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 static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 static __le32 ext4_superblock_csum(struct super_block *sb,
126 struct ext4_super_block *es)
128 struct ext4_sb_info *sbi = EXT4_SB(sb);
129 int offset = offsetof(struct ext4_super_block, s_checksum);
132 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
134 return cpu_to_le32(csum);
137 int ext4_superblock_csum_verify(struct super_block *sb,
138 struct ext4_super_block *es)
140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
141 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144 return es->s_checksum == ext4_superblock_csum(sb, es);
147 void ext4_superblock_csum_set(struct super_block *sb,
148 struct ext4_super_block *es)
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
154 es->s_checksum = ext4_superblock_csum(sb, es);
157 void *ext4_kvmalloc(size_t size, gfp_t flags)
161 ret = kmalloc(size, flags);
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
167 void *ext4_kvzalloc(size_t size, gfp_t flags)
171 ret = kzalloc(size, flags);
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 void ext4_kvfree(void *ptr)
179 if (is_vmalloc_addr(ptr))
186 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
194 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
202 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
205 return le32_to_cpu(bg->bg_inode_table_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
210 __u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
218 __u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
226 __u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
234 __u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
242 void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
250 void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
258 void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
266 void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
274 void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
282 void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
290 void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299 /* Just increment the non-pointer handle value */
300 static handle_t *ext4_get_nojournal(void)
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
308 handle = (handle_t *)ref_cnt;
310 current->journal_info = handle;
315 /* Decrement the non-pointer handle value */
316 static void ext4_put_nojournal(handle_t *handle)
318 unsigned long ref_cnt = (unsigned long)handle;
320 BUG_ON(ref_cnt == 0);
323 handle = (handle_t *)ref_cnt;
325 current->journal_info = handle;
329 * Wrappers for jbd2_journal_start/end.
331 * The only special thing we need to do here is to make sure that all
332 * journal_end calls result in the superblock being marked dirty, so
333 * that sync() will call the filesystem's write_super callback if
336 * To avoid j_barrier hold in userspace when a user calls freeze(),
337 * ext4 prevents a new handle from being started by s_frozen, which
338 * is in an upper layer.
340 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
345 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
346 if (sb->s_flags & MS_RDONLY)
347 return ERR_PTR(-EROFS);
349 journal = EXT4_SB(sb)->s_journal;
350 handle = ext4_journal_current_handle();
353 * If a handle has been started, it should be allowed to
354 * finish, otherwise deadlock could happen between freeze
355 * and others(e.g. truncate) due to the restart of the
356 * journal handle if the filesystem is forzen and active
357 * handles are not stopped.
360 vfs_check_frozen(sb, SB_FREEZE_TRANS);
363 return ext4_get_nojournal();
365 * Special case here: if the journal has aborted behind our
366 * backs (eg. EIO in the commit thread), then we still need to
367 * take the FS itself readonly cleanly.
369 if (is_journal_aborted(journal)) {
370 ext4_abort(sb, "Detected aborted journal");
371 return ERR_PTR(-EROFS);
373 return jbd2_journal_start(journal, nblocks);
377 * The only special thing we need to do here is to make sure that all
378 * jbd2_journal_stop calls result in the superblock being marked dirty, so
379 * that sync() will call the filesystem's write_super callback if
382 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
384 struct super_block *sb;
388 if (!ext4_handle_valid(handle)) {
389 ext4_put_nojournal(handle);
392 sb = handle->h_transaction->t_journal->j_private;
394 rc = jbd2_journal_stop(handle);
399 __ext4_std_error(sb, where, line, err);
403 void ext4_journal_abort_handle(const char *caller, unsigned int line,
404 const char *err_fn, struct buffer_head *bh,
405 handle_t *handle, int err)
408 const char *errstr = ext4_decode_error(NULL, err, nbuf);
410 BUG_ON(!ext4_handle_valid(handle));
413 BUFFER_TRACE(bh, "abort");
418 if (is_handle_aborted(handle))
421 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
422 caller, line, errstr, err_fn);
424 jbd2_journal_abort_handle(handle);
427 static void __save_error_info(struct super_block *sb, const char *func,
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
432 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
433 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
434 es->s_last_error_time = cpu_to_le32(get_seconds());
435 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
436 es->s_last_error_line = cpu_to_le32(line);
437 if (!es->s_first_error_time) {
438 es->s_first_error_time = es->s_last_error_time;
439 strncpy(es->s_first_error_func, func,
440 sizeof(es->s_first_error_func));
441 es->s_first_error_line = cpu_to_le32(line);
442 es->s_first_error_ino = es->s_last_error_ino;
443 es->s_first_error_block = es->s_last_error_block;
446 * Start the daily error reporting function if it hasn't been
449 if (!es->s_error_count)
450 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
451 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
454 static void save_error_info(struct super_block *sb, const char *func,
457 __save_error_info(sb, func, line);
458 ext4_commit_super(sb, 1);
462 * The del_gendisk() function uninitializes the disk-specific data
463 * structures, including the bdi structure, without telling anyone
464 * else. Once this happens, any attempt to call mark_buffer_dirty()
465 * (for example, by ext4_commit_super), will cause a kernel OOPS.
466 * This is a kludge to prevent these oops until we can put in a proper
467 * hook in del_gendisk() to inform the VFS and file system layers.
469 static int block_device_ejected(struct super_block *sb)
471 struct inode *bd_inode = sb->s_bdev->bd_inode;
472 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
474 return bdi->dev == NULL;
477 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
479 struct super_block *sb = journal->j_private;
480 struct ext4_sb_info *sbi = EXT4_SB(sb);
481 int error = is_journal_aborted(journal);
482 struct ext4_journal_cb_entry *jce, *tmp;
484 spin_lock(&sbi->s_md_lock);
485 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
486 list_del_init(&jce->jce_list);
487 spin_unlock(&sbi->s_md_lock);
488 jce->jce_func(sb, jce, error);
489 spin_lock(&sbi->s_md_lock);
491 spin_unlock(&sbi->s_md_lock);
494 /* Deal with the reporting of failure conditions on a filesystem such as
495 * inconsistencies detected or read IO failures.
497 * On ext2, we can store the error state of the filesystem in the
498 * superblock. That is not possible on ext4, because we may have other
499 * write ordering constraints on the superblock which prevent us from
500 * writing it out straight away; and given that the journal is about to
501 * be aborted, we can't rely on the current, or future, transactions to
502 * write out the superblock safely.
504 * We'll just use the jbd2_journal_abort() error code to record an error in
505 * the journal instead. On recovery, the journal will complain about
506 * that error until we've noted it down and cleared it.
509 static void ext4_handle_error(struct super_block *sb)
511 if (sb->s_flags & MS_RDONLY)
514 if (!test_opt(sb, ERRORS_CONT)) {
515 journal_t *journal = EXT4_SB(sb)->s_journal;
517 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
519 jbd2_journal_abort(journal, -EIO);
521 if (test_opt(sb, ERRORS_RO)) {
522 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
523 sb->s_flags |= MS_RDONLY;
525 if (test_opt(sb, ERRORS_PANIC))
526 panic("EXT4-fs (device %s): panic forced after error\n",
530 void __ext4_error(struct super_block *sb, const char *function,
531 unsigned int line, const char *fmt, ...)
533 struct va_format vaf;
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
540 sb->s_id, function, line, current->comm, &vaf);
542 save_error_info(sb, function, line);
544 ext4_handle_error(sb);
547 void ext4_error_inode(struct inode *inode, const char *function,
548 unsigned int line, ext4_fsblk_t block,
549 const char *fmt, ...)
552 struct va_format vaf;
553 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
555 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
556 es->s_last_error_block = cpu_to_le64(block);
557 save_error_info(inode->i_sb, function, line);
562 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
563 "inode #%lu: block %llu: comm %s: %pV\n",
564 inode->i_sb->s_id, function, line, inode->i_ino,
565 block, current->comm, &vaf);
567 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
568 "inode #%lu: comm %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, &vaf);
573 ext4_handle_error(inode->i_sb);
576 void ext4_error_file(struct file *file, const char *function,
577 unsigned int line, ext4_fsblk_t block,
578 const char *fmt, ...)
