2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 void *ext4_kvmalloc(size_t size, gfp_t flags)
118 ret = kmalloc(size, flags);
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
124 void *ext4_kvzalloc(size_t size, gfp_t flags)
128 ret = kzalloc(size, flags);
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
134 void ext4_kvfree(void *ptr)
136 if (is_vmalloc_addr(ptr))
143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
159 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
162 return le32_to_cpu(bg->bg_inode_table_lo) |
163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
167 __u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
175 __u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
183 __u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
191 __u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
199 void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
207 void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
215 void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
223 void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
231 void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
239 void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
247 void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
256 /* Just increment the non-pointer handle value */
257 static handle_t *ext4_get_nojournal(void)
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
265 handle = (handle_t *)ref_cnt;
267 current->journal_info = handle;
272 /* Decrement the non-pointer handle value */
273 static void ext4_put_nojournal(handle_t *handle)
275 unsigned long ref_cnt = (unsigned long)handle;
277 BUG_ON(ref_cnt == 0);
280 handle = (handle_t *)ref_cnt;
282 current->journal_info = handle;
286 * Wrappers for jbd2_journal_start/end.
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
306 journal = EXT4_SB(sb)->s_journal;
307 handle = ext4_journal_current_handle();
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
320 return ext4_get_nojournal();
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
330 return jbd2_journal_start(journal, nblocks);
334 * The only special thing we need to do here is to make sure that all
335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
336 * that sync() will call the filesystem's write_super callback if
339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
341 struct super_block *sb;
345 if (!ext4_handle_valid(handle)) {
346 ext4_put_nojournal(handle);
349 sb = handle->h_transaction->t_journal->j_private;
351 rc = jbd2_journal_stop(handle);
356 __ext4_std_error(sb, where, line, err);
360 void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
367 BUG_ON(!ext4_handle_valid(handle));
370 BUFFER_TRACE(bh, "abort");
375 if (is_handle_aborted(handle))
378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
381 jbd2_journal_abort_handle(handle);
384 static void __save_error_info(struct super_block *sb, const char *func,
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
403 * Start the daily error reporting function if it hasn't been
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
411 static void save_error_info(struct super_block *sb, const char *func,
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
426 static int block_device_ejected(struct super_block *sb)
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
431 return bdi->dev == NULL;
435 /* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
438 * On ext2, we can store the error state of the filesystem in the
439 * superblock. That is not possible on ext4, because we may have other
440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
445 * We'll just use the jbd2_journal_abort() error code to record an error in
446 * the journal instead. On recovery, the journal will complain about
447 * that error until we've noted it down and cleared it.
450 static void ext4_handle_error(struct super_block *sb)
452 if (sb->s_flags & MS_RDONLY)
455 if (!test_opt(sb, ERRORS_CONT)) {
456 journal_t *journal = EXT4_SB(sb)->s_journal;
458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
460 jbd2_journal_abort(journal, -EIO);
462 if (test_opt(sb, ERRORS_RO)) {
463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
464 sb->s_flags |= MS_RDONLY;
466 if (test_opt(sb, ERRORS_PANIC))
467 panic("EXT4-fs (device %s): panic forced after error\n",
471 void __ext4_error(struct super_block *sb, const char *function,
472 unsigned int line, const char *fmt, ...)
474 struct va_format vaf;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
483 save_error_info(sb, function, line);
485 ext4_handle_error(sb);
488 void ext4_error_inode(struct inode *inode, const char *function,
489 unsigned int line, ext4_fsblk_t block,
490 const char *fmt, ...)
493 struct va_format vaf;
494 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
496 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
497 es->s_last_error_block = cpu_to_le64(block);
498 save_error_info(inode->i_sb, function, line);
502 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
503 inode->i_sb->s_id, function, line, inode->i_ino);
505 printk(KERN_CONT "block %llu: ", block);
506 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
509 ext4_handle_error(inode->i_sb);
512 void ext4_error_file(struct file *file, const char *function,
513 unsigned int line, ext4_fsblk_t block,
514 const char *fmt, ...)
517 struct va_format vaf;
518 struct ext4_super_block *es;
519 struct inode *inode = file->f_dentry->d_inode;
520 char pathname[80], *path;
522 es = EXT4_SB(inode->i_sb)->s_es;
523 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
524 save_error_info(inode->i_sb, function, line);
525 path = d_path(&(file->f_path), pathname, sizeof(pathname));
529 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
530 inode->i_sb->s_id, function, line, inode->i_ino);
532 printk(KERN_CONT "block %llu: ", block);
536 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
539 ext4_handle_error(inode->i_sb);
542 static const char *ext4_decode_error(struct super_block *sb, int errno,
549 errstr = "IO failure";
552 errstr = "Out of memory";
555 if (!sb || (EXT4_SB(sb)->s_journal &&
556 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
557 errstr = "Journal has aborted";
559 errstr = "Readonly filesystem";
562 /* If the caller passed in an extra buffer for unknown
563 * errors, textualise them now. Else we just return
566 /* Check for truncated error codes... */
567 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
576 /* __ext4_std_error decodes expected errors from journaling functions
577 * automatically and invokes the appropriate error response. */
579 void __ext4_std_error(struct super_block *sb, const char *function,
580 unsigned int line, int errno)
585 /* Special case: if the error is EROFS, and we're not already
586 * inside a transaction, then there's really no point in logging
588 if (errno == -EROFS && journal_current_handle() == NULL &&
589 (sb->s_flags & MS_RDONLY))
592 errstr = ext4_decode_error(sb, errno, nbuf);
593 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
594 sb->s_id, function, line, errstr);
595 save_error_info(sb, function, line);
597 ext4_handle_error(sb);
601 * ext4_abort is a much stronger failure handler than ext4_error. The
602 * abort function may be used to deal with unrecoverable failures such
603 * as journal IO errors or ENOMEM at a critical moment in log management.
605 * We unconditionally force the filesystem into an ABORT|READONLY state,
606 * unless the error response on the fs has been set to panic in which
607 * case we take the easy way out and panic immediately.
610 void __ext4_abort(struct super_block *sb, const char *function,
611 unsigned int line, const char *fmt, ...)
615 save_error_info(sb, function, line);
617 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
623 if ((sb->s_flags & MS_RDONLY) == 0) {
624 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
625 sb->s_flags |= MS_RDONLY;
626 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
627 if (EXT4_SB(sb)->s_journal)
628 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
629 save_error_info(sb, function, line);
631 if (test_opt(sb, ERRORS_PANIC))
632 panic("EXT4-fs panic from previous error\n");
635 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
637 struct va_format vaf;
643 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
647 void __ext4_warning(struct super_block *sb, const char *function,
648 unsigned int line, const char *fmt, ...)
650 struct va_format vaf;
656 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
657 sb->s_id, function, line, &vaf);
661 void __ext4_grp_locked_error(const char *function, unsigned int line,
662 struct super_block *sb, ext4_group_t grp,
663 unsigned long ino, ext4_fsblk_t block,
664 const char *fmt, ...)
668 struct va_format vaf;
670 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
672 es->s_last_error_ino = cpu_to_le32(ino);
673 es->s_last_error_block = cpu_to_le64(block);
674 __save_error_info(sb, function, line);
680 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
681 sb->s_id, function, line, grp);
683 printk(KERN_CONT "inode %lu: ", ino);
685 printk(KERN_CONT "block %llu:", (unsigned long long) block);
686 printk(KERN_CONT "%pV\n", &vaf);
689 if (test_opt(sb, ERRORS_CONT)) {
690 ext4_commit_super(sb, 0);
694 ext4_unlock_group(sb, grp);
695 ext4_handle_error(sb);
697 * We only get here in the ERRORS_RO case; relocking the group
698 * may be dangerous, but nothing bad will happen since the
699 * filesystem will have already been marked read/only and the
700 * journal has been aborted. We return 1 as a hint to callers
701 * who might what to use the return value from
702 * ext4_grp_locked_error() to distinguish between the
703 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
704 * aggressively from the ext4 function in question, with a
705 * more appropriate error code.
707 ext4_lock_group(sb, grp);
711 void ext4_update_dynamic_rev(struct super_block *sb)
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
715 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
719 "updating to rev %d because of new feature flag, "
720 "running e2fsck is recommended",
723 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
724 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
725 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
726 /* leave es->s_feature_*compat flags alone */
727 /* es->s_uuid will be set by e2fsck if empty */
730 * The rest of the superblock fields should be zero, and if not it
731 * means they are likely already in use, so leave them alone. We
732 * can leave it up to e2fsck to clean up any inconsistencies there.
737 * Open the external journal device
739 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
741 struct block_device *bdev;
742 char b[BDEVNAME_SIZE];
744 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
750 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
751 __bdevname(dev, b), PTR_ERR(bdev));
756 * Release the journal device
758 static int ext4_blkdev_put(struct block_device *bdev)
760 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
763 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
765 struct block_device *bdev;
768 bdev = sbi->journal_bdev;
770 ret = ext4_blkdev_put(bdev);
771 sbi->journal_bdev = NULL;
776 static inline struct inode *orphan_list_entry(struct list_head *l)
778 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
781 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
785 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
786 le32_to_cpu(sbi->s_es->s_last_orphan));
788 printk(KERN_ERR "sb_info orphan list:\n");
789 list_for_each(l, &sbi->s_orphan) {
790 struct inode *inode = orphan_list_entry(l);
792 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
793 inode->i_sb->s_id, inode->i_ino, inode,
794 inode->i_mode, inode->i_nlink,
799 static void ext4_put_super(struct super_block *sb)
801 struct ext4_sb_info *sbi = EXT4_SB(sb);
802 struct ext4_super_block *es = sbi->s_es;
805 ext4_unregister_li_request(sb);
806 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
808 flush_workqueue(sbi->dio_unwritten_wq);
809 destroy_workqueue(sbi->dio_unwritten_wq);
813 ext4_commit_super(sb, 1);
815 if (sbi->s_journal) {
816 err = jbd2_journal_destroy(sbi->s_journal);
817 sbi->s_journal = NULL;
819 ext4_abort(sb, "Couldn't clean up the journal");
822 del_timer_sync(&sbi->s_err_report);
823 ext4_release_system_zone(sb);
825 ext4_ext_release(sb);
826 ext4_xattr_put_super(sb);
828 if (!(sb->s_flags & MS_RDONLY)) {
829 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
830 es->s_state = cpu_to_le16(sbi->s_mount_state);
831 ext4_commit_super(sb, 1);
834 remove_proc_entry(sb->s_id, ext4_proc_root);
836 kobject_del(&sbi->s_kobj);
838 for (i = 0; i < sbi->s_gdb_count; i++)
839 brelse(sbi->s_group_desc[i]);
840 ext4_kvfree(sbi->s_group_desc);
841 ext4_kvfree(sbi->s_flex_groups);
842 percpu_counter_destroy(&sbi->s_freeclusters_counter);
843 percpu_counter_destroy(&sbi->s_freeinodes_counter);
844 percpu_counter_destroy(&sbi->s_dirs_counter);
845 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
848 for (i = 0; i < MAXQUOTAS; i++)
849 kfree(sbi->s_qf_names[i]);
852 /* Debugging code just in case the in-memory inode orphan list
853 * isn't empty. The on-disk one can be non-empty if we've
854 * detected an error and taken the fs readonly, but the
855 * in-memory list had better be clean by this point. */
856 if (!list_empty(&sbi->s_orphan))
857 dump_orphan_list(sb, sbi);
858 J_ASSERT(list_empty(&sbi->s_orphan));
860 invalidate_bdev(sb->s_bdev);
861 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
863 * Invalidate the journal device's buffers. We don't want them
864 * floating about in memory - the physical journal device may
865 * hotswapped, and it breaks the `ro-after' testing code.
867 sync_blockdev(sbi->journal_bdev);
868 invalidate_bdev(sbi->journal_bdev);
869 ext4_blkdev_remove(sbi);
872 kthread_stop(sbi->s_mmp_tsk);
873 sb->s_fs_info = NULL;
875 * Now that we are completely done shutting down the
876 * superblock, we need to actually destroy the kobject.
