "IO failure";
550 errstr = "Out of memory";
553 if (!sb || (EXT4_SB(sb)->s_journal &&
554 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
555 errstr = "Journal has aborted";
557 errstr = "Readonly filesystem";
560 /* If the caller passed in an extra buffer for unknown
561 * errors, textualise them now. Else we just return
564 /* Check for truncated error codes... */
565 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
574 /* __ext4_std_error decodes expected errors from journaling functions
575 * automatically and invokes the appropriate error response. */
577 void __ext4_std_error(struct super_block *sb, const char *function,
578 unsigned int line, int errno)
583 /* Special case: if the error is EROFS, and we're not already
584 * inside a transaction, then there's really no point in logging
586 if (errno == -EROFS && journal_current_handle() == NULL &&
587 (sb->s_flags & MS_RDONLY))
590 errstr = ext4_decode_error(sb, errno, nbuf);
591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
593 save_error_info(sb, function, line);
595 ext4_handle_error(sb);
599 * ext4_abort is a much stronger failure handler than ext4_error. The
600 * abort function may be used to deal with unrecoverable failures such
601 * as journal IO errors or ENOMEM at a critical moment in log management.
603 * We unconditionally force the filesystem into an ABORT|READONLY state,
604 * unless the error response on the fs has been set to panic in which
605 * case we take the easy way out and panic immediately.
608 void __ext4_abort(struct super_block *sb, const char *function,
609 unsigned int line, const char *fmt, ...)
613 save_error_info(sb, function, line);
615 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
621 if ((sb->s_flags & MS_RDONLY) == 0) {
622 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
623 sb->s_flags |= MS_RDONLY;
624 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
625 if (EXT4_SB(sb)->s_journal)
626 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
627 save_error_info(sb, function, line);
629 if (test_opt(sb, ERRORS_PANIC))
630 panic("EXT4-fs panic from previous error\n");
633 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
635 struct va_format vaf;
641 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
645 void __ext4_warning(struct super_block *sb, const char *function,
646 unsigned int line, const char *fmt, ...)
648 struct va_format vaf;
654 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
655 sb->s_id, function, line, &vaf);
659 void __ext4_grp_locked_error(const char *function, unsigned int line,
660 struct super_block *sb, ext4_group_t grp,
661 unsigned long ino, ext4_fsblk_t block,
662 const char *fmt, ...)
666 struct va_format vaf;
668 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670 es->s_last_error_ino = cpu_to_le32(ino);
671 es->s_last_error_block = cpu_to_le64(block);
672 __save_error_info(sb, function, line);
678 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
679 sb->s_id, function, line, grp);
681 printk(KERN_CONT "inode %lu: ", ino);
683 printk(KERN_CONT "block %llu:", (unsigned long long) block);
684 printk(KERN_CONT "%pV\n", &vaf);
687 if (test_opt(sb, ERRORS_CONT)) {
688 ext4_commit_super(sb, 0);
692 ext4_unlock_group(sb, grp);
693 ext4_handle_error(sb);
695 * We only get here in the ERRORS_RO case; relocking the group
696 * may be dangerous, but nothing bad will happen since the
697 * filesystem will have already been marked read/only and the
698 * journal has been aborted. We return 1 as a hint to callers
699 * who might what to use the return value from
700 * ext4_grp_locked_error() to distinguish between the
701 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
702 * aggressively from the ext4 function in question, with a
703 * more appropriate error code.
705 ext4_lock_group(sb, grp);
709 void ext4_update_dynamic_rev(struct super_block *sb)
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
713 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
717 "updating to rev %d because of new feature flag, "
718 "running e2fsck is recommended",
721 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
722 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
723 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
724 /* leave es->s_feature_*compat flags alone */
725 /* es->s_uuid will be set by e2fsck if empty */
728 * The rest of the superblock fields should be zero, and if not it
729 * means they are likely already in use, so leave them alone. We
730 * can leave it up to e2fsck to clean up any inconsistencies there.
735 * Open the external journal device
737 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
739 struct block_device *bdev;
740 char b[BDEVNAME_SIZE];
742 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
748 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
749 __bdevname(dev, b), PTR_ERR(bdev));
754 * Release the journal device
756 static int ext4_blkdev_put(struct block_device *bdev)
758 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
761 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
763 struct block_device *bdev;
766 bdev = sbi->journal_bdev;
768 ret = ext4_blkdev_put(bdev);
769 sbi->journal_bdev = NULL;
774 static inline struct inode *orphan_list_entry(struct list_head *l)
776 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
779 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
783 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
784 le32_to_cpu(sbi->s_es->s_last_orphan));
786 printk(KERN_ERR "sb_info orphan list:\n");
787 list_for_each(l, &sbi->s_orphan) {
788 struct inode *inode = orphan_list_entry(l);
790 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
791 inode->i_sb->s_id, inode->i_ino, inode,
792 inode->i_mode, inode->i_nlink,
797 static void ext4_put_super(struct super_block *sb)
799 struct ext4_sb_info *sbi = EXT4_SB(sb);
800 struct ext4_super_block *es = sbi->s_es;
803 ext4_unregister_li_request(sb);
804 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
806 flush_workqueue(sbi->dio_unwritten_wq);
807 destroy_workqueue(sbi->dio_unwritten_wq);
811 ext4_commit_super(sb, 1);
813 if (sbi->s_journal) {
814 err = jbd2_journal_destroy(sbi->s_journal);
815 sbi->s_journal = NULL;
817 ext4_abort(sb, "Couldn't clean up the journal");
820 del_timer(&sbi->s_err_report);
821 ext4_release_system_zone(sb);
823 ext4_ext_release(sb);
824 ext4_xattr_put_super(sb);
826 if (!(sb->s_flags & MS_RDONLY)) {
827 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
828 es->s_state = cpu_to_le16(sbi->s_mount_state);
829 ext4_commit_super(sb, 1);
832 remove_proc_entry(sb->s_id, ext4_proc_root);
834 kobject_del(&sbi->s_kobj);
836 for (i = 0; i < sbi->s_gdb_count; i++)
837 brelse(sbi->s_group_desc[i]);
838 ext4_kvfree(sbi->s_group_desc);
839 ext4_kvfree(sbi->s_flex_groups);
840 percpu_counter_destroy(&sbi->s_freeclusters_counter);
841 percpu_counter_destroy(&sbi->s_freeinodes_counter);
842 percpu_counter_destroy(&sbi->s_dirs_counter);
843 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
846 for (i = 0; i < MAXQUOTAS; i++)
847 kfree(sbi->s_qf_names[i]);
850 /* Debugging code just in case the in-memory inode orphan list
851 * isn't empty. The on-disk one can be non-empty if we've
852 * detected an error and taken the fs readonly, but the
853 * in-memory list had better be clean by this point. */
854 if (!list_empty(&sbi->s_orphan))
855 dump_orphan_list(sb, sbi);
856 J_ASSERT(list_empty(&sbi->s_orphan));
858 invalidate_bdev(sb->s_bdev);
859 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
861 * Invalidate the journal device's buffers. We don't want them
862 * floating about in memory - the physical journal device may
863 * hotswapped, and it breaks the `ro-after' testing code.
865 sync_blockdev(sbi->journal_bdev);
866 invalidate_bdev(sbi->journal_bdev);
867 ext4_blkdev_remove(sbi);
870 kthread_stop(sbi->s_mmp_tsk);
871 sb->s_fs_info = NULL;
873 * Now that we are completely done shutting down the
874 * superblock, we need to actually destroy the kobject.
877 kobject_put(&sbi->s_kobj);
878 wait_for_completion(&sbi->s_kobj_unregister);
879 kfree(sbi->s_blockgroup_lock);
883 static struct kmem_cache *ext4_inode_cachep;
886 * Called inside transaction, so use GFP_NOFS
888 static struct inode *ext4_alloc_inode(struct super_block *sb)
890 struct ext4_inode_info *ei;
892 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
896 ei->vfs_inode.i_version = 1;
897 ei->vfs_inode.i_data.writeback_index = 0;
898 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
899 INIT_LIST_HEAD(&ei->i_prealloc_list);
900 spin_lock_init(&ei->i_prealloc_lock);
901 ei->i_reserved_data_blocks = 0;
902 ei->i_reserved_meta_blocks = 0;
903 ei->i_allocated_meta_blocks = 0;
904 ei->i_da_metadata_calc_len = 0;
905 spin_lock_init(&(ei->i_block_reservation_lock));
907 ei->i_reserved_quota = 0;
910 INIT_LIST_HEAD(&ei->i_completed_io_list);
911 spin_lock_init(&ei->i_completed_io_lock);
912 ei->cur_aio_dio = NULL;
914 ei->i_datasync_tid = 0;
915 atomic_set(&ei->i_ioend_count, 0);
916 atomic_set(&ei->i_aiodio_unwritten, 0);
918 return &ei->vfs_inode;
921 static int ext4_drop_inode(struct inode *inode)
923 int drop = generic_drop_inode(inode);
925 trace_ext4_drop_inode(inode, drop);
929 static void ext4_i_callback(struct rcu_head *head)
931 struct inode *inode = container_of(head, struct inode, i_rcu);
932 INIT_LIST_HEAD(&inode->i_dentry);
933 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
936 static void ext4_destroy_inode(struct inode *inode)
938 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
939 ext4_msg(inode->i_sb, KERN_ERR,
940 "Inode %lu (%p): orphan list check failed!",
941 inode->i_ino, EXT4_I(inode));
942 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
943 EXT4_I(inode), sizeof(struct ext4_inode_info),
947 call_rcu(&inode->i_rcu, ext4_i_callback);
950 static void init_once(void *foo)
952 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
954 INIT_LIST_HEAD(&ei->i_orphan);
955 #ifdef CONFIG_EXT4_FS_XATTR
956 init_rwsem(&ei->xattr_sem);
958 init_rwsem(&ei->i_data_sem);
959 inode_init_once(&ei->vfs_inode);
962 static int init_inodecache(void)
964 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
965 sizeof(struct ext4_inode_info),
966 0, (SLAB_RECLAIM_ACCOUNT|
969 if (ext4_inode_cachep == NULL)
974 static void destroy_inodecache(void)
976 kmem_cache_destroy(ext4_inode_cachep);
979 void ext4_clear_inode(struct inode *inode)
981 invalidate_inode_buffers(inode);
982 end_writeback(inode);
984 ext4_discard_preallocations(inode);
985 if (EXT4_I(inode)->jinode) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 EXT4_I(inode)->jinode);
988 jbd2_free_inode(EXT4_I(inode)->jinode);
989 EXT4_I(inode)->jinode = NULL;
993 static inline void ext4_show_quota_options(struct seq_file *seq,
994 struct super_block *sb)
996 #if defined(CONFIG_QUOTA)
997 struct ext4_sb_info *sbi = EXT4_SB(sb);
999 if (sbi->s_jquota_fmt) {
1002 switch (sbi->s_jquota_fmt) {
1013 seq_printf(seq, ",jqfmt=%s", fmtname);
1016 if (sbi->s_qf_names[USRQUOTA])
1017 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1019 if (sbi->s_qf_names[GRPQUOTA])
1020 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1022 if (test_opt(sb, USRQUOTA))
1023 seq_puts(seq, ",usrquota");
1025 if (test_opt(sb, GRPQUOTA))
1026 seq_puts(seq, ",grpquota");
1032 * - it's set to a non-default value OR
1033 * - if the per-sb default is different from the global default
1035 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1038 unsigned long def_mount_opts;
1039 struct super_block *sb = vfs->mnt_sb;
1040 struct ext4_sb_info *sbi = EXT4_SB(sb);
1041 struct ext4_super_block *es = sbi->s_es;
1043 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1044 def_errors = le16_to_cpu(es->s_errors);
1046 if (sbi->s_sb_block != 1)
1047 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1048 if (test_opt(sb, MINIX_DF))
1049 seq_puts(seq, ",minixdf");
1050 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1051 seq_puts(seq, ",grpid");
1052 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1053 seq_puts(seq, ",nogrpid");
1054 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1055 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1056 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1058 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1059 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1060 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1062 if (test_opt(sb, ERRORS_RO)) {
1063 if (def_errors == EXT4_ERRORS_PANIC ||
1064 def_errors == EXT4_ERRORS_CONTINUE) {
1065 seq_puts(seq, ",errors=remount-ro");
1068 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1069 seq_puts(seq, ",errors=continue");
1070 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1071 seq_puts(seq, ",errors=panic");
1072 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1073 seq_puts(seq, ",nouid32");
1074 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1075 seq_puts(seq, ",debug");
1076 if (test_opt(sb, OLDALLOC))
1077 seq_puts(seq, ",oldalloc");
1078 #ifdef CONFIG_EXT4_FS_XATTR
1079 if (test_opt(sb, XATTR_USER))
1080 seq_puts(seq, ",user_xattr");
1081 if (!test_opt(sb, XATTR_USER))
1082 seq_puts(seq, ",nouser_xattr");
1084 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1085 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1086 seq_puts(seq, ",acl");
1087 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1088 seq_puts(seq, ",noacl");
1090 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1091 seq_printf(seq, ",commit=%u",
1092 (unsigned) (sbi->s_commit_interval / HZ));
1094 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1095 seq_printf(seq, ",min_batch_time=%u",
1096 (unsigned) sbi->s_min_batch_time);
1098 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1099 seq_printf(seq, ",max_batch_time=%u",
1100 (unsigned) sbi->s_min_batch_time);
1104 * We're changing the default of barrier mount option, so
1105 * let's always display its mount state so it's clear what its
1108 seq_puts(seq, ",barrier=");
1109 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1110 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1111 seq_puts(seq, ",journal_async_commit");
1112 else if (test_opt(sb, JOURNAL_CHECKSUM))
1113 seq_puts(seq, ",journal_checksum");
1114 if (test_opt(sb, I_VERSION))
1115 seq_puts(seq, ",i_version");
1116 if (!test_opt(sb, DELALLOC) &&
1117 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1118 seq_puts(seq, ",nodelalloc");
1120 if (!test_opt(sb, MBLK_IO_SUBMIT))
1121 seq_puts(seq, ",nomblk_io_submit");
1123 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1125 * journal mode get enabled in different ways
1126 * So just print the value even if we didn't specify it
1128 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1129 seq_puts(seq, ",data=journal");
1130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1131 seq_puts(seq, ",data=ordered");
1132 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1133 seq_puts(seq, ",data=writeback");
1135 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1136 seq_printf(seq, ",inode_readahead_blks=%u",
1137 sbi->s_inode_readahead_blks);
1139 if (test_opt(sb, DATA_ERR_ABORT))
1140 seq_puts(seq, ",data_err=abort");
1142 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1143 seq_puts(seq, ",noauto_da_alloc");
1145 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1146 seq_puts(seq, ",discard");
1148 if (test_opt(sb, NOLOAD))
1149 seq_puts(seq, ",norecovery");
1151 if (test_opt(sb, DIOREAD_NOLOCK))
1152 seq_puts(seq, ",dioread_nolock");
1154 if (test_opt(sb, BLOCK_VALIDITY) &&
1155 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1156 seq_puts(seq, ",block_validity");
1158 if (!test_opt(sb, INIT_INODE_TABLE))
1159 seq_puts(seq, ",noinit_inode_table");
1160 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1161 seq_printf(seq, ",init_inode_table=%u",
1162 (unsigned) sbi->s_li_wait_mult);
1164 ext4_show_quota_options(seq, sb);
1169 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1170 u64 ino, u32 generation)
1172 struct inode *inode;
1174 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1175 return ERR_PTR(-ESTALE);
1176 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1177 return ERR_PTR(-ESTALE);
1179 /* iget isn't really right if the inode is currently unallocated!!