581 struct va_format vaf;
582 struct ext4_super_block *es;
583 struct inode *inode = file->f_dentry->d_inode;
584 char pathname[80], *path;
586 es = EXT4_SB(inode->i_sb)->s_es;
587 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
588 save_error_info(inode->i_sb, function, line);
589 path = d_path(&(file->f_path), pathname, sizeof(pathname));
597 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
598 "block %llu: comm %s: path %s: %pV\n",
599 inode->i_sb->s_id, function, line, inode->i_ino,
600 block, current->comm, path, &vaf);
603 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
604 "comm %s: path %s: %pV\n",
605 inode->i_sb->s_id, function, line, inode->i_ino,
606 current->comm, path, &vaf);
609 ext4_handle_error(inode->i_sb);
612 static const char *ext4_decode_error(struct super_block *sb, int errno,
619 errstr = "IO failure";
622 errstr = "Out of memory";
625 if (!sb || (EXT4_SB(sb)->s_journal &&
626 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
627 errstr = "Journal has aborted";
629 errstr = "Readonly filesystem";
632 /* If the caller passed in an extra buffer for unknown
633 * errors, textualise them now. Else we just return
636 /* Check for truncated error codes... */
637 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
646 /* __ext4_std_error decodes expected errors from journaling functions
647 * automatically and invokes the appropriate error response. */
649 void __ext4_std_error(struct super_block *sb, const char *function,
650 unsigned int line, int errno)
655 /* Special case: if the error is EROFS, and we're not already
656 * inside a transaction, then there's really no point in logging
658 if (errno == -EROFS && journal_current_handle() == NULL &&
659 (sb->s_flags & MS_RDONLY))
662 errstr = ext4_decode_error(sb, errno, nbuf);
663 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
664 sb->s_id, function, line, errstr);
665 save_error_info(sb, function, line);
667 ext4_handle_error(sb);
671 * ext4_abort is a much stronger failure handler than ext4_error. The
672 * abort function may be used to deal with unrecoverable failures such
673 * as journal IO errors or ENOMEM at a critical moment in log management.
675 * We unconditionally force the filesystem into an ABORT|READONLY state,
676 * unless the error response on the fs has been set to panic in which
677 * case we take the easy way out and panic immediately.
680 void __ext4_abort(struct super_block *sb, const char *function,
681 unsigned int line, const char *fmt, ...)
685 save_error_info(sb, function, line);
687 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
693 if ((sb->s_flags & MS_RDONLY) == 0) {
694 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
695 sb->s_flags |= MS_RDONLY;
696 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
697 if (EXT4_SB(sb)->s_journal)
698 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
699 save_error_info(sb, function, line);
701 if (test_opt(sb, ERRORS_PANIC))
702 panic("EXT4-fs panic from previous error\n");
705 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
707 struct va_format vaf;
713 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
717 void __ext4_warning(struct super_block *sb, const char *function,
718 unsigned int line, const char *fmt, ...)
720 struct va_format vaf;
726 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
727 sb->s_id, function, line, &vaf);
731 void __ext4_grp_locked_error(const char *function, unsigned int line,
732 struct super_block *sb, ext4_group_t grp,
733 unsigned long ino, ext4_fsblk_t block,
734 const char *fmt, ...)
738 struct va_format vaf;
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
742 es->s_last_error_ino = cpu_to_le32(ino);
743 es->s_last_error_block = cpu_to_le64(block);
744 __save_error_info(sb, function, line);
750 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
751 sb->s_id, function, line, grp);
753 printk(KERN_CONT "inode %lu: ", ino);
755 printk(KERN_CONT "block %llu:", (unsigned long long) block);
756 printk(KERN_CONT "%pV\n", &vaf);
759 if (test_opt(sb, ERRORS_CONT)) {
760 ext4_commit_super(sb, 0);
764 ext4_unlock_group(sb, grp);
765 ext4_handle_error(sb);
767 * We only get here in the ERRORS_RO case; relocking the group
768 * may be dangerous, but nothing bad will happen since the
769 * filesystem will have already been marked read/only and the
770 * journal has been aborted. We return 1 as a hint to callers
771 * who might what to use the return value from
772 * ext4_grp_locked_error() to distinguish between the
773 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
774 * aggressively from the ext4 function in question, with a
775 * more appropriate error code.
777 ext4_lock_group(sb, grp);
781 void ext4_update_dynamic_rev(struct super_block *sb)
783 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
785 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
789 "updating to rev %d because of new feature flag, "
790 "running e2fsck is recommended",
793 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
794 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
795 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
796 /* leave es->s_feature_*compat flags alone */
797 /* es->s_uuid will be set by e2fsck if empty */
800 * The rest of the superblock fields should be zero, and if not it
801 * means they are likely already in use, so leave them alone. We
802 * can leave it up to e2fsck to clean up any inconsistencies there.
807 * Open the external journal device
809 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
811 struct block_device *bdev;
812 char b[BDEVNAME_SIZE];
814 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
820 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
821 __bdevname(dev, b), PTR_ERR(bdev));
826 * Release the journal device
828 static int ext4_blkdev_put(struct block_device *bdev)
830 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
833 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
835 struct block_device *bdev;
838 bdev = sbi->journal_bdev;
840 ret = ext4_blkdev_put(bdev);
841 sbi->journal_bdev = NULL;
846 static inline struct inode *orphan_list_entry(struct list_head *l)
848 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
851 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
855 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
856 le32_to_cpu(sbi->s_es->s_last_orphan));
858 printk(KERN_ERR "sb_info orphan list:\n");
859 list_for_each(l, &sbi->s_orphan) {
860 struct inode *inode = orphan_list_entry(l);
862 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
863 inode->i_sb->s_id, inode->i_ino, inode,
864 inode->i_mode, inode->i_nlink,
869 static void ext4_put_super(struct super_block *sb)
871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
875 ext4_unregister_li_request(sb);
876 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
878 flush_workqueue(sbi->dio_unwritten_wq);
879 destroy_workqueue(sbi->dio_unwritten_wq);
882 if (sbi->s_journal) {
883 err = jbd2_journal_destroy(sbi->s_journal);
884 sbi->s_journal = NULL;
886 ext4_abort(sb, "Couldn't clean up the journal");
889 del_timer(&sbi->s_err_report);
890 ext4_release_system_zone(sb);
892 ext4_ext_release(sb);
893 ext4_xattr_put_super(sb);
895 if (!(sb->s_flags & MS_RDONLY)) {
896 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
897 es->s_state = cpu_to_le16(sbi->s_mount_state);
899 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
900 ext4_commit_super(sb, 1);
903 remove_proc_entry("options", sbi->s_proc);
904 remove_proc_entry(sb->s_id, ext4_proc_root);
906 kobject_del(&sbi->s_kobj);
908 for (i = 0; i < sbi->s_gdb_count; i++)
909 brelse(sbi->s_group_desc[i]);
910 ext4_kvfree(sbi->s_group_desc);
911 ext4_kvfree(sbi->s_flex_groups);
912 percpu_counter_destroy(&sbi->s_freeclusters_counter);
913 percpu_counter_destroy(&sbi->s_freeinodes_counter);
914 percpu_counter_destroy(&sbi->s_dirs_counter);
915 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
918 for (i = 0; i < MAXQUOTAS; i++)
919 kfree(sbi->s_qf_names[i]);
922 /* Debugging code just in case the in-memory inode orphan list
923 * isn't empty. The on-disk one can be non-empty if we've
924 * detected an error and taken the fs readonly, but the
925 * in-memory list had better be clean by this point. */
926 if (!list_empty(&sbi->s_orphan))
927 dump_orphan_list(sb, sbi);
928 J_ASSERT(list_empty(&sbi->s_orphan));
930 invalidate_bdev(sb->s_bdev);
931 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
933 * Invalidate the journal device's buffers. We don't want them
934 * floating about in memory - the physical journal device may
935 * hotswapped, and it breaks the `ro-after' testing code.
937 sync_blockdev(sbi->journal_bdev);
938 invalidate_bdev(sbi->journal_bdev);
939 ext4_blkdev_remove(sbi);
942 kthread_stop(sbi->s_mmp_tsk);
943 sb->s_fs_info = NULL;
945 * Now that we are completely done shutting down the
946 * superblock, we need to actually destroy the kobject.