879 kobject_put(&sbi->s_kobj);
880 wait_for_completion(&sbi->s_kobj_unregister);
881 kfree(sbi->s_blockgroup_lock);
885 static struct kmem_cache *ext4_inode_cachep;
888 * Called inside transaction, so use GFP_NOFS
890 static struct inode *ext4_alloc_inode(struct super_block *sb)
892 struct ext4_inode_info *ei;
894 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
898 ei->vfs_inode.i_version = 1;
899 ei->vfs_inode.i_data.writeback_index = 0;
900 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
901 INIT_LIST_HEAD(&ei->i_prealloc_list);
902 spin_lock_init(&ei->i_prealloc_lock);
903 ei->i_reserved_data_blocks = 0;
904 ei->i_reserved_meta_blocks = 0;
905 ei->i_allocated_meta_blocks = 0;
906 ei->i_da_metadata_calc_len = 0;
907 ei->i_da_metadata_calc_last_lblock = 0;
908 spin_lock_init(&(ei->i_block_reservation_lock));
910 ei->i_reserved_quota = 0;
913 INIT_LIST_HEAD(&ei->i_completed_io_list);
914 spin_lock_init(&ei->i_completed_io_lock);
915 ei->cur_aio_dio = NULL;
917 ei->i_datasync_tid = 0;
918 atomic_set(&ei->i_ioend_count, 0);
919 atomic_set(&ei->i_aiodio_unwritten, 0);
921 return &ei->vfs_inode;
924 static int ext4_drop_inode(struct inode *inode)
926 int drop = generic_drop_inode(inode);
928 trace_ext4_drop_inode(inode, drop);
932 static void ext4_i_callback(struct rcu_head *head)
934 struct inode *inode = container_of(head, struct inode, i_rcu);
935 INIT_LIST_HEAD(&inode->i_dentry);
936 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
939 static void ext4_destroy_inode(struct inode *inode)
941 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
942 ext4_msg(inode->i_sb, KERN_ERR,
943 "Inode %lu (%p): orphan list check failed!",
944 inode->i_ino, EXT4_I(inode));
945 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
946 EXT4_I(inode), sizeof(struct ext4_inode_info),
950 call_rcu(&inode->i_rcu, ext4_i_callback);
953 static void init_once(void *foo)
955 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
957 INIT_LIST_HEAD(&ei->i_orphan);
958 #ifdef CONFIG_EXT4_FS_XATTR
959 init_rwsem(&ei->xattr_sem);
961 init_rwsem(&ei->i_data_sem);
962 inode_init_once(&ei->vfs_inode);
965 static int init_inodecache(void)
967 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
968 sizeof(struct ext4_inode_info),
969 0, (SLAB_RECLAIM_ACCOUNT|
972 if (ext4_inode_cachep == NULL)
977 static void destroy_inodecache(void)
979 kmem_cache_destroy(ext4_inode_cachep);
982 void ext4_clear_inode(struct inode *inode)
984 invalidate_inode_buffers(inode);
985 end_writeback(inode);
987 ext4_discard_preallocations(inode);
988 if (EXT4_I(inode)->jinode) {
989 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
990 EXT4_I(inode)->jinode);
991 jbd2_free_inode(EXT4_I(inode)->jinode);
992 EXT4_I(inode)->jinode = NULL;
996 static inline void ext4_show_quota_options(struct seq_file *seq,
997 struct super_block *sb)
999 #if defined(CONFIG_QUOTA)
1000 struct ext4_sb_info *sbi = EXT4_SB(sb);
1002 if (sbi->s_jquota_fmt) {
1005 switch (sbi->s_jquota_fmt) {
1016 seq_printf(seq, ",jqfmt=%s", fmtname);
1019 if (sbi->s_qf_names[USRQUOTA])
1020 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1022 if (sbi->s_qf_names[GRPQUOTA])
1023 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1025 if (test_opt(sb, USRQUOTA))
1026 seq_puts(seq, ",usrquota");
1028 if (test_opt(sb, GRPQUOTA))
1029 seq_puts(seq, ",grpquota");
1035 * - it's set to a non-default value OR
1036 * - if the per-sb default is different from the global default
1038 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1041 unsigned long def_mount_opts;
1042 struct super_block *sb = vfs->mnt_sb;
1043 struct ext4_sb_info *sbi = EXT4_SB(sb);
1044 struct ext4_super_block *es = sbi->s_es;
1046 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1047 def_errors = le16_to_cpu(es->s_errors);
1049 if (sbi->s_sb_block != 1)
1050 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1051 if (test_opt(sb, MINIX_DF))
1052 seq_puts(seq, ",minixdf");
1053 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1054 seq_puts(seq, ",grpid");
1055 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1056 seq_puts(seq, ",nogrpid");
1057 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1058 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1059 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1061 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1062 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1063 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1065 if (test_opt(sb, ERRORS_RO)) {
1066 if (def_errors == EXT4_ERRORS_PANIC ||
1067 def_errors == EXT4_ERRORS_CONTINUE) {
1068 seq_puts(seq, ",errors=remount-ro");
1071 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1072 seq_puts(seq, ",errors=continue");
1073 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1074 seq_puts(seq, ",errors=panic");
1075 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1076 seq_puts(seq, ",nouid32");
1077 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1078 seq_puts(seq, ",debug");
1079 #ifdef CONFIG_EXT4_FS_XATTR
1080 if (test_opt(sb, XATTR_USER))
1081 seq_puts(seq, ",user_xattr");
1082 if (!test_opt(sb, XATTR_USER))
1083 seq_puts(seq, ",nouser_xattr");
1085 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1086 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1087 seq_puts(seq, ",acl");
1088 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1089 seq_puts(seq, ",noacl");
1091 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1092 seq_printf(seq, ",commit=%u",
1093 (unsigned) (sbi->s_commit_interval / HZ));
1095 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1096 seq_printf(seq, ",min_batch_time=%u",
1097 (unsigned) sbi->s_min_batch_time);
1099 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1100 seq_printf(seq, ",max_batch_time=%u",
1101 (unsigned) sbi->s_max_batch_time);
1105 * We're changing the default of barrier mount option, so
1106 * let's always display its mount state so it's clear what its
1109 seq_puts(seq, ",barrier=");
1110 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1111 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1112 seq_puts(seq, ",journal_async_commit");
1113 else if (test_opt(sb, JOURNAL_CHECKSUM))
1114 seq_puts(seq, ",journal_checksum");
1115 if (test_opt(sb, I_VERSION))
1116 seq_puts(seq, ",i_version");
1117 if (!test_opt(sb, DELALLOC) &&
1118 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1119 seq_puts(seq, ",nodelalloc");
1121 if (!test_opt(sb, MBLK_IO_SUBMIT))
1122 seq_puts(seq, ",nomblk_io_submit");
1124 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1126 * journal mode get enabled in different ways
1127 * So just print the value even if we didn't specify it
1129 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1130 seq_puts(seq, ",data=journal");
1131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1132 seq_puts(seq, ",data=ordered");
1133 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1134 seq_puts(seq, ",data=writeback");
1136 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1137 seq_printf(seq, ",inode_readahead_blks=%u",
1138 sbi->s_inode_readahead_blks);
1140 if (test_opt(sb, DATA_ERR_ABORT))
1141 seq_puts(seq, ",data_err=abort");
1143 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1144 seq_puts(seq, ",noauto_da_alloc");
1146 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1147 seq_puts(seq, ",discard");
1149 if (test_opt(sb, NOLOAD))
1150 seq_puts(seq, ",norecovery");
1152 if (test_opt(sb, DIOREAD_NOLOCK))
1153 seq_puts(seq, ",dioread_nolock");
1155 if (test_opt(sb, BLOCK_VALIDITY) &&
1156 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1157 seq_puts(seq, ",block_validity");
1159 if (!test_opt(sb, INIT_INODE_TABLE))
1160 seq_puts(seq, ",noinit_itable");
1161 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1162 seq_printf(seq, ",init_itable=%u",
1163 (unsigned) sbi->s_li_wait_mult);
1165 ext4_show_quota_options(seq, sb);
1170 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1171 u64 ino, u32 generation)
1173 struct inode *inode;
1175 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1176 return ERR_PTR(-ESTALE);
1177 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1178 return ERR_PTR(-ESTALE);
1180 /* iget isn't really right if the inode is currently unallocated!!
1182 * ext4_read_inode will return a bad_inode if the inode had been
1183 * deleted, so we should be safe.
1185 * Currently we don't know the generation for parent directory, so
1186 * a generation of 0 means "accept any"
1188 inode = ext4_iget(sb, ino);
1190 return ERR_CAST(inode);
1191 if (generation && inode->i_generation != generation) {
1193 return ERR_PTR(-ESTALE);
1199 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1200 int fh_len, int fh_type)
1202 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1203 ext4_nfs_get_inode);
1206 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1207 int fh_len, int fh_type)
1209 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1210 ext4_nfs_get_inode);
1214 * Try to release metadata pages (indirect blocks, directories) which are
1215 * mapped via the block device. Since these pages could have journal heads
1216 * which would prevent try_to_free_buffers() from freeing them, we must use
1217 * jbd2 layer's try_to_free_buffers() function to release them.
1219 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1222 journal_t *journal = EXT4_SB(sb)->s_journal;
1224 WARN_ON(PageChecked(page));
1225 if (!page_has_buffers(page))
1228 return jbd2_journal_try_to_free_buffers(journal, page,
1229 wait & ~__GFP_WAIT);
1230 return try_to_free_buffers(page);
1234 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1235 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1237 static int ext4_write_dquot(struct dquot *dquot);
1238 static int ext4_acquire_dquot(struct dquot *dquot);
1239 static int ext4_release_dquot(struct dquot *dquot);
1240 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1241 static int ext4_write_info(struct super_block *sb, int type);
1242 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1244 static int ext4_quota_off(struct super_block *sb, int type);
1245 static int ext4_quota_on_mount(struct super_block *sb, int type);
1246 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1247 size_t len, loff_t off);
1248 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1249 const char *data, size_t len, loff_t off);
1251 static const struct dquot_operations ext4_quota_operations = {
1252 .get_reserved_space = ext4_get_reserved_space,
1253 .write_dquot = ext4_write_dquot,
1254 .acquire_dquot = ext4_acquire_dquot,
1255 .release_dquot = ext4_release_dquot,
1256 .mark_dirty = ext4_mark_dquot_dirty,
1257 .write_info = ext4_write_info,
1258 .alloc_dquot = dquot_alloc,
1259 .destroy_dquot = dquot_destroy,
1262 static const struct quotactl_ops ext4_qctl_operations = {
1263 .quota_on = ext4_quota_on,
1264 .quota_off = ext4_quota_off,
1265 .quota_sync = dquot_quota_sync,
1266 .get_info = dquot_get_dqinfo,
1267 .set_info = dquot_set_dqinfo,
1268 .get_dqblk = dquot_get_dqblk,
1269 .set_dqblk = dquot_set_dqblk
1273 static const struct super_operations ext4_sops = {
1274 .alloc_inode = ext4_alloc_inode,
1275 .destroy_inode = ext4_destroy_inode,
1276 .write_inode = ext4_write_inode,
1277 .dirty_inode = ext4_dirty_inode,
1278 .drop_inode = ext4_drop_inode,
1279 .evict_inode = ext4_evict_inode,
1280 .put_super = ext4_put_super,
1281 .sync_fs = ext4_sync_fs,
1282 .freeze_fs = ext4_freeze,
1283 .unfreeze_fs = ext4_unfreeze,
1284 .statfs = ext4_statfs,
1285 .remount_fs = ext4_remount,
1286 .show_options = ext4_show_options,
1288 .quota_read = ext4_quota_read,
1289 .quota_write = ext4_quota_write,
1291 .bdev_try_to_free_page = bdev_try_to_free_page,
1294 static const struct super_operations ext4_nojournal_sops = {
1295 .alloc_inode = ext4_alloc_inode,
1296 .destroy_inode = ext4_destroy_inode,
1297 .write_inode = ext4_write_inode,
1298 .dirty_inode = ext4_dirty_inode,
1299 .drop_inode = ext4_drop_inode,
1300 .evict_inode = ext4_evict_inode,
1301 .write_super = ext4_write_super,
1302 .put_super = ext4_put_super,
1303 .statfs = ext4_statfs,
1304 .remount_fs = ext4_remount,
1305 .show_options = ext4_show_options,
1307 .quota_read = ext4_quota_read,
1308 .quota_write = ext4_quota_write,
1310 .bdev_try_to_free_page = bdev_try_to_free_page,
1313 static const struct export_operations ext4_export_ops = {
1314 .fh_to_dentry = ext4_fh_to_dentry,
1315 .fh_to_parent = ext4_fh_to_parent,
1316 .get_parent = ext4_get_parent,
1320 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1321 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1322 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1323 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1324 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1325 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1326 Opt_journal_update, Opt_journal_dev,
1327 Opt_journal_checksum, Opt_journal_async_commit,
1328 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1329 Opt_data_err_abort, Opt_data_err_ignore,
1330 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1331 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1332 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1333 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1334 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1335 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1336 Opt_inode_readahead_blks, Opt_journal_ioprio,
1337 Opt_dioread_nolock, Opt_dioread_lock,
1338 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
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_itable, "init_itable=%u"},
1412 {Opt_init_itable, "init_itable"},
1413 {Opt_noinit_itable, "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 ext4_msg(sb, KERN_WARNING,
1587 "Ignoring deprecated oldalloc option");
