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_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static void ext4_write_super(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #define IS_EXT2_SB(sb) (0)
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
112 #define IS_EXT3_SB(sb) (0)
115 void *ext4_kvmalloc(size_t size, gfp_t flags)
119 ret = kmalloc(size, flags);
121 ret = __vmalloc(size, flags, PAGE_KERNEL);
125 void *ext4_kvzalloc(size_t size, gfp_t flags)
129 ret = kzalloc(size, flags);
131 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
135 void ext4_kvfree(void *ptr)
137 if (is_vmalloc_addr(ptr))
144 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
145 struct ext4_group_desc *bg)
147 return le32_to_cpu(bg->bg_block_bitmap_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
149 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
152 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
153 struct ext4_group_desc *bg)
155 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
156 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
157 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
160 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
161 struct ext4_group_desc *bg)
163 return le32_to_cpu(bg->bg_inode_table_lo) |
164 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
165 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
168 __u32 ext4_free_group_clusters(struct super_block *sb,
169 struct ext4_group_desc *bg)
171 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
172 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
173 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
176 __u32 ext4_free_inodes_count(struct super_block *sb,
177 struct ext4_group_desc *bg)
179 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
180 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
181 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
184 __u32 ext4_used_dirs_count(struct super_block *sb,
185 struct ext4_group_desc *bg)
187 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
189 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
192 __u32 ext4_itable_unused_count(struct super_block *sb,
193 struct ext4_group_desc *bg)
195 return le16_to_cpu(bg->bg_itable_unused_lo) |
196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
197 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
200 void ext4_block_bitmap_set(struct super_block *sb,
201 struct ext4_group_desc *bg, ext4_fsblk_t blk)
203 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
208 void ext4_inode_bitmap_set(struct super_block *sb,
209 struct ext4_group_desc *bg, ext4_fsblk_t blk)
211 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
212 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
213 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
216 void ext4_inode_table_set(struct super_block *sb,
217 struct ext4_group_desc *bg, ext4_fsblk_t blk)
219 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
220 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
221 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
224 void ext4_free_group_clusters_set(struct super_block *sb,
225 struct ext4_group_desc *bg, __u32 count)
227 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
228 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
229 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
232 void ext4_free_inodes_set(struct super_block *sb,
233 struct ext4_group_desc *bg, __u32 count)
235 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
236 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
237 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
240 void ext4_used_dirs_set(struct super_block *sb,
241 struct ext4_group_desc *bg, __u32 count)
243 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
248 void ext4_itable_unused_set(struct super_block *sb,
249 struct ext4_group_desc *bg, __u32 count)
251 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
257 /* Just increment the non-pointer handle value */
258 static handle_t *ext4_get_nojournal(void)
260 handle_t *handle = current->journal_info;
261 unsigned long ref_cnt = (unsigned long)handle;
263 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
266 handle = (handle_t *)ref_cnt;
268 current->journal_info = handle;
273 /* Decrement the non-pointer handle value */
274 static void ext4_put_nojournal(handle_t *handle)
276 unsigned long ref_cnt = (unsigned long)handle;
278 BUG_ON(ref_cnt == 0);
281 handle = (handle_t *)ref_cnt;
283 current->journal_info = handle;
287 * Wrappers for jbd2_journal_start/end.
289 * The only special thing we need to do here is to make sure that all
290 * journal_end calls result in the superblock being marked dirty, so
291 * that sync() will call the filesystem's write_super callback if
294 * To avoid j_barrier hold in userspace when a user calls freeze(),
295 * ext4 prevents a new handle from being started by s_frozen, which
296 * is in an upper layer.
298 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
303 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
304 if (sb->s_flags & MS_RDONLY)
305 return ERR_PTR(-EROFS);
307 journal = EXT4_SB(sb)->s_journal;
308 handle = ext4_journal_current_handle();
311 * If a handle has been started, it should be allowed to
312 * finish, otherwise deadlock could happen between freeze
313 * and others(e.g. truncate) due to the restart of the
314 * journal handle if the filesystem is forzen and active
315 * handles are not stopped.
318 vfs_check_frozen(sb, SB_FREEZE_TRANS);
321 return ext4_get_nojournal();
323 * Special case here: if the journal has aborted behind our
324 * backs (eg. EIO in the commit thread), then we still need to
325 * take the FS itself readonly cleanly.
327 if (is_journal_aborted(journal)) {
328 ext4_abort(sb, "Detected aborted journal");
329 return ERR_PTR(-EROFS);
331 return jbd2_journal_start(journal, nblocks);
335 * The only special thing we need to do here is to make sure that all
336 * jbd2_journal_stop calls result in the superblock being marked dirty, so
337 * that sync() will call the filesystem's write_super callback if
340 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
342 struct super_block *sb;
346 if (!ext4_handle_valid(handle)) {
347 ext4_put_nojournal(handle);
350 sb = handle->h_transaction->t_journal->j_private;
352 rc = jbd2_journal_stop(handle);
357 __ext4_std_error(sb, where, line, err);
361 void ext4_journal_abort_handle(const char *caller, unsigned int line,
362 const char *err_fn, struct buffer_head *bh,
363 handle_t *handle, int err)
366 const char *errstr = ext4_decode_error(NULL, err, nbuf);
368 BUG_ON(!ext4_handle_valid(handle));
371 BUFFER_TRACE(bh, "abort");
376 if (is_handle_aborted(handle))
379 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
380 caller, line, errstr, err_fn);
382 jbd2_journal_abort_handle(handle);
385 static void __save_error_info(struct super_block *sb, const char *func,
388 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
390 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
391 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
392 es->s_last_error_time = cpu_to_le32(get_seconds());
393 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
394 es->s_last_error_line = cpu_to_le32(line);
395 if (!es->s_first_error_time) {
396 es->s_first_error_time = es->s_last_error_time;
397 strncpy(es->s_first_error_func, func,
398 sizeof(es->s_first_error_func));
399 es->s_first_error_line = cpu_to_le32(line);
400 es->s_first_error_ino = es->s_last_error_ino;
401 es->s_first_error_block = es->s_last_error_block;
404 * Start the daily error reporting function if it hasn't been
407 if (!es->s_error_count)
408 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
409 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
412 static void save_error_info(struct super_block *sb, const char *func,
415 __save_error_info(sb, func, line);
416 ext4_commit_super(sb, 1);
420 * The del_gendisk() function uninitializes the disk-specific data
421 * structures, including the bdi structure, without telling anyone
422 * else. Once this happens, any attempt to call mark_buffer_dirty()
423 * (for example, by ext4_commit_super), will cause a kernel OOPS.
424 * This is a kludge to prevent these oops until we can put in a proper
425 * hook in del_gendisk() to inform the VFS and file system layers.
427 static int block_device_ejected(struct super_block *sb)
429 struct inode *bd_inode = sb->s_bdev->bd_inode;
430 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
432 return bdi->dev == NULL;
435 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
437 struct super_block *sb = journal->j_private;
438 struct ext4_sb_info *sbi = EXT4_SB(sb);
439 int error = is_journal_aborted(journal);
440 struct ext4_journal_cb_entry *jce, *tmp;
442 spin_lock(&sbi->s_md_lock);
443 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
444 list_del_init(&jce->jce_list);
445 spin_unlock(&sbi->s_md_lock);
446 jce->jce_func(sb, jce, error);
447 spin_lock(&sbi->s_md_lock);
449 spin_unlock(&sbi->s_md_lock);
452 /* Deal with the reporting of failure conditions on a filesystem such as
453 * inconsistencies detected or read IO failures.
455 * On ext2, we can store the error state of the filesystem in the
456 * superblock. That is not possible on ext4, because we may have other
457 * write ordering constraints on the superblock which prevent us from
458 * writing it out straight away; and given that the journal is about to
459 * be aborted, we can't rely on the current, or future, transactions to
460 * write out the superblock safely.
462 * We'll just use the jbd2_journal_abort() error code to record an error in
463 * the journal instead. On recovery, the journal will complain about
464 * that error until we've noted it down and cleared it.
467 static void ext4_handle_error(struct super_block *sb)
469 if (sb->s_flags & MS_RDONLY)
472 if (!test_opt(sb, ERRORS_CONT)) {
473 journal_t *journal = EXT4_SB(sb)->s_journal;
475 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
477 jbd2_journal_abort(journal, -EIO);
479 if (test_opt(sb, ERRORS_RO)) {
480 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
481 sb->s_flags |= MS_RDONLY;
483 if (test_opt(sb, ERRORS_PANIC))
484 panic("EXT4-fs (device %s): panic forced after error\n",
488 void __ext4_error(struct super_block *sb, const char *function,
489 unsigned int line, const char *fmt, ...)
491 struct va_format vaf;
497 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
498 sb->s_id, function, line, current->comm, &vaf);
501 ext4_handle_error(sb);
504 void ext4_error_inode(struct inode *inode, const char *function,
505 unsigned int line, ext4_fsblk_t block,
506 const char *fmt, ...)
509 struct va_format vaf;
510 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
512 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
513 es->s_last_error_block = cpu_to_le64(block);
514 save_error_info(inode->i_sb, function, line);
519 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
520 "inode #%lu: block %llu: comm %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, &vaf);
524 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
525 "inode #%lu: comm %s: %pV\n",
526 inode->i_sb->s_id, function, line, inode->i_ino,
527 current->comm, &vaf);
530 ext4_handle_error(inode->i_sb);
533 void ext4_error_file(struct file *file, const char *function,
534 unsigned int line, ext4_fsblk_t block,
535 const char *fmt, ...)
538 struct va_format vaf;
539 struct ext4_super_block *es;
540 struct inode *inode = file->f_dentry->d_inode;
541 char pathname[80], *path;
543 es = EXT4_SB(inode->i_sb)->s_es;
544 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
545 save_error_info(inode->i_sb, function, line);
546 path = d_path(&(file->f_path), pathname, sizeof(pathname));
554 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
555 "block %llu: comm %s: path %s: %pV\n",
556 inode->i_sb->s_id, function, line, inode->i_ino,
557 block, current->comm, path, &vaf);
560 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
561 "comm %s: path %s: %pV\n",
562 inode->i_sb->s_id, function, line, inode->i_ino,
563 current->comm, path, &vaf);
566 ext4_handle_error(inode->i_sb);
569 static const char *ext4_decode_error(struct super_block *sb, int errno,
576 errstr = "IO failure";
579 errstr = "Out of memory";
582 if (!sb || (EXT4_SB(sb)->s_journal &&
583 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
584 errstr = "Journal has aborted";
586 errstr = "Readonly filesystem";
589 /* If the caller passed in an extra buffer for unknown
590 * errors, textualise them now. Else we just return
593 /* Check for truncated error codes... */
594 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
603 /* __ext4_std_error decodes expected errors from journaling functions
604 * automatically and invokes the appropriate error response. */
606 void __ext4_std_error(struct super_block *sb, const char *function,
607 unsigned int line, int errno)
612 /* Special case: if the error is EROFS, and we're not already
613 * inside a transaction, then there's really no point in logging
615 if (errno == -EROFS && journal_current_handle() == NULL &&
616 (sb->s_flags & MS_RDONLY))
619 errstr = ext4_decode_error(sb, errno, nbuf);
620 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
621 sb->s_id, function, line, errstr);
622 save_error_info(sb, function, line);
624 ext4_handle_error(sb);
628 * ext4_abort is a much stronger failure handler than ext4_error. The
629 * abort function may be used to deal with unrecoverable failures such
630 * as journal IO errors or ENOMEM at a critical moment in log management.
632 * We unconditionally force the filesystem into an ABORT|READONLY state,
633 * unless the error response on the fs has been set to panic in which
634 * case we take the easy way out and panic immediately.
637 void __ext4_abort(struct super_block *sb, const char *function,
638 unsigned int line, const char *fmt, ...)
642 save_error_info(sb, function, line);
644 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
650 if ((sb->s_flags & MS_RDONLY) == 0) {
651 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
652 sb->s_flags |= MS_RDONLY;
653 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
654 if (EXT4_SB(sb)->s_journal)
655 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
656 save_error_info(sb, function, line);
658 if (test_opt(sb, ERRORS_PANIC))
659 panic("EXT4-fs panic from previous error\n");
662 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
664 struct va_format vaf;
670 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
674 void __ext4_warning(struct super_block *sb, const char *function,
675 unsigned int line, const char *fmt, ...)
677 struct va_format vaf;
683 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
684 sb->s_id, function, line, &vaf);
688 void __ext4_grp_locked_error(const char *function, unsigned int line,
689 struct super_block *sb, ext4_group_t grp,
690 unsigned long ino, ext4_fsblk_t block,
691 const char *fmt, ...)
695 struct va_format vaf;
697 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
699 es->s_last_error_ino = cpu_to_le32(ino);
700 es->s_last_error_block = cpu_to_le64(block);
701 __save_error_info(sb, function, line);
707 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
708 sb->s_id, function, line, grp);
710 printk(KERN_CONT "inode %lu: ", ino);
712 printk(KERN_CONT "block %llu:", (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
716 if (test_opt(sb, ERRORS_CONT)) {
717 ext4_commit_super(sb, 0);
721 ext4_unlock_group(sb, grp);
722 ext4_handle_error(sb);
724 * We only get here in the ERRORS_RO case; relocking the group
725 * may be dangerous, but nothing bad will happen since the
726 * filesystem will have already been marked read/only and the
727 * journal has been aborted. We return 1 as a hint to callers
728 * who might what to use the return value from
729 * ext4_grp_locked_error() to distinguish between the
730 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
731 * aggressively from the ext4 function in question, with a
732 * more appropriate error code.
734 ext4_lock_group(sb, grp);
738 void ext4_update_dynamic_rev(struct super_block *sb)
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
742 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
746 "updating to rev %d because of new feature flag, "
747 "running e2fsck is recommended",
750 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
751 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
752 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
753 /* leave es->s_feature_*compat flags alone */
754 /* es->s_uuid will be set by e2fsck if empty */
757 * The rest of the superblock fields should be zero, and if not it
758 * means they are likely already in use, so leave them alone. We
759 * can leave it up to e2fsck to clean up any inconsistencies there.