949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
953 kfree(sbi->s_blockgroup_lock);
957 static struct kmem_cache *ext4_inode_cachep;
960 * Called inside transaction, so use GFP_NOFS
962 static struct inode *ext4_alloc_inode(struct super_block *sb)
964 struct ext4_inode_info *ei;
966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
970 ei->vfs_inode.i_version = 1;
971 ei->vfs_inode.i_data.writeback_index = 0;
972 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
973 INIT_LIST_HEAD(&ei->i_prealloc_list);
974 spin_lock_init(&ei->i_prealloc_lock);
975 ei->i_reserved_data_blocks = 0;
976 ei->i_reserved_meta_blocks = 0;
977 ei->i_allocated_meta_blocks = 0;
978 ei->i_da_metadata_calc_len = 0;
979 spin_lock_init(&(ei->i_block_reservation_lock));
981 ei->i_reserved_quota = 0;
984 INIT_LIST_HEAD(&ei->i_completed_io_list);
985 spin_lock_init(&ei->i_completed_io_lock);
986 ei->cur_aio_dio = NULL;
988 ei->i_datasync_tid = 0;
989 atomic_set(&ei->i_ioend_count, 0);
990 atomic_set(&ei->i_aiodio_unwritten, 0);
992 return &ei->vfs_inode;
995 static int ext4_drop_inode(struct inode *inode)
997 int drop = generic_drop_inode(inode);
999 trace_ext4_drop_inode(inode, drop);
1003 static void ext4_i_callback(struct rcu_head *head)
1005 struct inode *inode = container_of(head, struct inode, i_rcu);
1006 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1009 static void ext4_destroy_inode(struct inode *inode)
1011 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1012 ext4_msg(inode->i_sb, KERN_ERR,
1013 "Inode %lu (%p): orphan list check failed!",
1014 inode->i_ino, EXT4_I(inode));
1015 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1016 EXT4_I(inode), sizeof(struct ext4_inode_info),
1020 call_rcu(&inode->i_rcu, ext4_i_callback);
1023 static void init_once(void *foo)
1025 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1027 INIT_LIST_HEAD(&ei->i_orphan);
1028 #ifdef CONFIG_EXT4_FS_XATTR
1029 init_rwsem(&ei->xattr_sem);
1031 init_rwsem(&ei->i_data_sem);
1032 inode_init_once(&ei->vfs_inode);
1035 static int init_inodecache(void)
1037 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1038 sizeof(struct ext4_inode_info),
1039 0, (SLAB_RECLAIM_ACCOUNT|
1042 if (ext4_inode_cachep == NULL)
1047 static void destroy_inodecache(void)
1049 kmem_cache_destroy(ext4_inode_cachep);
1052 void ext4_clear_inode(struct inode *inode)
1054 invalidate_inode_buffers(inode);
1055 end_writeback(inode);
1057 ext4_discard_preallocations(inode);
1058 if (EXT4_I(inode)->jinode) {
1059 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1060 EXT4_I(inode)->jinode);
1061 jbd2_free_inode(EXT4_I(inode)->jinode);
1062 EXT4_I(inode)->jinode = NULL;
1066 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1067 u64 ino, u32 generation)
1069 struct inode *inode;
1071 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1072 return ERR_PTR(-ESTALE);
1073 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1074 return ERR_PTR(-ESTALE);
1076 /* iget isn't really right if the inode is currently unallocated!!
1078 * ext4_read_inode will return a bad_inode if the inode had been
1079 * deleted, so we should be safe.
1081 * Currently we don't know the generation for parent directory, so
1082 * a generation of 0 means "accept any"
1084 inode = ext4_iget(sb, ino);
1086 return ERR_CAST(inode);
1087 if (generation && inode->i_generation != generation) {
1089 return ERR_PTR(-ESTALE);
1095 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1096 int fh_len, int fh_type)
1098 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1099 ext4_nfs_get_inode);
1102 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1103 int fh_len, int fh_type)
1105 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1106 ext4_nfs_get_inode);
1110 * Try to release metadata pages (indirect blocks, directories) which are
1111 * mapped via the block device. Since these pages could have journal heads
1112 * which would prevent try_to_free_buffers() from freeing them, we must use
1113 * jbd2 layer's try_to_free_buffers() function to release them.
1115 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1118 journal_t *journal = EXT4_SB(sb)->s_journal;
1120 WARN_ON(PageChecked(page));
1121 if (!page_has_buffers(page))
1124 return jbd2_journal_try_to_free_buffers(journal, page,
1125 wait & ~__GFP_WAIT);
1126 return try_to_free_buffers(page);
1130 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1131 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1133 static int ext4_write_dquot(struct dquot *dquot);
1134 static int ext4_acquire_dquot(struct dquot *dquot);
1135 static int ext4_release_dquot(struct dquot *dquot);
1136 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1137 static int ext4_write_info(struct super_block *sb, int type);
1138 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1140 static int ext4_quota_off(struct super_block *sb, int type);
1141 static int ext4_quota_on_mount(struct super_block *sb, int type);
1142 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1143 size_t len, loff_t off);
1144 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1145 const char *data, size_t len, loff_t off);
1147 static const struct dquot_operations ext4_quota_operations = {
1148 .get_reserved_space = ext4_get_reserved_space,
1149 .write_dquot = ext4_write_dquot,
1150 .acquire_dquot = ext4_acquire_dquot,
1151 .release_dquot = ext4_release_dquot,
1152 .mark_dirty = ext4_mark_dquot_dirty,
1153 .write_info = ext4_write_info,
1154 .alloc_dquot = dquot_alloc,
1155 .destroy_dquot = dquot_destroy,
1158 static const struct quotactl_ops ext4_qctl_operations = {
1159 .quota_on = ext4_quota_on,
1160 .quota_off = ext4_quota_off,
1161 .quota_sync = dquot_quota_sync,
1162 .get_info = dquot_get_dqinfo,
1163 .set_info = dquot_set_dqinfo,
1164 .get_dqblk = dquot_get_dqblk,
1165 .set_dqblk = dquot_set_dqblk
1169 static const struct super_operations ext4_sops = {
1170 .alloc_inode = ext4_alloc_inode,
1171 .destroy_inode = ext4_destroy_inode,
1172 .write_inode = ext4_write_inode,
1173 .dirty_inode = ext4_dirty_inode,
1174 .drop_inode = ext4_drop_inode,
1175 .evict_inode = ext4_evict_inode,
1176 .put_super = ext4_put_super,
1177 .sync_fs = ext4_sync_fs,
1178 .freeze_fs = ext4_freeze,
1179 .unfreeze_fs = ext4_unfreeze,
1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
1182 .show_options = ext4_show_options,
1184 .quota_read = ext4_quota_read,
1185 .quota_write = ext4_quota_write,
1187 .bdev_try_to_free_page = bdev_try_to_free_page,
1190 static const struct super_operations ext4_nojournal_sops = {
1191 .alloc_inode = ext4_alloc_inode,
1192 .destroy_inode = ext4_destroy_inode,
1193 .write_inode = ext4_write_inode,
1194 .dirty_inode = ext4_dirty_inode,
1195 .drop_inode = ext4_drop_inode,
1196 .evict_inode = ext4_evict_inode,
1197 .write_super = ext4_write_super,
1198 .put_super = ext4_put_super,
1199 .statfs = ext4_statfs,
1200 .remount_fs = ext4_remount,
1201 .show_options = ext4_show_options,
1203 .quota_read = ext4_quota_read,
1204 .quota_write = ext4_quota_write,
1206 .bdev_try_to_free_page = bdev_try_to_free_page,
1209 static const struct export_operations ext4_export_ops = {
1210 .fh_to_dentry = ext4_fh_to_dentry,
1211 .fh_to_parent = ext4_fh_to_parent,
1212 .get_parent = ext4_get_parent,
1216 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1217 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1218 Opt_nouid32, Opt_debug, Opt_removed,
1219 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1220 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1221 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1222 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1223 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1224 Opt_data_err_abort, Opt_data_err_ignore,
1225 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1226 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1227 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1228 Opt_usrquota, Opt_grpquota, Opt_i_version,
1229 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1230 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1231 Opt_inode_readahead_blks, Opt_journal_ioprio,
1232 Opt_dioread_nolock, Opt_dioread_lock,
1233 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1236 static const match_table_t tokens = {
1237 {Opt_bsd_df, "bsddf"},
1238 {Opt_minix_df, "minixdf"},
1239 {Opt_grpid, "grpid"},
1240 {Opt_grpid, "bsdgroups"},
1241 {Opt_nogrpid, "nogrpid"},
1242 {Opt_nogrpid, "sysvgroups"},
1243 {Opt_resgid, "resgid=%u"},
1244 {Opt_resuid, "resuid=%u"},
1246 {Opt_err_cont, "errors=continue"},
1247 {Opt_err_panic, "errors=panic"},
1248 {Opt_err_ro, "errors=remount-ro"},
1249 {Opt_nouid32, "nouid32"},
1250 {Opt_debug, "debug"},
1251 {Opt_removed, "oldalloc"},
1252 {Opt_removed, "orlov"},
1253 {Opt_user_xattr, "user_xattr"},
1254 {Opt_nouser_xattr, "nouser_xattr"},
1256 {Opt_noacl, "noacl"},
1257 {Opt_noload, "norecovery"},
1258 {Opt_noload, "noload"},
1259 {Opt_removed, "nobh"},
1260 {Opt_removed, "bh"},
1261 {Opt_commit, "commit=%u"},
1262 {Opt_min_batch_time, "min_batch_time=%u"},
1263 {Opt_max_batch_time, "max_batch_time=%u"},