1590 ext4_msg(sb, KERN_WARNING,
1591 "Ignoring deprecated orlov option");
1593 #ifdef CONFIG_EXT4_FS_XATTR
1594 case Opt_user_xattr:
1595 set_opt(sb, XATTR_USER);
1597 case Opt_nouser_xattr:
1598 clear_opt(sb, XATTR_USER);
1601 case Opt_user_xattr:
1602 case Opt_nouser_xattr:
1603 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1606 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1608 set_opt(sb, POSIX_ACL);
1611 clear_opt(sb, POSIX_ACL);
1616 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1619 case Opt_journal_update:
1621 /* Eventually we will want to be able to create
1622 a journal file here. For now, only allow the
1623 user to specify an existing inode to be the
1626 ext4_msg(sb, KERN_ERR,
1627 "Cannot specify journal on remount");
1630 set_opt(sb, UPDATE_JOURNAL);
1632 case Opt_journal_dev:
1634 ext4_msg(sb, KERN_ERR,
1635 "Cannot specify journal on remount");
1638 if (match_int(&args[0], &option))
1640 *journal_devnum = option;
1642 case Opt_journal_checksum:
1643 set_opt(sb, JOURNAL_CHECKSUM);
1645 case Opt_journal_async_commit:
1646 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1647 set_opt(sb, JOURNAL_CHECKSUM);
1650 set_opt(sb, NOLOAD);
1653 if (match_int(&args[0], &option))
1658 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1659 sbi->s_commit_interval = HZ * option;
1661 case Opt_max_batch_time:
1662 if (match_int(&args[0], &option))
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 (!sbi->s_journal)
1686 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1687 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1688 ext4_msg(sb, KERN_ERR,
1689 "Cannot change data mode on remount");
1693 clear_opt(sb, DATA_FLAGS);
1694 sbi->s_mount_opt |= data_opt;
1697 case Opt_data_err_abort:
1698 set_opt(sb, DATA_ERR_ABORT);
1700 case Opt_data_err_ignore:
1701 clear_opt(sb, DATA_ERR_ABORT);
1705 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1709 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1712 case Opt_offusrjquota:
1713 if (!clear_qf_name(sb, USRQUOTA))
1716 case Opt_offgrpjquota:
1717 if (!clear_qf_name(sb, GRPQUOTA))
1721 case Opt_jqfmt_vfsold:
1722 qfmt = QFMT_VFS_OLD;
1724 case Opt_jqfmt_vfsv0:
1727 case Opt_jqfmt_vfsv1:
1730 if (sb_any_quota_loaded(sb) &&
1731 sbi->s_jquota_fmt != qfmt) {
1732 ext4_msg(sb, KERN_ERR, "Cannot change "
1733 "journaled quota options when "
1737 sbi->s_jquota_fmt = qfmt;
1742 set_opt(sb, USRQUOTA);
1746 set_opt(sb, GRPQUOTA);
1749 if (sb_any_quota_loaded(sb)) {
1750 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1751 "options when quota turned on");
1754 clear_opt(sb, QUOTA);
1755 clear_opt(sb, USRQUOTA);
1756 clear_opt(sb, GRPQUOTA);
1762 ext4_msg(sb, KERN_ERR,
1763 "quota options not supported");
1767 case Opt_offusrjquota:
1768 case Opt_offgrpjquota:
1769 case Opt_jqfmt_vfsold:
1770 case Opt_jqfmt_vfsv0:
1771 case Opt_jqfmt_vfsv1:
1772 ext4_msg(sb, KERN_ERR,
1773 "journaled quota options not supported");
1779 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1782 clear_opt(sb, BARRIER);
1786 if (match_int(&args[0], &option))
1789 option = 1; /* No argument, default to 1 */
1791 set_opt(sb, BARRIER);
1793 clear_opt(sb, BARRIER);
1799 ext4_msg(sb, KERN_ERR,
1800 "resize option only available "
1804 if (match_int(&args[0], &option) != 0)
1806 *n_blocks_count = option;
1809 ext4_msg(sb, KERN_WARNING,
1810 "Ignoring deprecated nobh option");
1813 ext4_msg(sb, KERN_WARNING,
1814 "Ignoring deprecated bh option");
1817 set_opt(sb, I_VERSION);
1818 sb->s_flags |= MS_I_VERSION;
1820 case Opt_nodelalloc:
1821 clear_opt(sb, DELALLOC);
1822 clear_opt2(sb, EXPLICIT_DELALLOC);
1824 case Opt_mblk_io_submit:
1825 set_opt(sb, MBLK_IO_SUBMIT);
1827 case Opt_nomblk_io_submit:
1828 clear_opt(sb, MBLK_IO_SUBMIT);
1831 if (match_int(&args[0], &option))
1835 sbi->s_stripe = option;
1838 set_opt(sb, DELALLOC);
1839 set_opt2(sb, EXPLICIT_DELALLOC);
1841 case Opt_block_validity:
1842 set_opt(sb, BLOCK_VALIDITY);
1844 case Opt_noblock_validity:
1845 clear_opt(sb, BLOCK_VALIDITY);
1847 case Opt_inode_readahead_blks:
1848 if (match_int(&args[0], &option))
1850 if (option < 0 || option > (1 << 30))
1852 if (option && !is_power_of_2(option)) {
1853 ext4_msg(sb, KERN_ERR,
1854 "EXT4-fs: inode_readahead_blks"
1855 " must be a power of 2");
1858 sbi->s_inode_readahead_blks = option;
1860 case Opt_journal_ioprio:
1861 if (match_int(&args[0], &option))
1863 if (option < 0 || option > 7)
1865 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1868 case Opt_noauto_da_alloc:
1869 set_opt(sb, NO_AUTO_DA_ALLOC);
1871 case Opt_auto_da_alloc:
1873 if (match_int(&args[0], &option))
1876 option = 1; /* No argument, default to 1 */
1878 clear_opt(sb, NO_AUTO_DA_ALLOC);
1880 set_opt(sb,NO_AUTO_DA_ALLOC);
1883 set_opt(sb, DISCARD);
1886 clear_opt(sb, DISCARD);
1888 case Opt_dioread_nolock:
1889 set_opt(sb, DIOREAD_NOLOCK);
1891 case Opt_dioread_lock:
1892 clear_opt(sb, DIOREAD_NOLOCK);
1894 case Opt_init_itable:
1895 set_opt(sb, INIT_INODE_TABLE);
1897 if (match_int(&args[0], &option))
1900 option = EXT4_DEF_LI_WAIT_MULT;
1903 sbi->s_li_wait_mult = option;
1905 case Opt_noinit_itable:
1906 clear_opt(sb, INIT_INODE_TABLE);
1909 ext4_msg(sb, KERN_ERR,
1910 "Unrecognized mount option \"%s\" "
1911 "or missing value", p);
1916 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1917 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1918 clear_opt(sb, USRQUOTA);
1920 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1921 clear_opt(sb, GRPQUOTA);
1923 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1924 ext4_msg(sb, KERN_ERR, "old and new quota "
1929 if (!sbi->s_jquota_fmt) {
1930 ext4_msg(sb, KERN_ERR, "journaled quota format "
1935 if (sbi->s_jquota_fmt) {
1936 ext4_msg(sb, KERN_ERR, "journaled quota format "
1937 "specified with no journaling "
1943 if (test_opt(sb, DIOREAD_NOLOCK)) {
1945 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1947 if (blocksize < PAGE_CACHE_SIZE) {
1948 ext4_msg(sb, KERN_ERR, "can't mount with "
1949 "dioread_nolock if block size != PAGE_SIZE");
1956 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1959 struct ext4_sb_info *sbi = EXT4_SB(sb);
1962 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1963 ext4_msg(sb, KERN_ERR, "revision level too high, "
1964 "forcing read-only mode");
1969 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1970 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1971 "running e2fsck is recommended");
1972 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1973 ext4_msg(sb, KERN_WARNING,
1974 "warning: mounting fs with errors, "
1975 "running e2fsck is recommended");
1976 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1977 le16_to_cpu(es->s_mnt_count) >=
1978 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1979 ext4_msg(sb, KERN_WARNING,
1980 "warning: maximal mount count reached, "
1981 "running e2fsck is recommended");
1982 else if (le32_to_cpu(es->s_checkinterval) &&
1983 (le32_to_cpu(es->s_lastcheck) +
1984 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1985 ext4_msg(sb, KERN_WARNING,
1986 "warning: checktime reached, "
1987 "running e2fsck is recommended");
1988 if (!sbi->s_journal)
1989 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1990 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1991 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1992 le16_add_cpu(&es->s_mnt_count, 1);
1993 es->s_mtime = cpu_to_le32(get_seconds());
1994 ext4_update_dynamic_rev(sb);
1996 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1998 ext4_commit_super(sb, 1);
2000 if (test_opt(sb, DEBUG))
2001 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2002 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2004 sbi->s_groups_count,
2005 EXT4_BLOCKS_PER_GROUP(sb),
2006 EXT4_INODES_PER_GROUP(sb),
2007 sbi->s_mount_opt, sbi->s_mount_opt2);
2009 cleancache_init_fs(sb);
2013 static int ext4_fill_flex_info(struct super_block *sb)
2015 struct ext4_sb_info *sbi = EXT4_SB(sb);
2016 struct ext4_group_desc *gdp = NULL;
2017 ext4_group_t flex_group_count;
2018 ext4_group_t flex_group;
2019 unsigned int groups_per_flex = 0;
2023 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2024 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2025 sbi->s_log_groups_per_flex = 0;
2028 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2030 /* We allocate both existing and potentially added groups */
2031 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2032 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2033 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2034 size = flex_group_count * sizeof(struct flex_groups);
2035 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2036 if (sbi->s_flex_groups == NULL) {
2037 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2042 for (i = 0; i < sbi->s_groups_count; i++) {
2043 gdp = ext4_get_group_desc(sb, i, NULL);
2045 flex_group = ext4_flex_group(sbi, i);
2046 atomic_add(ext4_free_inodes_count(sb, gdp),
2047 &sbi->s_flex_groups[flex_group].free_inodes);
2048 atomic64_add(ext4_free_group_clusters(sb, gdp),
2049 &sbi->s_flex_groups[flex_group].free_clusters);
2050 atomic_add(ext4_used_dirs_count(sb, gdp),
2051 &sbi->s_flex_groups[flex_group].used_dirs);
2059 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2060 struct ext4_group_desc *gdp)
2064 if (sbi->s_es->s_feature_ro_compat &
2065 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2066 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2067 __le32 le_group = cpu_to_le32(block_group);
2069 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2070 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2071 crc = crc16(crc, (__u8 *)gdp, offset);
2072 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2073 /* for checksum of struct ext4_group_desc do the rest...*/
2074 if ((sbi->s_es->s_feature_incompat &
2075 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2076 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2077 crc = crc16(crc, (__u8 *)gdp + offset,
2078 le16_to_cpu(sbi->s_es->s_desc_size) -
2082 return cpu_to_le16(crc);
2085 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2086 struct ext4_group_desc *gdp)
2088 if ((sbi->s_es->s_feature_ro_compat &
2089 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2090 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2096 /* Called at mount-time, super-block is locked */
2097 static int ext4_check_descriptors(struct super_block *sb,
2098 ext4_group_t *first_not_zeroed)
2100 struct ext4_sb_info *sbi = EXT4_SB(sb);
2101 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2102 ext4_fsblk_t last_block;
2103 ext4_fsblk_t block_bitmap;
2104 ext4_fsblk_t inode_bitmap;
2105 ext4_fsblk_t inode_table;
2106 int flexbg_flag = 0;
2107 ext4_group_t i, grp = sbi->s_groups_count;
2109 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2112 ext4_debug("Checking group descriptors");
2114 for (i = 0; i < sbi->s_groups_count; i++) {
2115 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2117 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2118 last_block = ext4_blocks_count(sbi->s_es) - 1;
2120 last_block = first_block +
2121 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2123 if ((grp == sbi->s_groups_count) &&
2124 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2127 block_bitmap = ext4_block_bitmap(sb, gdp);
2128 if (block_bitmap < first_block || block_bitmap > last_block) {
2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2130 "Block bitmap for group %u not in group "
2131 "(block %llu)!", i, block_bitmap);
2134 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2135 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2136 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2137 "Inode bitmap for group %u not in group "
2138 "(block %llu)!", i, inode_bitmap);
2141 inode_table = ext4_inode_table(sb, gdp);
2142 if (inode_table < first_block ||
2143 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2144 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2145 "Inode table for group %u not in group "
2146 "(block %llu)!", i, inode_table);
2149 ext4_lock_group(sb, i);
2150 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2151 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2152 "Checksum for group %u failed (%u!=%u)",
2153 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2154 gdp)), le16_to_cpu(gdp->bg_checksum));
2155 if (!(sb->s_flags & MS_RDONLY)) {
2156 ext4_unlock_group(sb, i);
2160 ext4_unlock_group(sb, i);
2162 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2164 if (NULL != first_not_zeroed)
2165 *first_not_zeroed = grp;
2167 ext4_free_blocks_count_set(sbi->s_es,
2168 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2169 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2173 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2174 * the superblock) which were deleted from all directories, but held open by
2175 * a process at the time of a crash. We walk the list and try to delete these
2176 * inodes at recovery time (only with a read-write filesystem).
2178 * In order to keep the orphan inode chain consistent during traversal (in
2179 * case of crash during recovery), we link each inode into the superblock
2180 * orphan list_head and handle it the same way as an inode deletion during
2181 * normal operation (which journals the operations for us).
2183 * We only do an iget() and an iput() on each inode, which is very safe if we
2184 * accidentally point at an in-use or already deleted inode. The worst that
2185 * can happen in this case is that we get a "bit already cleared" message from
2186 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2187 * e2fsck was run on this filesystem, and it must have already done the orphan
2188 * inode cleanup for us, so we can safely abort without any further action.