1181 * ext4_read_inode will return a bad_inode if the inode had been
1182 * deleted, so we should be safe.
1184 * Currently we don't know the generation for parent directory, so
1185 * a generation of 0 means "accept any"
1187 inode = ext4_iget(sb, ino);
1189 return ERR_CAST(inode);
1190 if (generation && inode->i_generation != generation) {
1192 return ERR_PTR(-ESTALE);
1198 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1199 int fh_len, int fh_type)
1201 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1202 ext4_nfs_get_inode);
1205 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1206 int fh_len, int fh_type)
1208 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1209 ext4_nfs_get_inode);
1213 * Try to release metadata pages (indirect blocks, directories) which are
1214 * mapped via the block device. Since these pages could have journal heads
1215 * which would prevent try_to_free_buffers() from freeing them, we must use
1216 * jbd2 layer's try_to_free_buffers() function to release them.
1218 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1221 journal_t *journal = EXT4_SB(sb)->s_journal;
1223 WARN_ON(PageChecked(page));
1224 if (!page_has_buffers(page))
1227 return jbd2_journal_try_to_free_buffers(journal, page,
1228 wait & ~__GFP_WAIT);
1229 return try_to_free_buffers(page);
1233 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1234 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1236 static int ext4_write_dquot(struct dquot *dquot);
1237 static int ext4_acquire_dquot(struct dquot *dquot);
1238 static int ext4_release_dquot(struct dquot *dquot);
1239 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1240 static int ext4_write_info(struct super_block *sb, int type);
1241 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1243 static int ext4_quota_off(struct super_block *sb, int type);
1244 static int ext4_quota_on_mount(struct super_block *sb, int type);
1245 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1246 size_t len, loff_t off);
1247 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1248 const char *data, size_t len, loff_t off);
1250 static const struct dquot_operations ext4_quota_operations = {
1251 .get_reserved_space = ext4_get_reserved_space,
1252 .write_dquot = ext4_write_dquot,
1253 .acquire_dquot = ext4_acquire_dquot,
1254 .release_dquot = ext4_release_dquot,
1255 .mark_dirty = ext4_mark_dquot_dirty,
1256 .write_info = ext4_write_info,
1257 .alloc_dquot = dquot_alloc,
1258 .destroy_dquot = dquot_destroy,
1261 static const struct quotactl_ops ext4_qctl_operations = {
1262 .quota_on = ext4_quota_on,
1263 .quota_off = ext4_quota_off,
1264 .quota_sync = dquot_quota_sync,
1265 .get_info = dquot_get_dqinfo,
1266 .set_info = dquot_set_dqinfo,
1267 .get_dqblk = dquot_get_dqblk,
1268 .set_dqblk = dquot_set_dqblk
1272 static const struct super_operations ext4_sops = {
1273 .alloc_inode = ext4_alloc_inode,
1274 .destroy_inode = ext4_destroy_inode,
1275 .write_inode = ext4_write_inode,
1276 .dirty_inode = ext4_dirty_inode,
1277 .drop_inode = ext4_drop_inode,
1278 .evict_inode = ext4_evict_inode,
1279 .put_super = ext4_put_super,
1280 .sync_fs = ext4_sync_fs,
1281 .freeze_fs = ext4_freeze,
1282 .unfreeze_fs = ext4_unfreeze,
1283 .statfs = ext4_statfs,
1284 .remount_fs = ext4_remount,
1285 .show_options = ext4_show_options,
1287 .quota_read = ext4_quota_read,
1288 .quota_write = ext4_quota_write,
1290 .bdev_try_to_free_page = bdev_try_to_free_page,
1293 static const struct super_operations ext4_nojournal_sops = {
1294 .alloc_inode = ext4_alloc_inode,
1295 .destroy_inode = ext4_destroy_inode,
1296 .write_inode = ext4_write_inode,
1297 .dirty_inode = ext4_dirty_inode,
1298 .drop_inode = ext4_drop_inode,
1299 .evict_inode = ext4_evict_inode,
1300 .write_super = ext4_write_super,
1301 .put_super = ext4_put_super,
1302 .statfs = ext4_statfs,
1303 .remount_fs = ext4_remount,
1304 .show_options = ext4_show_options,
1306 .quota_read = ext4_quota_read,
1307 .quota_write = ext4_quota_write,
1309 .bdev_try_to_free_page = bdev_try_to_free_page,
1312 static const struct export_operations ext4_export_ops = {
1313 .fh_to_dentry = ext4_fh_to_dentry,
1314 .fh_to_parent = ext4_fh_to_parent,
1315 .get_parent = ext4_get_parent,
1319 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1320 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1321 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1322 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1323 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1324 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1325 Opt_journal_update, Opt_journal_dev,
1326 Opt_journal_checksum, Opt_journal_async_commit,
1327 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1328 Opt_data_err_abort, Opt_data_err_ignore,
1329 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1330 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1331 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1332 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1333 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1334 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1335 Opt_inode_readahead_blks, Opt_journal_ioprio,
1336 Opt_dioread_nolock, Opt_dioread_lock,
1337 Opt_discard, Opt_nodiscard,
1338 Opt_init_inode_table, Opt_noinit_inode_table,
1341 static const match_table_t tokens = {
1342 {Opt_bsd_df, "bsddf"},
1343 {Opt_minix_df, "minixdf"},
1344 {Opt_grpid, "grpid"},
1345 {Opt_grpid, "bsdgroups"},
1346 {Opt_nogrpid, "nogrpid"},
1347 {Opt_nogrpid, "sysvgroups"},
1348 {Opt_resgid, "resgid=%u"},
1349 {Opt_resuid, "resuid=%u"},
1351 {Opt_err_cont, "errors=continue"},
1352 {Opt_err_panic, "errors=panic"},
1353 {Opt_err_ro, "errors=remount-ro"},
1354 {Opt_nouid32, "nouid32"},
1355 {Opt_debug, "debug"},
1356 {Opt_oldalloc, "oldalloc"},
1357 {Opt_orlov, "orlov"},
1358 {Opt_user_xattr, "user_xattr"},
1359 {Opt_nouser_xattr, "nouser_xattr"},
1361 {Opt_noacl, "noacl"},
1362 {Opt_noload, "noload"},
1363 {Opt_noload, "norecovery"},
1366 {Opt_commit, "commit=%u"},
1367 {Opt_min_batch_time, "min_batch_time=%u"},
1368 {Opt_max_batch_time, "max_batch_time=%u"},
1369 {Opt_journal_update, "journal=update"},
1370 {Opt_journal_dev, "journal_dev=%u"},
1371 {Opt_journal_checksum, "journal_checksum"},
1372 {Opt_journal_async_commit, "journal_async_commit"},
1373 {Opt_abort, "abort"},
1374 {Opt_data_journal, "data=journal"},
1375 {Opt_data_ordered, "data=ordered"},
1376 {Opt_data_writeback, "data=writeback"},
1377 {Opt_data_err_abort, "data_err=abort"},
1378 {Opt_data_err_ignore, "data_err=ignore"},
1379 {Opt_offusrjquota, "usrjquota="},
1380 {Opt_usrjquota, "usrjquota=%s"},
1381 {Opt_offgrpjquota, "grpjquota="},
1382 {Opt_grpjquota, "grpjquota=%s"},
1383 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1384 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1385 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1386 {Opt_grpquota, "grpquota"},
1387 {Opt_noquota, "noquota"},
1388 {Opt_quota, "quota"},
1389 {Opt_usrquota, "usrquota"},
1390 {Opt_barrier, "barrier=%u"},
1391 {Opt_barrier, "barrier"},
1392 {Opt_nobarrier, "nobarrier"},
1393 {Opt_i_version, "i_version"},
1394 {Opt_stripe, "stripe=%u"},
1395 {Opt_resize, "resize"},
1396 {Opt_delalloc, "delalloc"},
1397 {Opt_nodelalloc, "nodelalloc"},
1398 {Opt_mblk_io_submit, "mblk_io_submit"},
1399 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1400 {Opt_block_validity, "block_validity"},
1401 {Opt_noblock_validity, "noblock_validity"},
1402 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1403 {Opt_journal_ioprio, "journal_ioprio=%u"},
1404 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1405 {Opt_auto_da_alloc, "auto_da_alloc"},
1406 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1407 {Opt_dioread_nolock, "dioread_nolock"},
1408 {Opt_dioread_lock, "dioread_lock"},
1409 {Opt_discard, "discard"},
1410 {Opt_nodiscard, "nodiscard"},
1411 {Opt_init_inode_table, "init_itable=%u"},
1412 {Opt_init_inode_table, "init_itable"},
1413 {Opt_noinit_inode_table, "noinit_itable"},
1417 static ext4_fsblk_t get_sb_block(void **data)
1419 ext4_fsblk_t sb_block;
1420 char *options = (char *) *data;
1422 if (!options || strncmp(options, "sb=", 3) != 0)
1423 return 1; /* Default location */
1426 /* TODO: use simple_strtoll with >32bit ext4 */
1427 sb_block = simple_strtoul(options, &options, 0);
1428 if (*options && *options != ',') {
1429 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1433 if (*options == ',')
1435 *data = (void *) options;
1440 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1441 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1442 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1445 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1450 if (sb_any_quota_loaded(sb) &&
1451 !sbi->s_qf_names[qtype]) {
1452 ext4_msg(sb, KERN_ERR,
1453 "Cannot change journaled "
1454 "quota options when quota turned on");
1457 qname = match_strdup(args);
1459 ext4_msg(sb, KERN_ERR,
1460 "Not enough memory for storing quotafile name");
1463 if (sbi->s_qf_names[qtype] &&
1464 strcmp(sbi->s_qf_names[qtype], qname)) {
1465 ext4_msg(sb, KERN_ERR,
1466 "%s quota file already specified", QTYPE2NAME(qtype));
1470 sbi->s_qf_names[qtype] = qname;
1471 if (strchr(sbi->s_qf_names[qtype], '/')) {
1472 ext4_msg(sb, KERN_ERR,
1473 "quotafile must be on filesystem root");
1474 kfree(sbi->s_qf_names[qtype]);
1475 sbi->s_qf_names[qtype] = NULL;
1482 static int clear_qf_name(struct super_block *sb, int qtype)
1485 struct ext4_sb_info *sbi = EXT4_SB(sb);
1487 if (sb_any_quota_loaded(sb) &&
1488 sbi->s_qf_names[qtype]) {
1489 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1490 " when quota turned on");
1494 * The space will be released later when all options are confirmed
1497 sbi->s_qf_names[qtype] = NULL;
1502 static int parse_options(char *options, struct super_block *sb,
1503 unsigned long *journal_devnum,
1504 unsigned int *journal_ioprio,
1505 ext4_fsblk_t *n_blocks_count, int is_remount)
1507 struct ext4_sb_info *sbi = EXT4_SB(sb);
1509 substring_t args[MAX_OPT_ARGS];
1519 while ((p = strsep(&options, ",")) != NULL) {
1525 * Initialize args struct so we know whether arg was
1526 * found; some options take optional arguments.