1264 {Opt_journal_dev, "journal_dev=%u"},
1265 {Opt_journal_checksum, "journal_checksum"},
1266 {Opt_journal_async_commit, "journal_async_commit"},
1267 {Opt_abort, "abort"},
1268 {Opt_data_journal, "data=journal"},
1269 {Opt_data_ordered, "data=ordered"},
1270 {Opt_data_writeback, "data=writeback"},
1271 {Opt_data_err_abort, "data_err=abort"},
1272 {Opt_data_err_ignore, "data_err=ignore"},
1273 {Opt_offusrjquota, "usrjquota="},
1274 {Opt_usrjquota, "usrjquota=%s"},
1275 {Opt_offgrpjquota, "grpjquota="},
1276 {Opt_grpjquota, "grpjquota=%s"},
1277 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1278 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1279 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1280 {Opt_grpquota, "grpquota"},
1281 {Opt_noquota, "noquota"},
1282 {Opt_quota, "quota"},
1283 {Opt_usrquota, "usrquota"},
1284 {Opt_barrier, "barrier=%u"},
1285 {Opt_barrier, "barrier"},
1286 {Opt_nobarrier, "nobarrier"},
1287 {Opt_i_version, "i_version"},
1288 {Opt_stripe, "stripe=%u"},
1289 {Opt_delalloc, "delalloc"},
1290 {Opt_nodelalloc, "nodelalloc"},
1291 {Opt_mblk_io_submit, "mblk_io_submit"},
1292 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1293 {Opt_block_validity, "block_validity"},
1294 {Opt_noblock_validity, "noblock_validity"},
1295 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1296 {Opt_journal_ioprio, "journal_ioprio=%u"},
1297 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1298 {Opt_auto_da_alloc, "auto_da_alloc"},
1299 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1300 {Opt_dioread_nolock, "dioread_nolock"},
1301 {Opt_dioread_lock, "dioread_lock"},
1302 {Opt_discard, "discard"},
1303 {Opt_nodiscard, "nodiscard"},
1304 {Opt_init_itable, "init_itable=%u"},
1305 {Opt_init_itable, "init_itable"},
1306 {Opt_noinit_itable, "noinit_itable"},
1307 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1308 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1309 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1310 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1311 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1315 static ext4_fsblk_t get_sb_block(void **data)
1317 ext4_fsblk_t sb_block;
1318 char *options = (char *) *data;
1320 if (!options || strncmp(options, "sb=", 3) != 0)
1321 return 1; /* Default location */
1324 /* TODO: use simple_strtoll with >32bit ext4 */
1325 sb_block = simple_strtoul(options, &options, 0);
1326 if (*options && *options != ',') {
1327 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1331 if (*options == ',')
1333 *data = (void *) options;
1338 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1339 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1340 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1343 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1345 struct ext4_sb_info *sbi = EXT4_SB(sb);
1348 if (sb_any_quota_loaded(sb) &&
1349 !sbi->s_qf_names[qtype]) {
1350 ext4_msg(sb, KERN_ERR,
1351 "Cannot change journaled "
1352 "quota options when quota turned on");
1355 qname = match_strdup(args);
1357 ext4_msg(sb, KERN_ERR,
1358 "Not enough memory for storing quotafile name");
1361 if (sbi->s_qf_names[qtype] &&
1362 strcmp(sbi->s_qf_names[qtype], qname)) {
1363 ext4_msg(sb, KERN_ERR,
1364 "%s quota file already specified", QTYPE2NAME(qtype));
1368 sbi->s_qf_names[qtype] = qname;
1369 if (strchr(sbi->s_qf_names[qtype], '/')) {
1370 ext4_msg(sb, KERN_ERR,
1371 "quotafile must be on filesystem root");
1372 kfree(sbi->s_qf_names[qtype]);
1373 sbi->s_qf_names[qtype] = NULL;
1380 static int clear_qf_name(struct super_block *sb, int qtype)
1383 struct ext4_sb_info *sbi = EXT4_SB(sb);
1385 if (sb_any_quota_loaded(sb) &&
1386 sbi->s_qf_names[qtype]) {
1387 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1388 " when quota turned on");
1392 * The space will be released later when all options are confirmed
1395 sbi->s_qf_names[qtype] = NULL;
1400 #define MOPT_SET 0x0001
1401 #define MOPT_CLEAR 0x0002
1402 #define MOPT_NOSUPPORT 0x0004
1403 #define MOPT_EXPLICIT 0x0008
1404 #define MOPT_CLEAR_ERR 0x0010
1405 #define MOPT_GTE0 0x0020
1408 #define MOPT_QFMT 0x0040
1410 #define MOPT_Q MOPT_NOSUPPORT
1411 #define MOPT_QFMT MOPT_NOSUPPORT
1413 #define MOPT_DATAJ 0x0080
1415 static const struct mount_opts {
1419 } ext4_mount_opts[] = {
1420 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1421 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1422 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1423 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1424 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1425 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1426 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1427 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1428 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1429 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1430 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1431 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1432 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1433 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1434 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1435 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1436 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1437 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1438 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1441 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1442 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1443 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1444 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1445 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1446 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1447 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1448 {Opt_commit, 0, MOPT_GTE0},
1449 {Opt_max_batch_time, 0, MOPT_GTE0},
1450 {Opt_min_batch_time, 0, MOPT_GTE0},
1451 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1452 {Opt_init_itable, 0, MOPT_GTE0},
1453 {Opt_stripe, 0, MOPT_GTE0},
1454 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1455 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1456 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1457 #ifdef CONFIG_EXT4_FS_XATTR
1458 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1459 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1461 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1462 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1464 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1465 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1466 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1468 {Opt_acl, 0, MOPT_NOSUPPORT},
1469 {Opt_noacl, 0, MOPT_NOSUPPORT},
1471 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1472 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1473 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1474 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1476 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1478 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1479 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1480 {Opt_usrjquota, 0, MOPT_Q},
1481 {Opt_grpjquota, 0, MOPT_Q},
1482 {Opt_offusrjquota, 0, MOPT_Q},
1483 {Opt_offgrpjquota, 0, MOPT_Q},
1484 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1485 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1486 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1490 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1491 substring_t *args, unsigned long *journal_devnum,
1492 unsigned int *journal_ioprio, int is_remount)
1494 struct ext4_sb_info *sbi = EXT4_SB(sb);
1495 const struct mount_opts *m;
1499 if (token == Opt_usrjquota)
1500 return set_qf_name(sb, USRQUOTA, &args[0]);
1501 else if (token == Opt_grpjquota)
1502 return set_qf_name(sb, GRPQUOTA, &args[0]);
1503 else if (token == Opt_offusrjquota)
1504 return clear_qf_name(sb, USRQUOTA);
1505 else if (token == Opt_offgrpjquota)
1506 return clear_qf_name(sb, GRPQUOTA);
1508 if (args->from && match_int(args, &arg))
1512 case Opt_nouser_xattr:
1513 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1516 return 1; /* handled by get_sb_block() */
1518 ext4_msg(sb, KERN_WARNING,
1519 "Ignoring removed %s option", opt);
1522 sbi->s_resuid = arg;
1525 sbi->s_resgid = arg;
1528 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1531 sb->s_flags |= MS_I_VERSION;
1533 case Opt_journal_dev:
1535 ext4_msg(sb, KERN_ERR,
1536 "Cannot specify journal on remount");
1539 *journal_devnum = arg;
1541 case Opt_journal_ioprio:
1542 if (arg < 0 || arg > 7)
1544 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1548 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1549 if (token != m->token)
1551 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1553 if (m->flags & MOPT_EXPLICIT)
1554 set_opt2(sb, EXPLICIT_DELALLOC);
1555 if (m->flags & MOPT_CLEAR_ERR)
1556 clear_opt(sb, ERRORS_MASK);
1557 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1558 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1559 "options when quota turned on");