2190 static void ext4_orphan_cleanup(struct super_block *sb,
2191 struct ext4_super_block *es)
2193 unsigned int s_flags = sb->s_flags;
2194 int nr_orphans = 0, nr_truncates = 0;
2198 if (!es->s_last_orphan) {
2199 jbd_debug(4, "no orphan inodes to clean up\n");
2203 if (bdev_read_only(sb->s_bdev)) {
2204 ext4_msg(sb, KERN_ERR, "write access "
2205 "unavailable, skipping orphan cleanup");
2209 /* Check if feature set would not allow a r/w mount */
2210 if (!ext4_feature_set_ok(sb, 0)) {
2211 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2212 "unknown ROCOMPAT features");
2216 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2217 if (es->s_last_orphan)
2218 jbd_debug(1, "Errors on filesystem, "
2219 "clearing orphan list.\n");
2220 es->s_last_orphan = 0;
2221 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2225 if (s_flags & MS_RDONLY) {
2226 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2227 sb->s_flags &= ~MS_RDONLY;
2230 /* Needed for iput() to work correctly and not trash data */
2231 sb->s_flags |= MS_ACTIVE;
2232 /* Turn on quotas so that they are updated correctly */
2233 for (i = 0; i < MAXQUOTAS; i++) {
2234 if (EXT4_SB(sb)->s_qf_names[i]) {
2235 int ret = ext4_quota_on_mount(sb, i);
2237 ext4_msg(sb, KERN_ERR,
2238 "Cannot turn on journaled "
2239 "quota: error %d", ret);
2244 while (es->s_last_orphan) {
2245 struct inode *inode;
2247 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2248 if (IS_ERR(inode)) {
2249 es->s_last_orphan = 0;
2253 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2254 dquot_initialize(inode);
2255 if (inode->i_nlink) {
2256 ext4_msg(sb, KERN_DEBUG,
2257 "%s: truncating inode %lu to %lld bytes",
2258 __func__, inode->i_ino, inode->i_size);
2259 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2260 inode->i_ino, inode->i_size);
2261 mutex_lock(&inode->i_mutex);
2262 ext4_truncate(inode);
2263 mutex_unlock(&inode->i_mutex);
2266 ext4_msg(sb, KERN_DEBUG,
2267 "%s: deleting unreferenced inode %lu",
2268 __func__, inode->i_ino);
2269 jbd_debug(2, "deleting unreferenced inode %lu\n",
2273 iput(inode); /* The delete magic happens here! */
2276 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2279 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2280 PLURAL(nr_orphans));
2282 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2283 PLURAL(nr_truncates));
2285 /* Turn quotas off */
2286 for (i = 0; i < MAXQUOTAS; i++) {
2287 if (sb_dqopt(sb)->files[i])
2288 dquot_quota_off(sb, i);
2291 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2295 * Maximal extent format file size.
2296 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2297 * extent format containers, within a sector_t, and within i_blocks
2298 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2299 * so that won't be a limiting factor.
2301 * However there is other limiting factor. We do store extents in the form
2302 * of starting block and length, hence the resulting length of the extent
2303 * covering maximum file size must fit into on-disk format containers as
2304 * well. Given that length is always by 1 unit bigger than max unit (because
2305 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2307 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2309 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2312 loff_t upper_limit = MAX_LFS_FILESIZE;
2314 /* small i_blocks in vfs inode? */
2315 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2317 * CONFIG_LBDAF is not enabled implies the inode
2318 * i_block represent total blocks in 512 bytes
2319 * 32 == size of vfs inode i_blocks * 8
2321 upper_limit = (1LL << 32) - 1;
2323 /* total blocks in file system block size */
2324 upper_limit >>= (blkbits - 9);
2325 upper_limit <<= blkbits;
2329 * 32-bit extent-start container, ee_block. We lower the maxbytes
2330 * by one fs block, so ee_len can cover the extent of maximum file
2333 res = (1LL << 32) - 1;
2336 /* Sanity check against vm- & vfs- imposed limits */
2337 if (res > upper_limit)
2344 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2345 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2346 * We need to be 1 filesystem block less than the 2^48 sector limit.
2348 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2350 loff_t res = EXT4_NDIR_BLOCKS;
2353 /* This is calculated to be the largest file size for a dense, block
2354 * mapped file such that the file's total number of 512-byte sectors,
2355 * including data and all indirect blocks, does not exceed (2^48 - 1).
2357 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2358 * number of 512-byte sectors of the file.
2361 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2363 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2364 * the inode i_block field represents total file blocks in
2365 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2367 upper_limit = (1LL << 32) - 1;
2369 /* total blocks in file system block size */
2370 upper_limit >>= (bits - 9);
2374 * We use 48 bit ext4_inode i_blocks
2375 * With EXT4_HUGE_FILE_FL set the i_blocks
2376 * represent total number of blocks in
2377 * file system block size
2379 upper_limit = (1LL << 48) - 1;
2383 /* indirect blocks */
2385 /* double indirect blocks */
2386 meta_blocks += 1 + (1LL << (bits-2));
2387 /* tripple indirect blocks */
2388 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2390 upper_limit -= meta_blocks;
2391 upper_limit <<= bits;
2393 res += 1LL << (bits-2);
2394 res += 1LL << (2*(bits-2));
2395 res += 1LL << (3*(bits-2));
2397 if (res > upper_limit)
2400 if (res > MAX_LFS_FILESIZE)
2401 res = MAX_LFS_FILESIZE;
2406 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2407 ext4_fsblk_t logical_sb_block, int nr)
2409 struct ext4_sb_info *sbi = EXT4_SB(sb);
2410 ext4_group_t bg, first_meta_bg;
2413 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2415 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2417 return logical_sb_block + nr + 1;
2418 bg = sbi->s_desc_per_block * nr;
2419 if (ext4_bg_has_super(sb, bg))
2422 return (has_super + ext4_group_first_block_no(sb, bg));
2426 * ext4_get_stripe_size: Get the stripe size.
2427 * @sbi: In memory super block info
2429 * If we have specified it via mount option, then
2430 * use the mount option value. If the value specified at mount time is
2431 * greater than the blocks per group use the super block value.
2432 * If the super block value is greater than blocks per group return 0.
2433 * Allocator needs it be less than blocks per group.
2436 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2438 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2439 unsigned long stripe_width =
2440 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2443 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2444 ret = sbi->s_stripe;
2445 else if (stripe_width <= sbi->s_blocks_per_group)
2447 else if (stride <= sbi->s_blocks_per_group)
2453 * If the stripe width is 1, this makes no sense and
2454 * we set it to 0 to turn off stripe handling code.
2465 struct attribute attr;
2466 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2467 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2468 const char *, size_t);
2472 static int parse_strtoul(const char *buf,
2473 unsigned long max, unsigned long *value)
2477 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2478 endp = skip_spaces(endp);
2479 if (*endp || *value > max)
2485 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2486 struct ext4_sb_info *sbi,
2489 return snprintf(buf, PAGE_SIZE, "%llu\n",
2491 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2494 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2495 struct ext4_sb_info *sbi, char *buf)
2497 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2499 if (!sb->s_bdev->bd_part)
2500 return snprintf(buf, PAGE_SIZE, "0\n");
2501 return snprintf(buf, PAGE_SIZE, "%lu\n",
2502 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2503 sbi->s_sectors_written_start) >> 1);
2506 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi, char *buf)
2509 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2511 if (!sb->s_bdev->bd_part)
2512 return snprintf(buf, PAGE_SIZE, "0\n");
2513 return snprintf(buf, PAGE_SIZE, "%llu\n",
2514 (unsigned long long)(sbi->s_kbytes_written +
2515 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2516 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2519 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2520 struct ext4_sb_info *sbi,
2521 const char *buf, size_t count)
2525 if (parse_strtoul(buf, 0x40000000, &t))
2528 if (t && !is_power_of_2(t))
2531 sbi->s_inode_readahead_blks = t;
2535 static ssize_t sbi_ui_show(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi, char *buf)
2538 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2540 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2543 static ssize_t sbi_ui_store(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2547 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2550 if (parse_strtoul(buf, 0xffffffff, &t))
2556 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2557 static struct ext4_attr ext4_attr_##_name = { \
2558 .attr = {.name = __stringify(_name), .mode = _mode }, \
2561 .offset = offsetof(struct ext4_sb_info, _elname), \
2563 #define EXT4_ATTR(name, mode, show, store) \
2564 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2566 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2567 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2568 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2569 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2570 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2571 #define ATTR_LIST(name) &ext4_attr_##name.attr
2573 EXT4_RO_ATTR(delayed_allocation_blocks);
2574 EXT4_RO_ATTR(session_write_kbytes);
2575 EXT4_RO_ATTR(lifetime_write_kbytes);
2576 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2577 inode_readahead_blks_store, s_inode_readahead_blks);
2578 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2579 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2580 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2581 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2582 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2583 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2584 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2585 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2587 static struct attribute *ext4_attrs[] = {
2588 ATTR_LIST(delayed_allocation_blocks),
2589 ATTR_LIST(session_write_kbytes),
2590 ATTR_LIST(lifetime_write_kbytes),
2591 ATTR_LIST(inode_readahead_blks),
2592 ATTR_LIST(inode_goal),
2593 ATTR_LIST(mb_stats),
2594 ATTR_LIST(mb_max_to_scan),
2595 ATTR_LIST(mb_min_to_scan),
2596 ATTR_LIST(mb_order2_req),
2597 ATTR_LIST(mb_stream_req),
2598 ATTR_LIST(mb_group_prealloc),
2599 ATTR_LIST(max_writeback_mb_bump),
2603 /* Features this copy of ext4 supports */
2604 EXT4_INFO_ATTR(lazy_itable_init);
2605 EXT4_INFO_ATTR(batched_discard);
2607 static struct attribute *ext4_feat_attrs[] = {
2608 ATTR_LIST(lazy_itable_init),
2609 ATTR_LIST(batched_discard),
2613 static ssize_t ext4_attr_show(struct kobject *kobj,
2614 struct attribute *attr, char *buf)
2616 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2618 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2620 return a->show ? a->show(a, sbi, buf) : 0;
2623 static ssize_t ext4_attr_store(struct kobject *kobj,
2624 struct attribute *attr,
2625 const char *buf, size_t len)
2627 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2629 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2631 return a->store ? a->store(a, sbi, buf, len) : 0;
2634 static void ext4_sb_release(struct kobject *kobj)
2636 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2638 complete(&sbi->s_kobj_unregister);
2641 static const struct sysfs_ops ext4_attr_ops = {
2642 .show = ext4_attr_show,
2643 .store = ext4_attr_store,
2646 static struct kobj_type ext4_ktype = {
2647 .default_attrs = ext4_attrs,
2648 .sysfs_ops = &ext4_attr_ops,
2649 .release = ext4_sb_release,
2652 static void ext4_feat_release(struct kobject *kobj)
2654 complete(&ext4_feat->f_kobj_unregister);
2657 static struct kobj_type ext4_feat_ktype = {
2658 .default_attrs = ext4_feat_attrs,
2659 .sysfs_ops = &ext4_attr_ops,
2660 .release = ext4_feat_release,
2664 * Check whether this filesystem can be mounted based on
2665 * the features present and the RDONLY/RDWR mount requested.
2666 * Returns 1 if this filesystem can be mounted as requested,
2667 * 0 if it cannot be.
2669 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2671 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2672 ext4_msg(sb, KERN_ERR,
2673 "Couldn't mount because of "
2674 "unsupported optional features (%x)",
2675 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2676 ~EXT4_FEATURE_INCOMPAT_SUPP));
2683 /* Check that feature set is OK for a read-write mount */
2684 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2685 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2686 "unsupported optional features (%x)",
2687 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2688 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2692 * Large file size enabled file system can only be mounted
2693 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2695 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2696 if (sizeof(blkcnt_t) < sizeof(u64)) {
2697 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2698 "cannot be mounted RDWR without "
2703 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2704 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2705 ext4_msg(sb, KERN_ERR,
2706 "Can't support bigalloc feature without "
2707 "extents feature\n");
2714 * This function is called once a day if we have errors logged
2715 * on the file system
2717 static void print_daily_error_info(unsigned long arg)
2719 struct super_block *sb = (struct super_block *) arg;
2720 struct ext4_sb_info *sbi;
2721 struct ext4_super_block *es;
2726 if (es->s_error_count)
2727 /* fsck newer than v1.41.13 is needed to clean this condition. */
2728 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2729 le32_to_cpu(es->s_error_count));
2730 if (es->s_first_error_time) {
2731 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2732 sb->s_id, le32_to_cpu(es->s_first_error_time),
2733 (int) sizeof(es->s_first_error_func),
2734 es->s_first_error_func,
2735 le32_to_cpu(es->s_first_error_line));
2736 if (es->s_first_error_ino)
2737 printk(": inode %u",
2738 le32_to_cpu(es->s_first_error_ino));
2739 if (es->s_first_error_block)
2740 printk(": block %llu", (unsigned long long)
2741 le64_to_cpu(es->s_first_error_block));
2744 if (es->s_last_error_time) {
2745 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2746 sb->s_id, le32_to_cpu(es->s_last_error_time),
2747 (int) sizeof(es->s_last_error_func),
2748 es->s_last_error_func,
2749 le32_to_cpu(es->s_last_error_line));
2750 if (es->s_last_error_ino)
2751 printk(": inode %u",
2752 le32_to_cpu(es->s_last_error_ino));
2753 if (es->s_last_error_block)
2754 printk(": block %llu", (unsigned long long)
2755 le64_to_cpu(es->s_last_error_block));
2758 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2761 /* Find next suitable group and run ext4_init_inode_table */
2762 static int ext4_run_li_request(struct ext4_li_request *elr)
2764 struct ext4_group_desc *gdp = NULL;
2765 ext4_group_t group, ngroups;
2766 struct super_block *sb;
2767 unsigned long timeout = 0;
2771 ngroups = EXT4_SB(sb)->s_groups_count;
2773 for (group = elr->lr_next_group; group < ngroups; group++) {
2774 gdp = ext4_get_group_desc(sb, group, NULL);
2780 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2784 if (group == ngroups)
2789 ret = ext4_init_inode_table(sb, group,
2790 elr->lr_timeout ? 0 : 1);
2791 if (elr->lr_timeout == 0) {
2792 timeout = (jiffies - timeout) *
2793 elr->lr_sbi->s_li_wait_mult;
2794 elr->lr_timeout = timeout;
2796 elr->lr_next_sched = jiffies + elr->lr_timeout;
2797 elr->lr_next_group = group + 1;
2804 * Remove lr_request from the list_request and free the
2805 * request structure. Should be called with li_list_mtx held
2807 static void ext4_remove_li_request(struct ext4_li_request *elr)
2809 struct ext4_sb_info *sbi;
2816 list_del(&elr->lr_request);
2817 sbi->s_li_request = NULL;
2821 static void ext4_unregister_li_request(struct super_block *sb)
2823 mutex_lock(&ext4_li_mtx);
2824 if (!ext4_li_info) {
2825 mutex_unlock(&ext4_li_mtx);
2829 mutex_lock(&ext4_li_info->li_list_mtx);
2830 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2831 mutex_unlock(&ext4_li_info->li_list_mtx);
2832 mutex_unlock(&ext4_li_mtx);
2835 static struct task_struct *ext4_lazyinit_task;
2838 * This is the function where ext4lazyinit thread lives. It walks
2839 * through the request list searching for next scheduled filesystem.