1528 args[0].to = args[0].from = NULL;
1529 token = match_token(p, tokens, args);
1532 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1533 clear_opt(sb, MINIX_DF);
1536 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1537 set_opt(sb, MINIX_DF);
1541 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1546 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1547 clear_opt(sb, GRPID);
1551 if (match_int(&args[0], &option))
1553 sbi->s_resuid = option;
1556 if (match_int(&args[0], &option))
1558 sbi->s_resgid = option;
1561 /* handled by get_sb_block() instead of here */
1562 /* *sb_block = match_int(&args[0]); */
1565 clear_opt(sb, ERRORS_CONT);
1566 clear_opt(sb, ERRORS_RO);
1567 set_opt(sb, ERRORS_PANIC);
1570 clear_opt(sb, ERRORS_CONT);
1571 clear_opt(sb, ERRORS_PANIC);
1572 set_opt(sb, ERRORS_RO);
1575 clear_opt(sb, ERRORS_RO);
1576 clear_opt(sb, ERRORS_PANIC);
1577 set_opt(sb, ERRORS_CONT);
1580 set_opt(sb, NO_UID32);
1586 set_opt(sb, OLDALLOC);
1589 clear_opt(sb, OLDALLOC);
1591 #ifdef CONFIG_EXT4_FS_XATTR
1592 case Opt_user_xattr:
1593 set_opt(sb, XATTR_USER);
1595 case Opt_nouser_xattr:
1596 clear_opt(sb, XATTR_USER);
1599 case Opt_user_xattr:
1600 case Opt_nouser_xattr:
1601 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1604 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1606 set_opt(sb, POSIX_ACL);
1609 clear_opt(sb, POSIX_ACL);
1614 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1617 case Opt_journal_update:
1619 /* Eventually we will want to be able to create
1620 a journal file here. For now, only allow the
1621 user to specify an existing inode to be the
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot specify journal on remount");
1628 set_opt(sb, UPDATE_JOURNAL);
1630 case Opt_journal_dev:
1632 ext4_msg(sb, KERN_ERR,
1633 "Cannot specify journal on remount");
1636 if (match_int(&args[0], &option))
1638 *journal_devnum = option;
1640 case Opt_journal_checksum:
1641 set_opt(sb, JOURNAL_CHECKSUM);
1643 case Opt_journal_async_commit:
1644 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1645 set_opt(sb, JOURNAL_CHECKSUM);
1648 set_opt(sb, NOLOAD);
1651 if (match_int(&args[0], &option))
1656 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1657 sbi->s_commit_interval = HZ * option;
1659 case Opt_max_batch_time:
1660 if (match_int(&args[0], &option))
1665 option = EXT4_DEF_MAX_BATCH_TIME;
1666 sbi->s_max_batch_time = option;
1668 case Opt_min_batch_time:
1669 if (match_int(&args[0], &option))
1673 sbi->s_min_batch_time = option;
1675 case Opt_data_journal:
1676 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1678 case Opt_data_ordered:
1679 data_opt = EXT4_MOUNT_ORDERED_DATA;
1681 case Opt_data_writeback:
1682 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1685 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1686 ext4_msg(sb, KERN_ERR,
1687 "Cannot change data mode on remount");
1691 clear_opt(sb, DATA_FLAGS);
1692 sbi->s_mount_opt |= data_opt;
1695 case Opt_data_err_abort:
1696 set_opt(sb, DATA_ERR_ABORT);
1698 case Opt_data_err_ignore:
1699 clear_opt(sb, DATA_ERR_ABORT);
1703 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1707 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1710 case Opt_offusrjquota:
1711 if (!clear_qf_name(sb, USRQUOTA))
1714 case Opt_offgrpjquota:
1715 if (!clear_qf_name(sb, GRPQUOTA))
1719 case Opt_jqfmt_vfsold:
1720 qfmt = QFMT_VFS_OLD;
1722 case Opt_jqfmt_vfsv0:
1725 case Opt_jqfmt_vfsv1:
1728 if (sb_any_quota_loaded(sb) &&
1729 sbi->s_jquota_fmt != qfmt) {
1730 ext4_msg(sb, KERN_ERR, "Cannot change "
1731 "journaled quota options when "
1735 sbi->s_jquota_fmt = qfmt;
1740 set_opt(sb, USRQUOTA);
1744 set_opt(sb, GRPQUOTA);
1747 if (sb_any_quota_loaded(sb)) {
1748 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1749 "options when quota turned on");
1752 clear_opt(sb, QUOTA);
1753 clear_opt(sb, USRQUOTA);
1754 clear_opt(sb, GRPQUOTA);
1760 ext4_msg(sb, KERN_ERR,
1761 "quota options not supported");
1765 case Opt_offusrjquota:
1766 case Opt_offgrpjquota:
1767 case Opt_jqfmt_vfsold:
1768 case Opt_jqfmt_vfsv0:
1769 case Opt_jqfmt_vfsv1:
1770 ext4_msg(sb, KERN_ERR,
1771 "journaled quota options not supported");
1777 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1780 clear_opt(sb, BARRIER);
1784 if (match_int(&args[0], &option))
1787 option = 1; /* No argument, default to 1 */
1789 set_opt(sb, BARRIER);
1791 clear_opt(sb, BARRIER);
1797 ext4_msg(sb, KERN_ERR,
1798 "resize option only available "
1802 if (match_int(&args[0], &option) != 0)
1804 *n_blocks_count = option;
1807 ext4_msg(sb, KERN_WARNING,
1808 "Ignoring deprecated nobh option");
1811 ext4_msg(sb, KERN_WARNING,
1812 "Ignoring deprecated bh option");
1815 set_opt(sb, I_VERSION);
1816 sb->s_flags |= MS_I_VERSION;
1818 case Opt_nodelalloc:
1819 clear_opt(sb, DELALLOC);
1820 clear_opt2(sb, EXPLICIT_DELALLOC);
1822 case Opt_mblk_io_submit:
1823 set_opt(sb, MBLK_IO_SUBMIT);
1825 case Opt_nomblk_io_submit:
1826 clear_opt(sb, MBLK_IO_SUBMIT);
1829 if (match_int(&args[0], &option))
1833 sbi->s_stripe = option;
1836 set_opt(sb, DELALLOC);
1837 set_opt2(sb, EXPLICIT_DELALLOC);
1839 case Opt_block_validity:
1840 set_opt(sb, BLOCK_VALIDITY);
1842 case Opt_noblock_validity:
1843 clear_opt(sb, BLOCK_VALIDITY);
1845 case Opt_inode_readahead_blks:
1846 if (match_int(&args[0], &option))
1848 if (option < 0 || option > (1 << 30))
1850 if (option && !is_power_of_2(option)) {
1851 ext4_msg(sb, KERN_ERR,
1852 "EXT4-fs: inode_readahead_blks"
1853 " must be a power of 2");
1856 sbi->s_inode_readahead_blks = option;
1858 case Opt_journal_ioprio:
1859 if (match_int(&args[0], &option))
1861 if (option < 0 || option > 7)
1863 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1866 case Opt_noauto_da_alloc:
1867 set_opt(sb, NO_AUTO_DA_ALLOC);
1869 case Opt_auto_da_alloc:
1871 if (match_int(&args[0], &option))
1874 option = 1; /* No argument, default to 1 */
1876 clear_opt(sb, NO_AUTO_DA_ALLOC);
1878 set_opt(sb,NO_AUTO_DA_ALLOC);
1881 set_opt(sb, DISCARD);
1884 clear_opt(sb, DISCARD);
1886 case Opt_dioread_nolock:
1887 set_opt(sb, DIOREAD_NOLOCK);
1889 case Opt_dioread_lock:
1890 clear_opt(sb, DIOREAD_NOLOCK);
1892 case Opt_init_inode_table:
1893 set_opt(sb, INIT_INODE_TABLE);
1895 if (match_int(&args[0], &option))
1898 option = EXT4_DEF_LI_WAIT_MULT;
1901 sbi->s_li_wait_mult = option;
1903 case Opt_noinit_inode_table:
1904 clear_opt(sb, INIT_INODE_TABLE);
1907 ext4_msg(sb, KERN_ERR,
1908 "Unrecognized mount option \"%s\" "
1909 "or missing value", p);
1914 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1915 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1916 clear_opt(sb, USRQUOTA);
1918 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1919 clear_opt(sb, GRPQUOTA);
1921 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1922 ext4_msg(sb, KERN_ERR, "old and new quota "
1927 if (!sbi->s_jquota_fmt) {
1928 ext4_msg(sb, KERN_ERR, "journaled quota format "
1933 if (sbi->s_jquota_fmt) {
1934 ext4_msg(sb, KERN_ERR, "journaled quota format "
1935 "specified with no journaling "
1944 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1947 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1951 ext4_msg(sb, KERN_ERR, "revision level too high, "
1952 "forcing read-only mode");
1957 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1958 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1959 "running e2fsck is recommended");
1960 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1961 ext4_msg(sb, KERN_WARNING,
1962 "warning: mounting fs with errors, "
1963 "running e2fsck is recommended");
1964 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1965 le16_to_cpu(es->s_mnt_count) >=
1966 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1967 ext4_msg(sb, KERN_WARNING,
1968 "warning: maximal mount count reached, "
1969 "running e2fsck is recommended");
1970 else if (le32_to_cpu(es->s_checkinterval) &&
1971 (le32_to_cpu(es->s_lastcheck) +
1972 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1973 ext4_msg(sb, KERN_WARNING,
1974 "warning: checktime reached, "
1975 "running e2fsck is recommended");
1976 if (!sbi->s_journal)
1977 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1978 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1979 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1980 le16_add_cpu(&es->s_mnt_count, 1);
1981 es->s_mtime = cpu_to_le32(get_seconds());
1982 ext4_update_dynamic_rev(sb);
1984 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1986 ext4_commit_super(sb, 1);
1988 if (test_opt(sb, DEBUG))
1989 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1990 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1992 sbi->s_groups_count,
1993 EXT4_BLOCKS_PER_GROUP(sb),
1994 EXT4_INODES_PER_GROUP(sb),
1995 sbi->s_mount_opt, sbi->s_mount_opt2);
1997 cleancache_init_fs(sb);
2001 static int ext4_fill_flex_info(struct super_block *sb)
2003 struct ext4_sb_info *sbi = EXT4_SB(sb);
2004 struct ext4_group_desc *gdp = NULL;
2005 ext4_group_t flex_group_count;
2006 ext4_group_t flex_group;
2007 int groups_per_flex = 0;
2011 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2012 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2014 if (groups_per_flex < 2) {
2015 sbi->s_log_groups_per_flex = 0;
2019 /* We allocate both existing and potentially added groups */
2020 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2021 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2022 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2023 size = flex_group_count * sizeof(struct flex_groups);
2024 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2025 if (sbi->s_flex_groups == NULL) {
2026 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2031 for (i = 0; i < sbi->s_groups_count; i++) {
2032 gdp = ext4_get_group_desc(sb, i, NULL);
2034 flex_group = ext4_flex_group(sbi, i);
2035 atomic_add(ext4_free_inodes_count(sb, gdp),
2036 &sbi->s_flex_groups[flex_group].free_inodes);
2037 atomic_add(ext4_free_group_clusters(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_clusters);
2039 atomic_add(ext4_used_dirs_count(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].used_dirs);
2048 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2053 if (sbi->s_es->s_feature_ro_compat &
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2055 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2056 __le32 le_group = cpu_to_le32(block_group);
2058 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2059 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2060 crc = crc16(crc, (__u8 *)gdp, offset);
2061 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi->s_es->s_feature_incompat &
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2065 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2066 crc = crc16(crc, (__u8 *)gdp + offset,
2067 le16_to_cpu(sbi->s_es->s_desc_size) -
2071 return cpu_to_le16(crc);
2074 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2075 struct ext4_group_desc *gdp)
2077 if ((sbi->s_es->s_feature_ro_compat &
2078 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2079 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2085 /* Called at mount-time, super-block is locked */
2086 static int ext4_check_descriptors(struct super_block *sb,
2087 ext4_group_t *first_not_zeroed)
2089 struct ext4_sb_info *sbi = EXT4_SB(sb);
2090 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2091 ext4_fsblk_t last_block;
2092 ext4_fsblk_t block_bitmap;
2093 ext4_fsblk_t inode_bitmap;
2094 ext4_fsblk_t inode_table;
2095 int flexbg_flag = 0;
2096 ext4_group_t i, grp = sbi->s_groups_count;
2098 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2101 ext4_debug("Checking group descriptors");
2103 for (i = 0; i < sbi->s_groups_count; i++) {
2104 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2106 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2107 last_block = ext4_blocks_count(sbi->s_es) - 1;
2109 last_block = first_block +
2110 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2112 if ((grp == sbi->s_groups_count) &&
2113 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2116 block_bitmap = ext4_block_bitmap(sb, gdp);
2117 if (block_bitmap < first_block || block_bitmap > last_block) {
2118 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2119 "Block bitmap for group %u not in group "
2120 "(block %llu)!", i, block_bitmap);
2123 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2124 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2126 "Inode bitmap for group %u not in group "
2127 "(block %llu)!", i, inode_bitmap);
2130 inode_table = ext4_inode_table(sb, gdp);
2131 if (inode_table < first_block ||
2132 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2134 "Inode table for group %u not in group "
2135 "(block %llu)!", i, inode_table);
2138 ext4_lock_group(sb, i);
2139 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2140 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2141 "Checksum for group %u failed (%u!=%u)",
2142 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2143 gdp)), le16_to_cpu(gdp->bg_checksum));
2144 if (!(sb->s_flags & MS_RDONLY)) {
2145 ext4_unlock_group(sb, i);
2149 ext4_unlock_group(sb, i);
2151 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2153 if (NULL != first_not_zeroed)
2154 *first_not_zeroed = grp;
2156 ext4_free_blocks_count_set(sbi->s_es,
2157 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2158 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2162 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2163 * the superblock) which were deleted from all directories, but held open by
2164 * a process at the time of a crash. We walk the list and try to delete these
2165 * inodes at recovery time (only with a read-write filesystem).