1563 if (m->flags & MOPT_NOSUPPORT) {
1564 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1565 } else if (token == Opt_commit) {
1567 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1568 sbi->s_commit_interval = HZ * arg;
1569 } else if (token == Opt_max_batch_time) {
1571 arg = EXT4_DEF_MAX_BATCH_TIME;
1572 sbi->s_max_batch_time = arg;
1573 } else if (token == Opt_min_batch_time) {
1574 sbi->s_min_batch_time = arg;
1575 } else if (token == Opt_inode_readahead_blks) {
1576 if (arg > (1 << 30))
1578 if (arg && !is_power_of_2(arg)) {
1579 ext4_msg(sb, KERN_ERR,
1580 "EXT4-fs: inode_readahead_blks"
1581 " must be a power of 2");
1584 sbi->s_inode_readahead_blks = arg;
1585 } else if (token == Opt_init_itable) {
1586 set_opt(sb, INIT_INODE_TABLE);
1588 arg = EXT4_DEF_LI_WAIT_MULT;
1589 sbi->s_li_wait_mult = arg;
1590 } else if (token == Opt_stripe) {
1591 sbi->s_stripe = arg;
1592 } else if (m->flags & MOPT_DATAJ) {
1594 if (!sbi->s_journal)
1595 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1596 else if (test_opt(sb, DATA_FLAGS) !=
1598 ext4_msg(sb, KERN_ERR,
1599 "Cannot change data mode on remount");
1603 clear_opt(sb, DATA_FLAGS);
1604 sbi->s_mount_opt |= m->mount_opt;
1607 } else if (m->flags & MOPT_QFMT) {
1608 if (sb_any_quota_loaded(sb) &&
1609 sbi->s_jquota_fmt != m->mount_opt) {
1610 ext4_msg(sb, KERN_ERR, "Cannot "
1611 "change journaled quota options "
1612 "when quota turned on");
1615 sbi->s_jquota_fmt = m->mount_opt;
1620 if (m->flags & MOPT_CLEAR)
1622 else if (unlikely(!(m->flags & MOPT_SET))) {
1623 ext4_msg(sb, KERN_WARNING,
1624 "buggy handling of option %s", opt);
1629 sbi->s_mount_opt |= m->mount_opt;
1631 sbi->s_mount_opt &= ~m->mount_opt;
1635 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1636 "or missing value", opt);
1640 static int parse_options(char *options, struct super_block *sb,
1641 unsigned long *journal_devnum,
1642 unsigned int *journal_ioprio,
1646 struct ext4_sb_info *sbi = EXT4_SB(sb);
1649 substring_t args[MAX_OPT_ARGS];
1655 while ((p = strsep(&options, ",")) != NULL) {
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1662 args[0].to = args[0].from = 0;
1663 token = match_token(p, tokens, args);
1664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
1669 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1670 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1671 clear_opt(sb, USRQUOTA);
1673 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1674 clear_opt(sb, GRPQUOTA);
1676 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1677 ext4_msg(sb, KERN_ERR, "old and new quota "
1682 if (!sbi->s_jquota_fmt) {
1683 ext4_msg(sb, KERN_ERR, "journaled quota format "
1688 if (sbi->s_jquota_fmt) {
1689 ext4_msg(sb, KERN_ERR, "journaled quota format "
1690 "specified with no journaling "
1699 static inline void ext4_show_quota_options(struct seq_file *seq,
1700 struct super_block *sb)
1702 #if defined(CONFIG_QUOTA)
1703 struct ext4_sb_info *sbi = EXT4_SB(sb);
1705 if (sbi->s_jquota_fmt) {
1708 switch (sbi->s_jquota_fmt) {
1719 seq_printf(seq, ",jqfmt=%s", fmtname);
1722 if (sbi->s_qf_names[USRQUOTA])
1723 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1725 if (sbi->s_qf_names[GRPQUOTA])
1726 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1728 if (test_opt(sb, USRQUOTA))
1729 seq_puts(seq, ",usrquota");
1731 if (test_opt(sb, GRPQUOTA))
1732 seq_puts(seq, ",grpquota");
1736 static const char *token2str(int token)
1738 static const struct match_token *t;
1740 for (t = tokens; t->token != Opt_err; t++)
1741 if (t->token == token && !strchr(t->pattern, '='))
1748 * - it's set to a non-default value OR
1749 * - if the per-sb default is different from the global default
1751 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1754 struct ext4_sb_info *sbi = EXT4_SB(sb);
1755 struct ext4_super_block *es = sbi->s_es;
1756 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1757 const struct mount_opts *m;
1758 char sep = nodefs ? '\n' : ',';
1760 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1761 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1763 if (sbi->s_sb_block != 1)
1764 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1766 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1767 int want_set = m->flags & MOPT_SET;
1768 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1769 (m->flags & MOPT_CLEAR_ERR))
1771 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1772 continue; /* skip if same as the default */
1774 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1775 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1776 continue; /* select Opt_noFoo vs Opt_Foo */
1777 SEQ_OPTS_PRINT("%s", token2str(m->token));
1780 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
1781 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1782 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
1783 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
1784 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1785 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
1786 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1787 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1788 SEQ_OPTS_PUTS("errors=remount-ro");
1789 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1790 SEQ_OPTS_PUTS("errors=continue");
1791 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1792 SEQ_OPTS_PUTS("errors=panic");
1793 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1794 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1795 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1796 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1797 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1798 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1799 if (sb->s_flags & MS_I_VERSION)
1800 SEQ_OPTS_PUTS("i_version");
1801 if (nodefs || sbi->s_stripe)
1802 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1803 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1804 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1805 SEQ_OPTS_PUTS("data=journal");
1806 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1807 SEQ_OPTS_PUTS("data=ordered");
1808 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1809 SEQ_OPTS_PUTS("data=writeback");
1812 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1813 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1814 sbi->s_inode_readahead_blks);
1816 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1817 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1818 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1820 ext4_show_quota_options(seq, sb);
1824 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1826 return _ext4_show_options(seq, root->d_sb, 0);
1829 static int options_seq_show(struct seq_file *seq, void *offset)
1831 struct super_block *sb = seq->private;
1834 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1835 rc = _ext4_show_options(seq, sb, 1);
1836 seq_puts(seq, "\n");
1840 static int options_open_fs(struct inode *inode, struct file *file)
1842 return single_open(file, options_seq_show, PDE(inode)->data);
1845 static const struct file_operations ext4_seq_options_fops = {
1846 .owner = THIS_MODULE,
1847 .open = options_open_fs,
1849 .llseek = seq_lseek,
1850 .release = single_release,
1853 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1856 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1860 ext4_msg(sb, KERN_ERR, "revision level too high, "
1861 "forcing read-only mode");
1866 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1867 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1868 "running e2fsck is recommended");
1869 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1870 ext4_msg(sb, KERN_WARNING,
1871 "warning: mounting fs with errors, "
1872 "running e2fsck is recommended");
1873 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1874 le16_to_cpu(es->s_mnt_count) >=
1875 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1876 ext4_msg(sb, KERN_WARNING,
1877 "warning: maximal mount count reached, "
1878 "running e2fsck is recommended");
1879 else if (le32_to_cpu(es->s_checkinterval) &&
1880 (le32_to_cpu(es->s_lastcheck) +
1881 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: checktime reached, "
1884 "running e2fsck is recommended");
1885 if (!sbi->s_journal)
1886 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1887 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1888 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1889 le16_add_cpu(&es->s_mnt_count, 1);
1890 es->s_mtime = cpu_to_le32(get_seconds());
1891 ext4_update_dynamic_rev(sb);
1893 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1895 ext4_commit_super(sb, 1);
1897 if (test_opt(sb, DEBUG))
1898 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1899 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1901 sbi->s_groups_count,
1902 EXT4_BLOCKS_PER_GROUP(sb),
1903 EXT4_INODES_PER_GROUP(sb),
1904 sbi->s_mount_opt, sbi->s_mount_opt2);
1906 cleancache_init_fs(sb);
1910 static int ext4_fill_flex_info(struct super_block *sb)