764 * Open the external journal device
766 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
768 struct block_device *bdev;
769 char b[BDEVNAME_SIZE];
771 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
777 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
778 __bdevname(dev, b), PTR_ERR(bdev));
783 * Release the journal device
785 static int ext4_blkdev_put(struct block_device *bdev)
787 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
790 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
792 struct block_device *bdev;
795 bdev = sbi->journal_bdev;
797 ret = ext4_blkdev_put(bdev);
798 sbi->journal_bdev = NULL;
803 static inline struct inode *orphan_list_entry(struct list_head *l)
805 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
808 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
812 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
813 le32_to_cpu(sbi->s_es->s_last_orphan));
815 printk(KERN_ERR "sb_info orphan list:\n");
816 list_for_each(l, &sbi->s_orphan) {
817 struct inode *inode = orphan_list_entry(l);
819 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
820 inode->i_sb->s_id, inode->i_ino, inode,
821 inode->i_mode, inode->i_nlink,
826 static void ext4_put_super(struct super_block *sb)
828 struct ext4_sb_info *sbi = EXT4_SB(sb);
829 struct ext4_super_block *es = sbi->s_es;
832 ext4_unregister_li_request(sb);
833 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
835 flush_workqueue(sbi->dio_unwritten_wq);
836 destroy_workqueue(sbi->dio_unwritten_wq);
839 if (sbi->s_journal) {
840 err = jbd2_journal_destroy(sbi->s_journal);
841 sbi->s_journal = NULL;
843 ext4_abort(sb, "Couldn't clean up the journal");
846 del_timer(&sbi->s_err_report);
847 ext4_release_system_zone(sb);
849 ext4_ext_release(sb);
850 ext4_xattr_put_super(sb);
852 if (!(sb->s_flags & MS_RDONLY)) {
853 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
854 es->s_state = cpu_to_le16(sbi->s_mount_state);
856 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
857 ext4_commit_super(sb, 1);
860 remove_proc_entry("options", sbi->s_proc);
861 remove_proc_entry(sb->s_id, ext4_proc_root);
863 kobject_del(&sbi->s_kobj);
865 for (i = 0; i < sbi->s_gdb_count; i++)
866 brelse(sbi->s_group_desc[i]);
867 ext4_kvfree(sbi->s_group_desc);
868 ext4_kvfree(sbi->s_flex_groups);
869 percpu_counter_destroy(&sbi->s_freeclusters_counter);
870 percpu_counter_destroy(&sbi->s_freeinodes_counter);
871 percpu_counter_destroy(&sbi->s_dirs_counter);
872 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
875 for (i = 0; i < MAXQUOTAS; i++)
876 kfree(sbi->s_qf_names[i]);
879 /* Debugging code just in case the in-memory inode orphan list
880 * isn't empty. The on-disk one can be non-empty if we've
881 * detected an error and taken the fs readonly, but the
882 * in-memory list had better be clean by this point. */
883 if (!list_empty(&sbi->s_orphan))
884 dump_orphan_list(sb, sbi);
885 J_ASSERT(list_empty(&sbi->s_orphan));
887 invalidate_bdev(sb->s_bdev);
888 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
890 * Invalidate the journal device's buffers. We don't want them
891 * floating about in memory - the physical journal device may
892 * hotswapped, and it breaks the `ro-after' testing code.
894 sync_blockdev(sbi->journal_bdev);
895 invalidate_bdev(sbi->journal_bdev);
896 ext4_blkdev_remove(sbi);
899 kthread_stop(sbi->s_mmp_tsk);
900 sb->s_fs_info = NULL;
902 * Now that we are completely done shutting down the
903 * superblock, we need to actually destroy the kobject.
906 kobject_put(&sbi->s_kobj);
907 wait_for_completion(&sbi->s_kobj_unregister);
908 kfree(sbi->s_blockgroup_lock);
912 static struct kmem_cache *ext4_inode_cachep;
915 * Called inside transaction, so use GFP_NOFS
917 static struct inode *ext4_alloc_inode(struct super_block *sb)
919 struct ext4_inode_info *ei;
921 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
925 ei->vfs_inode.i_version = 1;
926 ei->vfs_inode.i_data.writeback_index = 0;
927 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
928 INIT_LIST_HEAD(&ei->i_prealloc_list);
929 spin_lock_init(&ei->i_prealloc_lock);
930 ei->i_reserved_data_blocks = 0;
931 ei->i_reserved_meta_blocks = 0;
932 ei->i_allocated_meta_blocks = 0;
933 ei->i_da_metadata_calc_len = 0;
934 spin_lock_init(&(ei->i_block_reservation_lock));
936 ei->i_reserved_quota = 0;
939 INIT_LIST_HEAD(&ei->i_completed_io_list);
940 spin_lock_init(&ei->i_completed_io_lock);
941 ei->cur_aio_dio = NULL;
943 ei->i_datasync_tid = 0;
944 atomic_set(&ei->i_ioend_count, 0);
945 atomic_set(&ei->i_aiodio_unwritten, 0);
947 return &ei->vfs_inode;
950 static int ext4_drop_inode(struct inode *inode)
952 int drop = generic_drop_inode(inode);
954 trace_ext4_drop_inode(inode, drop);
958 static void ext4_i_callback(struct rcu_head *head)
960 struct inode *inode = container_of(head, struct inode, i_rcu);
961 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
964 static void ext4_destroy_inode(struct inode *inode)
966 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
967 ext4_msg(inode->i_sb, KERN_ERR,
968 "Inode %lu (%p): orphan list check failed!",
969 inode->i_ino, EXT4_I(inode));
970 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
971 EXT4_I(inode), sizeof(struct ext4_inode_info),
975 call_rcu(&inode->i_rcu, ext4_i_callback);
978 static void init_once(void *foo)
980 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
982 INIT_LIST_HEAD(&ei->i_orphan);
983 #ifdef CONFIG_EXT4_FS_XATTR
984 init_rwsem(&ei->xattr_sem);
986 init_rwsem(&ei->i_data_sem);
987 inode_init_once(&ei->vfs_inode);
990 static int init_inodecache(void)
992 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
993 sizeof(struct ext4_inode_info),
994 0, (SLAB_RECLAIM_ACCOUNT|
997 if (ext4_inode_cachep == NULL)
1002 static void destroy_inodecache(void)
1004 kmem_cache_destroy(ext4_inode_cachep);
1007 void ext4_clear_inode(struct inode *inode)
1009 invalidate_inode_buffers(inode);
1010 end_writeback(inode);
1012 ext4_discard_preallocations(inode);
1013 if (EXT4_I(inode)->jinode) {
1014 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1015 EXT4_I(inode)->jinode);
1016 jbd2_free_inode(EXT4_I(inode)->jinode);
1017 EXT4_I(inode)->jinode = NULL;
1021 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1022 u64 ino, u32 generation)
1024 struct inode *inode;
1026 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1027 return ERR_PTR(-ESTALE);
1028 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1029 return ERR_PTR(-ESTALE);
1031 /* iget isn't really right if the inode is currently unallocated!!
1033 * ext4_read_inode will return a bad_inode if the inode had been
1034 * deleted, so we should be safe.
1036 * Currently we don't know the generation for parent directory, so
1037 * a generation of 0 means "accept any"
1039 inode = ext4_iget(sb, ino);
1041 return ERR_CAST(inode);
1042 if (generation && inode->i_generation != generation) {
1044 return ERR_PTR(-ESTALE);
1050 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1051 int fh_len, int fh_type)
1053 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1054 ext4_nfs_get_inode);
1057 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1058 int fh_len, int fh_type)
1060 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1061 ext4_nfs_get_inode);
1065 * Try to release metadata pages (indirect blocks, directories) which are
1066 * mapped via the block device. Since these pages could have journal heads
1067 * which would prevent try_to_free_buffers() from freeing them, we must use
1068 * jbd2 layer's try_to_free_buffers() function to release them.
1070 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1073 journal_t *journal = EXT4_SB(sb)->s_journal;
1075 WARN_ON(PageChecked(page));
1076 if (!page_has_buffers(page))
1079 return jbd2_journal_try_to_free_buffers(journal, page,
1080 wait & ~__GFP_WAIT);
1081 return try_to_free_buffers(page);
1085 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1086 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1088 static int ext4_write_dquot(struct dquot *dquot);
1089 static int ext4_acquire_dquot(struct dquot *dquot);
1090 static int ext4_release_dquot(struct dquot *dquot);
1091 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1092 static int ext4_write_info(struct super_block *sb, int type);
1093 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1095 static int ext4_quota_off(struct super_block *sb, int type);
1096 static int ext4_quota_on_mount(struct super_block *sb, int type);
1097 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1098 size_t len, loff_t off);
1099 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1100 const char *data, size_t len, loff_t off);
1102 static const struct dquot_operations ext4_quota_operations = {
1103 .get_reserved_space = ext4_get_reserved_space,
1104 .write_dquot = ext4_write_dquot,
1105 .acquire_dquot = ext4_acquire_dquot,
1106 .release_dquot = ext4_release_dquot,
1107 .mark_dirty = ext4_mark_dquot_dirty,
1108 .write_info = ext4_write_info,
1109 .alloc_dquot = dquot_alloc,
1110 .destroy_dquot = dquot_destroy,
1113 static const struct quotactl_ops ext4_qctl_operations = {
1114 .quota_on = ext4_quota_on,
1115 .quota_off = ext4_quota_off,
1116 .quota_sync = dquot_quota_sync,
1117 .get_info = dquot_get_dqinfo,
1118 .set_info = dquot_set_dqinfo,
1119 .get_dqblk = dquot_get_dqblk,
1120 .set_dqblk = dquot_set_dqblk
1124 static const struct super_operations ext4_sops = {
1125 .alloc_inode = ext4_alloc_inode,
1126 .destroy_inode = ext4_destroy_inode,
1127 .write_inode = ext4_write_inode,
1128 .dirty_inode = ext4_dirty_inode,
1129 .drop_inode = ext4_drop_inode,
1130 .evict_inode = ext4_evict_inode,
1131 .put_super = ext4_put_super,
1132 .sync_fs = ext4_sync_fs,
1133 .freeze_fs = ext4_freeze,
1134 .unfreeze_fs = ext4_unfreeze,
1135 .statfs = ext4_statfs,
1136 .remount_fs = ext4_remount,
1137 .show_options = ext4_show_options,
1139 .quota_read = ext4_quota_read,
1140 .quota_write = ext4_quota_write,
1142 .bdev_try_to_free_page = bdev_try_to_free_page,
1145 static const struct super_operations ext4_nojournal_sops = {
1146 .alloc_inode = ext4_alloc_inode,
1147 .destroy_inode = ext4_destroy_inode,
1148 .write_inode = ext4_write_inode,
1149 .dirty_inode = ext4_dirty_inode,
1150 .drop_inode = ext4_drop_inode,
1151 .evict_inode = ext4_evict_inode,
1152 .write_super = ext4_write_super,
1153 .put_super = ext4_put_super,
1154 .statfs = ext4_statfs,
1155 .remount_fs = ext4_remount,
1156 .show_options = ext4_show_options,
1158 .quota_read = ext4_quota_read,
1159 .quota_write = ext4_quota_write,
1161 .bdev_try_to_free_page = bdev_try_to_free_page,
1164 static const struct export_operations ext4_export_ops = {
1165 .fh_to_dentry = ext4_fh_to_dentry,
1166 .fh_to_parent = ext4_fh_to_parent,
1167 .get_parent = ext4_get_parent,
1171 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1172 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1173 Opt_nouid32, Opt_debug, Opt_removed,
1174 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1175 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1176 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1177 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1178 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1179 Opt_data_err_abort, Opt_data_err_ignore,
1180 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1181 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1182 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1183 Opt_usrquota, Opt_grpquota, Opt_i_version,
1184 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1185 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1186 Opt_inode_readahead_blks, Opt_journal_ioprio,
1187 Opt_dioread_nolock, Opt_dioread_lock,
1188 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1191 static const match_table_t tokens = {
1192 {Opt_bsd_df, "bsddf"},
1193 {Opt_minix_df, "minixdf"},
1194 {Opt_grpid, "grpid"},
1195 {Opt_grpid, "bsdgroups"},
1196 {Opt_nogrpid, "nogrpid"},
1197 {Opt_nogrpid, "sysvgroups"},
1198 {Opt_resgid, "resgid=%u"},
1199 {Opt_resuid, "resuid=%u"},
1201 {Opt_err_cont, "errors=continue"},
1202 {Opt_err_panic, "errors=panic"},
1203 {Opt_err_ro, "errors=remount-ro"},
1204 {Opt_nouid32, "nouid32"},
1205 {Opt_debug, "debug"},
1206 {Opt_removed, "oldalloc"},
1207 {Opt_removed, "orlov"},
1208 {Opt_user_xattr, "user_xattr"},
1209 {Opt_nouser_xattr, "nouser_xattr"},
1211 {Opt_noacl, "noacl"},
1212 {Opt_noload, "norecovery"},
1213 {Opt_noload, "noload"},
1214 {Opt_removed, "nobh"},
1215 {Opt_removed, "bh"},
1216 {Opt_commit, "commit=%u"},
1217 {Opt_min_batch_time, "min_batch_time=%u"},
1218 {Opt_max_batch_time, "max_batch_time=%u"},
1219 {Opt_journal_dev, "journal_dev=%u"},
1220 {Opt_journal_checksum, "journal_checksum"},
1221 {Opt_journal_async_commit, "journal_async_commit"},
1222 {Opt_abort, "abort"},
1223 {Opt_data_journal, "data=journal"},
1224 {Opt_data_ordered, "data=ordered"},
1225 {Opt_data_writeback, "data=writeback"},
1226 {Opt_data_err_abort, "data_err=abort"},
1227 {Opt_data_err_ignore, "data_err=ignore"},
1228 {Opt_offusrjquota, "usrjquota="},
1229 {Opt_usrjquota, "usrjquota=%s"},
1230 {Opt_offgrpjquota, "grpjquota="},
1231 {Opt_grpjquota, "grpjquota=%s"},
1232 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1233 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1234 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1235 {Opt_grpquota, "grpquota"},
1236 {Opt_noquota, "noquota"},
1237 {Opt_quota, "quota"},
1238 {Opt_usrquota, "usrquota"},
1239 {Opt_barrier, "barrier=%u"},
1240 {Opt_barrier, "barrier"},
1241 {Opt_nobarrier, "nobarrier"},
1242 {Opt_i_version, "i_version"},
1243 {Opt_stripe, "stripe=%u"},
1244 {Opt_delalloc, "delalloc"},
1245 {Opt_nodelalloc, "nodelalloc"},
1246 {Opt_mblk_io_submit, "mblk_io_submit"},
1247 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1248 {Opt_block_validity, "block_validity"},
1249 {Opt_noblock_validity, "noblock_validity"},
1250 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1251 {Opt_journal_ioprio, "journal_ioprio=%u"},
1252 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1253 {Opt_auto_da_alloc, "auto_da_alloc"},
1254 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1255 {Opt_dioread_nolock, "dioread_nolock"},
1256 {Opt_dioread_lock, "dioread_lock"},
1257 {Opt_discard, "discard"},
1258 {Opt_nodiscard, "nodiscard"},
1259 {Opt_init_itable, "init_itable=%u"},
1260 {Opt_init_itable, "init_itable"},
1261 {Opt_noinit_itable, "noinit_itable"},
1262 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1263 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1264 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1265 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1266 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1270 static ext4_fsblk_t get_sb_block(void **data)
1272 ext4_fsblk_t sb_block;
1273 char *options = (char *) *data;
1275 if (!options || strncmp(options, "sb=", 3) != 0)
1276 return 1; /* Default location */
1279 /* TODO: use simple_strtoll with >32bit ext4 */
1280 sb_block = simple_strtoul(options, &options, 0);
1281 if (*options && *options != ',') {
1282 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1286 if (*options == ',')
1288 *data = (void *) options;
1293 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1294 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1295 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1298 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1303 if (sb_any_quota_loaded(sb) &&
1304 !sbi->s_qf_names[qtype]) {
1305 ext4_msg(sb, KERN_ERR,
1306 "Cannot change journaled "
1307 "quota options when quota turned on");
1310 qname = match_strdup(args);
1312 ext4_msg(sb, KERN_ERR,
1313 "Not enough memory for storing quotafile name");
1316 if (sbi->s_qf_names[qtype] &&
1317 strcmp(sbi->s_qf_names[qtype], qname)) {
1318 ext4_msg(sb, KERN_ERR,
1319 "%s quota file already specified", QTYPE2NAME(qtype));
1323 sbi->s_qf_names[qtype] = qname;
1324 if (strchr(sbi->s_qf_names[qtype], '/')) {
1325 ext4_msg(sb, KERN_ERR,
1326 "quotafile must be on filesystem root");
1327 kfree(sbi->s_qf_names[qtype]);
1328 sbi->s_qf_names[qtype] = NULL;
1335 static int clear_qf_name(struct super_block *sb, int qtype)
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1340 if (sb_any_quota_loaded(sb) &&
1341 sbi->s_qf_names[qtype]) {
1342 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1343 " when quota turned on");
1347 * The space will be released later when all options are confirmed
1350 sbi->s_qf_names[qtype] = NULL;
1355 #define MOPT_SET 0x0001
1356 #define MOPT_CLEAR 0x0002
1357 #define MOPT_NOSUPPORT 0x0004
1358 #define MOPT_EXPLICIT 0x0008
1359 #define MOPT_CLEAR_ERR 0x0010
1360 #define MOPT_GTE0 0x0020
1363 #define MOPT_QFMT 0x0040
1365 #define MOPT_Q MOPT_NOSUPPORT
1366 #define MOPT_QFMT MOPT_NOSUPPORT
1368 #define MOPT_DATAJ 0x0080
1370 static const struct mount_opts {
1374 } ext4_mount_opts[] = {
1375 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1376 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1377 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1378 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1379 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1380 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1381 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1382 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1383 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1384 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1385 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1386 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1387 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1388 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1389 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1390 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1391 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1392 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1393 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1398 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1399 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1400 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1401 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1402 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1403 {Opt_commit, 0, MOPT_GTE0},