2167 * In order to keep the orphan inode chain consistent during traversal (in
2168 * case of crash during recovery), we link each inode into the superblock
2169 * orphan list_head and handle it the same way as an inode deletion during
2170 * normal operation (which journals the operations for us).
2172 * We only do an iget() and an iput() on each inode, which is very safe if we
2173 * accidentally point at an in-use or already deleted inode. The worst that
2174 * can happen in this case is that we get a "bit already cleared" message from
2175 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2176 * e2fsck was run on this filesystem, and it must have already done the orphan
2177 * inode cleanup for us, so we can safely abort without any further action.
2179 static void ext4_orphan_cleanup(struct super_block *sb,
2180 struct ext4_super_block *es)
2182 unsigned int s_flags = sb->s_flags;
2183 int nr_orphans = 0, nr_truncates = 0;
2187 if (!es->s_last_orphan) {
2188 jbd_debug(4, "no orphan inodes to clean up\n");
2192 if (bdev_read_only(sb->s_bdev)) {
2193 ext4_msg(sb, KERN_ERR, "write access "
2194 "unavailable, skipping orphan cleanup");
2198 /* Check if feature set would not allow a r/w mount */
2199 if (!ext4_feature_set_ok(sb, 0)) {
2200 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2201 "unknown ROCOMPAT features");
2205 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2206 if (es->s_last_orphan)
2207 jbd_debug(1, "Errors on filesystem, "
2208 "clearing orphan list.\n");
2209 es->s_last_orphan = 0;
2210 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2214 if (s_flags & MS_RDONLY) {
2215 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2216 sb->s_flags &= ~MS_RDONLY;
2219 /* Needed for iput() to work correctly and not trash data */
2220 sb->s_flags |= MS_ACTIVE;
2221 /* Turn on quotas so that they are updated correctly */
2222 for (i = 0; i < MAXQUOTAS; i++) {
2223 if (EXT4_SB(sb)->s_qf_names[i]) {
2224 int ret = ext4_quota_on_mount(sb, i);
2226 ext4_msg(sb, KERN_ERR,
2227 "Cannot turn on journaled "
2228 "quota: error %d", ret);
2233 while (es->s_last_orphan) {
2234 struct inode *inode;
2236 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2237 if (IS_ERR(inode)) {
2238 es->s_last_orphan = 0;
2242 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2243 dquot_initialize(inode);
2244 if (inode->i_nlink) {
2245 ext4_msg(sb, KERN_DEBUG,
2246 "%s: truncating inode %lu to %lld bytes",
2247 __func__, inode->i_ino, inode->i_size);
2248 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2249 inode->i_ino, inode->i_size);
2250 ext4_truncate(inode);
2253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: deleting unreferenced inode %lu",
2255 __func__, inode->i_ino);
2256 jbd_debug(2, "deleting unreferenced inode %lu\n",
2260 iput(inode); /* The delete magic happens here! */
2263 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2266 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2267 PLURAL(nr_orphans));
2269 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2270 PLURAL(nr_truncates));
2272 /* Turn quotas off */
2273 for (i = 0; i < MAXQUOTAS; i++) {
2274 if (sb_dqopt(sb)->files[i])
2275 dquot_quota_off(sb, i);
2278 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2282 * Maximal extent format file size.
2283 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2284 * extent format containers, within a sector_t, and within i_blocks
2285 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2286 * so that won't be a limiting factor.
2288 * However there is other limiting factor. We do store extents in the form
2289 * of starting block and length, hence the resulting length of the extent
2290 * covering maximum file size must fit into on-disk format containers as
2291 * well. Given that length is always by 1 unit bigger than max unit (because
2292 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2294 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2296 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2299 loff_t upper_limit = MAX_LFS_FILESIZE;
2301 /* small i_blocks in vfs inode? */
2302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2304 * CONFIG_LBDAF is not enabled implies the inode
2305 * i_block represent total blocks in 512 bytes
2306 * 32 == size of vfs inode i_blocks * 8
2308 upper_limit = (1LL << 32) - 1;
2310 /* total blocks in file system block size */
2311 upper_limit >>= (blkbits - 9);
2312 upper_limit <<= blkbits;
2316 * 32-bit extent-start container, ee_block. We lower the maxbytes
2317 * by one fs block, so ee_len can cover the extent of maximum file
2320 res = (1LL << 32) - 1;
2323 /* Sanity check against vm- & vfs- imposed limits */
2324 if (res > upper_limit)
2331 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2332 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2333 * We need to be 1 filesystem block less than the 2^48 sector limit.
2335 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2337 loff_t res = EXT4_NDIR_BLOCKS;
2340 /* This is calculated to be the largest file size for a dense, block
2341 * mapped file such that the file's total number of 512-byte sectors,
2342 * including data and all indirect blocks, does not exceed (2^48 - 1).
2344 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2345 * number of 512-byte sectors of the file.
2348 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2350 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2351 * the inode i_block field represents total file blocks in
2352 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2354 upper_limit = (1LL << 32) - 1;
2356 /* total blocks in file system block size */
2357 upper_limit >>= (bits - 9);
2361 * We use 48 bit ext4_inode i_blocks
2362 * With EXT4_HUGE_FILE_FL set the i_blocks
2363 * represent total number of blocks in
2364 * file system block size
2366 upper_limit = (1LL << 48) - 1;
2370 /* indirect blocks */
2372 /* double indirect blocks */
2373 meta_blocks += 1 + (1LL << (bits-2));
2374 /* tripple indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2377 upper_limit -= meta_blocks;
2378 upper_limit <<= bits;
2380 res += 1LL << (bits-2);
2381 res += 1LL << (2*(bits-2));
2382 res += 1LL << (3*(bits-2));
2384 if (res > upper_limit)
2387 if (res > MAX_LFS_FILESIZE)
2388 res = MAX_LFS_FILESIZE;
2393 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2394 ext4_fsblk_t logical_sb_block, int nr)
2396 struct ext4_sb_info *sbi = EXT4_SB(sb);
2397 ext4_group_t bg, first_meta_bg;
2400 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2404 return logical_sb_block + nr + 1;
2405 bg = sbi->s_desc_per_block * nr;
2406 if (ext4_bg_has_super(sb, bg))
2409 return (has_super + ext4_group_first_block_no(sb, bg));
2413 * ext4_get_stripe_size: Get the stripe size.
2414 * @sbi: In memory super block info
2416 * If we have specified it via mount option, then
2417 * use the mount option value. If the value specified at mount time is
2418 * greater than the blocks per group use the super block value.
2419 * If the super block value is greater than blocks per group return 0.
2420 * Allocator needs it be less than blocks per group.
2423 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2425 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2426 unsigned long stripe_width =
2427 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2430 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2431 ret = sbi->s_stripe;
2432 else if (stripe_width <= sbi->s_blocks_per_group)
2434 else if (stride <= sbi->s_blocks_per_group)
2440 * If the stripe width is 1, this makes no sense and
2441 * we set it to 0 to turn off stripe handling code.
2452 struct attribute attr;
2453 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2454 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2455 const char *, size_t);
2459 static int parse_strtoul(const char *buf,
2460 unsigned long max, unsigned long *value)
2464 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2465 endp = skip_spaces(endp);
2466 if (*endp || *value > max)
2472 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi,
2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
2478 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2481 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi, char *buf)
2484 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2486 if (!sb->s_bdev->bd_part)
2487 return snprintf(buf, PAGE_SIZE, "0\n");
2488 return snprintf(buf, PAGE_SIZE, "%lu\n",
2489 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2490 sbi->s_sectors_written_start) >> 1);
2493 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
2500 return snprintf(buf, PAGE_SIZE, "%llu\n",
2501 (unsigned long long)(sbi->s_kbytes_written +
2502 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2503 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2506 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi, char *buf)
2509 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2512 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2513 struct ext4_sb_info *sbi, char *buf)
2515 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2518 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2524 if (parse_strtoul(buf, 0x40000000, &t))
2527 if (t && !is_power_of_2(t))
2530 sbi->s_inode_readahead_blks = t;
2534 static ssize_t sbi_ui_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2537 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2539 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2542 static ssize_t sbi_ui_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2546 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2549 if (parse_strtoul(buf, 0xffffffff, &t))
2555 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2556 static struct ext4_attr ext4_attr_##_name = { \
2557 .attr = {.name = __stringify(_name), .mode = _mode }, \
2560 .offset = offsetof(struct ext4_sb_info, _elname), \
2562 #define EXT4_ATTR(name, mode, show, store) \
2563 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2565 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2566 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2567 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2568 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2569 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2570 #define ATTR_LIST(name) &ext4_attr_##name.attr
2572 EXT4_RO_ATTR(delayed_allocation_blocks);
2573 EXT4_RO_ATTR(session_write_kbytes);
2574 EXT4_RO_ATTR(lifetime_write_kbytes);
2575 EXT4_RO_ATTR(extent_cache_hits);
2576 EXT4_RO_ATTR(extent_cache_misses);
2577 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2578 inode_readahead_blks_store, s_inode_readahead_blks);
2579 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2580 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2581 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2582 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2583 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2584 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2585 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2586 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2588 static struct attribute *ext4_attrs[] = {
2589 ATTR_LIST(delayed_allocation_blocks),
2590 ATTR_LIST(session_write_kbytes),
2591 ATTR_LIST(lifetime_write_kbytes),
2592 ATTR_LIST(extent_cache_hits),
2593 ATTR_LIST(extent_cache_misses),
2594 ATTR_LIST(inode_readahead_blks),
2595 ATTR_LIST(inode_goal),
2596 ATTR_LIST(mb_stats),
2597 ATTR_LIST(mb_max_to_scan),
2598 ATTR_LIST(mb_min_to_scan),
2599 ATTR_LIST(mb_order2_req),
2600 ATTR_LIST(mb_stream_req),
2601 ATTR_LIST(mb_group_prealloc),
2602 ATTR_LIST(max_writeback_mb_bump),
2606 /* Features this copy of ext4 supports */
2607 EXT4_INFO_ATTR(lazy_itable_init);
2608 EXT4_INFO_ATTR(batched_discard);
2610 static struct attribute *ext4_feat_attrs[] = {
2611 ATTR_LIST(lazy_itable_init),
2612 ATTR_LIST(batched_discard),
2616 static ssize_t ext4_attr_show(struct kobject *kobj,
2617 struct attribute *attr, char *buf)
2619 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2621 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2623 return a->show ? a->show(a, sbi, buf) : 0;
2626 static ssize_t ext4_attr_store(struct kobject *kobj,
2627 struct attribute *attr,
2628 const char *buf, size_t len)
2630 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2632 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2634 return a->store ? a->store(a, sbi, buf, len) : 0;
2637 static void ext4_sb_release(struct kobject *kobj)
2639 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2641 complete(&sbi->s_kobj_unregister);
2644 static const struct sysfs_ops ext4_attr_ops = {
2645 .show = ext4_attr_show,
2646 .store = ext4_attr_store,
2649 static struct kobj_type ext4_ktype = {
2650 .default_attrs = ext4_attrs,
2651 .sysfs_ops = &ext4_attr_ops,
2652 .release = ext4_sb_release,
2655 static void ext4_feat_release(struct kobject *kobj)
2657 complete(&ext4_feat->f_kobj_unregister);
2660 static struct kobj_type ext4_feat_ktype = {
2661 .default_attrs = ext4_feat_attrs,
2662 .sysfs_ops = &ext4_attr_ops,
2663 .release = ext4_feat_release,
2667 * Check whether this filesystem can be mounted based on
2668 * the features present and the RDONLY/RDWR mount requested.
2669 * Returns 1 if this filesystem can be mounted as requested,
2670 * 0 if it cannot be.