1912 struct ext4_sb_info *sbi = EXT4_SB(sb);
1913 struct ext4_group_desc *gdp = NULL;
1914 ext4_group_t flex_group_count;
1915 ext4_group_t flex_group;
1916 unsigned int groups_per_flex = 0;
1920 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1921 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1922 sbi->s_log_groups_per_flex = 0;
1925 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1927 /* We allocate both existing and potentially added groups */
1928 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1929 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1930 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1931 size = flex_group_count * sizeof(struct flex_groups);
1932 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1933 if (sbi->s_flex_groups == NULL) {
1934 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1939 for (i = 0; i < sbi->s_groups_count; i++) {
1940 gdp = ext4_get_group_desc(sb, i, NULL);
1942 flex_group = ext4_flex_group(sbi, i);
1943 atomic_add(ext4_free_inodes_count(sb, gdp),
1944 &sbi->s_flex_groups[flex_group].free_inodes);
1945 atomic_add(ext4_free_group_clusters(sb, gdp),
1946 &sbi->s_flex_groups[flex_group].free_clusters);
1947 atomic_add(ext4_used_dirs_count(sb, gdp),
1948 &sbi->s_flex_groups[flex_group].used_dirs);
1956 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1957 struct ext4_group_desc *gdp)
1961 __le32 le_group = cpu_to_le32(block_group);
1963 if ((sbi->s_es->s_feature_ro_compat &
1964 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1965 /* Use new metadata_csum algorithm */
1969 old_csum = gdp->bg_checksum;
1970 gdp->bg_checksum = 0;
1971 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1973 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1975 gdp->bg_checksum = old_csum;
1977 crc = csum32 & 0xFFFF;
1981 /* old crc16 code */
1982 offset = offsetof(struct ext4_group_desc, bg_checksum);
1984 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1985 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1986 crc = crc16(crc, (__u8 *)gdp, offset);
1987 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1988 /* for checksum of struct ext4_group_desc do the rest...*/
1989 if ((sbi->s_es->s_feature_incompat &
1990 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1991 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1992 crc = crc16(crc, (__u8 *)gdp + offset,
1993 le16_to_cpu(sbi->s_es->s_desc_size) -
1997 return cpu_to_le16(crc);
2000 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2001 struct ext4_group_desc *gdp)
2003 if (ext4_has_group_desc_csum(sb) &&
2004 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2011 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2012 struct ext4_group_desc *gdp)
2014 if (!ext4_has_group_desc_csum(sb))
2016 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2019 /* Called at mount-time, super-block is locked */
2020 static int ext4_check_descriptors(struct super_block *sb,
2021 ext4_group_t *first_not_zeroed)
2023 struct ext4_sb_info *sbi = EXT4_SB(sb);
2024 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2025 ext4_fsblk_t last_block;
2026 ext4_fsblk_t block_bitmap;
2027 ext4_fsblk_t inode_bitmap;
2028 ext4_fsblk_t inode_table;
2029 int flexbg_flag = 0;
2030 ext4_group_t i, grp = sbi->s_groups_count;
2032 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2035 ext4_debug("Checking group descriptors");
2037 for (i = 0; i < sbi->s_groups_count; i++) {
2038 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2040 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2041 last_block = ext4_blocks_count(sbi->s_es) - 1;
2043 last_block = first_block +
2044 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2046 if ((grp == sbi->s_groups_count) &&
2047 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2050 block_bitmap = ext4_block_bitmap(sb, gdp);
2051 if (block_bitmap < first_block || block_bitmap > last_block) {
2052 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2053 "Block bitmap for group %u not in group "
2054 "(block %llu)!", i, block_bitmap);
2057 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2058 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2059 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2060 "Inode bitmap for group %u not in group "
2061 "(block %llu)!", i, inode_bitmap);
2064 inode_table = ext4_inode_table(sb, gdp);
2065 if (inode_table < first_block ||
2066 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2067 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2068 "Inode table for group %u not in group "
2069 "(block %llu)!", i, inode_table);
2072 ext4_lock_group(sb, i);
2073 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2074 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2075 "Checksum for group %u failed (%u!=%u)",
2076 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2077 gdp)), le16_to_cpu(gdp->bg_checksum));
2078 if (!(sb->s_flags & MS_RDONLY)) {
2079 ext4_unlock_group(sb, i);
2083 ext4_unlock_group(sb, i);
2085 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2087 if (NULL != first_not_zeroed)
2088 *first_not_zeroed = grp;
2090 ext4_free_blocks_count_set(sbi->s_es,
2091 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2092 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2096 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2097 * the superblock) which were deleted from all directories, but held open by
2098 * a process at the time of a crash. We walk the list and try to delete these
2099 * inodes at recovery time (only with a read-write filesystem).
2101 * In order to keep the orphan inode chain consistent during traversal (in
2102 * case of crash during recovery), we link each inode into the superblock
2103 * orphan list_head and handle it the same way as an inode deletion during
2104 * normal operation (which journals the operations for us).
2106 * We only do an iget() and an iput() on each inode, which is very safe if we
2107 * accidentally point at an in-use or already deleted inode. The worst that
2108 * can happen in this case is that we get a "bit already cleared" message from
2109 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2110 * e2fsck was run on this filesystem, and it must have already done the orphan
2111 * inode cleanup for us, so we can safely abort without any further action.
2113 static void ext4_orphan_cleanup(struct super_block *sb,
2114 struct ext4_super_block *es)
2116 unsigned int s_flags = sb->s_flags;
2117 int nr_orphans = 0, nr_truncates = 0;
2121 if (!es->s_last_orphan) {
2122 jbd_debug(4, "no orphan inodes to clean up\n");
2126 if (bdev_read_only(sb->s_bdev)) {
2127 ext4_msg(sb, KERN_ERR, "write access "
2128 "unavailable, skipping orphan cleanup");
2132 /* Check if feature set would not allow a r/w mount */
2133 if (!ext4_feature_set_ok(sb, 0)) {
2134 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2135 "unknown ROCOMPAT features");
2139 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2140 if (es->s_last_orphan)
2141 jbd_debug(1, "Errors on filesystem, "
2142 "clearing orphan list.\n");
2143 es->s_last_orphan = 0;
2144 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2148 if (s_flags & MS_RDONLY) {
2149 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2150 sb->s_flags &= ~MS_RDONLY;
2153 /* Needed for iput() to work correctly and not trash data */
2154 sb->s_flags |= MS_ACTIVE;
2155 /* Turn on quotas so that they are updated correctly */
2156 for (i = 0; i < MAXQUOTAS; i++) {
2157 if (EXT4_SB(sb)->s_qf_names[i]) {
2158 int ret = ext4_quota_on_mount(sb, i);
2160 ext4_msg(sb, KERN_ERR,
2161 "Cannot turn on journaled "
2162 "quota: error %d", ret);
2167 while (es->s_last_orphan) {
2168 struct inode *inode;
2170 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2171 if (IS_ERR(inode)) {
2172 es->s_last_orphan = 0;
2176 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2177 dquot_initialize(inode);
2178 if (inode->i_nlink) {
2179 ext4_msg(sb, KERN_DEBUG,
2180 "%s: truncating inode %lu to %lld bytes",
2181 __func__, inode->i_ino, inode->i_size);
2182 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2183 inode->i_ino, inode->i_size);
2184 ext4_truncate(inode);
2187 ext4_msg(sb, KERN_DEBUG,
2188 "%s: deleting unreferenced inode %lu",
2189 __func__, inode->i_ino);
2190 jbd_debug(2, "deleting unreferenced inode %lu\n",
2194 iput(inode); /* The delete magic happens here! */
2197 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2200 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2201 PLURAL(nr_orphans));
2203 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2204 PLURAL(nr_truncates));
2206 /* Turn quotas off */
2207 for (i = 0; i < MAXQUOTAS; i++) {
2208 if (sb_dqopt(sb)->files[i])
2209 dquot_quota_off(sb, i);
2212 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2216 * Maximal extent format file size.
2217 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2218 * extent format containers, within a sector_t, and within i_blocks
2219 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2220 * so that won't be a limiting factor.