1404 {Opt_max_batch_time, 0, MOPT_GTE0},
1405 {Opt_min_batch_time, 0, MOPT_GTE0},
1406 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1407 {Opt_init_itable, 0, MOPT_GTE0},
1408 {Opt_stripe, 0, MOPT_GTE0},
1409 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1410 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1411 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1412 #ifdef CONFIG_EXT4_FS_XATTR
1413 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1414 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1416 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1417 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1419 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1420 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1421 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1423 {Opt_acl, 0, MOPT_NOSUPPORT},
1424 {Opt_noacl, 0, MOPT_NOSUPPORT},
1426 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1427 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1428 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1429 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1431 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1433 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1434 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1435 {Opt_usrjquota, 0, MOPT_Q},
1436 {Opt_grpjquota, 0, MOPT_Q},
1437 {Opt_offusrjquota, 0, MOPT_Q},
1438 {Opt_offgrpjquota, 0, MOPT_Q},
1439 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1440 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1441 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1445 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1446 substring_t *args, unsigned long *journal_devnum,
1447 unsigned int *journal_ioprio, int is_remount)
1449 struct ext4_sb_info *sbi = EXT4_SB(sb);
1450 const struct mount_opts *m;
1455 if (args->from && match_int(args, &arg))
1459 case Opt_nouser_xattr:
1460 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1463 return 1; /* handled by get_sb_block() */
1465 ext4_msg(sb, KERN_WARNING,
1466 "Ignoring removed %s option", opt);
1469 uid = make_kuid(current_user_ns(), arg);
1470 if (!uid_valid(uid)) {
1471 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1474 sbi->s_resuid = uid;
1477 gid = make_kgid(current_user_ns(), arg);
1478 if (!gid_valid(gid)) {
1479 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1482 sbi->s_resgid = gid;
1485 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1488 sb->s_flags |= MS_I_VERSION;
1490 case Opt_journal_dev:
1492 ext4_msg(sb, KERN_ERR,
1493 "Cannot specify journal on remount");
1496 *journal_devnum = arg;
1498 case Opt_journal_ioprio:
1499 if (arg < 0 || arg > 7)
1501 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1505 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1506 if (token != m->token)
1508 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 if (m->flags & MOPT_EXPLICIT)
1511 set_opt2(sb, EXPLICIT_DELALLOC);
1512 if (m->flags & MOPT_CLEAR_ERR)
1513 clear_opt(sb, ERRORS_MASK);
1514 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1515 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1516 "options when quota turned on");
1520 if (m->flags & MOPT_NOSUPPORT) {
1521 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1522 } else if (token == Opt_commit) {
1524 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1525 sbi->s_commit_interval = HZ * arg;
1526 } else if (token == Opt_max_batch_time) {
1528 arg = EXT4_DEF_MAX_BATCH_TIME;
1529 sbi->s_max_batch_time = arg;
1530 } else if (token == Opt_min_batch_time) {
1531 sbi->s_min_batch_time = arg;
1532 } else if (token == Opt_inode_readahead_blks) {
1533 if (arg > (1 << 30))
1535 if (arg && !is_power_of_2(arg)) {
1536 ext4_msg(sb, KERN_ERR,
1537 "EXT4-fs: inode_readahead_blks"
1538 " must be a power of 2");
1541 sbi->s_inode_readahead_blks = arg;
1542 } else if (token == Opt_init_itable) {
1543 set_opt(sb, INIT_INODE_TABLE);
1545 arg = EXT4_DEF_LI_WAIT_MULT;
1546 sbi->s_li_wait_mult = arg;
1547 } else if (token == Opt_stripe) {
1548 sbi->s_stripe = arg;
1549 } else if (m->flags & MOPT_DATAJ) {
1551 if (!sbi->s_journal)
1552 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1553 else if (test_opt(sb, DATA_FLAGS) !=
1555 ext4_msg(sb, KERN_ERR,
1556 "Cannot change data mode on remount");
1560 clear_opt(sb, DATA_FLAGS);
1561 sbi->s_mount_opt |= m->mount_opt;
1564 } else if (token == Opt_usrjquota) {
1565 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1567 } else if (token == Opt_grpjquota) {
1568 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1570 } else if (token == Opt_offusrjquota) {
1571 if (!clear_qf_name(sb, USRQUOTA))
1573 } else if (token == Opt_offgrpjquota) {
1574 if (!clear_qf_name(sb, GRPQUOTA))
1576 } else if (m->flags & MOPT_QFMT) {
1577 if (sb_any_quota_loaded(sb) &&
1578 sbi->s_jquota_fmt != m->mount_opt) {
1579 ext4_msg(sb, KERN_ERR, "Cannot "
1580 "change journaled quota options "
1581 "when quota turned on");
1584 sbi->s_jquota_fmt = m->mount_opt;
1589 if (m->flags & MOPT_CLEAR)
1591 else if (unlikely(!(m->flags & MOPT_SET))) {
1592 ext4_msg(sb, KERN_WARNING,
1593 "buggy handling of option %s", opt);
1598 sbi->s_mount_opt |= m->mount_opt;
1600 sbi->s_mount_opt &= ~m->mount_opt;
1604 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1605 "or missing value", opt);
1609 static int parse_options(char *options, struct super_block *sb,
1610 unsigned long *journal_devnum,
1611 unsigned int *journal_ioprio,
1614 struct ext4_sb_info *sbi = EXT4_SB(sb);
1616 substring_t args[MAX_OPT_ARGS];
1622 while ((p = strsep(&options, ",")) != NULL) {
1626 * Initialize args struct so we know whether arg was
1627 * found; some options take optional arguments.
1629 args[0].to = args[0].from = 0;
1630 token = match_token(p, tokens, args);
1631 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1632 journal_ioprio, is_remount) < 0)
1636 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1637 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1638 clear_opt(sb, USRQUOTA);
1640 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1641 clear_opt(sb, GRPQUOTA);
1643 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1644 ext4_msg(sb, KERN_ERR, "old and new quota "
1649 if (!sbi->s_jquota_fmt) {
1650 ext4_msg(sb, KERN_ERR, "journaled quota format "
1655 if (sbi->s_jquota_fmt) {
1656 ext4_msg(sb, KERN_ERR, "journaled quota format "
1657 "specified with no journaling "
1666 static inline void ext4_show_quota_options(struct seq_file *seq,
1667 struct super_block *sb)
1669 #if defined(CONFIG_QUOTA)
1670 struct ext4_sb_info *sbi = EXT4_SB(sb);
1672 if (sbi->s_jquota_fmt) {
1675 switch (sbi->s_jquota_fmt) {
1686 seq_printf(seq, ",jqfmt=%s", fmtname);
1689 if (sbi->s_qf_names[USRQUOTA])
1690 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1692 if (sbi->s_qf_names[GRPQUOTA])
1693 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1695 if (test_opt(sb, USRQUOTA))
1696 seq_puts(seq, ",usrquota");
1698 if (test_opt(sb, GRPQUOTA))
1699 seq_puts(seq, ",grpquota");
1703 static const char *token2str(int token)
1705 static const struct match_token *t;
1707 for (t = tokens; t->token != Opt_err; t++)
1708 if (t->token == token && !strchr(t->pattern, '='))
1715 * - it's set to a non-default value OR
1716 * - if the per-sb default is different from the global default
1718 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1721 struct ext4_sb_info *sbi = EXT4_SB(sb);
1722 struct ext4_super_block *es = sbi->s_es;
1723 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1724 const struct mount_opts *m;
1725 char sep = nodefs ? '\n' : ',';
1727 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1728 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1730 if (sbi->s_sb_block != 1)
1731 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1733 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1734 int want_set = m->flags & MOPT_SET;
1735 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1736 (m->flags & MOPT_CLEAR_ERR))
1738 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1739 continue; /* skip if same as the default */
1741 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1742 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1743 continue; /* select Opt_noFoo vs Opt_Foo */
1744 SEQ_OPTS_PRINT("%s", token2str(m->token));
1747 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1748 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1749 SEQ_OPTS_PRINT("resuid=%u",
1750 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1751 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1752 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1753 SEQ_OPTS_PRINT("resgid=%u",
1754 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1755 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1756 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1757 SEQ_OPTS_PUTS("errors=remount-ro");
1758 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1759 SEQ_OPTS_PUTS("errors=continue");
1760 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1761 SEQ_OPTS_PUTS("errors=panic");
1762 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1763 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1764 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1765 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1766 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1767 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1768 if (sb->s_flags & MS_I_VERSION)
1769 SEQ_OPTS_PUTS("i_version");
1770 if (nodefs || sbi->s_stripe)
1771 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1772 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1773 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1774 SEQ_OPTS_PUTS("data=journal");
1775 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1776 SEQ_OPTS_PUTS("data=ordered");
1777 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1778 SEQ_OPTS_PUTS("data=writeback");
1781 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1782 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1783 sbi->s_inode_readahead_blks);
1785 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1786 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1787 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1789 ext4_show_quota_options(seq, sb);
1793 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1795 return _ext4_show_options(seq, root->d_sb, 0);
1798 static int options_seq_show(struct seq_file *seq, void *offset)
1800 struct super_block *sb = seq->private;
1803 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1804 rc = _ext4_show_options(seq, sb, 1);
1805 seq_puts(seq, "\n");
1809 static int options_open_fs(struct inode *inode, struct file *file)
1811 return single_open(file, options_seq_show, PDE(inode)->data);
1814 static const struct file_operations ext4_seq_options_fops = {
1815 .owner = THIS_MODULE,
1816 .open = options_open_fs,
1818 .llseek = seq_lseek,
1819 .release = single_release,
1822 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1825 struct ext4_sb_info *sbi = EXT4_SB(sb);
1828 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1829 ext4_msg(sb, KERN_ERR, "revision level too high, "
1830 "forcing read-only mode");
1835 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1836 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1837 "running e2fsck is recommended");
1838 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1839 ext4_msg(sb, KERN_WARNING,
1840 "warning: mounting fs with errors, "
1841 "running e2fsck is recommended");
1842 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1843 le16_to_cpu(es->s_mnt_count) >=
1844 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1845 ext4_msg(sb, KERN_WARNING,
1846 "warning: maximal mount count reached, "
1847 "running e2fsck is recommended");
1848 else if (le32_to_cpu(es->s_checkinterval) &&
1849 (le32_to_cpu(es->s_lastcheck) +
1850 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1851 ext4_msg(sb, KERN_WARNING,
1852 "warning: checktime reached, "
1853 "running e2fsck is recommended");
1854 if (!sbi->s_journal)
1855 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1856 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1857 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1858 le16_add_cpu(&es->s_mnt_count, 1);
1859 es->s_mtime = cpu_to_le32(get_seconds());
1860 ext4_update_dynamic_rev(sb);
1862 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1864 ext4_commit_super(sb, 1);
1866 if (test_opt(sb, DEBUG))
1867 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1868 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1870 sbi->s_groups_count,
1871 EXT4_BLOCKS_PER_GROUP(sb),
1872 EXT4_INODES_PER_GROUP(sb),
1873 sbi->s_mount_opt, sbi->s_mount_opt2);
1875 cleancache_init_fs(sb);
1879 static int ext4_fill_flex_info(struct super_block *sb)
1881 struct ext4_sb_info *sbi = EXT4_SB(sb);
1882 struct ext4_group_desc *gdp = NULL;
1883 ext4_group_t flex_group_count;
1884 ext4_group_t flex_group;
1885 unsigned int groups_per_flex = 0;
1889 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1890 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1891 sbi->s_log_groups_per_flex = 0;
1894 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1896 /* We allocate both existing and potentially added groups */
1897 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1898 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1899 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1900 size = flex_group_count * sizeof(struct flex_groups);
1901 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1902 if (sbi->s_flex_groups == NULL) {
1903 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1908 for (i = 0; i < sbi->s_groups_count; i++) {
1909 gdp = ext4_get_group_desc(sb, i, NULL);
1911 flex_group = ext4_flex_group(sbi, i);
1912 atomic_add(ext4_free_inodes_count(sb, gdp),
1913 &sbi->s_flex_groups[flex_group].free_inodes);
1914 atomic_add(ext4_free_group_clusters(sb, gdp),
1915 &sbi->s_flex_groups[flex_group].free_clusters);
1916 atomic_add(ext4_used_dirs_count(sb, gdp),
1917 &sbi->s_flex_groups[flex_group].used_dirs);
1925 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1926 struct ext4_group_desc *gdp)
1930 if (sbi->s_es->s_feature_ro_compat &
1931 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1932 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1933 __le32 le_group = cpu_to_le32(block_group);
1935 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1936 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1937 crc = crc16(crc, (__u8 *)gdp, offset);
1938 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1939 /* for checksum of struct ext4_group_desc do the rest...*/
1940 if ((sbi->s_es->s_feature_incompat &
1941 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1942 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1943 crc = crc16(crc, (__u8 *)gdp + offset,
1944 le16_to_cpu(sbi->s_es->s_desc_size) -
1948 return cpu_to_le16(crc);
1951 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1952 struct ext4_group_desc *gdp)
1954 if ((sbi->s_es->s_feature_ro_compat &
1955 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1956 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1962 /* Called at mount-time, super-block is locked */
1963 static int ext4_check_descriptors(struct super_block *sb,
1964 ext4_group_t *first_not_zeroed)
1966 struct ext4_sb_info *sbi = EXT4_SB(sb);
1967 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1968 ext4_fsblk_t last_block;
1969 ext4_fsblk_t block_bitmap;
1970 ext4_fsblk_t inode_bitmap;
1971 ext4_fsblk_t inode_table;
1972 int flexbg_flag = 0;
1973 ext4_group_t i, grp = sbi->s_groups_count;
1975 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1978 ext4_debug("Checking group descriptors");
1980 for (i = 0; i < sbi->s_groups_count; i++) {
1981 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1983 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1984 last_block = ext4_blocks_count(sbi->s_es) - 1;
1986 last_block = first_block +
1987 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1989 if ((grp == sbi->s_groups_count) &&
1990 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1993 block_bitmap = ext4_block_bitmap(sb, gdp);
1994 if (block_bitmap < first_block || block_bitmap > last_block) {
1995 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1996 "Block bitmap for group %u not in group "
1997 "(block %llu)!", i, block_bitmap);
2000 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2001 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2002 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2003 "Inode bitmap for group %u not in group "
2004 "(block %llu)!", i, inode_bitmap);
2007 inode_table = ext4_inode_table(sb, gdp);
2008 if (inode_table < first_block ||
2009 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2010 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2011 "Inode table for group %u not in group "
2012 "(block %llu)!", i, inode_table);
2015 ext4_lock_group(sb, i);
2016 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2017 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2018 "Checksum for group %u failed (%u!=%u)",
2019 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2020 gdp)), le16_to_cpu(gdp->bg_checksum));
2021 if (!(sb->s_flags & MS_RDONLY)) {
2022 ext4_unlock_group(sb, i);
2026 ext4_unlock_group(sb, i);
2028 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2030 if (NULL != first_not_zeroed)
2031 *first_not_zeroed = grp;
2033 ext4_free_blocks_count_set(sbi->s_es,
2034 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2035 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2039 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2040 * the superblock) which were deleted from all directories, but held open by
2041 * a process at the time of a crash. We walk the list and try to delete these
2042 * inodes at recovery time (only with a read-write filesystem).