2840 * When such a fs is found, run the lazy initialization request
2841 * (ext4_rn_li_request) and keep track of the time spend in this
2842 * function. Based on that time we compute next schedule time of
2843 * the request. When walking through the list is complete, compute
2844 * next waking time and put itself into sleep.
2846 static int ext4_lazyinit_thread(void *arg)
2848 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2849 struct list_head *pos, *n;
2850 struct ext4_li_request *elr;
2851 unsigned long next_wakeup, cur;
2853 BUG_ON(NULL == eli);
2857 next_wakeup = MAX_JIFFY_OFFSET;
2859 mutex_lock(&eli->li_list_mtx);
2860 if (list_empty(&eli->li_request_list)) {
2861 mutex_unlock(&eli->li_list_mtx);
2865 list_for_each_safe(pos, n, &eli->li_request_list) {
2866 elr = list_entry(pos, struct ext4_li_request,
2869 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2870 if (ext4_run_li_request(elr) != 0) {
2871 /* error, remove the lazy_init job */
2872 ext4_remove_li_request(elr);
2877 if (time_before(elr->lr_next_sched, next_wakeup))
2878 next_wakeup = elr->lr_next_sched;
2880 mutex_unlock(&eli->li_list_mtx);
2882 if (freezing(current))
2886 if ((time_after_eq(cur, next_wakeup)) ||
2887 (MAX_JIFFY_OFFSET == next_wakeup)) {
2892 schedule_timeout_interruptible(next_wakeup - cur);
2894 if (kthread_should_stop()) {
2895 ext4_clear_request_list();
2902 * It looks like the request list is empty, but we need
2903 * to check it under the li_list_mtx lock, to prevent any
2904 * additions into it, and of course we should lock ext4_li_mtx
2905 * to atomically free the list and ext4_li_info, because at
2906 * this point another ext4 filesystem could be registering
2909 mutex_lock(&ext4_li_mtx);
2910 mutex_lock(&eli->li_list_mtx);
2911 if (!list_empty(&eli->li_request_list)) {
2912 mutex_unlock(&eli->li_list_mtx);
2913 mutex_unlock(&ext4_li_mtx);
2916 mutex_unlock(&eli->li_list_mtx);
2917 kfree(ext4_li_info);
2918 ext4_li_info = NULL;
2919 mutex_unlock(&ext4_li_mtx);
2924 static void ext4_clear_request_list(void)
2926 struct list_head *pos, *n;
2927 struct ext4_li_request *elr;
2929 mutex_lock(&ext4_li_info->li_list_mtx);
2930 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2931 elr = list_entry(pos, struct ext4_li_request,
2933 ext4_remove_li_request(elr);
2935 mutex_unlock(&ext4_li_info->li_list_mtx);
2938 static int ext4_run_lazyinit_thread(void)
2940 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2941 ext4_li_info, "ext4lazyinit");
2942 if (IS_ERR(ext4_lazyinit_task)) {
2943 int err = PTR_ERR(ext4_lazyinit_task);
2944 ext4_clear_request_list();
2945 kfree(ext4_li_info);
2946 ext4_li_info = NULL;
2947 printk(KERN_CRIT "EXT4: error %d creating inode table "
2948 "initialization thread\n",
2952 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2957 * Check whether it make sense to run itable init. thread or not.
2958 * If there is at least one uninitialized inode table, return
2959 * corresponding group number, else the loop goes through all
2960 * groups and return total number of groups.
2962 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2964 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2965 struct ext4_group_desc *gdp = NULL;
2967 for (group = 0; group < ngroups; group++) {
2968 gdp = ext4_get_group_desc(sb, group, NULL);
2972 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2979 static int ext4_li_info_new(void)
2981 struct ext4_lazy_init *eli = NULL;
2983 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2987 INIT_LIST_HEAD(&eli->li_request_list);
2988 mutex_init(&eli->li_list_mtx);
2990 eli->li_state |= EXT4_LAZYINIT_QUIT;
2997 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3000 struct ext4_sb_info *sbi = EXT4_SB(sb);
3001 struct ext4_li_request *elr;
3004 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3010 elr->lr_next_group = start;
3013 * Randomize first schedule time of the request to
3014 * spread the inode table initialization requests
3017 get_random_bytes(&rnd, sizeof(rnd));
3018 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3019 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3024 static int ext4_register_li_request(struct super_block *sb,
3025 ext4_group_t first_not_zeroed)
3027 struct ext4_sb_info *sbi = EXT4_SB(sb);
3028 struct ext4_li_request *elr;
3029 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3032 if (sbi->s_li_request != NULL) {
3034 * Reset timeout so it can be computed again, because
3035 * s_li_wait_mult might have changed.
3037 sbi->s_li_request->lr_timeout = 0;
3041 if (first_not_zeroed == ngroups ||
3042 (sb->s_flags & MS_RDONLY) ||
3043 !test_opt(sb, INIT_INODE_TABLE))
3046 elr = ext4_li_request_new(sb, first_not_zeroed);
3050 mutex_lock(&ext4_li_mtx);
3052 if (NULL == ext4_li_info) {
3053 ret = ext4_li_info_new();
3058 mutex_lock(&ext4_li_info->li_list_mtx);
3059 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3060 mutex_unlock(&ext4_li_info->li_list_mtx);
3062 sbi->s_li_request = elr;
3064 * set elr to NULL here since it has been inserted to
3065 * the request_list and the removal and free of it is
3066 * handled by ext4_clear_request_list from now on.
3070 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3071 ret = ext4_run_lazyinit_thread();
3076 mutex_unlock(&ext4_li_mtx);
3083 * We do not need to lock anything since this is called on
3086 static void ext4_destroy_lazyinit_thread(void)
3089 * If thread exited earlier
3090 * there's nothing to be done.
3092 if (!ext4_li_info || !ext4_lazyinit_task)
3095 kthread_stop(ext4_lazyinit_task);
3099 * Note: calculating the overhead so we can be compatible with
3100 * historical BSD practice is quite difficult in the face of
3101 * clusters/bigalloc. This is because multiple metadata blocks from
3102 * different block group can end up in the same allocation cluster.
3103 * Calculating the exact overhead in the face of clustered allocation
3104 * requires either O(all block bitmaps) in memory or O(number of block
3105 * groups**2) in time. We will still calculate the superblock for
3106 * older file systems --- and if we come across with a bigalloc file
3107 * system with zero in s_overhead_clusters the estimate will be close to
3108 * correct especially for very large cluster sizes --- but for newer
3109 * file systems, it's better to calculate this figure once at mkfs
3110 * time, and store it in the superblock. If the superblock value is
3111 * present (even for non-bigalloc file systems), we will use it.
3113 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3116 struct ext4_sb_info *sbi = EXT4_SB(sb);
3117 struct ext4_group_desc *gdp;
3118 ext4_fsblk_t first_block, last_block, b;
3119 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3120 int s, j, count = 0;
3122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3123 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3124 sbi->s_itb_per_group + 2);
3126 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3127 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3128 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3129 for (i = 0; i < ngroups; i++) {
3130 gdp = ext4_get_group_desc(sb, i, NULL);
3131 b = ext4_block_bitmap(sb, gdp);
3132 if (b >= first_block && b <= last_block) {
3133 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3136 b = ext4_inode_bitmap(sb, gdp);
3137 if (b >= first_block && b <= last_block) {
3138 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3141 b = ext4_inode_table(sb, gdp);
3142 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3143 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3144 int c = EXT4_B2C(sbi, b - first_block);
3145 ext4_set_bit(c, buf);
3151 if (ext4_bg_has_super(sb, grp)) {
3152 ext4_set_bit(s++, buf);
3155 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3156 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3162 return EXT4_CLUSTERS_PER_GROUP(sb) -
3163 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3167 * Compute the overhead and stash it in sbi->s_overhead
3169 int ext4_calculate_overhead(struct super_block *sb)
3171 struct ext4_sb_info *sbi = EXT4_SB(sb);
3172 struct ext4_super_block *es = sbi->s_es;
3173 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3174 ext4_fsblk_t overhead = 0;
3175 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3177 memset(buf, 0, PAGE_SIZE);
3182 * Compute the overhead (FS structures). This is constant
3183 * for a given filesystem unless the number of block groups
3184 * changes so we cache the previous value until it does.
3188 * All of the blocks before first_data_block are overhead
3190 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3193 * Add the overhead found in each block group
3195 for (i = 0; i < ngroups; i++) {
3198 blks = count_overhead(sb, i, buf);
3201 memset(buf, 0, PAGE_SIZE);
3204 sbi->s_overhead = overhead;
3206 free_page((unsigned long) buf);
3210 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3212 char *orig_data = kstrdup(data, GFP_KERNEL);
3213 struct buffer_head *bh;
3214 struct ext4_super_block *es = NULL;
3215 struct ext4_sb_info *sbi;
3217 ext4_fsblk_t sb_block = get_sb_block(&data);
3218 ext4_fsblk_t logical_sb_block;
3219 unsigned long offset = 0;
3220 unsigned long journal_devnum = 0;
3221 unsigned long def_mount_opts;
3226 int blocksize, clustersize;
3227 unsigned int db_count;
3229 int needs_recovery, has_huge_files, has_bigalloc;
3232 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3233 ext4_group_t first_not_zeroed;
3235 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3239 sbi->s_blockgroup_lock =
3240 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3241 if (!sbi->s_blockgroup_lock) {
3245 sb->s_fs_info = sbi;
3246 sbi->s_mount_opt = 0;
3247 sbi->s_resuid = EXT4_DEF_RESUID;
3248 sbi->s_resgid = EXT4_DEF_RESGID;
3249 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3250 sbi->s_sb_block = sb_block;
3251 if (sb->s_bdev->bd_part)
3252 sbi->s_sectors_written_start =
3253 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3255 /* Cleanup superblock name */
3256 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3260 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3262 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3267 * The ext4 superblock will not be buffer aligned for other than 1kB
3268 * block sizes. We need to calculate the offset from buffer start.
3270 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3271 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3272 offset = do_div(logical_sb_block, blocksize);
3274 logical_sb_block = sb_block;
3277 if (!(bh = sb_bread(sb, logical_sb_block))) {
3278 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3282 * Note: s_es must be initialized as soon as possible because
3283 * some ext4 macro-instructions depend on its value
3285 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3287 sb->s_magic = le16_to_cpu(es->s_magic);
3288 if (sb->s_magic != EXT4_SUPER_MAGIC)
3290 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3292 /* Set defaults before we parse the mount options */
3293 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3294 set_opt(sb, INIT_INODE_TABLE);
3295 if (def_mount_opts & EXT4_DEFM_DEBUG)
3297 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3298 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3302 if (def_mount_opts & EXT4_DEFM_UID16)
3303 set_opt(sb, NO_UID32);
3304 /* xattr user namespace & acls are now defaulted on */
3305 #ifdef CONFIG_EXT4_FS_XATTR
3306 set_opt(sb, XATTR_USER);
3308 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3309 set_opt(sb, POSIX_ACL);
3311 set_opt(sb, MBLK_IO_SUBMIT);
3312 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3313 set_opt(sb, JOURNAL_DATA);
3314 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3315 set_opt(sb, ORDERED_DATA);
3316 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3317 set_opt(sb, WRITEBACK_DATA);
3319 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3320 set_opt(sb, ERRORS_PANIC);
3321 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3322 set_opt(sb, ERRORS_CONT);
3324 set_opt(sb, ERRORS_RO);
3325 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3326 set_opt(sb, BLOCK_VALIDITY);
3327 if (def_mount_opts & EXT4_DEFM_DISCARD)
3328 set_opt(sb, DISCARD);
3330 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3331 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3332 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3333 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3334 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3336 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3337 set_opt(sb, BARRIER);
3340 * enable delayed allocation by default
3341 * Use -o nodelalloc to turn it off
3343 if (!IS_EXT3_SB(sb) &&
3344 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3345 set_opt(sb, DELALLOC);
3348 * set default s_li_wait_mult for lazyinit, for the case there is
3349 * no mount option specified.
3351 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3353 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3354 &journal_devnum, &journal_ioprio, NULL, 0)) {
3355 ext4_msg(sb, KERN_WARNING,
3356 "failed to parse options in superblock: %s",
3357 sbi->s_es->s_mount_opts);
3359 if (!parse_options((char *) data, sb, &journal_devnum,
3360 &journal_ioprio, NULL, 0))
3363 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3364 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3365 "with data=journal disables delayed "
3366 "allocation and O_DIRECT support!\n");
3367 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3368 ext4_msg(sb, KERN_ERR, "can't mount with "
3369 "both data=journal and delalloc");
3372 if (test_opt(sb, DIOREAD_NOLOCK)) {
3373 ext4_msg(sb, KERN_ERR, "can't mount with "
3374 "both data=journal and dioread_nolock");
3377 if (test_opt(sb, DELALLOC))
3378 clear_opt(sb, DELALLOC);
3381 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3382 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3384 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3385 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3386 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3387 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3388 ext4_msg(sb, KERN_WARNING,
3389 "feature flags set on rev 0 fs, "
3390 "running e2fsck is recommended");
3392 if (IS_EXT2_SB(sb)) {
3393 if (ext2_feature_set_ok(sb))
3394 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3395 "using the ext4 subsystem");
3397 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3398 "to feature incompatibilities");
3403 if (IS_EXT3_SB(sb)) {
3404 if (ext3_feature_set_ok(sb))
3405 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3406 "using the ext4 subsystem");
3408 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3409 "to feature incompatibilities");
3415 * Check feature flags regardless of the revision level, since we
3416 * previously didn't change the revision level when setting the flags,
3417 * so there is a chance incompat flags are set on a rev 0 filesystem.