2222 * However there is other limiting factor. We do store extents in the form
2223 * of starting block and length, hence the resulting length of the extent
2224 * covering maximum file size must fit into on-disk format containers as
2225 * well. Given that length is always by 1 unit bigger than max unit (because
2226 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2228 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2230 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2233 loff_t upper_limit = MAX_LFS_FILESIZE;
2235 /* small i_blocks in vfs inode? */
2236 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2238 * CONFIG_LBDAF is not enabled implies the inode
2239 * i_block represent total blocks in 512 bytes
2240 * 32 == size of vfs inode i_blocks * 8
2242 upper_limit = (1LL << 32) - 1;
2244 /* total blocks in file system block size */
2245 upper_limit >>= (blkbits - 9);
2246 upper_limit <<= blkbits;
2250 * 32-bit extent-start container, ee_block. We lower the maxbytes
2251 * by one fs block, so ee_len can cover the extent of maximum file
2254 res = (1LL << 32) - 1;
2257 /* Sanity check against vm- & vfs- imposed limits */
2258 if (res > upper_limit)
2265 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2266 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2267 * We need to be 1 filesystem block less than the 2^48 sector limit.
2269 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2271 loff_t res = EXT4_NDIR_BLOCKS;
2274 /* This is calculated to be the largest file size for a dense, block
2275 * mapped file such that the file's total number of 512-byte sectors,
2276 * including data and all indirect blocks, does not exceed (2^48 - 1).
2278 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2279 * number of 512-byte sectors of the file.
2282 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2284 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2285 * the inode i_block field represents total file blocks in
2286 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2288 upper_limit = (1LL << 32) - 1;
2290 /* total blocks in file system block size */
2291 upper_limit >>= (bits - 9);
2295 * We use 48 bit ext4_inode i_blocks
2296 * With EXT4_HUGE_FILE_FL set the i_blocks
2297 * represent total number of blocks in
2298 * file system block size
2300 upper_limit = (1LL << 48) - 1;
2304 /* indirect blocks */
2306 /* double indirect blocks */
2307 meta_blocks += 1 + (1LL << (bits-2));
2308 /* tripple indirect blocks */
2309 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2311 upper_limit -= meta_blocks;
2312 upper_limit <<= bits;
2314 res += 1LL << (bits-2);
2315 res += 1LL << (2*(bits-2));
2316 res += 1LL << (3*(bits-2));
2318 if (res > upper_limit)
2321 if (res > MAX_LFS_FILESIZE)
2322 res = MAX_LFS_FILESIZE;
2327 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2328 ext4_fsblk_t logical_sb_block, int nr)
2330 struct ext4_sb_info *sbi = EXT4_SB(sb);
2331 ext4_group_t bg, first_meta_bg;
2334 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2336 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2338 return logical_sb_block + nr + 1;
2339 bg = sbi->s_desc_per_block * nr;
2340 if (ext4_bg_has_super(sb, bg))
2343 return (has_super + ext4_group_first_block_no(sb, bg));
2347 * ext4_get_stripe_size: Get the stripe size.
2348 * @sbi: In memory super block info
2350 * If we have specified it via mount option, then
2351 * use the mount option value. If the value specified at mount time is
2352 * greater than the blocks per group use the super block value.
2353 * If the super block value is greater than blocks per group return 0.
2354 * Allocator needs it be less than blocks per group.
2357 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2359 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2360 unsigned long stripe_width =
2361 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2364 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2365 ret = sbi->s_stripe;
2366 else if (stripe_width <= sbi->s_blocks_per_group)
2368 else if (stride <= sbi->s_blocks_per_group)
2374 * If the stripe width is 1, this makes no sense and
2375 * we set it to 0 to turn off stripe handling code.
2386 struct attribute attr;
2387 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2388 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2389 const char *, size_t);
2393 static int parse_strtoul(const char *buf,
2394 unsigned long max, unsigned long *value)
2398 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2399 endp = skip_spaces(endp);
2400 if (*endp || *value > max)
2406 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2410 return snprintf(buf, PAGE_SIZE, "%llu\n",
2412 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2415 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2416 struct ext4_sb_info *sbi, char *buf)
2418 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2420 if (!sb->s_bdev->bd_part)
2421 return snprintf(buf, PAGE_SIZE, "0\n");
2422 return snprintf(buf, PAGE_SIZE, "%lu\n",
2423 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2424 sbi->s_sectors_written_start) >> 1);
2427 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi, char *buf)
2430 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2432 if (!sb->s_bdev->bd_part)
2433 return snprintf(buf, PAGE_SIZE, "0\n");
2434 return snprintf(buf, PAGE_SIZE, "%llu\n",
2435 (unsigned long long)(sbi->s_kbytes_written +
2436 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2437 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2440 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2441 struct ext4_sb_info *sbi,
2442 const char *buf, size_t count)
2446 if (parse_strtoul(buf, 0x40000000, &t))
2449 if (t && !is_power_of_2(t))
2452 sbi->s_inode_readahead_blks = t;
2456 static ssize_t sbi_ui_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi, char *buf)
2459 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2461 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2464 static ssize_t sbi_ui_store(struct ext4_attr *a,
2465 struct ext4_sb_info *sbi,
2466 const char *buf, size_t count)
2468 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2471 if (parse_strtoul(buf, 0xffffffff, &t))
2477 static ssize_t trigger_test_error(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi,
2479 const char *buf, size_t count)
2483 if (!capable(CAP_SYS_ADMIN))
2486 if (len && buf[len-1] == '\n')
2490 ext4_error(sbi->s_sb, "%.*s", len, buf);
2494 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2495 static struct ext4_attr ext4_attr_##_name = { \
2496 .attr = {.name = __stringify(_name), .mode = _mode }, \
2499 .offset = offsetof(struct ext4_sb_info, _elname), \
2501 #define EXT4_ATTR(name, mode, show, store) \
2502 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2504 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2505 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2506 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2507 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2508 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2509 #define ATTR_LIST(name) &ext4_attr_##name.attr
2511 EXT4_RO_ATTR(delayed_allocation_blocks);
2512 EXT4_RO_ATTR(session_write_kbytes);
2513 EXT4_RO_ATTR(lifetime_write_kbytes);
2514 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2515 inode_readahead_blks_store, s_inode_readahead_blks);
2516 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2517 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2518 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2519 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2520 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2521 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2522 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2523 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2524 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2526 static struct attribute *ext4_attrs[] = {
2527 ATTR_LIST(delayed_allocation_blocks),
2528 ATTR_LIST(session_write_kbytes),
2529 ATTR_LIST(lifetime_write_kbytes),
2530 ATTR_LIST(inode_readahead_blks),
2531 ATTR_LIST(inode_goal),
2532 ATTR_LIST(mb_stats),
2533 ATTR_LIST(mb_max_to_scan),
2534 ATTR_LIST(mb_min_to_scan),
2535 ATTR_LIST(mb_order2_req),
2536 ATTR_LIST(mb_stream_req),
2537 ATTR_LIST(mb_group_prealloc),
2538 ATTR_LIST(max_writeback_mb_bump),
2539 ATTR_LIST(trigger_fs_error),
2543 /* Features this copy of ext4 supports */
2544 EXT4_INFO_ATTR(lazy_itable_init);
2545 EXT4_INFO_ATTR(batched_discard);
2547 static struct attribute *ext4_feat_attrs[] = {
2548 ATTR_LIST(lazy_itable_init),
2549 ATTR_LIST(batched_discard),
2553 static ssize_t ext4_attr_show(struct kobject *kobj,
2554 struct attribute *attr, char *buf)
2556 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2558 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2560 return a->show ? a->show(a, sbi, buf) : 0;
2563 static ssize_t ext4_attr_store(struct kobject *kobj,
2564 struct attribute *attr,
2565 const char *buf, size_t len)
2567 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2569 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2571 return a->store ? a->store(a, sbi, buf, len) : 0;
2574 static void ext4_sb_release(struct kobject *kobj)
2576 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2578 complete(&sbi->s_kobj_unregister);
2581 static const struct sysfs_ops ext4_attr_ops = {
2582 .show = ext4_attr_show,
2583 .store = ext4_attr_store,
2586 static struct kobj_type ext4_ktype = {
2587 .default_attrs = ext4_attrs,
2588 .sysfs_ops = &ext4_attr_ops,
2589 .release = ext4_sb_release,
2592 static void ext4_feat_release(struct kobject *kobj)
2594 complete(&ext4_feat->f_kobj_unregister);
2597 static struct kobj_type ext4_feat_ktype = {
2598 .default_attrs = ext4_feat_attrs,
2599 .sysfs_ops = &ext4_attr_ops,
2600 .release = ext4_feat_release,
2604 * Check whether this filesystem can be mounted based on
2605 * the features present and the RDONLY/RDWR mount requested.
2606 * Returns 1 if this filesystem can be mounted as requested,
2607 * 0 if it cannot be.