2672 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2674 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2675 ext4_msg(sb, KERN_ERR,
2676 "Couldn't mount because of "
2677 "unsupported optional features (%x)",
2678 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2679 ~EXT4_FEATURE_INCOMPAT_SUPP));
2686 /* Check that feature set is OK for a read-write mount */
2687 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2688 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2689 "unsupported optional features (%x)",
2690 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2691 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2695 * Large file size enabled file system can only be mounted
2696 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2698 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2699 if (sizeof(blkcnt_t) < sizeof(u64)) {
2700 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2701 "cannot be mounted RDWR without "
2706 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2707 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2708 ext4_msg(sb, KERN_ERR,
2709 "Can't support bigalloc feature without "
2710 "extents feature\n");
2717 * This function is called once a day if we have errors logged
2718 * on the file system
2720 static void print_daily_error_info(unsigned long arg)
2722 struct super_block *sb = (struct super_block *) arg;
2723 struct ext4_sb_info *sbi;
2724 struct ext4_super_block *es;
2729 if (es->s_error_count)
2730 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2731 le32_to_cpu(es->s_error_count));
2732 if (es->s_first_error_time) {
2733 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2734 sb->s_id, le32_to_cpu(es->s_first_error_time),
2735 (int) sizeof(es->s_first_error_func),
2736 es->s_first_error_func,
2737 le32_to_cpu(es->s_first_error_line));
2738 if (es->s_first_error_ino)
2739 printk(": inode %u",
2740 le32_to_cpu(es->s_first_error_ino));
2741 if (es->s_first_error_block)
2742 printk(": block %llu", (unsigned long long)
2743 le64_to_cpu(es->s_first_error_block));
2746 if (es->s_last_error_time) {
2747 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2748 sb->s_id, le32_to_cpu(es->s_last_error_time),
2749 (int) sizeof(es->s_last_error_func),
2750 es->s_last_error_func,
2751 le32_to_cpu(es->s_last_error_line));
2752 if (es->s_last_error_ino)
2753 printk(": inode %u",
2754 le32_to_cpu(es->s_last_error_ino));
2755 if (es->s_last_error_block)
2756 printk(": block %llu", (unsigned long long)
2757 le64_to_cpu(es->s_last_error_block));
2760 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2763 /* Find next suitable group and run ext4_init_inode_table */
2764 static int ext4_run_li_request(struct ext4_li_request *elr)
2766 struct ext4_group_desc *gdp = NULL;
2767 ext4_group_t group, ngroups;
2768 struct super_block *sb;
2769 unsigned long timeout = 0;
2773 ngroups = EXT4_SB(sb)->s_groups_count;
2775 for (group = elr->lr_next_group; group < ngroups; group++) {
2776 gdp = ext4_get_group_desc(sb, group, NULL);
2782 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2786 if (group == ngroups)
2791 ret = ext4_init_inode_table(sb, group,
2792 elr->lr_timeout ? 0 : 1);
2793 if (elr->lr_timeout == 0) {
2794 timeout = (jiffies - timeout) *
2795 elr->lr_sbi->s_li_wait_mult;
2796 elr->lr_timeout = timeout;
2798 elr->lr_next_sched = jiffies + elr->lr_timeout;
2799 elr->lr_next_group = group + 1;
2806 * Remove lr_request from the list_request and free the
2807 * request structure. Should be called with li_list_mtx held
2809 static void ext4_remove_li_request(struct ext4_li_request *elr)
2811 struct ext4_sb_info *sbi;
2818 list_del(&elr->lr_request);
2819 sbi->s_li_request = NULL;
2823 static void ext4_unregister_li_request(struct super_block *sb)
2825 mutex_lock(&ext4_li_mtx);
2826 if (!ext4_li_info) {
2827 mutex_unlock(&ext4_li_mtx);
2831 mutex_lock(&ext4_li_info->li_list_mtx);
2832 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2833 mutex_unlock(&ext4_li_info->li_list_mtx);
2834 mutex_unlock(&ext4_li_mtx);
2837 static struct task_struct *ext4_lazyinit_task;
2840 * This is the function where ext4lazyinit thread lives. It walks
2841 * through the request list searching for next scheduled filesystem.
2842 * When such a fs is found, run the lazy initialization request
2843 * (ext4_rn_li_request) and keep track of the time spend in this
2844 * function. Based on that time we compute next schedule time of
2845 * the request. When walking through the list is complete, compute
2846 * next waking time and put itself into sleep.
2848 static int ext4_lazyinit_thread(void *arg)
2850 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2851 struct list_head *pos, *n;
2852 struct ext4_li_request *elr;
2853 unsigned long next_wakeup, cur;
2855 BUG_ON(NULL == eli);
2859 next_wakeup = MAX_JIFFY_OFFSET;
2861 mutex_lock(&eli->li_list_mtx);
2862 if (list_empty(&eli->li_request_list)) {
2863 mutex_unlock(&eli->li_list_mtx);
2867 list_for_each_safe(pos, n, &eli->li_request_list) {
2868 elr = list_entry(pos, struct ext4_li_request,
2871 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2872 if (ext4_run_li_request(elr) != 0) {
2873 /* error, remove the lazy_init job */
2874 ext4_remove_li_request(elr);
2879 if (time_before(elr->lr_next_sched, next_wakeup))
2880 next_wakeup = elr->lr_next_sched;
2882 mutex_unlock(&eli->li_list_mtx);
2884 if (freezing(current))
2888 if ((time_after_eq(cur, next_wakeup)) ||
2889 (MAX_JIFFY_OFFSET == next_wakeup)) {
2894 schedule_timeout_interruptible(next_wakeup - cur);
2896 if (kthread_should_stop()) {
2897 ext4_clear_request_list();
2904 * It looks like the request list is empty, but we need
2905 * to check it under the li_list_mtx lock, to prevent any
2906 * additions into it, and of course we should lock ext4_li_mtx
2907 * to atomically free the list and ext4_li_info, because at
2908 * this point another ext4 filesystem could be registering
2911 mutex_lock(&ext4_li_mtx);
2912 mutex_lock(&eli->li_list_mtx);
2913 if (!list_empty(&eli->li_request_list)) {
2914 mutex_unlock(&eli->li_list_mtx);
2915 mutex_unlock(&ext4_li_mtx);
2918 mutex_unlock(&eli->li_list_mtx);
2919 kfree(ext4_li_info);
2920 ext4_li_info = NULL;
2921 mutex_unlock(&ext4_li_mtx);
2926 static void ext4_clear_request_list(void)
2928 struct list_head *pos, *n;
2929 struct ext4_li_request *elr;
2931 mutex_lock(&ext4_li_info->li_list_mtx);
2932 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2933 elr = list_entry(pos, struct ext4_li_request,
2935 ext4_remove_li_request(elr);
2937 mutex_unlock(&ext4_li_info->li_list_mtx);
2940 static int ext4_run_lazyinit_thread(void)
2942 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2943 ext4_li_info, "ext4lazyinit");
2944 if (IS_ERR(ext4_lazyinit_task)) {
2945 int err = PTR_ERR(ext4_lazyinit_task);
2946 ext4_clear_request_list();
2947 kfree(ext4_li_info);
2948 ext4_li_info = NULL;
2949 printk(KERN_CRIT "EXT4: error %d creating inode table "
2950 "initialization thread\n",
2954 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2959 * Check whether it make sense to run itable init. thread or not.
2960 * If there is at least one uninitialized inode table, return
2961 * corresponding group number, else the loop goes through all
2962 * groups and return total number of groups.
2964 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2966 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2967 struct ext4_group_desc *gdp = NULL;
2969 for (group = 0; group < ngroups; group++) {
2970 gdp = ext4_get_group_desc(sb, group, NULL);
2974 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2981 static int ext4_li_info_new(void)
2983 struct ext4_lazy_init *eli = NULL;
2985 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2989 INIT_LIST_HEAD(&eli->li_request_list);
2990 mutex_init(&eli->li_list_mtx);
2992 eli->li_state |= EXT4_LAZYINIT_QUIT;
2999 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3002 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 struct ext4_li_request *elr;
3006 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3012 elr->lr_next_group = start;
3015 * Randomize first schedule time of the request to
3016 * spread the inode table initialization requests
3019 get_random_bytes(&rnd, sizeof(rnd));
3020 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3021 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3026 static int ext4_register_li_request(struct super_block *sb,
3027 ext4_group_t first_not_zeroed)
3029 struct ext4_sb_info *sbi = EXT4_SB(sb);
3030 struct ext4_li_request *elr;
3031 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3034 if (sbi->s_li_request != NULL) {
3036 * Reset timeout so it can be computed again, because
3037 * s_li_wait_mult might have changed.
3039 sbi->s_li_request->lr_timeout = 0;
3043 if (first_not_zeroed == ngroups ||
3044 (sb->s_flags & MS_RDONLY) ||
3045 !test_opt(sb, INIT_INODE_TABLE))
3048 elr = ext4_li_request_new(sb, first_not_zeroed);
3052 mutex_lock(&ext4_li_mtx);
3054 if (NULL == ext4_li_info) {
3055 ret = ext4_li_info_new();
3060 mutex_lock(&ext4_li_info->li_list_mtx);
3061 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3062 mutex_unlock(&ext4_li_info->li_list_mtx);
3064 sbi->s_li_request = elr;
3066 * set elr to NULL here since it has been inserted to
3067 * the request_list and the removal and free of it is
3068 * handled by ext4_clear_request_list from now on.
3072 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3073 ret = ext4_run_lazyinit_thread();
3078 mutex_unlock(&ext4_li_mtx);
3085 * We do not need to lock anything since this is called on
3088 static void ext4_destroy_lazyinit_thread(void)
3091 * If thread exited earlier
3092 * there's nothing to be done.
3094 if (!ext4_li_info || !ext4_lazyinit_task)
3097 kthread_stop(ext4_lazyinit_task);
3100 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3101 __releases(kernel_lock)
3102 __acquires(kernel_lock)
3104 char *orig_data = kstrdup(data, GFP_KERNEL);
3105 struct buffer_head *bh;
3106 struct ext4_super_block *es = NULL;
3107 struct ext4_sb_info *sbi;
3109 ext4_fsblk_t sb_block = get_sb_block(&data);
3110 ext4_fsblk_t logical_sb_block;
3111 unsigned long offset = 0;
3112 unsigned long journal_devnum = 0;
3113 unsigned long def_mount_opts;
3118 int blocksize, clustersize;
3119 unsigned int db_count;
3121 int needs_recovery, has_huge_files, has_bigalloc;
3124 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3125 ext4_group_t first_not_zeroed;
3127 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3131 sbi->s_blockgroup_lock =
3132 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3133 if (!sbi->s_blockgroup_lock) {
3137 sb->s_fs_info = sbi;
3138 sbi->s_mount_opt = 0;
3139 sbi->s_resuid = EXT4_DEF_RESUID;
3140 sbi->s_resgid = EXT4_DEF_RESGID;
3141 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3142 sbi->s_sb_block = sb_block;
3143 if (sb->s_bdev->bd_part)
3144 sbi->s_sectors_written_start =
3145 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3147 /* Cleanup superblock name */
3148 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3152 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3154 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3159 * The ext4 superblock will not be buffer aligned for other than 1kB
3160 * block sizes. We need to calculate the offset from buffer start.
3162 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3163 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3164 offset = do_div(logical_sb_block, blocksize);
3166 logical_sb_block = sb_block;
3169 if (!(bh = sb_bread(sb, logical_sb_block))) {
3170 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3174 * Note: s_es must be initialized as soon as possible because
3175 * some ext4 macro-instructions depend on its value
3177 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3179 sb->s_magic = le16_to_cpu(es->s_magic);
3180 if (sb->s_magic != EXT4_SUPER_MAGIC)
3182 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3184 /* Set defaults before we parse the mount options */
3185 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3186 set_opt(sb, INIT_INODE_TABLE);
3187 if (def_mount_opts & EXT4_DEFM_DEBUG)
3189 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3190 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3194 if (def_mount_opts & EXT4_DEFM_UID16)
3195 set_opt(sb, NO_UID32);
3196 /* xattr user namespace & acls are now defaulted on */
3197 #ifdef CONFIG_EXT4_FS_XATTR
3198 set_opt(sb, XATTR_USER);
3200 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3201 set_opt(sb, POSIX_ACL);
3203 set_opt(sb, MBLK_IO_SUBMIT);
3204 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3205 set_opt(sb, JOURNAL_DATA);
3206 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3207 set_opt(sb, ORDERED_DATA);
3208 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3209 set_opt(sb, WRITEBACK_DATA);
3211 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3212 set_opt(sb, ERRORS_PANIC);
3213 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3214 set_opt(sb, ERRORS_CONT);
3216 set_opt(sb, ERRORS_RO);
3217 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3218 set_opt(sb, BLOCK_VALIDITY);
3219 if (def_mount_opts & EXT4_DEFM_DISCARD)
3220 set_opt(sb, DISCARD);
3222 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3223 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3224 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3225 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3226 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3228 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3229 set_opt(sb, BARRIER);
3232 * enable delayed allocation by default
3233 * Use -o nodelalloc to turn it off
3235 if (!IS_EXT3_SB(sb) &&
3236 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3237 set_opt(sb, DELALLOC);
3240 * set default s_li_wait_mult for lazyinit, for the case there is
3241 * no mount option specified.
3243 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3245 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3246 &journal_devnum, &journal_ioprio, NULL, 0)) {
3247 ext4_msg(sb, KERN_WARNING,
3248 "failed to parse options in superblock: %s",
3249 sbi->s_es->s_mount_opts);
3251 if (!parse_options((char *) data, sb, &journal_devnum,
3252 &journal_ioprio, NULL, 0))
3255 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3256 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3257 "with data=journal disables delayed "
3258 "allocation and O_DIRECT support!\n");
3259 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3260 ext4_msg(sb, KERN_ERR, "can't mount with "
3261 "both data=journal and delalloc");
3264 if (test_opt(sb, DIOREAD_NOLOCK)) {
3265 ext4_msg(sb, KERN_ERR, "can't mount with "
3266 "both data=journal and delalloc");
3269 if (test_opt(sb, DELALLOC))
3270 clear_opt(sb, DELALLOC);
3273 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3274 if (test_opt(sb, DIOREAD_NOLOCK)) {
3275 if (blocksize < PAGE_SIZE) {
3276 ext4_msg(sb, KERN_ERR, "can't mount with "
3277 "dioread_nolock if block size != PAGE_SIZE");
3282 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3283 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3285 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3286 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3287 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3288 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3289 ext4_msg(sb, KERN_WARNING,
3290 "feature flags set on rev 0 fs, "
3291 "running e2fsck is recommended");
3293 if (IS_EXT2_SB(sb)) {
3294 if (ext2_feature_set_ok(sb))
3295 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3296 "using the ext4 subsystem");
3298 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3299 "to feature incompatibilities");
3304 if (IS_EXT3_SB(sb)) {
3305 if (ext3_feature_set_ok(sb))
3306 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3307 "using the ext4 subsystem");
3309 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3310 "to feature incompatibilities");
3316 * Check feature flags regardless of the revision level, since we
3317 * previously didn't change the revision level when setting the flags,
3318 * so there is a chance incompat flags are set on a rev 0 filesystem.