2044 * In order to keep the orphan inode chain consistent during traversal (in
2045 * case of crash during recovery), we link each inode into the superblock
2046 * orphan list_head and handle it the same way as an inode deletion during
2047 * normal operation (which journals the operations for us).
2049 * We only do an iget() and an iput() on each inode, which is very safe if we
2050 * accidentally point at an in-use or already deleted inode. The worst that
2051 * can happen in this case is that we get a "bit already cleared" message from
2052 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2053 * e2fsck was run on this filesystem, and it must have already done the orphan
2054 * inode cleanup for us, so we can safely abort without any further action.
2056 static void ext4_orphan_cleanup(struct super_block *sb,
2057 struct ext4_super_block *es)
2059 unsigned int s_flags = sb->s_flags;
2060 int nr_orphans = 0, nr_truncates = 0;
2064 if (!es->s_last_orphan) {
2065 jbd_debug(4, "no orphan inodes to clean up\n");
2069 if (bdev_read_only(sb->s_bdev)) {
2070 ext4_msg(sb, KERN_ERR, "write access "
2071 "unavailable, skipping orphan cleanup");
2075 /* Check if feature set would not allow a r/w mount */
2076 if (!ext4_feature_set_ok(sb, 0)) {
2077 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2078 "unknown ROCOMPAT features");
2082 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2083 if (es->s_last_orphan)
2084 jbd_debug(1, "Errors on filesystem, "
2085 "clearing orphan list.\n");
2086 es->s_last_orphan = 0;
2087 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2091 if (s_flags & MS_RDONLY) {
2092 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2093 sb->s_flags &= ~MS_RDONLY;
2096 /* Needed for iput() to work correctly and not trash data */
2097 sb->s_flags |= MS_ACTIVE;
2098 /* Turn on quotas so that they are updated correctly */
2099 for (i = 0; i < MAXQUOTAS; i++) {
2100 if (EXT4_SB(sb)->s_qf_names[i]) {
2101 int ret = ext4_quota_on_mount(sb, i);
2103 ext4_msg(sb, KERN_ERR,
2104 "Cannot turn on journaled "
2105 "quota: error %d", ret);
2110 while (es->s_last_orphan) {
2111 struct inode *inode;
2113 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2114 if (IS_ERR(inode)) {
2115 es->s_last_orphan = 0;
2119 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2120 dquot_initialize(inode);
2121 if (inode->i_nlink) {
2122 ext4_msg(sb, KERN_DEBUG,
2123 "%s: truncating inode %lu to %lld bytes",
2124 __func__, inode->i_ino, inode->i_size);
2125 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2126 inode->i_ino, inode->i_size);
2127 ext4_truncate(inode);
2130 ext4_msg(sb, KERN_DEBUG,
2131 "%s: deleting unreferenced inode %lu",
2132 __func__, inode->i_ino);
2133 jbd_debug(2, "deleting unreferenced inode %lu\n",
2137 iput(inode); /* The delete magic happens here! */
2140 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2143 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2144 PLURAL(nr_orphans));
2146 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2147 PLURAL(nr_truncates));
2149 /* Turn quotas off */
2150 for (i = 0; i < MAXQUOTAS; i++) {
2151 if (sb_dqopt(sb)->files[i])
2152 dquot_quota_off(sb, i);
2155 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2159 * Maximal extent format file size.
2160 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2161 * extent format containers, within a sector_t, and within i_blocks
2162 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2163 * so that won't be a limiting factor.
2165 * However there is other limiting factor. We do store extents in the form
2166 * of starting block and length, hence the resulting length of the extent
2167 * covering maximum file size must fit into on-disk format containers as
2168 * well. Given that length is always by 1 unit bigger than max unit (because
2169 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2171 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2173 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2176 loff_t upper_limit = MAX_LFS_FILESIZE;
2178 /* small i_blocks in vfs inode? */
2179 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2181 * CONFIG_LBDAF is not enabled implies the inode
2182 * i_block represent total blocks in 512 bytes
2183 * 32 == size of vfs inode i_blocks * 8
2185 upper_limit = (1LL << 32) - 1;
2187 /* total blocks in file system block size */
2188 upper_limit >>= (blkbits - 9);
2189 upper_limit <<= blkbits;
2193 * 32-bit extent-start container, ee_block. We lower the maxbytes
2194 * by one fs block, so ee_len can cover the extent of maximum file
2197 res = (1LL << 32) - 1;
2200 /* Sanity check against vm- & vfs- imposed limits */
2201 if (res > upper_limit)
2208 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2209 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2210 * We need to be 1 filesystem block less than the 2^48 sector limit.
2212 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2214 loff_t res = EXT4_NDIR_BLOCKS;
2217 /* This is calculated to be the largest file size for a dense, block
2218 * mapped file such that the file's total number of 512-byte sectors,
2219 * including data and all indirect blocks, does not exceed (2^48 - 1).
2221 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2222 * number of 512-byte sectors of the file.
2225 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2227 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2228 * the inode i_block field represents total file blocks in
2229 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2231 upper_limit = (1LL << 32) - 1;
2233 /* total blocks in file system block size */
2234 upper_limit >>= (bits - 9);
2238 * We use 48 bit ext4_inode i_blocks
2239 * With EXT4_HUGE_FILE_FL set the i_blocks
2240 * represent total number of blocks in
2241 * file system block size
2243 upper_limit = (1LL << 48) - 1;
2247 /* indirect blocks */
2249 /* double indirect blocks */
2250 meta_blocks += 1 + (1LL << (bits-2));
2251 /* tripple indirect blocks */
2252 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2254 upper_limit -= meta_blocks;
2255 upper_limit <<= bits;
2257 res += 1LL << (bits-2);
2258 res += 1LL << (2*(bits-2));
2259 res += 1LL << (3*(bits-2));
2261 if (res > upper_limit)
2264 if (res > MAX_LFS_FILESIZE)
2265 res = MAX_LFS_FILESIZE;
2270 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2271 ext4_fsblk_t logical_sb_block, int nr)
2273 struct ext4_sb_info *sbi = EXT4_SB(sb);
2274 ext4_group_t bg, first_meta_bg;
2277 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2279 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2281 return logical_sb_block + nr + 1;
2282 bg = sbi->s_desc_per_block * nr;
2283 if (ext4_bg_has_super(sb, bg))
2286 return (has_super + ext4_group_first_block_no(sb, bg));
2290 * ext4_get_stripe_size: Get the stripe size.
2291 * @sbi: In memory super block info
2293 * If we have specified it via mount option, then
2294 * use the mount option value. If the value specified at mount time is
2295 * greater than the blocks per group use the super block value.
2296 * If the super block value is greater than blocks per group return 0.
2297 * Allocator needs it be less than blocks per group.
2300 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2302 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2303 unsigned long stripe_width =
2304 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2307 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2308 ret = sbi->s_stripe;
2309 else if (stripe_width <= sbi->s_blocks_per_group)
2311 else if (stride <= sbi->s_blocks_per_group)
2317 * If the stripe width is 1, this makes no sense and
2318 * we set it to 0 to turn off stripe handling code.
2329 struct attribute attr;
2330 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2331 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2332 const char *, size_t);
2336 static int parse_strtoul(const char *buf,
2337 unsigned long max, unsigned long *value)
2341 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2342 endp = skip_spaces(endp);
2343 if (*endp || *value > max)
2349 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2350 struct ext4_sb_info *sbi,
2353 return snprintf(buf, PAGE_SIZE, "%llu\n",
2355 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2358 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2359 struct ext4_sb_info *sbi, char *buf)
2361 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2363 if (!sb->s_bdev->bd_part)
2364 return snprintf(buf, PAGE_SIZE, "0\n");
2365 return snprintf(buf, PAGE_SIZE, "%lu\n",
2366 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2367 sbi->s_sectors_written_start) >> 1);
2370 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2371 struct ext4_sb_info *sbi, char *buf)
2373 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2375 if (!sb->s_bdev->bd_part)
2376 return snprintf(buf, PAGE_SIZE, "0\n");
2377 return snprintf(buf, PAGE_SIZE, "%llu\n",
2378 (unsigned long long)(sbi->s_kbytes_written +
2379 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2380 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2383 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2384 struct ext4_sb_info *sbi, char *buf)
2386 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2389 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2390 struct ext4_sb_info *sbi, char *buf)
2392 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2395 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2396 struct ext4_sb_info *sbi,
2397 const char *buf, size_t count)
2401 if (parse_strtoul(buf, 0x40000000, &t))
2404 if (t && !is_power_of_2(t))
2407 sbi->s_inode_readahead_blks = t;
2411 static ssize_t sbi_ui_show(struct ext4_attr *a,
2412 struct ext4_sb_info *sbi, char *buf)
2414 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2416 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2419 static ssize_t sbi_ui_store(struct ext4_attr *a,
2420 struct ext4_sb_info *sbi,
2421 const char *buf, size_t count)
2423 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2426 if (parse_strtoul(buf, 0xffffffff, &t))
2432 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2433 static struct ext4_attr ext4_attr_##_name = { \
2434 .attr = {.name = __stringify(_name), .mode = _mode }, \
2437 .offset = offsetof(struct ext4_sb_info, _elname), \
2439 #define EXT4_ATTR(name, mode, show, store) \
2440 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2442 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2443 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2444 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2445 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2446 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2447 #define ATTR_LIST(name) &ext4_attr_##name.attr
2449 EXT4_RO_ATTR(delayed_allocation_blocks);
2450 EXT4_RO_ATTR(session_write_kbytes);
2451 EXT4_RO_ATTR(lifetime_write_kbytes);
2452 EXT4_RO_ATTR(extent_cache_hits);
2453 EXT4_RO_ATTR(extent_cache_misses);
2454 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2455 inode_readahead_blks_store, s_inode_readahead_blks);
2456 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2457 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2458 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2459 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2460 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2461 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2462 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2463 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2465 static struct attribute *ext4_attrs[] = {
2466 ATTR_LIST(delayed_allocation_blocks),
2467 ATTR_LIST(session_write_kbytes),
2468 ATTR_LIST(lifetime_write_kbytes),
2469 ATTR_LIST(extent_cache_hits),
2470 ATTR_LIST(extent_cache_misses),
2471 ATTR_LIST(inode_readahead_blks),
2472 ATTR_LIST(inode_goal),
2473 ATTR_LIST(mb_stats),
2474 ATTR_LIST(mb_max_to_scan),
2475 ATTR_LIST(mb_min_to_scan),
2476 ATTR_LIST(mb_order2_req),
2477 ATTR_LIST(mb_stream_req),
2478 ATTR_LIST(mb_group_prealloc),
2479 ATTR_LIST(max_writeback_mb_bump),
2483 /* Features this copy of ext4 supports */
2484 EXT4_INFO_ATTR(lazy_itable_init);
2485 EXT4_INFO_ATTR(batched_discard);
2487 static struct attribute *ext4_feat_attrs[] = {
2488 ATTR_LIST(lazy_itable_init),
2489 ATTR_LIST(batched_discard),
2493 static ssize_t ext4_attr_show(struct kobject *kobj,
2494 struct attribute *attr, char *buf)
2496 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2498 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2500 return a->show ? a->show(a, sbi, buf) : 0;
2503 static ssize_t ext4_attr_store(struct kobject *kobj,
2504 struct attribute *attr,
2505 const char *buf, size_t len)
2507 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2509 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2511 return a->store ? a->store(a, sbi, buf, len) : 0;
2514 static void ext4_sb_release(struct kobject *kobj)
2516 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2518 complete(&sbi->s_kobj_unregister);
2521 static const struct sysfs_ops ext4_attr_ops = {
2522 .show = ext4_attr_show,
2523 .store = ext4_attr_store,
2526 static struct kobj_type ext4_ktype = {
2527 .default_attrs = ext4_attrs,
2528 .sysfs_ops = &ext4_attr_ops,
2529 .release = ext4_sb_release,
2532 static void ext4_feat_release(struct kobject *kobj)
2534 complete(&ext4_feat->f_kobj_unregister);
2537 static struct kobj_type ext4_feat_ktype = {
2538 .default_attrs = ext4_feat_attrs,
2539 .sysfs_ops = &ext4_attr_ops,
2540 .release = ext4_feat_release,
2544 * Check whether this filesystem can be mounted based on
2545 * the features present and the RDONLY/RDWR mount requested.