3419 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3422 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3423 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3424 blocksize > EXT4_MAX_BLOCK_SIZE) {
3425 ext4_msg(sb, KERN_ERR,
3426 "Unsupported filesystem blocksize %d", blocksize);
3430 if (sb->s_blocksize != blocksize) {
3431 /* Validate the filesystem blocksize */
3432 if (!sb_set_blocksize(sb, blocksize)) {
3433 ext4_msg(sb, KERN_ERR, "bad block size %d",
3439 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3440 offset = do_div(logical_sb_block, blocksize);
3441 bh = sb_bread(sb, logical_sb_block);
3443 ext4_msg(sb, KERN_ERR,
3444 "Can't read superblock on 2nd try");
3447 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3449 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3450 ext4_msg(sb, KERN_ERR,
3451 "Magic mismatch, very weird!");
3456 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3457 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3458 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3460 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3462 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3463 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3464 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3466 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3467 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3468 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3469 (!is_power_of_2(sbi->s_inode_size)) ||
3470 (sbi->s_inode_size > blocksize)) {
3471 ext4_msg(sb, KERN_ERR,
3472 "unsupported inode size: %d",
3476 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3477 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3480 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3482 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3483 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3484 !is_power_of_2(sbi->s_desc_size)) {
3485 ext4_msg(sb, KERN_ERR,
3486 "unsupported descriptor size %lu",
3491 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3493 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3494 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3495 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3498 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3499 if (sbi->s_inodes_per_block == 0)
3501 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3502 sbi->s_inodes_per_block;
3503 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3505 sbi->s_mount_state = le16_to_cpu(es->s_state);
3506 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3507 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3509 for (i = 0; i < 4; i++)
3510 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3511 sbi->s_def_hash_version = es->s_def_hash_version;
3512 i = le32_to_cpu(es->s_flags);
3513 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3514 sbi->s_hash_unsigned = 3;
3515 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3516 #ifdef __CHAR_UNSIGNED__
3517 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3518 sbi->s_hash_unsigned = 3;
3520 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3525 /* Handle clustersize */
3526 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3527 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3528 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3530 if (clustersize < blocksize) {
3531 ext4_msg(sb, KERN_ERR,
3532 "cluster size (%d) smaller than "
3533 "block size (%d)", clustersize, blocksize);
3536 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3537 le32_to_cpu(es->s_log_block_size);
3538 sbi->s_clusters_per_group =
3539 le32_to_cpu(es->s_clusters_per_group);
3540 if (sbi->s_clusters_per_group > blocksize * 8) {
3541 ext4_msg(sb, KERN_ERR,
3542 "#clusters per group too big: %lu",
3543 sbi->s_clusters_per_group);
3546 if (sbi->s_blocks_per_group !=
3547 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3548 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3549 "clusters per group (%lu) inconsistent",
3550 sbi->s_blocks_per_group,
3551 sbi->s_clusters_per_group);
3555 if (clustersize != blocksize) {
3556 ext4_warning(sb, "fragment/cluster size (%d) != "
3557 "block size (%d)", clustersize,
3559 clustersize = blocksize;
3561 if (sbi->s_blocks_per_group > blocksize * 8) {
3562 ext4_msg(sb, KERN_ERR,
3563 "#blocks per group too big: %lu",
3564 sbi->s_blocks_per_group);
3567 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3568 sbi->s_cluster_bits = 0;
3570 sbi->s_cluster_ratio = clustersize / blocksize;
3572 if (sbi->s_inodes_per_group > blocksize * 8) {
3573 ext4_msg(sb, KERN_ERR,
3574 "#inodes per group too big: %lu",
3575 sbi->s_inodes_per_group);
3580 * Test whether we have more sectors than will fit in sector_t,
3581 * and whether the max offset is addressable by the page cache.
3583 err = generic_check_addressable(sb->s_blocksize_bits,
3584 ext4_blocks_count(es));
3586 ext4_msg(sb, KERN_ERR, "filesystem"
3587 " too large to mount safely on this system");
3588 if (sizeof(sector_t) < 8)
3589 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3594 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3597 /* check blocks count against device size */
3598 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3599 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3600 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3601 "exceeds size of device (%llu blocks)",
3602 ext4_blocks_count(es), blocks_count);
3607 * It makes no sense for the first data block to be beyond the end
3608 * of the filesystem.
3610 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3611 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3612 "block %u is beyond end of filesystem (%llu)",
3613 le32_to_cpu(es->s_first_data_block),
3614 ext4_blocks_count(es));
3617 blocks_count = (ext4_blocks_count(es) -
3618 le32_to_cpu(es->s_first_data_block) +
3619 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3620 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3621 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3622 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3623 "(block count %llu, first data block %u, "
3624 "blocks per group %lu)", sbi->s_groups_count,
3625 ext4_blocks_count(es),
3626 le32_to_cpu(es->s_first_data_block),
3627 EXT4_BLOCKS_PER_GROUP(sb));
3630 sbi->s_groups_count = blocks_count;
3631 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3632 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3633 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3634 EXT4_DESC_PER_BLOCK(sb);
3635 sbi->s_group_desc = ext4_kvmalloc(db_count *
3636 sizeof(struct buffer_head *),
3638 if (sbi->s_group_desc == NULL) {
3639 ext4_msg(sb, KERN_ERR, "not enough memory");
3644 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3646 bgl_lock_init(sbi->s_blockgroup_lock);
3648 for (i = 0; i < db_count; i++) {
3649 block = descriptor_loc(sb, logical_sb_block, i);
3650 sbi->s_group_desc[i] = sb_bread(sb, block);
3651 if (!sbi->s_group_desc[i]) {
3652 ext4_msg(sb, KERN_ERR,
3653 "can't read group descriptor %d", i);
3658 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3659 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3662 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3663 if (!ext4_fill_flex_info(sb)) {
3664 ext4_msg(sb, KERN_ERR,
3665 "unable to initialize "
3666 "flex_bg meta info!");
3670 sbi->s_gdb_count = db_count;
3671 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3672 spin_lock_init(&sbi->s_next_gen_lock);
3674 init_timer(&sbi->s_err_report);
3675 sbi->s_err_report.function = print_daily_error_info;
3676 sbi->s_err_report.data = (unsigned long) sb;
3678 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3679 ext4_count_free_clusters(sb));
3681 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3682 ext4_count_free_inodes(sb));
3685 err = percpu_counter_init(&sbi->s_dirs_counter,
3686 ext4_count_dirs(sb));
3689 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3692 ext4_msg(sb, KERN_ERR, "insufficient memory");
3696 sbi->s_stripe = ext4_get_stripe_size(sbi);
3697 sbi->s_max_writeback_mb_bump = 128;
3700 * set up enough so that it can read an inode
3702 if (!test_opt(sb, NOLOAD) &&
3703 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3704 sb->s_op = &ext4_sops;
3706 sb->s_op = &ext4_nojournal_sops;
3707 sb->s_export_op = &ext4_export_ops;
3708 sb->s_xattr = ext4_xattr_handlers;
3710 sb->s_qcop = &ext4_qctl_operations;
3711 sb->dq_op = &ext4_quota_operations;
3713 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3715 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3716 mutex_init(&sbi->s_orphan_lock);
3717 sbi->s_resize_flags = 0;
3721 needs_recovery = (es->s_last_orphan != 0 ||
3722 EXT4_HAS_INCOMPAT_FEATURE(sb,
3723 EXT4_FEATURE_INCOMPAT_RECOVER));
3725 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3726 !(sb->s_flags & MS_RDONLY))
3727 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3731 * The first inode we look at is the journal inode. Don't try
3732 * root first: it may be modified in the journal!
3734 if (!test_opt(sb, NOLOAD) &&
3735 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3736 if (ext4_load_journal(sb, es, journal_devnum))
3738 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3739 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3740 ext4_msg(sb, KERN_ERR, "required journal recovery "
3741 "suppressed and not mounted read-only");
3742 goto failed_mount_wq;
3744 clear_opt(sb, DATA_FLAGS);
3745 sbi->s_journal = NULL;
3750 if (ext4_blocks_count(es) > 0xffffffffULL &&
3751 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3752 JBD2_FEATURE_INCOMPAT_64BIT)) {
3753 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3754 goto failed_mount_wq;
3757 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3758 jbd2_journal_set_features(sbi->s_journal,
3759 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3760 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3761 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3762 jbd2_journal_set_features(sbi->s_journal,
3763 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3764 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3765 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3767 jbd2_journal_clear_features(sbi->s_journal,
3768 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3769 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3772 /* We have now updated the journal if required, so we can
3773 * validate the data journaling mode. */
3774 switch (test_opt(sb, DATA_FLAGS)) {
3776 /* No mode set, assume a default based on the journal
3777 * capabilities: ORDERED_DATA if the journal can
3778 * cope, else JOURNAL_DATA
3780 if (jbd2_journal_check_available_features
3781 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3782 set_opt(sb, ORDERED_DATA);
3784 set_opt(sb, JOURNAL_DATA);
3787 case EXT4_MOUNT_ORDERED_DATA:
3788 case EXT4_MOUNT_WRITEBACK_DATA:
3789 if (!jbd2_journal_check_available_features
3790 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3791 ext4_msg(sb, KERN_ERR, "Journal does not support "
3792 "requested data journaling mode");
3793 goto failed_mount_wq;
3798 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3801 * The journal may have updated the bg summary counts, so we
3802 * need to update the global counters.
3804 percpu_counter_set(&sbi->s_freeclusters_counter,
3805 ext4_count_free_clusters(sb));
3806 percpu_counter_set(&sbi->s_freeinodes_counter,
3807 ext4_count_free_inodes(sb));
3808 percpu_counter_set(&sbi->s_dirs_counter,
3809 ext4_count_dirs(sb));
3810 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3814 * Get the # of file system overhead blocks from the
3815 * superblock if present.
3817 if (es->s_overhead_clusters)
3818 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3820 ret = ext4_calculate_overhead(sb);
3822 goto failed_mount_wq;
3826 * The maximum number of concurrent works can be high and
3827 * concurrency isn't really necessary. Limit it to 1.
3829 EXT4_SB(sb)->dio_unwritten_wq =
3830 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3831 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3832 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3833 goto failed_mount_wq;
3837 * The jbd2_journal_load will have done any necessary log recovery,
3838 * so we can safely mount the rest of the filesystem now.
3841 root = ext4_iget(sb, EXT4_ROOT_INO);
3843 ext4_msg(sb, KERN_ERR, "get root inode failed");
3844 ret = PTR_ERR(root);
3848 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3849 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3852 sb->s_root = d_alloc_root(root);
3854 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3859 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3860 sb->s_flags |= MS_RDONLY;
3862 /* determine the minimum size of new large inodes, if present */
3863 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3864 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3865 EXT4_GOOD_OLD_INODE_SIZE;
3866 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3867 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3868 if (sbi->s_want_extra_isize <
3869 le16_to_cpu(es->s_want_extra_isize))
3870 sbi->s_want_extra_isize =
3871 le16_to_cpu(es->s_want_extra_isize);
3872 if (sbi->s_want_extra_isize <
3873 le16_to_cpu(es->s_min_extra_isize))
3874 sbi->s_want_extra_isize =
3875 le16_to_cpu(es->s_min_extra_isize);
3878 /* Check if enough inode space is available */
3879 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3880 sbi->s_inode_size) {
3881 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3882 EXT4_GOOD_OLD_INODE_SIZE;
3883 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3887 err = ext4_setup_system_zone(sb);
3889 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3895 err = ext4_mb_init(sb, needs_recovery);
3897 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3902 err = ext4_register_li_request(sb, first_not_zeroed);
3906 sbi->s_kobj.kset = ext4_kset;
3907 init_completion(&sbi->s_kobj_unregister);
3908 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3913 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3914 ext4_orphan_cleanup(sb, es);
3915 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3916 if (needs_recovery) {
3917 ext4_msg(sb, KERN_INFO, "recovery complete");
3918 ext4_mark_recovery_complete(sb, es);
3920 if (EXT4_SB(sb)->s_journal) {
3921 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3922 descr = " journalled data mode";
3923 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3924 descr = " ordered data mode";
3926 descr = " writeback data mode";
3928 descr = "out journal";
3930 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3931 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3932 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3934 if (es->s_error_count)
3935 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3942 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3946 ext4_unregister_li_request(sb);
3948 ext4_ext_release(sb);
3950 ext4_mb_release(sb);
3951 ext4_release_system_zone(sb);
3955 ext4_msg(sb, KERN_ERR, "mount failed");
3956 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3958 if (sbi->s_journal) {
3959 jbd2_journal_destroy(sbi->s_journal);
3960 sbi->s_journal = NULL;
3963 del_timer_sync(&sbi->s_err_report);
3964 if (sbi->s_flex_groups)
3965 ext4_kvfree(sbi->s_flex_groups);
3966 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3967 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3968 percpu_counter_destroy(&sbi->s_dirs_counter);
3969 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3971 kthread_stop(sbi->s_mmp_tsk);
3973 for (i = 0; i < db_count; i++)
3974 brelse(sbi->s_group_desc[i]);
3975 ext4_kvfree(sbi->s_group_desc);
3978 remove_proc_entry(sb->s_id, ext4_proc_root);
3981 for (i = 0; i < MAXQUOTAS; i++)
3982 kfree(sbi->s_qf_names[i]);
3984 ext4_blkdev_remove(sbi);
3987 sb->s_fs_info = NULL;
3988 kfree(sbi->s_blockgroup_lock);
3996 * Setup any per-fs journal parameters now. We'll do this both on
3997 * initial mount, once the journal has been initialised but before we've
3998 * done any recovery; and again on any subsequent remount.