3320 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3323 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3324 blocksize > EXT4_MAX_BLOCK_SIZE) {
3325 ext4_msg(sb, KERN_ERR,
3326 "Unsupported filesystem blocksize %d", blocksize);
3330 if (sb->s_blocksize != blocksize) {
3331 /* Validate the filesystem blocksize */
3332 if (!sb_set_blocksize(sb, blocksize)) {
3333 ext4_msg(sb, KERN_ERR, "bad block size %d",
3339 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3340 offset = do_div(logical_sb_block, blocksize);
3341 bh = sb_bread(sb, logical_sb_block);
3343 ext4_msg(sb, KERN_ERR,
3344 "Can't read superblock on 2nd try");
3347 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3349 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3350 ext4_msg(sb, KERN_ERR,
3351 "Magic mismatch, very weird!");
3356 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3357 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3358 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3360 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3362 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3363 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3364 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3366 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3367 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3368 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3369 (!is_power_of_2(sbi->s_inode_size)) ||
3370 (sbi->s_inode_size > blocksize)) {
3371 ext4_msg(sb, KERN_ERR,
3372 "unsupported inode size: %d",
3376 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3377 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3380 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3381 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3382 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3383 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3384 !is_power_of_2(sbi->s_desc_size)) {
3385 ext4_msg(sb, KERN_ERR,
3386 "unsupported descriptor size %lu",
3391 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3393 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3394 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3395 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3398 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3399 if (sbi->s_inodes_per_block == 0)
3401 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3402 sbi->s_inodes_per_block;
3403 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3405 sbi->s_mount_state = le16_to_cpu(es->s_state);
3406 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3407 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3409 for (i = 0; i < 4; i++)
3410 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3411 sbi->s_def_hash_version = es->s_def_hash_version;
3412 i = le32_to_cpu(es->s_flags);
3413 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3414 sbi->s_hash_unsigned = 3;
3415 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3416 #ifdef __CHAR_UNSIGNED__
3417 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3418 sbi->s_hash_unsigned = 3;
3420 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3425 /* Handle clustersize */
3426 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3427 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3428 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3430 if (clustersize < blocksize) {
3431 ext4_msg(sb, KERN_ERR,
3432 "cluster size (%d) smaller than "
3433 "block size (%d)", clustersize, blocksize);
3436 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3437 le32_to_cpu(es->s_log_block_size);
3438 sbi->s_clusters_per_group =
3439 le32_to_cpu(es->s_clusters_per_group);
3440 if (sbi->s_clusters_per_group > blocksize * 8) {
3441 ext4_msg(sb, KERN_ERR,
3442 "#clusters per group too big: %lu",
3443 sbi->s_clusters_per_group);
3446 if (sbi->s_blocks_per_group !=
3447 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3448 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3449 "clusters per group (%lu) inconsistent",
3450 sbi->s_blocks_per_group,
3451 sbi->s_clusters_per_group);
3455 if (clustersize != blocksize) {
3456 ext4_warning(sb, "fragment/cluster size (%d) != "
3457 "block size (%d)", clustersize,
3459 clustersize = blocksize;
3461 if (sbi->s_blocks_per_group > blocksize * 8) {
3462 ext4_msg(sb, KERN_ERR,
3463 "#blocks per group too big: %lu",
3464 sbi->s_blocks_per_group);
3467 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3468 sbi->s_cluster_bits = 0;
3470 sbi->s_cluster_ratio = clustersize / blocksize;
3472 if (sbi->s_inodes_per_group > blocksize * 8) {
3473 ext4_msg(sb, KERN_ERR,
3474 "#inodes per group too big: %lu",
3475 sbi->s_inodes_per_group);
3480 * Test whether we have more sectors than will fit in sector_t,
3481 * and whether the max offset is addressable by the page cache.
3483 err = generic_check_addressable(sb->s_blocksize_bits,
3484 ext4_blocks_count(es));
3486 ext4_msg(sb, KERN_ERR, "filesystem"
3487 " too large to mount safely on this system");
3488 if (sizeof(sector_t) < 8)
3489 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3494 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3497 /* check blocks count against device size */
3498 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3499 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3500 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3501 "exceeds size of device (%llu blocks)",
3502 ext4_blocks_count(es), blocks_count);
3507 * It makes no sense for the first data block to be beyond the end
3508 * of the filesystem.
3510 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3511 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3512 "block %u is beyond end of filesystem (%llu)",
3513 le32_to_cpu(es->s_first_data_block),
3514 ext4_blocks_count(es));
3517 blocks_count = (ext4_blocks_count(es) -
3518 le32_to_cpu(es->s_first_data_block) +
3519 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3520 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3521 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3522 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3523 "(block count %llu, first data block %u, "
3524 "blocks per group %lu)", sbi->s_groups_count,
3525 ext4_blocks_count(es),
3526 le32_to_cpu(es->s_first_data_block),
3527 EXT4_BLOCKS_PER_GROUP(sb));
3530 sbi->s_groups_count = blocks_count;
3531 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3532 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3533 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3534 EXT4_DESC_PER_BLOCK(sb);
3535 sbi->s_group_desc = ext4_kvmalloc(db_count *
3536 sizeof(struct buffer_head *),
3538 if (sbi->s_group_desc == NULL) {
3539 ext4_msg(sb, KERN_ERR, "not enough memory");
3543 #ifdef CONFIG_PROC_FS
3545 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3548 bgl_lock_init(sbi->s_blockgroup_lock);
3550 for (i = 0; i < db_count; i++) {
3551 block = descriptor_loc(sb, logical_sb_block, i);
3552 sbi->s_group_desc[i] = sb_bread(sb, block);
3553 if (!sbi->s_group_desc[i]) {
3554 ext4_msg(sb, KERN_ERR,
3555 "can't read group descriptor %d", i);
3560 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3561 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3564 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3565 if (!ext4_fill_flex_info(sb)) {
3566 ext4_msg(sb, KERN_ERR,
3567 "unable to initialize "
3568 "flex_bg meta info!");
3572 sbi->s_gdb_count = db_count;
3573 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3574 spin_lock_init(&sbi->s_next_gen_lock);
3576 init_timer(&sbi->s_err_report);
3577 sbi->s_err_report.function = print_daily_error_info;
3578 sbi->s_err_report.data = (unsigned long) sb;
3580 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3581 ext4_count_free_clusters(sb));
3583 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3584 ext4_count_free_inodes(sb));
3587 err = percpu_counter_init(&sbi->s_dirs_counter,
3588 ext4_count_dirs(sb));
3591 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3594 ext4_msg(sb, KERN_ERR, "insufficient memory");
3598 sbi->s_stripe = ext4_get_stripe_size(sbi);
3599 sbi->s_max_writeback_mb_bump = 128;
3602 * set up enough so that it can read an inode
3604 if (!test_opt(sb, NOLOAD) &&
3605 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3606 sb->s_op = &ext4_sops;
3608 sb->s_op = &ext4_nojournal_sops;
3609 sb->s_export_op = &ext4_export_ops;
3610 sb->s_xattr = ext4_xattr_handlers;
3612 sb->s_qcop = &ext4_qctl_operations;
3613 sb->dq_op = &ext4_quota_operations;
3615 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3617 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3618 mutex_init(&sbi->s_orphan_lock);
3619 sbi->s_resize_flags = 0;
3623 needs_recovery = (es->s_last_orphan != 0 ||
3624 EXT4_HAS_INCOMPAT_FEATURE(sb,
3625 EXT4_FEATURE_INCOMPAT_RECOVER));
3627 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3628 !(sb->s_flags & MS_RDONLY))
3629 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3633 * The first inode we look at is the journal inode. Don't try
3634 * root first: it may be modified in the journal!
3636 if (!test_opt(sb, NOLOAD) &&
3637 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3638 if (ext4_load_journal(sb, es, journal_devnum))
3640 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3641 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3642 ext4_msg(sb, KERN_ERR, "required journal recovery "
3643 "suppressed and not mounted read-only");
3644 goto failed_mount_wq;
3646 clear_opt(sb, DATA_FLAGS);
3647 sbi->s_journal = NULL;
3652 if (ext4_blocks_count(es) > 0xffffffffULL &&
3653 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3654 JBD2_FEATURE_INCOMPAT_64BIT)) {
3655 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3656 goto failed_mount_wq;
3659 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3660 jbd2_journal_set_features(sbi->s_journal,
3661 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3662 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3663 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3664 jbd2_journal_set_features(sbi->s_journal,
3665 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3666 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3667 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3669 jbd2_journal_clear_features(sbi->s_journal,
3670 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3671 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3674 /* We have now updated the journal if required, so we can
3675 * validate the data journaling mode. */
3676 switch (test_opt(sb, DATA_FLAGS)) {
3678 /* No mode set, assume a default based on the journal
3679 * capabilities: ORDERED_DATA if the journal can
3680 * cope, else JOURNAL_DATA
3682 if (jbd2_journal_check_available_features
3683 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3684 set_opt(sb, ORDERED_DATA);
3686 set_opt(sb, JOURNAL_DATA);
3689 case EXT4_MOUNT_ORDERED_DATA:
3690 case EXT4_MOUNT_WRITEBACK_DATA:
3691 if (!jbd2_journal_check_available_features
3692 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3693 ext4_msg(sb, KERN_ERR, "Journal does not support "
3694 "requested data journaling mode");
3695 goto failed_mount_wq;
3700 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3703 * The journal may have updated the bg summary counts, so we
3704 * need to update the global counters.
3706 percpu_counter_set(&sbi->s_freeclusters_counter,
3707 ext4_count_free_clusters(sb));
3708 percpu_counter_set(&sbi->s_freeinodes_counter,
3709 ext4_count_free_inodes(sb));
3710 percpu_counter_set(&sbi->s_dirs_counter,
3711 ext4_count_dirs(sb));
3712 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3716 * The maximum number of concurrent works can be high and
3717 * concurrency isn't really necessary. Limit it to 1.
3719 EXT4_SB(sb)->dio_unwritten_wq =
3720 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3721 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3722 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3723 goto failed_mount_wq;
3727 * The jbd2_journal_load will have done any necessary log recovery,
3728 * so we can safely mount the rest of the filesystem now.
3731 root = ext4_iget(sb, EXT4_ROOT_INO);
3733 ext4_msg(sb, KERN_ERR, "get root inode failed");
3734 ret = PTR_ERR(root);
3738 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3739 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3742 sb->s_root = d_alloc_root(root);
3744 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3749 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3751 /* determine the minimum size of new large inodes, if present */
3752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3753 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3754 EXT4_GOOD_OLD_INODE_SIZE;
3755 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3756 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3757 if (sbi->s_want_extra_isize <
3758 le16_to_cpu(es->s_want_extra_isize))
3759 sbi->s_want_extra_isize =
3760 le16_to_cpu(es->s_want_extra_isize);
3761 if (sbi->s_want_extra_isize <
3762 le16_to_cpu(es->s_min_extra_isize))
3763 sbi->s_want_extra_isize =
3764 le16_to_cpu(es->s_min_extra_isize);
3767 /* Check if enough inode space is available */
3768 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3769 sbi->s_inode_size) {
3770 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3771 EXT4_GOOD_OLD_INODE_SIZE;
3772 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3776 err = ext4_setup_system_zone(sb);
3778 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3784 err = ext4_mb_init(sb, needs_recovery);
3786 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3791 err = ext4_register_li_request(sb, first_not_zeroed);
3795 sbi->s_kobj.kset = ext4_kset;
3796 init_completion(&sbi->s_kobj_unregister);
3797 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3800 ext4_mb_release(sb);
3801 ext4_ext_release(sb);
3805 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3806 ext4_orphan_cleanup(sb, es);
3807 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3808 if (needs_recovery) {
3809 ext4_msg(sb, KERN_INFO, "recovery complete");
3810 ext4_mark_recovery_complete(sb, es);
3812 if (EXT4_SB(sb)->s_journal) {
3813 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3814 descr = " journalled data mode";
3815 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3816 descr = " ordered data mode";
3818 descr = " writeback data mode";
3820 descr = "out journal";
3822 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3823 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3824 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3826 if (es->s_error_count)
3827 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3834 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3840 ext4_msg(sb, KERN_ERR, "mount failed");
3841 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3843 ext4_release_system_zone(sb);
3844 if (sbi->s_journal) {
3845 jbd2_journal_destroy(sbi->s_journal);
3846 sbi->s_journal = NULL;
3849 del_timer(&sbi->s_err_report);
3850 if (sbi->s_flex_groups)
3851 ext4_kvfree(sbi->s_flex_groups);
3852 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3853 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3854 percpu_counter_destroy(&sbi->s_dirs_counter);
3855 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3857 kthread_stop(sbi->s_mmp_tsk);
3859 for (i = 0; i < db_count; i++)
3860 brelse(sbi->s_group_desc[i]);
3861 ext4_kvfree(sbi->s_group_desc);
3864 remove_proc_entry(sb->s_id, ext4_proc_root);
3867 for (i = 0; i < MAXQUOTAS; i++)
3868 kfree(sbi->s_qf_names[i]);
3870 ext4_blkdev_remove(sbi);
3873 sb->s_fs_info = NULL;
3874 kfree(sbi->s_blockgroup_lock);
3882 * Setup any per-fs journal parameters now. We'll do this both on
3883 * initial mount, once the journal has been initialised but before we've
3884 * done any recovery; and again on any subsequent remount.