2546 * Returns 1 if this filesystem can be mounted as requested,
2547 * 0 if it cannot be.
2549 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2552 ext4_msg(sb, KERN_ERR,
2553 "Couldn't mount because of "
2554 "unsupported optional features (%x)",
2555 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2556 ~EXT4_FEATURE_INCOMPAT_SUPP));
2563 /* Check that feature set is OK for a read-write mount */
2564 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2565 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2566 "unsupported optional features (%x)",
2567 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2568 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2572 * Large file size enabled file system can only be mounted
2573 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2575 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2576 if (sizeof(blkcnt_t) < sizeof(u64)) {
2577 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2578 "cannot be mounted RDWR without "
2583 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2584 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2585 ext4_msg(sb, KERN_ERR,
2586 "Can't support bigalloc feature without "
2587 "extents feature\n");
2594 * This function is called once a day if we have errors logged
2595 * on the file system
2597 static void print_daily_error_info(unsigned long arg)
2599 struct super_block *sb = (struct super_block *) arg;
2600 struct ext4_sb_info *sbi;
2601 struct ext4_super_block *es;
2606 if (es->s_error_count)
2607 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2608 le32_to_cpu(es->s_error_count));
2609 if (es->s_first_error_time) {
2610 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2611 sb->s_id, le32_to_cpu(es->s_first_error_time),
2612 (int) sizeof(es->s_first_error_func),
2613 es->s_first_error_func,
2614 le32_to_cpu(es->s_first_error_line));
2615 if (es->s_first_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_first_error_ino));
2618 if (es->s_first_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_first_error_block));
2623 if (es->s_last_error_time) {
2624 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2625 sb->s_id, le32_to_cpu(es->s_last_error_time),
2626 (int) sizeof(es->s_last_error_func),
2627 es->s_last_error_func,
2628 le32_to_cpu(es->s_last_error_line));
2629 if (es->s_last_error_ino)
2630 printk(": inode %u",
2631 le32_to_cpu(es->s_last_error_ino));
2632 if (es->s_last_error_block)
2633 printk(": block %llu", (unsigned long long)
2634 le64_to_cpu(es->s_last_error_block));
2637 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2640 /* Find next suitable group and run ext4_init_inode_table */
2641 static int ext4_run_li_request(struct ext4_li_request *elr)
2643 struct ext4_group_desc *gdp = NULL;
2644 ext4_group_t group, ngroups;
2645 struct super_block *sb;
2646 unsigned long timeout = 0;
2650 ngroups = EXT4_SB(sb)->s_groups_count;
2652 for (group = elr->lr_next_group; group < ngroups; group++) {
2653 gdp = ext4_get_group_desc(sb, group, NULL);
2659 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2663 if (group == ngroups)
2668 ret = ext4_init_inode_table(sb, group,
2669 elr->lr_timeout ? 0 : 1);
2670 if (elr->lr_timeout == 0) {
2671 timeout = (jiffies - timeout) *
2672 elr->lr_sbi->s_li_wait_mult;
2673 elr->lr_timeout = timeout;
2675 elr->lr_next_sched = jiffies + elr->lr_timeout;
2676 elr->lr_next_group = group + 1;
2683 * Remove lr_request from the list_request and free the
2684 * request structure. Should be called with li_list_mtx held
2686 static void ext4_remove_li_request(struct ext4_li_request *elr)
2688 struct ext4_sb_info *sbi;
2695 list_del(&elr->lr_request);
2696 sbi->s_li_request = NULL;
2700 static void ext4_unregister_li_request(struct super_block *sb)
2702 mutex_lock(&ext4_li_mtx);
2703 if (!ext4_li_info) {
2704 mutex_unlock(&ext4_li_mtx);
2708 mutex_lock(&ext4_li_info->li_list_mtx);
2709 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2710 mutex_unlock(&ext4_li_info->li_list_mtx);
2711 mutex_unlock(&ext4_li_mtx);
2714 static struct task_struct *ext4_lazyinit_task;
2717 * This is the function where ext4lazyinit thread lives. It walks
2718 * through the request list searching for next scheduled filesystem.
2719 * When such a fs is found, run the lazy initialization request
2720 * (ext4_rn_li_request) and keep track of the time spend in this
2721 * function. Based on that time we compute next schedule time of
2722 * the request. When walking through the list is complete, compute
2723 * next waking time and put itself into sleep.
2725 static int ext4_lazyinit_thread(void *arg)
2727 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2728 struct list_head *pos, *n;
2729 struct ext4_li_request *elr;
2730 unsigned long next_wakeup, cur;
2732 BUG_ON(NULL == eli);
2736 next_wakeup = MAX_JIFFY_OFFSET;
2738 mutex_lock(&eli->li_list_mtx);
2739 if (list_empty(&eli->li_request_list)) {
2740 mutex_unlock(&eli->li_list_mtx);
2744 list_for_each_safe(pos, n, &eli->li_request_list) {
2745 elr = list_entry(pos, struct ext4_li_request,
2748 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2749 if (ext4_run_li_request(elr) != 0) {
2750 /* error, remove the lazy_init job */
2751 ext4_remove_li_request(elr);
2756 if (time_before(elr->lr_next_sched, next_wakeup))
2757 next_wakeup = elr->lr_next_sched;
2759 mutex_unlock(&eli->li_list_mtx);
2764 if ((time_after_eq(cur, next_wakeup)) ||
2765 (MAX_JIFFY_OFFSET == next_wakeup)) {
2770 schedule_timeout_interruptible(next_wakeup - cur);
2772 if (kthread_should_stop()) {
2773 ext4_clear_request_list();
2780 * It looks like the request list is empty, but we need
2781 * to check it under the li_list_mtx lock, to prevent any
2782 * additions into it, and of course we should lock ext4_li_mtx
2783 * to atomically free the list and ext4_li_info, because at
2784 * this point another ext4 filesystem could be registering
2787 mutex_lock(&ext4_li_mtx);
2788 mutex_lock(&eli->li_list_mtx);
2789 if (!list_empty(&eli->li_request_list)) {
2790 mutex_unlock(&eli->li_list_mtx);
2791 mutex_unlock(&ext4_li_mtx);
2794 mutex_unlock(&eli->li_list_mtx);
2795 kfree(ext4_li_info);
2796 ext4_li_info = NULL;
2797 mutex_unlock(&ext4_li_mtx);
2802 static void ext4_clear_request_list(void)
2804 struct list_head *pos, *n;
2805 struct ext4_li_request *elr;
2807 mutex_lock(&ext4_li_info->li_list_mtx);
2808 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2809 elr = list_entry(pos, struct ext4_li_request,
2811 ext4_remove_li_request(elr);
2813 mutex_unlock(&ext4_li_info->li_list_mtx);
2816 static int ext4_run_lazyinit_thread(void)
2818 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2819 ext4_li_info, "ext4lazyinit");
2820 if (IS_ERR(ext4_lazyinit_task)) {
2821 int err = PTR_ERR(ext4_lazyinit_task);
2822 ext4_clear_request_list();
2823 kfree(ext4_li_info);
2824 ext4_li_info = NULL;
2825 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2826 "initialization thread\n",
2830 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2835 * Check whether it make sense to run itable init. thread or not.
2836 * If there is at least one uninitialized inode table, return
2837 * corresponding group number, else the loop goes through all
2838 * groups and return total number of groups.
2840 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2842 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2843 struct ext4_group_desc *gdp = NULL;
2845 for (group = 0; group < ngroups; group++) {
2846 gdp = ext4_get_group_desc(sb, group, NULL);
2850 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2857 static int ext4_li_info_new(void)
2859 struct ext4_lazy_init *eli = NULL;
2861 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2865 INIT_LIST_HEAD(&eli->li_request_list);
2866 mutex_init(&eli->li_list_mtx);
2868 eli->li_state |= EXT4_LAZYINIT_QUIT;
2875 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2878 struct ext4_sb_info *sbi = EXT4_SB(sb);
2879 struct ext4_li_request *elr;
2882 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2888 elr->lr_next_group = start;
2891 * Randomize first schedule time of the request to
2892 * spread the inode table initialization requests
2895 get_random_bytes(&rnd, sizeof(rnd));
2896 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2897 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2902 static int ext4_register_li_request(struct super_block *sb,
2903 ext4_group_t first_not_zeroed)
2905 struct ext4_sb_info *sbi = EXT4_SB(sb);
2906 struct ext4_li_request *elr;
2907 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2910 if (sbi->s_li_request != NULL) {
2912 * Reset timeout so it can be computed again, because
2913 * s_li_wait_mult might have changed.
2915 sbi->s_li_request->lr_timeout = 0;
2919 if (first_not_zeroed == ngroups ||
2920 (sb->s_flags & MS_RDONLY) ||
2921 !test_opt(sb, INIT_INODE_TABLE))
2924 elr = ext4_li_request_new(sb, first_not_zeroed);
2928 mutex_lock(&ext4_li_mtx);
2930 if (NULL == ext4_li_info) {
2931 ret = ext4_li_info_new();
2936 mutex_lock(&ext4_li_info->li_list_mtx);
2937 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2938 mutex_unlock(&ext4_li_info->li_list_mtx);
2940 sbi->s_li_request = elr;
2942 * set elr to NULL here since it has been inserted to
2943 * the request_list and the removal and free of it is
2944 * handled by ext4_clear_request_list from now on.
2948 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2949 ret = ext4_run_lazyinit_thread();
2954 mutex_unlock(&ext4_li_mtx);
2961 * We do not need to lock anything since this is called on
2964 static void ext4_destroy_lazyinit_thread(void)
2967 * If thread exited earlier
2968 * there's nothing to be done.
2970 if (!ext4_li_info || !ext4_lazyinit_task)
2973 kthread_stop(ext4_lazyinit_task);
2976 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2978 char *orig_data = kstrdup(data, GFP_KERNEL);
2979 struct buffer_head *bh;
2980 struct ext4_super_block *es = NULL;
2981 struct ext4_sb_info *sbi;
2983 ext4_fsblk_t sb_block = get_sb_block(&data);
2984 ext4_fsblk_t logical_sb_block;
2985 unsigned long offset = 0;
2986 unsigned long journal_devnum = 0;
2987 unsigned long def_mount_opts;
2992 int blocksize, clustersize;
2993 unsigned int db_count;
2995 int needs_recovery, has_huge_files, has_bigalloc;
2998 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2999 ext4_group_t first_not_zeroed;
3001 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3005 sbi->s_blockgroup_lock =
3006 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3007 if (!sbi->s_blockgroup_lock) {
3011 sb->s_fs_info = sbi;
3012 sbi->s_mount_opt = 0;
3013 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3014 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3015 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3016 sbi->s_sb_block = sb_block;
3017 if (sb->s_bdev->bd_part)
3018 sbi->s_sectors_written_start =
3019 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3021 /* Cleanup superblock name */
3022 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3026 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3028 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3033 * The ext4 superblock will not be buffer aligned for other than 1kB
3034 * block sizes. We need to calculate the offset from buffer start.
3036 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3037 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3038 offset = do_div(logical_sb_block, blocksize);
3040 logical_sb_block = sb_block;
3043 if (!(bh = sb_bread(sb, logical_sb_block))) {
3044 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3048 * Note: s_es must be initialized as soon as possible because
3049 * some ext4 macro-instructions depend on its value
3051 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3053 sb->s_magic = le16_to_cpu(es->s_magic);
3054 if (sb->s_magic != EXT4_SUPER_MAGIC)
3056 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3058 /* Set defaults before we parse the mount options */
3059 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3060 set_opt(sb, INIT_INODE_TABLE);
3061 if (def_mount_opts & EXT4_DEFM_DEBUG)
3063 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3065 if (def_mount_opts & EXT4_DEFM_UID16)
3066 set_opt(sb, NO_UID32);
3067 /* xattr user namespace & acls are now defaulted on */
3068 #ifdef CONFIG_EXT4_FS_XATTR
3069 set_opt(sb, XATTR_USER);
3071 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3072 set_opt(sb, POSIX_ACL);
3074 set_opt(sb, MBLK_IO_SUBMIT);
3075 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3076 set_opt(sb, JOURNAL_DATA);
3077 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3078 set_opt(sb, ORDERED_DATA);
3079 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3080 set_opt(sb, WRITEBACK_DATA);
3082 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3083 set_opt(sb, ERRORS_PANIC);
3084 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3085 set_opt(sb, ERRORS_CONT);
3087 set_opt(sb, ERRORS_RO);
3088 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3089 set_opt(sb, BLOCK_VALIDITY);
3090 if (def_mount_opts & EXT4_DEFM_DISCARD)
3091 set_opt(sb, DISCARD);
3093 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3094 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3095 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3096 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3097 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3099 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3100 set_opt(sb, BARRIER);
3103 * enable delayed allocation by default
3104 * Use -o nodelalloc to turn it off
3106 if (!IS_EXT3_SB(sb) &&
3107 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3108 set_opt(sb, DELALLOC);
3111 * set default s_li_wait_mult for lazyinit, for the case there is
3112 * no mount option specified.