4000 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4002 struct ext4_sb_info *sbi = EXT4_SB(sb);
4004 journal->j_commit_interval = sbi->s_commit_interval;
4005 journal->j_min_batch_time = sbi->s_min_batch_time;
4006 journal->j_max_batch_time = sbi->s_max_batch_time;
4008 write_lock(&journal->j_state_lock);
4009 if (test_opt(sb, BARRIER))
4010 journal->j_flags |= JBD2_BARRIER;
4012 journal->j_flags &= ~JBD2_BARRIER;
4013 if (test_opt(sb, DATA_ERR_ABORT))
4014 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4016 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4017 write_unlock(&journal->j_state_lock);
4020 static journal_t *ext4_get_journal(struct super_block *sb,
4021 unsigned int journal_inum)
4023 struct inode *journal_inode;
4026 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4028 /* First, test for the existence of a valid inode on disk. Bad
4029 * things happen if we iget() an unused inode, as the subsequent
4030 * iput() will try to delete it. */
4032 journal_inode = ext4_iget(sb, journal_inum);
4033 if (IS_ERR(journal_inode)) {
4034 ext4_msg(sb, KERN_ERR, "no journal found");
4037 if (!journal_inode->i_nlink) {
4038 make_bad_inode(journal_inode);
4039 iput(journal_inode);
4040 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4044 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4045 journal_inode, journal_inode->i_size);
4046 if (!S_ISREG(journal_inode->i_mode)) {
4047 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4048 iput(journal_inode);
4052 journal = jbd2_journal_init_inode(journal_inode);
4054 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4055 iput(journal_inode);
4058 journal->j_private = sb;
4059 ext4_init_journal_params(sb, journal);
4063 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4066 struct buffer_head *bh;
4070 int hblock, blocksize;
4071 ext4_fsblk_t sb_block;
4072 unsigned long offset;
4073 struct ext4_super_block *es;
4074 struct block_device *bdev;
4076 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4078 bdev = ext4_blkdev_get(j_dev, sb);
4082 blocksize = sb->s_blocksize;
4083 hblock = bdev_logical_block_size(bdev);
4084 if (blocksize < hblock) {
4085 ext4_msg(sb, KERN_ERR,
4086 "blocksize too small for journal device");
4090 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4091 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4092 set_blocksize(bdev, blocksize);
4093 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4094 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4095 "external journal");
4099 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
4100 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4101 !(le32_to_cpu(es->s_feature_incompat) &
4102 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4103 ext4_msg(sb, KERN_ERR, "external journal has "
4109 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4110 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4115 len = ext4_blocks_count(es);
4116 start = sb_block + 1;
4117 brelse(bh); /* we're done with the superblock */
4119 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4120 start, len, blocksize);
4122 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4125 journal->j_private = sb;
4126 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4127 wait_on_buffer(journal->j_sb_buffer);
4128 if (!buffer_uptodate(journal->j_sb_buffer)) {
4129 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4132 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4133 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4134 "user (unsupported) - %d",
4135 be32_to_cpu(journal->j_superblock->s_nr_users));
4138 EXT4_SB(sb)->journal_bdev = bdev;
4139 ext4_init_journal_params(sb, journal);
4143 jbd2_journal_destroy(journal);
4145 ext4_blkdev_put(bdev);
4149 static int ext4_load_journal(struct super_block *sb,
4150 struct ext4_super_block *es,
4151 unsigned long journal_devnum)
4154 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4157 int really_read_only;
4159 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4161 if (journal_devnum &&
4162 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4163 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4164 "numbers have changed");
4165 journal_dev = new_decode_dev(journal_devnum);
4167 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4169 really_read_only = bdev_read_only(sb->s_bdev);
4172 * Are we loading a blank journal or performing recovery after a
4173 * crash? For recovery, we need to check in advance whether we
4174 * can get read-write access to the device.
4176 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4177 if (sb->s_flags & MS_RDONLY) {
4178 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4179 "required on readonly filesystem");
4180 if (really_read_only) {
4181 ext4_msg(sb, KERN_ERR, "write access "
4182 "unavailable, cannot proceed");
4185 ext4_msg(sb, KERN_INFO, "write access will "
4186 "be enabled during recovery");
4190 if (journal_inum && journal_dev) {
4191 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4192 "and inode journals!");
4197 if (!(journal = ext4_get_journal(sb, journal_inum)))
4200 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4204 if (!(journal->j_flags & JBD2_BARRIER))
4205 ext4_msg(sb, KERN_INFO, "barriers disabled");
4207 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4208 err = jbd2_journal_update_format(journal);
4210 ext4_msg(sb, KERN_ERR, "error updating journal");
4211 jbd2_journal_destroy(journal);
4216 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4217 err = jbd2_journal_wipe(journal, !really_read_only);
4219 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4221 memcpy(save, ((char *) es) +
4222 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4223 err = jbd2_journal_load(journal);
4225 memcpy(((char *) es) + EXT4_S_ERR_START,
4226 save, EXT4_S_ERR_LEN);
4231 ext4_msg(sb, KERN_ERR, "error loading journal");
4232 jbd2_journal_destroy(journal);
4236 EXT4_SB(sb)->s_journal = journal;
4237 ext4_clear_journal_err(sb, es);
4239 if (!really_read_only && journal_devnum &&
4240 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4241 es->s_journal_dev = cpu_to_le32(journal_devnum);
4243 /* Make sure we flush the recovery flag to disk. */
4244 ext4_commit_super(sb, 1);
4250 static int ext4_commit_super(struct super_block *sb, int sync)
4252 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4253 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4256 if (!sbh || block_device_ejected(sb))
4258 if (buffer_write_io_error(sbh)) {
4260 * Oh, dear. A previous attempt to write the
4261 * superblock failed. This could happen because the
4262 * USB device was yanked out. Or it could happen to
4263 * be a transient write error and maybe the block will
4264 * be remapped. Nothing we can do but to retry the
4265 * write and hope for the best.
4267 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4268 "superblock detected");
4269 clear_buffer_write_io_error(sbh);
4270 set_buffer_uptodate(sbh);
4273 * If the file system is mounted read-only, don't update the
4274 * superblock write time. This avoids updating the superblock
4275 * write time when we are mounting the root file system
4276 * read/only but we need to replay the journal; at that point,
4277 * for people who are east of GMT and who make their clock
4278 * tick in localtime for Windows bug-for-bug compatibility,
4279 * the clock is set in the future, and this will cause e2fsck
4280 * to complain and force a full file system check.
4282 if (!(sb->s_flags & MS_RDONLY))
4283 es->s_wtime = cpu_to_le32(get_seconds());
4284 if (sb->s_bdev->bd_part)
4285 es->s_kbytes_written =
4286 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4287 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4288 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4290 es->s_kbytes_written =
4291 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4292 ext4_free_blocks_count_set(es,
4293 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4294 &EXT4_SB(sb)->s_freeclusters_counter)));
4295 es->s_free_inodes_count =
4296 cpu_to_le32(percpu_counter_sum_positive(
4297 &EXT4_SB(sb)->s_freeinodes_counter));
4299 BUFFER_TRACE(sbh, "marking dirty");
4300 mark_buffer_dirty(sbh);
4302 error = sync_dirty_buffer(sbh);
4306 error = buffer_write_io_error(sbh);
4308 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4310 clear_buffer_write_io_error(sbh);
4311 set_buffer_uptodate(sbh);
4318 * Have we just finished recovery? If so, and if we are mounting (or
4319 * remounting) the filesystem readonly, then we will end up with a
4320 * consistent fs on disk. Record that fact.
4322 static void ext4_mark_recovery_complete(struct super_block *sb,
4323 struct ext4_super_block *es)
4325 journal_t *journal = EXT4_SB(sb)->s_journal;
4327 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4328 BUG_ON(journal != NULL);
4331 jbd2_journal_lock_updates(journal);
4332 if (jbd2_journal_flush(journal) < 0)
4335 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4336 sb->s_flags & MS_RDONLY) {
4337 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4338 ext4_commit_super(sb, 1);
4342 jbd2_journal_unlock_updates(journal);
4346 * If we are mounting (or read-write remounting) a filesystem whose journal
4347 * has recorded an error from a previous lifetime, move that error to the
4348 * main filesystem now.
4350 static void ext4_clear_journal_err(struct super_block *sb,
4351 struct ext4_super_block *es)
4357 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4359 journal = EXT4_SB(sb)->s_journal;
4362 * Now check for any error status which may have been recorded in the
4363 * journal by a prior ext4_error() or ext4_abort()
4366 j_errno = jbd2_journal_errno(journal);
4370 errstr = ext4_decode_error(sb, j_errno, nbuf);
4371 ext4_warning(sb, "Filesystem error recorded "
4372 "from previous mount: %s", errstr);
4373 ext4_warning(sb, "Marking fs in need of filesystem check.");
4375 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4376 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4377 ext4_commit_super(sb, 1);
4379 jbd2_journal_clear_err(journal);
4384 * Force the running and committing transactions to commit,
4385 * and wait on the commit.
4387 int ext4_force_commit(struct super_block *sb)
4392 if (sb->s_flags & MS_RDONLY)
4395 journal = EXT4_SB(sb)->s_journal;
4397 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4398 ret = ext4_journal_force_commit(journal);
4404 static void ext4_write_super(struct super_block *sb)
4407 ext4_commit_super(sb, 1);
4411 static int ext4_sync_fs(struct super_block *sb, int wait)
4415 struct ext4_sb_info *sbi = EXT4_SB(sb);
4417 trace_ext4_sync_fs(sb, wait);
4418 flush_workqueue(sbi->dio_unwritten_wq);
4419 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4421 jbd2_log_wait_commit(sbi->s_journal, target);
4427 * LVM calls this function before a (read-only) snapshot is created. This
4428 * gives us a chance to flush the journal completely and mark the fs clean.
4430 * Note that only this function cannot bring a filesystem to be in a clean
4431 * state independently, because ext4 prevents a new handle from being started
4432 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4435 static int ext4_freeze(struct super_block *sb)
4440 if (sb->s_flags & MS_RDONLY)
4443 journal = EXT4_SB(sb)->s_journal;
4445 /* Now we set up the journal barrier. */
4446 jbd2_journal_lock_updates(journal);
4449 * Don't clear the needs_recovery flag if we failed to flush
4452 error = jbd2_journal_flush(journal);
4456 /* Journal blocked and flushed, clear needs_recovery flag. */
4457 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4458 error = ext4_commit_super(sb, 1);
4460 /* we rely on s_frozen to stop further updates */
4461 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4466 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4467 * flag here, even though the filesystem is not technically dirty yet.
4469 static int ext4_unfreeze(struct super_block *sb)
4471 if (sb->s_flags & MS_RDONLY)
4475 /* Reset the needs_recovery flag before the fs is unlocked. */
4476 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4477 ext4_commit_super(sb, 1);
4483 * Structure to save mount options for ext4_remount's benefit
4485 struct ext4_mount_options {
4486 unsigned long s_mount_opt;
4487 unsigned long s_mount_opt2;
4490 unsigned long s_commit_interval;
4491 u32 s_min_batch_time, s_max_batch_time;
4494 char *s_qf_names[MAXQUOTAS];
4498 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4500 struct ext4_super_block *es;
4501 struct ext4_sb_info *sbi = EXT4_SB(sb);
4502 ext4_fsblk_t n_blocks_count = 0;
4503 unsigned long old_sb_flags;
4504 struct ext4_mount_options old_opts;
4505 int enable_quota = 0;
4507 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4512 char *orig_data = kstrdup(data, GFP_KERNEL);
4514 /* Store the original options */
4516 old_sb_flags = sb->s_flags;
4517 old_opts.s_mount_opt = sbi->s_mount_opt;
4518 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4519 old_opts.s_resuid = sbi->s_resuid;
4520 old_opts.s_resgid = sbi->s_resgid;
4521 old_opts.s_commit_interval = sbi->s_commit_interval;
4522 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4523 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4525 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4526 for (i = 0; i < MAXQUOTAS; i++)
4527 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4529 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4530 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4533 * Allow the "check" option to be passed as a remount option.
4535 if (!parse_options(data, sb, NULL, &journal_ioprio,
4536 &n_blocks_count, 1)) {
4541 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4542 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4543 ext4_msg(sb, KERN_ERR, "can't mount with "
4544 "both data=journal and delalloc");
4548 if (test_opt(sb, DIOREAD_NOLOCK)) {
4549 ext4_msg(sb, KERN_ERR, "can't mount with "
4550 "both data=journal and dioread_nolock");
4556 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4557 ext4_abort(sb, "Abort forced by user");
4559 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4560 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4564 if (sbi->s_journal) {
4565 ext4_init_journal_params(sb, sbi->s_journal);
4566 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4569 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4570 n_blocks_count > ext4_blocks_count(es)) {
4571 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4576 if (*flags & MS_RDONLY) {
4577 err = dquot_suspend(sb, -1);
4582 * First of all, the unconditional stuff we have to do
4583 * to disable replay of the journal when we next remount
4585 sb->s_flags |= MS_RDONLY;
4588 * OK, test if we are remounting a valid rw partition
4589 * readonly, and if so set the rdonly flag and then
4590 * mark the partition as valid again.