3886 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3888 struct ext4_sb_info *sbi = EXT4_SB(sb);
3890 journal->j_commit_interval = sbi->s_commit_interval;
3891 journal->j_min_batch_time = sbi->s_min_batch_time;
3892 journal->j_max_batch_time = sbi->s_max_batch_time;
3894 write_lock(&journal->j_state_lock);
3895 if (test_opt(sb, BARRIER))
3896 journal->j_flags |= JBD2_BARRIER;
3898 journal->j_flags &= ~JBD2_BARRIER;
3899 if (test_opt(sb, DATA_ERR_ABORT))
3900 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3902 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3903 write_unlock(&journal->j_state_lock);
3906 static journal_t *ext4_get_journal(struct super_block *sb,
3907 unsigned int journal_inum)
3909 struct inode *journal_inode;
3912 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3914 /* First, test for the existence of a valid inode on disk. Bad
3915 * things happen if we iget() an unused inode, as the subsequent
3916 * iput() will try to delete it. */
3918 journal_inode = ext4_iget(sb, journal_inum);
3919 if (IS_ERR(journal_inode)) {
3920 ext4_msg(sb, KERN_ERR, "no journal found");
3923 if (!journal_inode->i_nlink) {
3924 make_bad_inode(journal_inode);
3925 iput(journal_inode);
3926 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3930 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3931 journal_inode, journal_inode->i_size);
3932 if (!S_ISREG(journal_inode->i_mode)) {
3933 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3934 iput(journal_inode);
3938 journal = jbd2_journal_init_inode(journal_inode);
3940 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3941 iput(journal_inode);
3944 journal->j_private = sb;
3945 ext4_init_journal_params(sb, journal);
3949 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3952 struct buffer_head *bh;
3956 int hblock, blocksize;
3957 ext4_fsblk_t sb_block;
3958 unsigned long offset;
3959 struct ext4_super_block *es;
3960 struct block_device *bdev;
3962 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3964 bdev = ext4_blkdev_get(j_dev, sb);
3968 blocksize = sb->s_blocksize;
3969 hblock = bdev_logical_block_size(bdev);
3970 if (blocksize < hblock) {
3971 ext4_msg(sb, KERN_ERR,
3972 "blocksize too small for journal device");
3976 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3977 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3978 set_blocksize(bdev, blocksize);
3979 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3980 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3981 "external journal");
3985 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3986 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3987 !(le32_to_cpu(es->s_feature_incompat) &
3988 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3989 ext4_msg(sb, KERN_ERR, "external journal has "
3995 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3996 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4001 len = ext4_blocks_count(es);
4002 start = sb_block + 1;
4003 brelse(bh); /* we're done with the superblock */
4005 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4006 start, len, blocksize);
4008 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4011 journal->j_private = sb;
4012 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4013 wait_on_buffer(journal->j_sb_buffer);
4014 if (!buffer_uptodate(journal->j_sb_buffer)) {
4015 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4018 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4019 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4020 "user (unsupported) - %d",
4021 be32_to_cpu(journal->j_superblock->s_nr_users));
4024 EXT4_SB(sb)->journal_bdev = bdev;
4025 ext4_init_journal_params(sb, journal);
4029 jbd2_journal_destroy(journal);
4031 ext4_blkdev_put(bdev);
4035 static int ext4_load_journal(struct super_block *sb,
4036 struct ext4_super_block *es,
4037 unsigned long journal_devnum)
4040 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4043 int really_read_only;
4045 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4047 if (journal_devnum &&
4048 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4049 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4050 "numbers have changed");
4051 journal_dev = new_decode_dev(journal_devnum);
4053 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4055 really_read_only = bdev_read_only(sb->s_bdev);
4058 * Are we loading a blank journal or performing recovery after a
4059 * crash? For recovery, we need to check in advance whether we
4060 * can get read-write access to the device.
4062 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4063 if (sb->s_flags & MS_RDONLY) {
4064 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4065 "required on readonly filesystem");
4066 if (really_read_only) {
4067 ext4_msg(sb, KERN_ERR, "write access "
4068 "unavailable, cannot proceed");
4071 ext4_msg(sb, KERN_INFO, "write access will "
4072 "be enabled during recovery");
4076 if (journal_inum && journal_dev) {
4077 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4078 "and inode journals!");
4083 if (!(journal = ext4_get_journal(sb, journal_inum)))
4086 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4090 if (!(journal->j_flags & JBD2_BARRIER))
4091 ext4_msg(sb, KERN_INFO, "barriers disabled");
4093 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4094 err = jbd2_journal_update_format(journal);
4096 ext4_msg(sb, KERN_ERR, "error updating journal");
4097 jbd2_journal_destroy(journal);
4102 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4103 err = jbd2_journal_wipe(journal, !really_read_only);
4105 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4107 memcpy(save, ((char *) es) +
4108 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4109 err = jbd2_journal_load(journal);
4111 memcpy(((char *) es) + EXT4_S_ERR_START,
4112 save, EXT4_S_ERR_LEN);
4117 ext4_msg(sb, KERN_ERR, "error loading journal");
4118 jbd2_journal_destroy(journal);
4122 EXT4_SB(sb)->s_journal = journal;
4123 ext4_clear_journal_err(sb, es);
4125 if (!really_read_only && journal_devnum &&
4126 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4127 es->s_journal_dev = cpu_to_le32(journal_devnum);
4129 /* Make sure we flush the recovery flag to disk. */
4130 ext4_commit_super(sb, 1);
4136 static int ext4_commit_super(struct super_block *sb, int sync)
4138 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4139 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4142 if (!sbh || block_device_ejected(sb))
4144 if (buffer_write_io_error(sbh)) {
4146 * Oh, dear. A previous attempt to write the
4147 * superblock failed. This could happen because the
4148 * USB device was yanked out. Or it could happen to
4149 * be a transient write error and maybe the block will
4150 * be remapped. Nothing we can do but to retry the
4151 * write and hope for the best.
4153 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4154 "superblock detected");
4155 clear_buffer_write_io_error(sbh);
4156 set_buffer_uptodate(sbh);
4159 * If the file system is mounted read-only, don't update the
4160 * superblock write time. This avoids updating the superblock
4161 * write time when we are mounting the root file system
4162 * read/only but we need to replay the journal; at that point,
4163 * for people who are east of GMT and who make their clock
4164 * tick in localtime for Windows bug-for-bug compatibility,
4165 * the clock is set in the future, and this will cause e2fsck
4166 * to complain and force a full file system check.
4168 if (!(sb->s_flags & MS_RDONLY))
4169 es->s_wtime = cpu_to_le32(get_seconds());
4170 if (sb->s_bdev->bd_part)
4171 es->s_kbytes_written =
4172 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4176 es->s_kbytes_written =
4177 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4178 ext4_free_blocks_count_set(es,
4179 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4180 &EXT4_SB(sb)->s_freeclusters_counter)));
4181 es->s_free_inodes_count =
4182 cpu_to_le32(percpu_counter_sum_positive(
4183 &EXT4_SB(sb)->s_freeinodes_counter));
4185 BUFFER_TRACE(sbh, "marking dirty");
4186 mark_buffer_dirty(sbh);
4188 error = sync_dirty_buffer(sbh);
4192 error = buffer_write_io_error(sbh);
4194 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4196 clear_buffer_write_io_error(sbh);
4197 set_buffer_uptodate(sbh);
4204 * Have we just finished recovery? If so, and if we are mounting (or
4205 * remounting) the filesystem readonly, then we will end up with a
4206 * consistent fs on disk. Record that fact.
4208 static void ext4_mark_recovery_complete(struct super_block *sb,
4209 struct ext4_super_block *es)
4211 journal_t *journal = EXT4_SB(sb)->s_journal;
4213 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4214 BUG_ON(journal != NULL);
4217 jbd2_journal_lock_updates(journal);
4218 if (jbd2_journal_flush(journal) < 0)
4221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4222 sb->s_flags & MS_RDONLY) {
4223 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4224 ext4_commit_super(sb, 1);
4228 jbd2_journal_unlock_updates(journal);
4232 * If we are mounting (or read-write remounting) a filesystem whose journal
4233 * has recorded an error from a previous lifetime, move that error to the
4234 * main filesystem now.
4236 static void ext4_clear_journal_err(struct super_block *sb,
4237 struct ext4_super_block *es)
4243 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4245 journal = EXT4_SB(sb)->s_journal;
4248 * Now check for any error status which may have been recorded in the
4249 * journal by a prior ext4_error() or ext4_abort()
4252 j_errno = jbd2_journal_errno(journal);
4256 errstr = ext4_decode_error(sb, j_errno, nbuf);
4257 ext4_warning(sb, "Filesystem error recorded "
4258 "from previous mount: %s", errstr);
4259 ext4_warning(sb, "Marking fs in need of filesystem check.");
4261 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4262 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4263 ext4_commit_super(sb, 1);
4265 jbd2_journal_clear_err(journal);
4270 * Force the running and committing transactions to commit,
4271 * and wait on the commit.
4273 int ext4_force_commit(struct super_block *sb)
4278 if (sb->s_flags & MS_RDONLY)
4281 journal = EXT4_SB(sb)->s_journal;
4283 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4284 ret = ext4_journal_force_commit(journal);
4290 static void ext4_write_super(struct super_block *sb)
4293 ext4_commit_super(sb, 1);
4297 static int ext4_sync_fs(struct super_block *sb, int wait)
4301 struct ext4_sb_info *sbi = EXT4_SB(sb);
4303 trace_ext4_sync_fs(sb, wait);
4304 flush_workqueue(sbi->dio_unwritten_wq);
4305 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4307 jbd2_log_wait_commit(sbi->s_journal, target);
4313 * LVM calls this function before a (read-only) snapshot is created. This
4314 * gives us a chance to flush the journal completely and mark the fs clean.
4316 * Note that only this function cannot bring a filesystem to be in a clean
4317 * state independently, because ext4 prevents a new handle from being started
4318 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4321 static int ext4_freeze(struct super_block *sb)
4326 if (sb->s_flags & MS_RDONLY)
4329 journal = EXT4_SB(sb)->s_journal;
4331 /* Now we set up the journal barrier. */
4332 jbd2_journal_lock_updates(journal);
4335 * Don't clear the needs_recovery flag if we failed to flush
4338 error = jbd2_journal_flush(journal);
4342 /* Journal blocked and flushed, clear needs_recovery flag. */
4343 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4344 error = ext4_commit_super(sb, 1);
4346 /* we rely on s_frozen to stop further updates */
4347 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4352 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4353 * flag here, even though the filesystem is not technically dirty yet.
4355 static int ext4_unfreeze(struct super_block *sb)
4357 if (sb->s_flags & MS_RDONLY)
4361 /* Reset the needs_recovery flag before the fs is unlocked. */
4362 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4363 ext4_commit_super(sb, 1);
4369 * Structure to save mount options for ext4_remount's benefit
4371 struct ext4_mount_options {
4372 unsigned long s_mount_opt;
4373 unsigned long s_mount_opt2;
4376 unsigned long s_commit_interval;
4377 u32 s_min_batch_time, s_max_batch_time;
4380 char *s_qf_names[MAXQUOTAS];
4384 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4386 struct ext4_super_block *es;
4387 struct ext4_sb_info *sbi = EXT4_SB(sb);
4388 ext4_fsblk_t n_blocks_count = 0;
4389 unsigned long old_sb_flags;
4390 struct ext4_mount_options old_opts;
4391 int enable_quota = 0;
4393 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4398 char *orig_data = kstrdup(data, GFP_KERNEL);
4400 /* Store the original options */
4402 old_sb_flags = sb->s_flags;
4403 old_opts.s_mount_opt = sbi->s_mount_opt;
4404 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4405 old_opts.s_resuid = sbi->s_resuid;
4406 old_opts.s_resgid = sbi->s_resgid;
4407 old_opts.s_commit_interval = sbi->s_commit_interval;
4408 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4409 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4411 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4412 for (i = 0; i < MAXQUOTAS; i++)
4413 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4415 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4416 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4419 * Allow the "check" option to be passed as a remount option.
4421 if (!parse_options(data, sb, NULL, &journal_ioprio,
4422 &n_blocks_count, 1)) {
4427 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4428 ext4_abort(sb, "Abort forced by user");
4430 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4431 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4435 if (sbi->s_journal) {
4436 ext4_init_journal_params(sb, sbi->s_journal);
4437 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4440 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4441 n_blocks_count > ext4_blocks_count(es)) {
4442 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4447 if (*flags & MS_RDONLY) {
4448 err = dquot_suspend(sb, -1);
4453 * First of all, the unconditional stuff we have to do
4454 * to disable replay of the journal when we next remount
4456 sb->s_flags |= MS_RDONLY;
4459 * OK, test if we are remounting a valid rw partition
4460 * readonly, and if so set the rdonly flag and then
4461 * mark the partition as valid again.
4463 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4464 (sbi->s_mount_state & EXT4_VALID_FS))
4465 es->s_state = cpu_to_le16(sbi->s_mount_state);
4468 ext4_mark_recovery_complete(sb, es);
4470 /* Make sure we can mount this feature set readwrite */
4471 if (!ext4_feature_set_ok(sb, 0)) {
4476 * Make sure the group descriptor checksums
4477 * are sane. If they aren't, refuse to remount r/w.
4479 for (g = 0; g < sbi->s_groups_count; g++) {
4480 struct ext4_group_desc *gdp =
4481 ext4_get_group_desc(sb, g, NULL);
4483 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4484 ext4_msg(sb, KERN_ERR,
4485 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4486 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4487 le16_to_cpu(gdp->bg_checksum));
4494 * If we have an unprocessed orphan list hanging
4495 * around from a previously readonly bdev mount,
4496 * require a full umount/remount for now.