3114 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3116 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3117 &journal_devnum, &journal_ioprio, 0)) {
3118 ext4_msg(sb, KERN_WARNING,
3119 "failed to parse options in superblock: %s",
3120 sbi->s_es->s_mount_opts);
3122 sbi->s_def_mount_opt = sbi->s_mount_opt;
3123 if (!parse_options((char *) data, sb, &journal_devnum,
3124 &journal_ioprio, 0))
3127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3128 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3129 "with data=journal disables delayed "
3130 "allocation and O_DIRECT support!\n");
3131 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3132 ext4_msg(sb, KERN_ERR, "can't mount with "
3133 "both data=journal and delalloc");
3136 if (test_opt(sb, DIOREAD_NOLOCK)) {
3137 ext4_msg(sb, KERN_ERR, "can't mount with "
3138 "both data=journal and delalloc");
3141 if (test_opt(sb, DELALLOC))
3142 clear_opt(sb, DELALLOC);
3145 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3146 if (test_opt(sb, DIOREAD_NOLOCK)) {
3147 if (blocksize < PAGE_SIZE) {
3148 ext4_msg(sb, KERN_ERR, "can't mount with "
3149 "dioread_nolock if block size != PAGE_SIZE");
3154 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3155 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3157 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3158 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3159 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3160 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3161 ext4_msg(sb, KERN_WARNING,
3162 "feature flags set on rev 0 fs, "
3163 "running e2fsck is recommended");
3165 if (IS_EXT2_SB(sb)) {
3166 if (ext2_feature_set_ok(sb))
3167 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3168 "using the ext4 subsystem");
3170 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3171 "to feature incompatibilities");
3176 if (IS_EXT3_SB(sb)) {
3177 if (ext3_feature_set_ok(sb))
3178 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3179 "using the ext4 subsystem");
3181 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3182 "to feature incompatibilities");
3188 * Check feature flags regardless of the revision level, since we
3189 * previously didn't change the revision level when setting the flags,
3190 * so there is a chance incompat flags are set on a rev 0 filesystem.
3192 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3195 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3196 blocksize > EXT4_MAX_BLOCK_SIZE) {
3197 ext4_msg(sb, KERN_ERR,
3198 "Unsupported filesystem blocksize %d", blocksize);
3202 if (sb->s_blocksize != blocksize) {
3203 /* Validate the filesystem blocksize */
3204 if (!sb_set_blocksize(sb, blocksize)) {
3205 ext4_msg(sb, KERN_ERR, "bad block size %d",
3211 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3212 offset = do_div(logical_sb_block, blocksize);
3213 bh = sb_bread(sb, logical_sb_block);
3215 ext4_msg(sb, KERN_ERR,
3216 "Can't read superblock on 2nd try");
3219 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3221 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3222 ext4_msg(sb, KERN_ERR,
3223 "Magic mismatch, very weird!");
3228 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3229 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3230 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3232 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3234 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3235 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3236 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3238 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3239 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3240 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3241 (!is_power_of_2(sbi->s_inode_size)) ||
3242 (sbi->s_inode_size > blocksize)) {
3243 ext4_msg(sb, KERN_ERR,
3244 "unsupported inode size: %d",
3248 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3249 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3252 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3253 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3254 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3255 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3256 !is_power_of_2(sbi->s_desc_size)) {
3257 ext4_msg(sb, KERN_ERR,
3258 "unsupported descriptor size %lu",
3263 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3265 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3266 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3267 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3270 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3271 if (sbi->s_inodes_per_block == 0)
3273 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3274 sbi->s_inodes_per_block;
3275 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3277 sbi->s_mount_state = le16_to_cpu(es->s_state);
3278 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3279 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3281 for (i = 0; i < 4; i++)
3282 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3283 sbi->s_def_hash_version = es->s_def_hash_version;
3284 i = le32_to_cpu(es->s_flags);
3285 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3286 sbi->s_hash_unsigned = 3;
3287 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3288 #ifdef __CHAR_UNSIGNED__
3289 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3290 sbi->s_hash_unsigned = 3;
3292 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3296 /* Handle clustersize */
3297 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3298 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3299 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3301 if (clustersize < blocksize) {
3302 ext4_msg(sb, KERN_ERR,
3303 "cluster size (%d) smaller than "
3304 "block size (%d)", clustersize, blocksize);
3307 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3308 le32_to_cpu(es->s_log_block_size);
3309 sbi->s_clusters_per_group =
3310 le32_to_cpu(es->s_clusters_per_group);
3311 if (sbi->s_clusters_per_group > blocksize * 8) {
3312 ext4_msg(sb, KERN_ERR,
3313 "#clusters per group too big: %lu",
3314 sbi->s_clusters_per_group);
3317 if (sbi->s_blocks_per_group !=
3318 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3319 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3320 "clusters per group (%lu) inconsistent",
3321 sbi->s_blocks_per_group,
3322 sbi->s_clusters_per_group);
3326 if (clustersize != blocksize) {
3327 ext4_warning(sb, "fragment/cluster size (%d) != "
3328 "block size (%d)", clustersize,
3330 clustersize = blocksize;
3332 if (sbi->s_blocks_per_group > blocksize * 8) {
3333 ext4_msg(sb, KERN_ERR,
3334 "#blocks per group too big: %lu",
3335 sbi->s_blocks_per_group);
3338 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3339 sbi->s_cluster_bits = 0;
3341 sbi->s_cluster_ratio = clustersize / blocksize;
3343 if (sbi->s_inodes_per_group > blocksize * 8) {
3344 ext4_msg(sb, KERN_ERR,
3345 "#inodes per group too big: %lu",
3346 sbi->s_inodes_per_group);
3351 * Test whether we have more sectors than will fit in sector_t,
3352 * and whether the max offset is addressable by the page cache.
3354 err = generic_check_addressable(sb->s_blocksize_bits,
3355 ext4_blocks_count(es));
3357 ext4_msg(sb, KERN_ERR, "filesystem"
3358 " too large to mount safely on this system");
3359 if (sizeof(sector_t) < 8)
3360 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3365 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3368 /* check blocks count against device size */
3369 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3370 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3371 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3372 "exceeds size of device (%llu blocks)",
3373 ext4_blocks_count(es), blocks_count);
3378 * It makes no sense for the first data block to be beyond the end
3379 * of the filesystem.
3381 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3382 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3383 "block %u is beyond end of filesystem (%llu)",
3384 le32_to_cpu(es->s_first_data_block),
3385 ext4_blocks_count(es));
3388 blocks_count = (ext4_blocks_count(es) -
3389 le32_to_cpu(es->s_first_data_block) +
3390 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3391 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3392 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3393 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3394 "(block count %llu, first data block %u, "
3395 "blocks per group %lu)", sbi->s_groups_count,
3396 ext4_blocks_count(es),
3397 le32_to_cpu(es->s_first_data_block),
3398 EXT4_BLOCKS_PER_GROUP(sb));
3401 sbi->s_groups_count = blocks_count;
3402 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3403 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3404 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3405 EXT4_DESC_PER_BLOCK(sb);
3406 sbi->s_group_desc = ext4_kvmalloc(db_count *
3407 sizeof(struct buffer_head *),
3409 if (sbi->s_group_desc == NULL) {
3410 ext4_msg(sb, KERN_ERR, "not enough memory");
3415 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3418 proc_create_data("options", S_IRUGO, sbi->s_proc,
3419 &ext4_seq_options_fops, sb);
3421 bgl_lock_init(sbi->s_blockgroup_lock);
3423 for (i = 0; i < db_count; i++) {
3424 block = descriptor_loc(sb, logical_sb_block, i);
3425 sbi->s_group_desc[i] = sb_bread(sb, block);
3426 if (!sbi->s_group_desc[i]) {
3427 ext4_msg(sb, KERN_ERR,
3428 "can't read group descriptor %d", i);
3433 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3434 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3437 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3438 if (!ext4_fill_flex_info(sb)) {
3439 ext4_msg(sb, KERN_ERR,
3440 "unable to initialize "
3441 "flex_bg meta info!");
3445 sbi->s_gdb_count = db_count;
3446 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3447 spin_lock_init(&sbi->s_next_gen_lock);
3449 init_timer(&sbi->s_err_report);
3450 sbi->s_err_report.function = print_daily_error_info;
3451 sbi->s_err_report.data = (unsigned long) sb;
3453 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3454 ext4_count_free_clusters(sb));
3456 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3457 ext4_count_free_inodes(sb));
3460 err = percpu_counter_init(&sbi->s_dirs_counter,
3461 ext4_count_dirs(sb));
3464 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3467 ext4_msg(sb, KERN_ERR, "insufficient memory");
3471 sbi->s_stripe = ext4_get_stripe_size(sbi);
3472 sbi->s_max_writeback_mb_bump = 128;
3475 * set up enough so that it can read an inode
3477 if (!test_opt(sb, NOLOAD) &&
3478 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3479 sb->s_op = &ext4_sops;
3481 sb->s_op = &ext4_nojournal_sops;
3482 sb->s_export_op = &ext4_export_ops;
3483 sb->s_xattr = ext4_xattr_handlers;
3485 sb->s_qcop = &ext4_qctl_operations;
3486 sb->dq_op = &ext4_quota_operations;
3488 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3490 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3491 mutex_init(&sbi->s_orphan_lock);
3492 sbi->s_resize_flags = 0;
3496 needs_recovery = (es->s_last_orphan != 0 ||
3497 EXT4_HAS_INCOMPAT_FEATURE(sb,
3498 EXT4_FEATURE_INCOMPAT_RECOVER));
3500 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3501 !(sb->s_flags & MS_RDONLY))
3502 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3506 * The first inode we look at is the journal inode. Don't try
3507 * root first: it may be modified in the journal!
3509 if (!test_opt(sb, NOLOAD) &&
3510 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3511 if (ext4_load_journal(sb, es, journal_devnum))
3513 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3514 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3515 ext4_msg(sb, KERN_ERR, "required journal recovery "
3516 "suppressed and not mounted read-only");
3517 goto failed_mount_wq;
3519 clear_opt(sb, DATA_FLAGS);
3520 sbi->s_journal = NULL;
3525 if (ext4_blocks_count(es) > 0xffffffffULL &&
3526 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3527 JBD2_FEATURE_INCOMPAT_64BIT)) {
3528 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3529 goto failed_mount_wq;
3532 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3533 jbd2_journal_set_features(sbi->s_journal,
3534 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3535 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3536 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3537 jbd2_journal_set_features(sbi->s_journal,
3538 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3539 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3540 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3542 jbd2_journal_clear_features(sbi->s_journal,
3543 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3544 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3547 /* We have now updated the journal if required, so we can
3548 * validate the data journaling mode. */
3549 switch (test_opt(sb, DATA_FLAGS)) {
3551 /* No mode set, assume a default based on the journal
3552 * capabilities: ORDERED_DATA if the journal can
3553 * cope, else JOURNAL_DATA
3555 if (jbd2_journal_check_available_features
3556 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3557 set_opt(sb, ORDERED_DATA);
3559 set_opt(sb, JOURNAL_DATA);
3562 case EXT4_MOUNT_ORDERED_DATA:
3563 case EXT4_MOUNT_WRITEBACK_DATA:
3564 if (!jbd2_journal_check_available_features
3565 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3566 ext4_msg(sb, KERN_ERR, "Journal does not support "
3567 "requested data journaling mode");
3568 goto failed_mount_wq;
3573 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3575 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3578 * The journal may have updated the bg summary counts, so we
3579 * need to update the global counters.
3581 percpu_counter_set(&sbi->s_freeclusters_counter,
3582 ext4_count_free_clusters(sb));
3583 percpu_counter_set(&sbi->s_freeinodes_counter,
3584 ext4_count_free_inodes(sb));
3585 percpu_counter_set(&sbi->s_dirs_counter,
3586 ext4_count_dirs(sb));
3587 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3591 * The maximum number of concurrent works can be high and
3592 * concurrency isn't really necessary. Limit it to 1.
3594 EXT4_SB(sb)->dio_unwritten_wq =
3595 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3596 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3597 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3598 goto failed_mount_wq;
3602 * The jbd2_journal_load will have done any necessary log recovery,
3603 * so we can safely mount the rest of the filesystem now.