4592 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4593 (sbi->s_mount_state & EXT4_VALID_FS))
4594 es->s_state = cpu_to_le16(sbi->s_mount_state);
4597 ext4_mark_recovery_complete(sb, es);
4599 /* Make sure we can mount this feature set readwrite */
4600 if (!ext4_feature_set_ok(sb, 0)) {
4605 * Make sure the group descriptor checksums
4606 * are sane. If they aren't, refuse to remount r/w.
4608 for (g = 0; g < sbi->s_groups_count; g++) {
4609 struct ext4_group_desc *gdp =
4610 ext4_get_group_desc(sb, g, NULL);
4612 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4613 ext4_msg(sb, KERN_ERR,
4614 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4615 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4616 le16_to_cpu(gdp->bg_checksum));
4623 * If we have an unprocessed orphan list hanging
4624 * around from a previously readonly bdev mount,
4625 * require a full umount/remount for now.
4627 if (es->s_last_orphan) {
4628 ext4_msg(sb, KERN_WARNING, "Couldn't "
4629 "remount RDWR because of unprocessed "
4630 "orphan inode list. Please "
4631 "umount/remount instead");
4637 * Mounting a RDONLY partition read-write, so reread
4638 * and store the current valid flag. (It may have
4639 * been changed by e2fsck since we originally mounted
4643 ext4_clear_journal_err(sb, es);
4644 sbi->s_mount_state = le16_to_cpu(es->s_state);
4645 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4647 if (!ext4_setup_super(sb, es, 0))
4648 sb->s_flags &= ~MS_RDONLY;
4649 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4650 EXT4_FEATURE_INCOMPAT_MMP))
4651 if (ext4_multi_mount_protect(sb,
4652 le64_to_cpu(es->s_mmp_block))) {
4661 * Reinitialize lazy itable initialization thread based on
4664 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4665 ext4_unregister_li_request(sb);
4667 ext4_group_t first_not_zeroed;
4668 first_not_zeroed = ext4_has_uninit_itable(sb);
4669 ext4_register_li_request(sb, first_not_zeroed);
4672 ext4_setup_system_zone(sb);
4673 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4674 ext4_commit_super(sb, 1);
4677 /* Release old quota file names */
4678 for (i = 0; i < MAXQUOTAS; i++)
4679 if (old_opts.s_qf_names[i] &&
4680 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4681 kfree(old_opts.s_qf_names[i]);
4685 dquot_resume(sb, -1);
4687 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4692 sb->s_flags = old_sb_flags;
4693 sbi->s_mount_opt = old_opts.s_mount_opt;
4694 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4695 sbi->s_resuid = old_opts.s_resuid;
4696 sbi->s_resgid = old_opts.s_resgid;
4697 sbi->s_commit_interval = old_opts.s_commit_interval;
4698 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4699 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4701 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4702 for (i = 0; i < MAXQUOTAS; i++) {
4703 if (sbi->s_qf_names[i] &&
4704 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4705 kfree(sbi->s_qf_names[i]);
4706 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4714 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4716 struct super_block *sb = dentry->d_sb;
4717 struct ext4_sb_info *sbi = EXT4_SB(sb);
4718 struct ext4_super_block *es = sbi->s_es;
4719 ext4_fsblk_t overhead = 0;
4723 if (!test_opt(sb, MINIX_DF))
4724 overhead = sbi->s_overhead;
4726 buf->f_type = EXT4_SUPER_MAGIC;
4727 buf->f_bsize = sb->s_blocksize;
4728 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4729 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4730 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4731 /* prevent underflow in case that few free space is available */
4732 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4733 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4734 if (buf->f_bfree < ext4_r_blocks_count(es))
4736 buf->f_files = le32_to_cpu(es->s_inodes_count);
4737 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4738 buf->f_namelen = EXT4_NAME_LEN;
4739 fsid = le64_to_cpup((void *)es->s_uuid) ^
4740 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4741 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4742 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4747 /* Helper function for writing quotas on sync - we need to start transaction
4748 * before quota file is locked for write. Otherwise the are possible deadlocks:
4749 * Process 1 Process 2
4750 * ext4_create() quota_sync()
4751 * jbd2_journal_start() write_dquot()
4752 * dquot_initialize() down(dqio_mutex)
4753 * down(dqio_mutex) jbd2_journal_start()
4759 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4761 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4764 static int ext4_write_dquot(struct dquot *dquot)
4768 struct inode *inode;
4770 inode = dquot_to_inode(dquot);
4771 handle = ext4_journal_start(inode,
4772 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4774 return PTR_ERR(handle);
4775 ret = dquot_commit(dquot);
4776 err = ext4_journal_stop(handle);
4782 static int ext4_acquire_dquot(struct dquot *dquot)
4787 handle = ext4_journal_start(dquot_to_inode(dquot),
4788 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4790 return PTR_ERR(handle);
4791 ret = dquot_acquire(dquot);
4792 err = ext4_journal_stop(handle);
4798 static int ext4_release_dquot(struct dquot *dquot)
4803 handle = ext4_journal_start(dquot_to_inode(dquot),
4804 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4805 if (IS_ERR(handle)) {
4806 /* Release dquot anyway to avoid endless cycle in dqput() */
4807 dquot_release(dquot);
4808 return PTR_ERR(handle);
4810 ret = dquot_release(dquot);
4811 err = ext4_journal_stop(handle);
4817 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4819 /* Are we journaling quotas? */
4820 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4821 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4822 dquot_mark_dquot_dirty(dquot);
4823 return ext4_write_dquot(dquot);
4825 return dquot_mark_dquot_dirty(dquot);
4829 static int ext4_write_info(struct super_block *sb, int type)
4834 /* Data block + inode block */
4835 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4837 return PTR_ERR(handle);
4838 ret = dquot_commit_info(sb, type);
4839 err = ext4_journal_stop(handle);
4846 * Turn on quotas during mount time - we need to find
4847 * the quota file and such...
4849 static int ext4_quota_on_mount(struct super_block *sb, int type)
4851 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4852 EXT4_SB(sb)->s_jquota_fmt, type);
4856 * Standard function to be called on quota_on
4858 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4863 if (!test_opt(sb, QUOTA))
4866 /* Quotafile not on the same filesystem? */
4867 if (path->mnt->mnt_sb != sb)
4869 /* Journaling quota? */
4870 if (EXT4_SB(sb)->s_qf_names[type]) {
4871 /* Quotafile not in fs root? */
4872 if (path->dentry->d_parent != sb->s_root)
4873 ext4_msg(sb, KERN_WARNING,
4874 "Quota file not on filesystem root. "
4875 "Journaled quota will not work");
4879 * When we journal data on quota file, we have to flush journal to see
4880 * all updates to the file when we bypass pagecache...
4882 if (EXT4_SB(sb)->s_journal &&
4883 ext4_should_journal_data(path->dentry->d_inode)) {
4885 * We don't need to lock updates but journal_flush() could
4886 * otherwise be livelocked...
4888 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4889 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4890 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4895 return dquot_quota_on(sb, type, format_id, path);
4898 static int ext4_quota_off(struct super_block *sb, int type)
4900 struct inode *inode = sb_dqopt(sb)->files[type];
4903 /* Force all delayed allocation blocks to be allocated.
4904 * Caller already holds s_umount sem */
4905 if (test_opt(sb, DELALLOC))
4906 sync_filesystem(sb);
4911 /* Update modification times of quota files when userspace can
4912 * start looking at them */
4913 handle = ext4_journal_start(inode, 1);
4916 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4917 ext4_mark_inode_dirty(handle, inode);
4918 ext4_journal_stop(handle);
4921 return dquot_quota_off(sb, type);
4924 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4925 * acquiring the locks... As quota files are never truncated and quota code
4926 * itself serializes the operations (and no one else should touch the files)
4927 * we don't have to be afraid of races */
4928 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4929 size_t len, loff_t off)
4931 struct inode *inode = sb_dqopt(sb)->files[type];
4932 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4934 int offset = off & (sb->s_blocksize - 1);
4937 struct buffer_head *bh;
4938 loff_t i_size = i_size_read(inode);
4942 if (off+len > i_size)
4945 while (toread > 0) {
4946 tocopy = sb->s_blocksize - offset < toread ?
4947 sb->s_blocksize - offset : toread;
4948 bh = ext4_bread(NULL, inode, blk, 0, &err);
4951 if (!bh) /* A hole? */
4952 memset(data, 0, tocopy);
4954 memcpy(data, bh->b_data+offset, tocopy);
4964 /* Write to quotafile (we know the transaction is already started and has
4965 * enough credits) */
4966 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4967 const char *data, size_t len, loff_t off)
4969 struct inode *inode = sb_dqopt(sb)->files[type];
4970 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4972 int offset = off & (sb->s_blocksize - 1);
4973 struct buffer_head *bh;
4974 handle_t *handle = journal_current_handle();
4976 if (EXT4_SB(sb)->s_journal && !handle) {
4977 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4978 " cancelled because transaction is not started",
4979 (unsigned long long)off, (unsigned long long)len);
4983 * Since we account only one data block in transaction credits,
4984 * then it is impossible to cross a block boundary.
4986 if (sb->s_blocksize - offset < len) {
4987 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4988 " cancelled because not block aligned",
4989 (unsigned long long)off, (unsigned long long)len);
4993 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4994 bh = ext4_bread(handle, inode, blk, 1, &err);
4997 err = ext4_journal_get_write_access(handle, bh);
5003 memcpy(bh->b_data+offset, data, len);
5004 flush_dcache_page(bh->b_page);
5006 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5010 mutex_unlock(&inode->i_mutex);
5013 if (inode->i_size < off + len) {
5014 i_size_write(inode, off + len);
5015 EXT4_I(inode)->i_disksize = inode->i_size;
5016 ext4_mark_inode_dirty(handle, inode);
5018 mutex_unlock(&inode->i_mutex);
5024 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5025 const char *dev_name, void *data)
5027 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5030 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5031 static inline void register_as_ext2(void)
5033 int err = register_filesystem(&ext2_fs_type);
5036 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5039 static inline void unregister_as_ext2(void)
5041 unregister_filesystem(&ext2_fs_type);
5044 static inline int ext2_feature_set_ok(struct super_block *sb)
5046 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5048 if (sb->s_flags & MS_RDONLY)
5050 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5054 MODULE_ALIAS("ext2");
5056 static inline void register_as_ext2(void) { }
5057 static inline void unregister_as_ext2(void) { }
5058 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5061 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5062 static inline void register_as_ext3(void)
5064 int err = register_filesystem(&ext3_fs_type);
5067 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5070 static inline void unregister_as_ext3(void)
5072 unregister_filesystem(&ext3_fs_type);
5075 static inline int ext3_feature_set_ok(struct super_block *sb)
5077 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5079 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5081 if (sb->s_flags & MS_RDONLY)
5083 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5087 MODULE_ALIAS("ext3");
5089 static inline void register_as_ext3(void) { }
5090 static inline void unregister_as_ext3(void) { }
5091 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5094 static struct file_system_type ext4_fs_type = {
5095 .owner = THIS_MODULE,
5097 .mount = ext4_mount,
5098 .kill_sb = kill_block_super,
5099 .fs_flags = FS_REQUIRES_DEV,
5102 static int __init ext4_init_feat_adverts(void)
5104 struct ext4_features *ef;
5107 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5111 ef->f_kobj.kset = ext4_kset;
5112 init_completion(&ef->f_kobj_unregister);
5113 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5126 static void ext4_exit_feat_adverts(void)
5128 kobject_put(&ext4_feat->f_kobj);
5129 wait_for_completion(&ext4_feat->f_kobj_unregister);
5133 /* Shared across all ext4 file systems */
5134 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5135 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5137 static int __init ext4_init_fs(void)
5141 ext4_check_flag_values();
5143 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5144 mutex_init(&ext4__aio_mutex[i]);
5145 init_waitqueue_head(&ext4__ioend_wq[i]);
5148 err = ext4_init_pageio();
5151 err = ext4_init_system_zone();
5154 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5157 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5159 err = ext4_init_feat_adverts();
5163 err = ext4_init_mballoc();
5167 err = ext4_init_xattr();
5170 err = init_inodecache();
5175 err = register_filesystem(&ext4_fs_type);
5179 ext4_li_info = NULL;
5180 mutex_init(&ext4_li_mtx);
5183 unregister_as_ext2();
5184 unregister_as_ext3();
5185 destroy_inodecache();
5189 ext4_exit_mballoc();
5191 ext4_exit_feat_adverts();
5194 remove_proc_entry("fs/ext4", NULL);
5195 kset_unregister(ext4_kset);
5197 ext4_exit_system_zone();
5203 static void __exit ext4_exit_fs(void)
5205 ext4_destroy_lazyinit_thread();
5206 unregister_as_ext2();
5207 unregister_as_ext3();
5208 unregister_filesystem(&ext4_fs_type);
5209 destroy_inodecache();
5211 ext4_exit_mballoc();
5212 ext4_exit_feat_adverts();
5213 remove_proc_entry("fs/ext4", NULL);
5214 kset_unregister(ext4_kset);
5215 ext4_exit_system_zone();
5219 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5220 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5221 MODULE_LICENSE("GPL");
5222 module_init(ext4_init_fs)
5223 module_exit(ext4_exit_fs)