4498 if (es->s_last_orphan) {
4499 ext4_msg(sb, KERN_WARNING, "Couldn't "
4500 "remount RDWR because of unprocessed "
4501 "orphan inode list. Please "
4502 "umount/remount instead");
4508 * Mounting a RDONLY partition read-write, so reread
4509 * and store the current valid flag. (It may have
4510 * been changed by e2fsck since we originally mounted
4514 ext4_clear_journal_err(sb, es);
4515 sbi->s_mount_state = le16_to_cpu(es->s_state);
4516 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4518 if (!ext4_setup_super(sb, es, 0))
4519 sb->s_flags &= ~MS_RDONLY;
4520 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4521 EXT4_FEATURE_INCOMPAT_MMP))
4522 if (ext4_multi_mount_protect(sb,
4523 le64_to_cpu(es->s_mmp_block))) {
4532 * Reinitialize lazy itable initialization thread based on
4535 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4536 ext4_unregister_li_request(sb);
4538 ext4_group_t first_not_zeroed;
4539 first_not_zeroed = ext4_has_uninit_itable(sb);
4540 ext4_register_li_request(sb, first_not_zeroed);
4543 ext4_setup_system_zone(sb);
4544 if (sbi->s_journal == NULL)
4545 ext4_commit_super(sb, 1);
4548 /* Release old quota file names */
4549 for (i = 0; i < MAXQUOTAS; i++)
4550 if (old_opts.s_qf_names[i] &&
4551 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4552 kfree(old_opts.s_qf_names[i]);
4556 dquot_resume(sb, -1);
4558 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4563 sb->s_flags = old_sb_flags;
4564 sbi->s_mount_opt = old_opts.s_mount_opt;
4565 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4566 sbi->s_resuid = old_opts.s_resuid;
4567 sbi->s_resgid = old_opts.s_resgid;
4568 sbi->s_commit_interval = old_opts.s_commit_interval;
4569 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4570 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4572 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4573 for (i = 0; i < MAXQUOTAS; i++) {
4574 if (sbi->s_qf_names[i] &&
4575 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4576 kfree(sbi->s_qf_names[i]);
4577 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4586 * Note: calculating the overhead so we can be compatible with
4587 * historical BSD practice is quite difficult in the face of
4588 * clusters/bigalloc. This is because multiple metadata blocks from
4589 * different block group can end up in the same allocation cluster.
4590 * Calculating the exact overhead in the face of clustered allocation
4591 * requires either O(all block bitmaps) in memory or O(number of block
4592 * groups**2) in time. We will still calculate the superblock for
4593 * older file systems --- and if we come across with a bigalloc file
4594 * system with zero in s_overhead_clusters the estimate will be close to
4595 * correct especially for very large cluster sizes --- but for newer
4596 * file systems, it's better to calculate this figure once at mkfs
4597 * time, and store it in the superblock. If the superblock value is
4598 * present (even for non-bigalloc file systems), we will use it.
4600 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4602 struct super_block *sb = dentry->d_sb;
4603 struct ext4_sb_info *sbi = EXT4_SB(sb);
4604 struct ext4_super_block *es = sbi->s_es;
4605 struct ext4_group_desc *gdp;
4609 if (test_opt(sb, MINIX_DF)) {
4610 sbi->s_overhead_last = 0;
4611 } else if (es->s_overhead_clusters) {
4612 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4613 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4614 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4615 ext4_fsblk_t overhead = 0;
4618 * Compute the overhead (FS structures). This is constant
4619 * for a given filesystem unless the number of block groups
4620 * changes so we cache the previous value until it does.
4624 * All of the blocks before first_data_block are
4627 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4630 * Add the overhead found in each block group
4632 for (i = 0; i < ngroups; i++) {
4633 gdp = ext4_get_group_desc(sb, i, NULL);
4634 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4637 sbi->s_overhead_last = overhead;
4639 sbi->s_blocks_last = ext4_blocks_count(es);
4642 buf->f_type = EXT4_SUPER_MAGIC;
4643 buf->f_bsize = sb->s_blocksize;
4644 buf->f_blocks = (ext4_blocks_count(es) -
4645 EXT4_C2B(sbi, sbi->s_overhead_last));
4646 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4647 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4648 /* prevent underflow in case that few free space is available */
4649 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4650 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4651 if (buf->f_bfree < ext4_r_blocks_count(es))
4653 buf->f_files = le32_to_cpu(es->s_inodes_count);
4654 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4655 buf->f_namelen = EXT4_NAME_LEN;
4656 fsid = le64_to_cpup((void *)es->s_uuid) ^
4657 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4658 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4659 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4664 /* Helper function for writing quotas on sync - we need to start transaction
4665 * before quota file is locked for write. Otherwise the are possible deadlocks:
4666 * Process 1 Process 2
4667 * ext4_create() quota_sync()
4668 * jbd2_journal_start() write_dquot()
4669 * dquot_initialize() down(dqio_mutex)
4670 * down(dqio_mutex) jbd2_journal_start()
4676 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4678 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4681 static int ext4_write_dquot(struct dquot *dquot)
4685 struct inode *inode;
4687 inode = dquot_to_inode(dquot);
4688 handle = ext4_journal_start(inode,
4689 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4691 return PTR_ERR(handle);
4692 ret = dquot_commit(dquot);
4693 err = ext4_journal_stop(handle);
4699 static int ext4_acquire_dquot(struct dquot *dquot)
4704 handle = ext4_journal_start(dquot_to_inode(dquot),
4705 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4707 return PTR_ERR(handle);
4708 ret = dquot_acquire(dquot);
4709 err = ext4_journal_stop(handle);
4715 static int ext4_release_dquot(struct dquot *dquot)
4720 handle = ext4_journal_start(dquot_to_inode(dquot),
4721 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4722 if (IS_ERR(handle)) {
4723 /* Release dquot anyway to avoid endless cycle in dqput() */
4724 dquot_release(dquot);
4725 return PTR_ERR(handle);
4727 ret = dquot_release(dquot);
4728 err = ext4_journal_stop(handle);
4734 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4736 /* Are we journaling quotas? */
4737 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4738 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4739 dquot_mark_dquot_dirty(dquot);
4740 return ext4_write_dquot(dquot);
4742 return dquot_mark_dquot_dirty(dquot);
4746 static int ext4_write_info(struct super_block *sb, int type)
4751 /* Data block + inode block */
4752 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4754 return PTR_ERR(handle);
4755 ret = dquot_commit_info(sb, type);
4756 err = ext4_journal_stop(handle);
4763 * Turn on quotas during mount time - we need to find
4764 * the quota file and such...
4766 static int ext4_quota_on_mount(struct super_block *sb, int type)
4768 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4769 EXT4_SB(sb)->s_jquota_fmt, type);
4773 * Standard function to be called on quota_on
4775 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4780 if (!test_opt(sb, QUOTA))
4783 /* Quotafile not on the same filesystem? */
4784 if (path->mnt->mnt_sb != sb)
4786 /* Journaling quota? */
4787 if (EXT4_SB(sb)->s_qf_names[type]) {
4788 /* Quotafile not in fs root? */
4789 if (path->dentry->d_parent != sb->s_root)
4790 ext4_msg(sb, KERN_WARNING,
4791 "Quota file not on filesystem root. "
4792 "Journaled quota will not work");
4796 * When we journal data on quota file, we have to flush journal to see
4797 * all updates to the file when we bypass pagecache...
4799 if (EXT4_SB(sb)->s_journal &&
4800 ext4_should_journal_data(path->dentry->d_inode)) {
4802 * We don't need to lock updates but journal_flush() could
4803 * otherwise be livelocked...
4805 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4806 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4807 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4812 return dquot_quota_on(sb, type, format_id, path);
4815 static int ext4_quota_off(struct super_block *sb, int type)
4817 struct inode *inode = sb_dqopt(sb)->files[type];
4820 /* Force all delayed allocation blocks to be allocated.
4821 * Caller already holds s_umount sem */
4822 if (test_opt(sb, DELALLOC))
4823 sync_filesystem(sb);
4828 /* Update modification times of quota files when userspace can
4829 * start looking at them */
4830 handle = ext4_journal_start(inode, 1);
4833 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4834 ext4_mark_inode_dirty(handle, inode);
4835 ext4_journal_stop(handle);
4838 return dquot_quota_off(sb, type);
4841 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4842 * acquiring the locks... As quota files are never truncated and quota code
4843 * itself serializes the operations (and no one else should touch the files)
4844 * we don't have to be afraid of races */
4845 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4846 size_t len, loff_t off)
4848 struct inode *inode = sb_dqopt(sb)->files[type];
4849 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4851 int offset = off & (sb->s_blocksize - 1);
4854 struct buffer_head *bh;
4855 loff_t i_size = i_size_read(inode);
4859 if (off+len > i_size)
4862 while (toread > 0) {
4863 tocopy = sb->s_blocksize - offset < toread ?
4864 sb->s_blocksize - offset : toread;
4865 bh = ext4_bread(NULL, inode, blk, 0, &err);
4868 if (!bh) /* A hole? */
4869 memset(data, 0, tocopy);
4871 memcpy(data, bh->b_data+offset, tocopy);
4881 /* Write to quotafile (we know the transaction is already started and has
4882 * enough credits) */
4883 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4884 const char *data, size_t len, loff_t off)
4886 struct inode *inode = sb_dqopt(sb)->files[type];
4887 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4889 int offset = off & (sb->s_blocksize - 1);
4890 struct buffer_head *bh;
4891 handle_t *handle = journal_current_handle();
4893 if (EXT4_SB(sb)->s_journal && !handle) {
4894 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4895 " cancelled because transaction is not started",
4896 (unsigned long long)off, (unsigned long long)len);
4900 * Since we account only one data block in transaction credits,
4901 * then it is impossible to cross a block boundary.
4903 if (sb->s_blocksize - offset < len) {
4904 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4905 " cancelled because not block aligned",
4906 (unsigned long long)off, (unsigned long long)len);
4910 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4911 bh = ext4_bread(handle, inode, blk, 1, &err);
4914 err = ext4_journal_get_write_access(handle, bh);
4920 memcpy(bh->b_data+offset, data, len);
4921 flush_dcache_page(bh->b_page);
4923 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4927 mutex_unlock(&inode->i_mutex);
4930 if (inode->i_size < off + len) {
4931 i_size_write(inode, off + len);
4932 EXT4_I(inode)->i_disksize = inode->i_size;
4933 ext4_mark_inode_dirty(handle, inode);
4935 mutex_unlock(&inode->i_mutex);
4941 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4942 const char *dev_name, void *data)
4944 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4947 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4948 static inline void register_as_ext2(void)
4950 int err = register_filesystem(&ext2_fs_type);
4953 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4956 static inline void unregister_as_ext2(void)
4958 unregister_filesystem(&ext2_fs_type);
4961 static inline int ext2_feature_set_ok(struct super_block *sb)
4963 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4965 if (sb->s_flags & MS_RDONLY)
4967 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4971 MODULE_ALIAS("ext2");
4973 static inline void register_as_ext2(void) { }
4974 static inline void unregister_as_ext2(void) { }
4975 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4978 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4979 static inline void register_as_ext3(void)
4981 int err = register_filesystem(&ext3_fs_type);
4984 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4987 static inline void unregister_as_ext3(void)
4989 unregister_filesystem(&ext3_fs_type);
4992 static inline int ext3_feature_set_ok(struct super_block *sb)
4994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4996 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4998 if (sb->s_flags & MS_RDONLY)
5000 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5004 MODULE_ALIAS("ext3");
5006 static inline void register_as_ext3(void) { }
5007 static inline void unregister_as_ext3(void) { }
5008 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5011 static struct file_system_type ext4_fs_type = {
5012 .owner = THIS_MODULE,
5014 .mount = ext4_mount,
5015 .kill_sb = kill_block_super,
5016 .fs_flags = FS_REQUIRES_DEV,
5019 static int __init ext4_init_feat_adverts(void)
5021 struct ext4_features *ef;
5024 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5028 ef->f_kobj.kset = ext4_kset;
5029 init_completion(&ef->f_kobj_unregister);
5030 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5043 static void ext4_exit_feat_adverts(void)
5045 kobject_put(&ext4_feat->f_kobj);
5046 wait_for_completion(&ext4_feat->f_kobj_unregister);
5050 /* Shared across all ext4 file systems */
5051 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5052 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5054 static int __init ext4_init_fs(void)
5058 ext4_check_flag_values();
5060 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5061 mutex_init(&ext4__aio_mutex[i]);
5062 init_waitqueue_head(&ext4__ioend_wq[i]);
5065 err = ext4_init_pageio();
5068 err = ext4_init_system_zone();
5071 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5074 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5075 if (!ext4_proc_root)
5078 err = ext4_init_feat_adverts();
5082 err = ext4_init_mballoc();
5086 err = ext4_init_xattr();
5089 err = init_inodecache();
5094 err = register_filesystem(&ext4_fs_type);
5098 ext4_li_info = NULL;
5099 mutex_init(&ext4_li_mtx);
5102 unregister_as_ext2();
5103 unregister_as_ext3();
5104 destroy_inodecache();
5108 ext4_exit_mballoc();
5110 ext4_exit_feat_adverts();
5112 remove_proc_entry("fs/ext4", NULL);
5114 kset_unregister(ext4_kset);
5116 ext4_exit_system_zone();
5122 static void __exit ext4_exit_fs(void)
5124 ext4_destroy_lazyinit_thread();
5125 unregister_as_ext2();
5126 unregister_as_ext3();
5127 unregister_filesystem(&ext4_fs_type);
5128 destroy_inodecache();
5130 ext4_exit_mballoc();
5131 ext4_exit_feat_adverts();
5132 remove_proc_entry("fs/ext4", NULL);
5133 kset_unregister(ext4_kset);
5134 ext4_exit_system_zone();
5138 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5139 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5140 MODULE_LICENSE("GPL");
5141 module_init(ext4_init_fs)
5142 module_exit(ext4_exit_fs)