3606 root = ext4_iget(sb, EXT4_ROOT_INO);
3608 ext4_msg(sb, KERN_ERR, "get root inode failed");
3609 ret = PTR_ERR(root);
3613 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3614 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3618 sb->s_root = d_make_root(root);
3620 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3625 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3627 /* determine the minimum size of new large inodes, if present */
3628 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3629 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3630 EXT4_GOOD_OLD_INODE_SIZE;
3631 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3632 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3633 if (sbi->s_want_extra_isize <
3634 le16_to_cpu(es->s_want_extra_isize))
3635 sbi->s_want_extra_isize =
3636 le16_to_cpu(es->s_want_extra_isize);
3637 if (sbi->s_want_extra_isize <
3638 le16_to_cpu(es->s_min_extra_isize))
3639 sbi->s_want_extra_isize =
3640 le16_to_cpu(es->s_min_extra_isize);
3643 /* Check if enough inode space is available */
3644 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3645 sbi->s_inode_size) {
3646 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3647 EXT4_GOOD_OLD_INODE_SIZE;
3648 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3652 err = ext4_setup_system_zone(sb);
3654 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3656 goto failed_mount4a;
3660 err = ext4_mb_init(sb, needs_recovery);
3662 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3667 err = ext4_register_li_request(sb, first_not_zeroed);
3671 sbi->s_kobj.kset = ext4_kset;
3672 init_completion(&sbi->s_kobj_unregister);
3673 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3678 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3679 ext4_orphan_cleanup(sb, es);
3680 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3681 if (needs_recovery) {
3682 ext4_msg(sb, KERN_INFO, "recovery complete");
3683 ext4_mark_recovery_complete(sb, es);
3685 if (EXT4_SB(sb)->s_journal) {
3686 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3687 descr = " journalled data mode";
3688 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3689 descr = " ordered data mode";
3691 descr = " writeback data mode";
3693 descr = "out journal";
3695 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3696 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3697 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3699 if (es->s_error_count)
3700 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3707 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3711 ext4_unregister_li_request(sb);
3713 ext4_mb_release(sb);
3715 ext4_ext_release(sb);
3716 ext4_release_system_zone(sb);
3721 ext4_msg(sb, KERN_ERR, "mount failed");
3722 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3724 if (sbi->s_journal) {
3725 jbd2_journal_destroy(sbi->s_journal);
3726 sbi->s_journal = NULL;
3729 del_timer(&sbi->s_err_report);
3730 if (sbi->s_flex_groups)
3731 ext4_kvfree(sbi->s_flex_groups);
3732 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3733 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3734 percpu_counter_destroy(&sbi->s_dirs_counter);
3735 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3737 kthread_stop(sbi->s_mmp_tsk);
3739 for (i = 0; i < db_count; i++)
3740 brelse(sbi->s_group_desc[i]);
3741 ext4_kvfree(sbi->s_group_desc);
3744 remove_proc_entry("options", sbi->s_proc);
3745 remove_proc_entry(sb->s_id, ext4_proc_root);
3748 for (i = 0; i < MAXQUOTAS; i++)
3749 kfree(sbi->s_qf_names[i]);
3751 ext4_blkdev_remove(sbi);
3754 sb->s_fs_info = NULL;
3755 kfree(sbi->s_blockgroup_lock);
3763 * Setup any per-fs journal parameters now. We'll do this both on
3764 * initial mount, once the journal has been initialised but before we've
3765 * done any recovery; and again on any subsequent remount.
3767 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3769 struct ext4_sb_info *sbi = EXT4_SB(sb);
3771 journal->j_commit_interval = sbi->s_commit_interval;
3772 journal->j_min_batch_time = sbi->s_min_batch_time;
3773 journal->j_max_batch_time = sbi->s_max_batch_time;
3775 write_lock(&journal->j_state_lock);
3776 if (test_opt(sb, BARRIER))
3777 journal->j_flags |= JBD2_BARRIER;
3779 journal->j_flags &= ~JBD2_BARRIER;
3780 if (test_opt(sb, DATA_ERR_ABORT))
3781 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3783 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3784 write_unlock(&journal->j_state_lock);
3787 static journal_t *ext4_get_journal(struct super_block *sb,
3788 unsigned int journal_inum)
3790 struct inode *journal_inode;
3793 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3795 /* First, test for the existence of a valid inode on disk. Bad
3796 * things happen if we iget() an unused inode, as the subsequent
3797 * iput() will try to delete it. */
3799 journal_inode = ext4_iget(sb, journal_inum);
3800 if (IS_ERR(journal_inode)) {
3801 ext4_msg(sb, KERN_ERR, "no journal found");
3804 if (!journal_inode->i_nlink) {
3805 make_bad_inode(journal_inode);
3806 iput(journal_inode);
3807 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3811 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3812 journal_inode, journal_inode->i_size);
3813 if (!S_ISREG(journal_inode->i_mode)) {
3814 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3815 iput(journal_inode);
3819 journal = jbd2_journal_init_inode(journal_inode);
3821 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3822 iput(journal_inode);
3825 journal->j_private = sb;
3826 ext4_init_journal_params(sb, journal);
3830 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3833 struct buffer_head *bh;
3837 int hblock, blocksize;
3838 ext4_fsblk_t sb_block;
3839 unsigned long offset;
3840 struct ext4_super_block *es;
3841 struct block_device *bdev;
3843 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3845 bdev = ext4_blkdev_get(j_dev, sb);
3849 blocksize = sb->s_blocksize;
3850 hblock = bdev_logical_block_size(bdev);
3851 if (blocksize < hblock) {
3852 ext4_msg(sb, KERN_ERR,
3853 "blocksize too small for journal device");
3857 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3858 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3859 set_blocksize(bdev, blocksize);
3860 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3861 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3862 "external journal");
3866 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3867 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3868 !(le32_to_cpu(es->s_feature_incompat) &
3869 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3870 ext4_msg(sb, KERN_ERR, "external journal has "
3876 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3877 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3882 len = ext4_blocks_count(es);
3883 start = sb_block + 1;
3884 brelse(bh); /* we're done with the superblock */
3886 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3887 start, len, blocksize);
3889 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3892 journal->j_private = sb;
3893 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3894 wait_on_buffer(journal->j_sb_buffer);
3895 if (!buffer_uptodate(journal->j_sb_buffer)) {
3896 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3899 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3900 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3901 "user (unsupported) - %d",
3902 be32_to_cpu(journal->j_superblock->s_nr_users));
3905 EXT4_SB(sb)->journal_bdev = bdev;
3906 ext4_init_journal_params(sb, journal);
3910 jbd2_journal_destroy(journal);
3912 ext4_blkdev_put(bdev);
3916 static int ext4_load_journal(struct super_block *sb,
3917 struct ext4_super_block *es,
3918 unsigned long journal_devnum)
3921 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3924 int really_read_only;
3926 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3928 if (journal_devnum &&
3929 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3930 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3931 "numbers have changed");
3932 journal_dev = new_decode_dev(journal_devnum);
3934 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3936 really_read_only = bdev_read_only(sb->s_bdev);
3939 * Are we loading a blank journal or performing recovery after a
3940 * crash? For recovery, we need to check in advance whether we
3941 * can get read-write access to the device.
3943 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3944 if (sb->s_flags & MS_RDONLY) {
3945 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3946 "required on readonly filesystem");
3947 if (really_read_only) {
3948 ext4_msg(sb, KERN_ERR, "write access "
3949 "unavailable, cannot proceed");
3952 ext4_msg(sb, KERN_INFO, "write access will "
3953 "be enabled during recovery");
3957 if (journal_inum && journal_dev) {
3958 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3959 "and inode journals!");
3964 if (!(journal = ext4_get_journal(sb, journal_inum)))
3967 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3971 if (!(journal->j_flags & JBD2_BARRIER))
3972 ext4_msg(sb, KERN_INFO, "barriers disabled");
3974 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3975 err = jbd2_journal_wipe(journal, !really_read_only);
3977 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3979 memcpy(save, ((char *) es) +
3980 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3981 err = jbd2_journal_load(journal);
3983 memcpy(((char *) es) + EXT4_S_ERR_START,
3984 save, EXT4_S_ERR_LEN);
3989 ext4_msg(sb, KERN_ERR, "error loading journal");
3990 jbd2_journal_destroy(journal);
3994 EXT4_SB(sb)->s_journal = journal;
3995 ext4_clear_journal_err(sb, es);
3997 if (!really_read_only && journal_devnum &&
3998 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3999 es->s_journal_dev = cpu_to_le32(journal_devnum);
4001 /* Make sure we flush the recovery flag to disk. */
4002 ext4_commit_super(sb, 1);
4008 static int ext4_commit_super(struct super_block *sb, int sync)
4010 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4011 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4014 if (!sbh || block_device_ejected(sb))
4016 if (buffer_write_io_error(sbh)) {
4018 * Oh, dear. A previous attempt to write the
4019 * superblock failed. This could happen because the
4020 * USB device was yanked out. Or it could happen to
4021 * be a transient write error and maybe the block will
4022 * be remapped. Nothing we can do but to retry the
4023 * write and hope for the best.
4025 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4026 "superblock detected");
4027 clear_buffer_write_io_error(sbh);
4028 set_buffer_uptodate(sbh);
4031 * If the file system is mounted read-only, don't update the
4032 * superblock write time. This avoids updating the superblock
4033 * write time when we are mounting the root file system
4034 * read/only but we need to replay the journal; at that point,
4035 * for people who are east of GMT and who make their clock
4036 * tick in localtime for Windows bug-for-bug compatibility,
4037 * the clock is set in the future, and this will cause e2fsck
4038 * to complain and force a full file system check.
4040 if (!(sb->s_flags & MS_RDONLY))
4041 es->s_wtime = cpu_to_le32(get_seconds());
4042 if (sb->s_bdev->bd_part)
4043 es->s_kbytes_written =
4044 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4045 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4046 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4048 es->s_kbytes_written =
4049 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4050 ext4_free_blocks_count_set(es,
4051 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4052 &EXT4_SB(sb)->s_freeclusters_counter)));
4053 es->s_free_inodes_count =
4054 cpu_to_le32(percpu_counter_sum_positive(
4055 &EXT4_SB(sb)->s_freeinodes_counter));
4057 BUFFER_TRACE(sbh, "marking dirty");
4058 mark_buffer_dirty(sbh);
4060 error = sync_dirty_buffer(sbh);
4064 error = buffer_write_io_error(sbh);
4066 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4068 clear_buffer_write_io_error(sbh);
4069 set_buffer_uptodate(sbh);
4076 * Have we just finished recovery? If so, and if we are mounting (or
4077 * remounting) the filesystem readonly, then we will end up with a
4078 * consistent fs on disk. Record that fact.
4080 static void ext4_mark_recovery_complete(struct super_block *sb,
4081 struct ext4_super_block *es)
4083 journal_t *journal = EXT4_SB(sb)->s_journal;
4085 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4086 BUG_ON(journal != NULL);
4089 jbd2_journal_lock_updates(journal);
4090 if (jbd2_journal_flush(journal) < 0)
4093 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4094 sb->s_flags & MS_RDONLY) {
4095 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4096 ext4_commit_super(sb, 1);
4100 jbd2_journal_unlock_updates(journal);
4104 * If we are mounting (or read-write remounting) a filesystem whose journal
4105 * has recorded an error from a previous lifetime, move that error to the
4106 * main filesystem now.
4108 static void ext4_clear_journal_err(struct super_block *sb,
4109 struct ext4_super_block *es)
4115 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4117 journal = EXT4_SB(sb)->s_journal;
4120 * Now check for any error status which may have been recorded in the
4121 * journal by a prior ext4_error() or ext4_abort()
4124 j_errno = jbd2_journal_errno(journal);
4128 errstr = ext4_decode_error(sb, j_errno, nbuf);
4129 ext4_warning(sb, "Filesystem error recorded "
4130 "from previous mount: %s", errstr);
4131 ext4_warning(sb, "Marking fs in need of filesystem check.");
4133 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4134 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4135 ext4_commit_super(sb, 1);
4137 jbd2_journal_clear_err(journal);
4142 * Force the running and committing transactions to commit,
4143 * and wait on the commit.
4145 int ext4_force_commit(struct super_block *sb)
4150 if (sb->s_flags & MS_RDONLY)
4153 journal = EXT4_SB(sb)->s_journal;
4155 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4156 ret = ext4_journal_force_commit(journal);
4162 static void ext4_write_super(struct super_block *sb)
4165 ext4_commit_super(sb, 1);
4169 static int ext4_sync_fs(struct super_block *sb, int wait)
4173 struct ext4_sb_info *sbi = EXT4_SB(sb);
4175 trace_ext4_sync_fs(sb, wait);
4176 flush_workqueue(sbi->dio_unwritten_wq);
4177 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4179 jbd2_log_wait_commit(sbi->s_journal, target);
4185 * LVM calls this function before a (read-only) snapshot is created. This
4186 * gives us a chance to flush the journal completely and mark the fs clean.
4188 * Note that only this function cannot bring a filesystem to be in a clean
4189 * state independently, because ext4 prevents a new handle from being started
4190 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4193 static int ext4_freeze(struct super_block *sb)
4198 if (sb->s_flags & MS_RDONLY)
4201 journal = EXT4_SB(sb)->s_journal;
4203 /* Now we set up the journal barrier. */
4204 jbd2_journal_lock_updates(journal);
4207 * Don't clear the needs_recovery flag if we failed to flush
4210 error = jbd2_journal_flush(journal);
4214 /* Journal blocked and flushed, clear needs_recovery flag. */
4215 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4216 error = ext4_commit_super(sb, 1);
4218 /* we rely on s_frozen to stop further updates */
4219 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4224 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4225 * flag here, even though the filesystem is not technically dirty yet.
4227 static int ext4_unfreeze(struct super_block *sb)
4229 if (sb->s_flags & MS_RDONLY)
4233 /* Reset the needs_recovery flag before the fs is unlocked. */
4234 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4235 ext4_commit_super(sb, 1);
4241 * Structure to save mount options for ext4_remount's benefit
4243 struct ext4_mount_options {
4244 unsigned long s_mount_opt;
4245 unsigned long s_mount_opt2;
4248 unsigned long s_commit_interval;
4249 u32 s_min_batch_time, s_max_batch_time;
4252 char *s_qf_names[MAXQUOTAS];
4256 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4258 struct ext4_super_block *es;
4259 struct ext4_sb_info *sbi = EXT4_SB(sb);
4260 unsigned long old_sb_flags;
4261 struct ext4_mount_options old_opts;
4262 int enable_quota = 0;
4264 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4269 char *orig_data = kstrdup(data, GFP_KERNEL);
4271 /* Store the original options */
4273 old_sb_flags = sb->s_flags;
4274 old_opts.s_mount_opt = sbi->s_mount_opt;
4275 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4276 old_opts.s_resuid = sbi->s_resuid;
4277 old_opts.s_resgid = sbi->s_resgid;
4278 old_opts.s_commit_interval = sbi->s_commit_interval;
4279 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4280 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4282 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4283 for (i = 0; i < MAXQUOTAS; i++)
4284 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4286 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4287 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4290 * Allow the "check" option to be passed as a remount option.
4292 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4297 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4298 ext4_abort(sb, "Abort forced by user");
4300 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4301 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4305 if (sbi->s_journal) {
4306 ext4_init_journal_params(sb, sbi->s_journal);
4307 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4310 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4311 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4316 if (*flags & MS_RDONLY) {
4317 err = dquot_suspend(sb, -1);
4322 * First of all, the unconditional stuff we have to do
4323 * to disable replay of the journal when we next remount
4325 sb->s_flags |= MS_RDONLY;
4328 * OK, test if we are remounting a valid rw partition
4329 * readonly, and if so set the rdonly flag and then
4330 * mark the partition as valid again.
4332 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4333 (sbi->s_mount_state & EXT4_VALID_FS))
4334 es->s_state = cpu_to_le16(sbi->s_mount_state);
4337 ext4_mark_recovery_complete(sb, es);
4339 /* Make sure we can mount this feature set readwrite */
4340 if (!ext4_feature_set_ok(sb, 0)) {
4345 * Make sure the group descriptor checksums
4346 * are sane. If they aren't, refuse to remount r/w.
4348 for (g = 0; g < sbi->s_groups_count; g++) {
4349 struct ext4_group_desc *gdp =
4350 ext4_get_group_desc(sb, g, NULL);
4352 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4353 ext4_msg(sb, KERN_ERR,
4354 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4355 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4356 le16_to_cpu(gdp->bg_checksum));
4363 * If we have an unprocessed orphan list hanging
4364 * around from a previously readonly bdev mount,
4365 * require a full umount/remount for now.
4367 if (es->s_last_orphan) {
4368 ext4_msg(sb, KERN_WARNING, "Couldn't "
4369 "remount RDWR because of unprocessed "
4370 "orphan inode list. Please "
4371 "umount/remount instead");
4377 * Mounting a RDONLY partition read-write, so reread
4378 * and store the current valid flag. (It may have
4379 * been changed by e2fsck since we originally mounted
4383 ext4_clear_journal_err(sb, es);
4384 sbi->s_mount_state = le16_to_cpu(es->s_state);
4385 if (!ext4_setup_super(sb, es, 0))
4386 sb->s_flags &= ~MS_RDONLY;
4387 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4388 EXT4_FEATURE_INCOMPAT_MMP))
4389 if (ext4_multi_mount_protect(sb,
4390 le64_to_cpu(es->s_mmp_block))) {
4399 * Reinitialize lazy itable initialization thread based on
4402 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4403 ext4_unregister_li_request(sb);
4405 ext4_group_t first_not_zeroed;
4406 first_not_zeroed = ext4_has_uninit_itable(sb);
4407 ext4_register_li_request(sb, first_not_zeroed);
4410 ext4_setup_system_zone(sb);
4411 if (sbi->s_journal == NULL)
4412 ext4_commit_super(sb, 1);
4415 /* Release old quota file names */
4416 for (i = 0; i < MAXQUOTAS; i++)
4417 if (old_opts.s_qf_names[i] &&
4418 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4419 kfree(old_opts.s_qf_names[i]);
4423 dquot_resume(sb, -1);
4425 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4430 sb->s_flags = old_sb_flags;
4431 sbi->s_mount_opt = old_opts.s_mount_opt;
4432 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4433 sbi->s_resuid = old_opts.s_resuid;
4434 sbi->s_resgid = old_opts.s_resgid;
4435 sbi->s_commit_interval = old_opts.s_commit_interval;
4436 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4437 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4439 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4440 for (i = 0; i < MAXQUOTAS; i++) {
4441 if (sbi->s_qf_names[i] &&
4442 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4443 kfree(sbi->s_qf_names[i]);
4444 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4453 * Note: calculating the overhead so we can be compatible with
4454 * historical BSD practice is quite difficult in the face of
4455 * clusters/bigalloc. This is because multiple metadata blocks from
4456 * different block group can end up in the same allocation cluster.
4457 * Calculating the exact overhead in the face of clustered allocation
4458 * requires either O(all block bitmaps) in memory or O(number of block
4459 * groups**2) in time. We will still calculate the superblock for
4460 * older file systems --- and if we come across with a bigalloc file
4461 * system with zero in s_overhead_clusters the estimate will be close to
4462 * correct especially for very large cluster sizes --- but for newer
4463 * file systems, it's better to calculate this figure once at mkfs
4464 * time, and store it in the superblock. If the superblock value is
4465 * present (even for non-bigalloc file systems), we will use it.
4467 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4469 struct super_block *sb = dentry->d_sb;
4470 struct ext4_sb_info *sbi = EXT4_SB(sb);
4471 struct ext4_super_block *es = sbi->s_es;
4472 struct ext4_group_desc *gdp;
4476 if (test_opt(sb, MINIX_DF)) {
4477 sbi->s_overhead_last = 0;
4478 } else if (es->s_overhead_clusters) {
4479 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4480 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4481 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4482 ext4_fsblk_t overhead = 0;
4485 * Compute the overhead (FS structures). This is constant
4486 * for a given filesystem unless the number of block groups
4487 * changes so we cache the previous value until it does.
4491 * All of the blocks before first_data_block are
4494 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4497 * Add the overhead found in each block group
4499 for (i = 0; i < ngroups; i++) {
4500 gdp = ext4_get_group_desc(sb, i, NULL);
4501 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4504 sbi->s_overhead_last = overhead;
4506 sbi->s_blocks_last = ext4_blocks_count(es);
4509 buf->f_type = EXT4_SUPER_MAGIC;
4510 buf->f_bsize = sb->s_blocksize;
4511 buf->f_blocks = (ext4_blocks_count(es) -
4512 EXT4_C2B(sbi, sbi->s_overhead_last));
4513 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4514 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4515 /* prevent underflow in case that few free space is available */
4516 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4517 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4518 if (buf->f_bfree < ext4_r_blocks_count(es))
4520 buf->f_files = le32_to_cpu(es->s_inodes_count);
4521 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4522 buf->f_namelen = EXT4_NAME_LEN;
4523 fsid = le64_to_cpup((void *)es->s_uuid) ^
4524 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4525 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4526 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4531 /* Helper function for writing quotas on sync - we need to start transaction
4532 * before quota file is locked for write. Otherwise the are possible deadlocks:
4533 * Process 1 Process 2
4534 * ext4_create() quota_sync()
4535 * jbd2_journal_start() write_dquot()
4536 * dquot_initialize() down(dqio_mutex)
4537 * down(dqio_mutex) jbd2_journal_start()
4543 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4545 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4548 static int ext4_write_dquot(struct dquot *dquot)
4552 struct inode *inode;
4554 inode = dquot_to_inode(dquot);
4555 handle = ext4_journal_start(inode,
4556 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4558 return PTR_ERR(handle);
4559 ret = dquot_commit(dquot);
4560 err = ext4_journal_stop(handle);
4566 static int ext4_acquire_dquot(struct dquot *dquot)
4571 handle = ext4_journal_start(dquot_to_inode(dquot),
4572 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4574 return PTR_ERR(handle);
4575 ret = dquot_acquire(dquot);
4576 err = ext4_journal_stop(handle);
4582 static int ext4_release_dquot(struct dquot *dquot)
4587 handle = ext4_journal_start(dquot_to_inode(dquot),
4588 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4589 if (IS_ERR(handle)) {
4590 /* Release dquot anyway to avoid endless cycle in dqput() */
4591 dquot_release(dquot);
4592 return PTR_ERR(handle);
4594 ret = dquot_release(dquot);
4595 err = ext4_journal_stop(handle);
4601 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4603 /* Are we journaling quotas? */
4604 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4605 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4606 dquot_mark_dquot_dirty(dquot);
4607 return ext4_write_dquot(dquot);
4609 return dquot_mark_dquot_dirty(dquot);
4613 static int ext4_write_info(struct super_block *sb, int type)
4618 /* Data block + inode block */
4619 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4621 return PTR_ERR(handle);
4622 ret = dquot_commit_info(sb, type);
4623 err = ext4_journal_stop(handle);
4630 * Turn on quotas during mount time - we need to find
4631 * the quota file and such...
4633 static int ext4_quota_on_mount(struct super_block *sb, int type)
4635 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4636 EXT4_SB(sb)->s_jquota_fmt, type);
4640 * Standard function to be called on quota_on
4642 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4647 if (!test_opt(sb, QUOTA))
4650 /* Quotafile not on the same filesystem? */
4651 if (path->dentry->d_sb != sb)
4653 /* Journaling quota? */
4654 if (EXT4_SB(sb)->s_qf_names[type]) {
4655 /* Quotafile not in fs root? */
4656 if (path->dentry->d_parent != sb->s_root)
4657 ext4_msg(sb, KERN_WARNING,
4658 "Quota file not on filesystem root. "
4659 "Journaled quota will not work");
4663 * When we journal data on quota file, we have to flush journal to see
4664 * all updates to the file when we bypass pagecache...
4666 if (EXT4_SB(sb)->s_journal &&
4667 ext4_should_journal_data(path->dentry->d_inode)) {
4669 * We don't need to lock updates but journal_flush() could
4670 * otherwise be livelocked...
4672 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4673 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4674 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4679 return dquot_quota_on(sb, type, format_id, path);
4682 static int ext4_quota_off(struct super_block *sb, int type)
4684 struct inode *inode = sb_dqopt(sb)->files[type];
4687 /* Force all delayed allocation blocks to be allocated.
4688 * Caller already holds s_umount sem */
4689 if (test_opt(sb, DELALLOC))
4690 sync_filesystem(sb);
4695 /* Update modification times of quota files when userspace can
4696 * start looking at them */
4697 handle = ext4_journal_start(inode, 1);
4700 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4701 ext4_mark_inode_dirty(handle, inode);
4702 ext4_journal_stop(handle);
4705 return dquot_quota_off(sb, type);
4708 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4709 * acquiring the locks... As quota files are never truncated and quota code
4710 * itself serializes the operations (and no one else should touch the files)
4711 * we don't have to be afraid of races */
4712 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4713 size_t len, loff_t off)
4715 struct inode *inode = sb_dqopt(sb)->files[type];
4716 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4718 int offset = off & (sb->s_blocksize - 1);
4721 struct buffer_head *bh;
4722 loff_t i_size = i_size_read(inode);
4726 if (off+len > i_size)
4729 while (toread > 0) {
4730 tocopy = sb->s_blocksize - offset < toread ?
4731 sb->s_blocksize - offset : toread;
4732 bh = ext4_bread(NULL, inode, blk, 0, &err);
4735 if (!bh) /* A hole? */
4736 memset(data, 0, tocopy);
4738 memcpy(data, bh->b_data+offset, tocopy);
4748 /* Write to quotafile (we know the transaction is already started and has
4749 * enough credits) */
4750 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4751 const char *data, size_t len, loff_t off)
4753 struct inode *inode = sb_dqopt(sb)->files[type];
4754 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4756 int offset = off & (sb->s_blocksize - 1);
4757 struct buffer_head *bh;
4758 handle_t *handle = journal_current_handle();
4760 if (EXT4_SB(sb)->s_journal && !handle) {
4761 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4762 " cancelled because transaction is not started",
4763 (unsigned long long)off, (unsigned long long)len);
4767 * Since we account only one data block in transaction credits,
4768 * then it is impossible to cross a block boundary.
4770 if (sb->s_blocksize - offset < len) {
4771 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4772 " cancelled because not block aligned",
4773 (unsigned long long)off, (unsigned long long)len);
4777 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4778 bh = ext4_bread(handle, inode, blk, 1, &err);
4781 err = ext4_journal_get_write_access(handle, bh);
4787 memcpy(bh->b_data+offset, data, len);
4788 flush_dcache_page(bh->b_page);
4790 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4794 mutex_unlock(&inode->i_mutex);
4797 if (inode->i_size < off + len) {
4798 i_size_write(inode, off + len);
4799 EXT4_I(inode)->i_disksize = inode->i_size;
4800 ext4_mark_inode_dirty(handle, inode);
4802 mutex_unlock(&inode->i_mutex);
4808 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4809 const char *dev_name, void *data)
4811 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4814 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4815 static inline void register_as_ext2(void)
4817 int err = register_filesystem(&ext2_fs_type);
4820 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4823 static inline void unregister_as_ext2(void)
4825 unregister_filesystem(&ext2_fs_type);
4828 static inline int ext2_feature_set_ok(struct super_block *sb)
4830 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4832 if (sb->s_flags & MS_RDONLY)
4834 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4838 MODULE_ALIAS("ext2");
4840 static inline void register_as_ext2(void) { }
4841 static inline void unregister_as_ext2(void) { }
4842 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4845 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4846 static inline void register_as_ext3(void)
4848 int err = register_filesystem(&ext3_fs_type);
4851 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4854 static inline void unregister_as_ext3(void)
4856 unregister_filesystem(&ext3_fs_type);
4859 static inline int ext3_feature_set_ok(struct super_block *sb)
4861 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4863 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4865 if (sb->s_flags & MS_RDONLY)
4867 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4871 MODULE_ALIAS("ext3");
4873 static inline void register_as_ext3(void) { }
4874 static inline void unregister_as_ext3(void) { }
4875 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4878 static struct file_system_type ext4_fs_type = {
4879 .owner = THIS_MODULE,
4881 .mount = ext4_mount,
4882 .kill_sb = kill_block_super,
4883 .fs_flags = FS_REQUIRES_DEV,
4886 static int __init ext4_init_feat_adverts(void)
4888 struct ext4_features *ef;
4891 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4895 ef->f_kobj.kset = ext4_kset;
4896 init_completion(&ef->f_kobj_unregister);
4897 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4910 static void ext4_exit_feat_adverts(void)
4912 kobject_put(&ext4_feat->f_kobj);
4913 wait_for_completion(&ext4_feat->f_kobj_unregister);
4917 /* Shared across all ext4 file systems */
4918 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4919 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4921 static int __init ext4_init_fs(void)
4925 ext4_li_info = NULL;
4926 mutex_init(&ext4_li_mtx);
4928 ext4_check_flag_values();
4930 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4931 mutex_init(&ext4__aio_mutex[i]);
4932 init_waitqueue_head(&ext4__ioend_wq[i]);
4935 err = ext4_init_pageio();
4938 err = ext4_init_system_zone();
4941 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4944 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4946 err = ext4_init_feat_adverts();
4950 err = ext4_init_mballoc();
4954 err = ext4_init_xattr();
4957 err = init_inodecache();
4962 err = register_filesystem(&ext4_fs_type);
4968 unregister_as_ext2();
4969 unregister_as_ext3();
4970 destroy_inodecache();
4974 ext4_exit_mballoc();
4976 ext4_exit_feat_adverts();
4979 remove_proc_entry("fs/ext4", NULL);
4980 kset_unregister(ext4_kset);
4982 ext4_exit_system_zone();
4988 static void __exit ext4_exit_fs(void)
4990 ext4_destroy_lazyinit_thread();
4991 unregister_as_ext2();
4992 unregister_as_ext3();
4993 unregister_filesystem(&ext4_fs_type);
4994 destroy_inodecache();
4996 ext4_exit_mballoc();
4997 ext4_exit_feat_adverts();
4998 remove_proc_entry("fs/ext4", NULL);
4999 kset_unregister(ext4_kset);
5000 ext4_exit_system_zone();
5004 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5005 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5006 MODULE_LICENSE("GPL");
5007 module_init(ext4_init_fs)
5008 module_exit(ext4_exit_fs)