2609 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2611 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2612 ext4_msg(sb, KERN_ERR,
2613 "Couldn't mount because of "
2614 "unsupported optional features (%x)",
2615 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2616 ~EXT4_FEATURE_INCOMPAT_SUPP));
2623 /* Check that feature set is OK for a read-write mount */
2624 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2625 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2626 "unsupported optional features (%x)",
2627 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2628 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2632 * Large file size enabled file system can only be mounted
2633 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2635 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2636 if (sizeof(blkcnt_t) < sizeof(u64)) {
2637 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2638 "cannot be mounted RDWR without "
2643 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2644 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2645 ext4_msg(sb, KERN_ERR,
2646 "Can't support bigalloc feature without "
2647 "extents feature\n");
2654 * This function is called once a day if we have errors logged
2655 * on the file system
2657 static void print_daily_error_info(unsigned long arg)
2659 struct super_block *sb = (struct super_block *) arg;
2660 struct ext4_sb_info *sbi;
2661 struct ext4_super_block *es;
2666 if (es->s_error_count)
2667 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2668 le32_to_cpu(es->s_error_count));
2669 if (es->s_first_error_time) {
2670 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2671 sb->s_id, le32_to_cpu(es->s_first_error_time),
2672 (int) sizeof(es->s_first_error_func),
2673 es->s_first_error_func,
2674 le32_to_cpu(es->s_first_error_line));
2675 if (es->s_first_error_ino)
2676 printk(": inode %u",
2677 le32_to_cpu(es->s_first_error_ino));
2678 if (es->s_first_error_block)
2679 printk(": block %llu", (unsigned long long)
2680 le64_to_cpu(es->s_first_error_block));
2683 if (es->s_last_error_time) {
2684 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2685 sb->s_id, le32_to_cpu(es->s_last_error_time),
2686 (int) sizeof(es->s_last_error_func),
2687 es->s_last_error_func,
2688 le32_to_cpu(es->s_last_error_line));
2689 if (es->s_last_error_ino)
2690 printk(": inode %u",
2691 le32_to_cpu(es->s_last_error_ino));
2692 if (es->s_last_error_block)
2693 printk(": block %llu", (unsigned long long)
2694 le64_to_cpu(es->s_last_error_block));
2697 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2700 /* Find next suitable group and run ext4_init_inode_table */
2701 static int ext4_run_li_request(struct ext4_li_request *elr)
2703 struct ext4_group_desc *gdp = NULL;
2704 ext4_group_t group, ngroups;
2705 struct super_block *sb;
2706 unsigned long timeout = 0;
2710 ngroups = EXT4_SB(sb)->s_groups_count;
2712 for (group = elr->lr_next_group; group < ngroups; group++) {
2713 gdp = ext4_get_group_desc(sb, group, NULL);
2719 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2723 if (group == ngroups)
2728 ret = ext4_init_inode_table(sb, group,
2729 elr->lr_timeout ? 0 : 1);
2730 if (elr->lr_timeout == 0) {
2731 timeout = (jiffies - timeout) *
2732 elr->lr_sbi->s_li_wait_mult;
2733 elr->lr_timeout = timeout;
2735 elr->lr_next_sched = jiffies + elr->lr_timeout;
2736 elr->lr_next_group = group + 1;
2743 * Remove lr_request from the list_request and free the
2744 * request structure. Should be called with li_list_mtx held
2746 static void ext4_remove_li_request(struct ext4_li_request *elr)
2748 struct ext4_sb_info *sbi;
2755 list_del(&elr->lr_request);
2756 sbi->s_li_request = NULL;
2760 static void ext4_unregister_li_request(struct super_block *sb)
2762 mutex_lock(&ext4_li_mtx);
2763 if (!ext4_li_info) {
2764 mutex_unlock(&ext4_li_mtx);
2768 mutex_lock(&ext4_li_info->li_list_mtx);
2769 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2770 mutex_unlock(&ext4_li_info->li_list_mtx);
2771 mutex_unlock(&ext4_li_mtx);
2774 static struct task_struct *ext4_lazyinit_task;
2777 * This is the function where ext4lazyinit thread lives. It walks
2778 * through the request list searching for next scheduled filesystem.
2779 * When such a fs is found, run the lazy initialization request
2780 * (ext4_rn_li_request) and keep track of the time spend in this
2781 * function. Based on that time we compute next schedule time of
2782 * the request. When walking through the list is complete, compute
2783 * next waking time and put itself into sleep.
2785 static int ext4_lazyinit_thread(void *arg)
2787 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2788 struct list_head *pos, *n;
2789 struct ext4_li_request *elr;
2790 unsigned long next_wakeup, cur;
2792 BUG_ON(NULL == eli);
2796 next_wakeup = MAX_JIFFY_OFFSET;
2798 mutex_lock(&eli->li_list_mtx);
2799 if (list_empty(&eli->li_request_list)) {
2800 mutex_unlock(&eli->li_list_mtx);
2804 list_for_each_safe(pos, n, &eli->li_request_list) {
2805 elr = list_entry(pos, struct ext4_li_request,
2808 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2809 if (ext4_run_li_request(elr) != 0) {
2810 /* error, remove the lazy_init job */
2811 ext4_remove_li_request(elr);
2816 if (time_before(elr->lr_next_sched, next_wakeup))
2817 next_wakeup = elr->lr_next_sched;
2819 mutex_unlock(&eli->li_list_mtx);
2824 if ((time_after_eq(cur, next_wakeup)) ||
2825 (MAX_JIFFY_OFFSET == next_wakeup)) {
2830 schedule_timeout_interruptible(next_wakeup - cur);
2832 if (kthread_should_stop()) {
2833 ext4_clear_request_list();
2840 * It looks like the request list is empty, but we need
2841 * to check it under the li_list_mtx lock, to prevent any
2842 * additions into it, and of course we should lock ext4_li_mtx
2843 * to atomically free the list and ext4_li_info, because at
2844 * this point another ext4 filesystem could be registering
2847 mutex_lock(&ext4_li_mtx);
2848 mutex_lock(&eli->li_list_mtx);
2849 if (!list_empty(&eli->li_request_list)) {
2850 mutex_unlock(&eli->li_list_mtx);
2851 mutex_unlock(&ext4_li_mtx);
2854 mutex_unlock(&eli->li_list_mtx);
2855 kfree(ext4_li_info);
2856 ext4_li_info = NULL;
2857 mutex_unlock(&ext4_li_mtx);
2862 static void ext4_clear_request_list(void)
2864 struct list_head *pos, *n;
2865 struct ext4_li_request *elr;
2867 mutex_lock(&ext4_li_info->li_list_mtx);
2868 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2869 elr = list_entry(pos, struct ext4_li_request,
2871 ext4_remove_li_request(elr);
2873 mutex_unlock(&ext4_li_info->li_list_mtx);
2876 static int ext4_run_lazyinit_thread(void)
2878 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2879 ext4_li_info, "ext4lazyinit");
2880 if (IS_ERR(ext4_lazyinit_task)) {
2881 int err = PTR_ERR(ext4_lazyinit_task);
2882 ext4_clear_request_list();
2883 kfree(ext4_li_info);
2884 ext4_li_info = NULL;
2885 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2886 "initialization thread\n",
2890 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2895 * Check whether it make sense to run itable init. thread or not.
2896 * If there is at least one uninitialized inode table, return
2897 * corresponding group number, else the loop goes through all
2898 * groups and return total number of groups.
2900 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2902 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2903 struct ext4_group_desc *gdp = NULL;
2905 for (group = 0; group < ngroups; group++) {
2906 gdp = ext4_get_group_desc(sb, group, NULL);
2910 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2917 static int ext4_li_info_new(void)
2919 struct ext4_lazy_init *eli = NULL;
2921 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2925 INIT_LIST_HEAD(&eli->li_request_list);
2926 mutex_init(&eli->li_list_mtx);
2928 eli->li_state |= EXT4_LAZYINIT_QUIT;
2935 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2938 struct ext4_sb_info *sbi = EXT4_SB(sb);
2939 struct ext4_li_request *elr;
2942 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2948 elr->lr_next_group = start;
2951 * Randomize first schedule time of the request to
2952 * spread the inode table initialization requests
2955 get_random_bytes(&rnd, sizeof(rnd));
2956 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2957 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2962 static int ext4_register_li_request(struct super_block *sb,
2963 ext4_group_t first_not_zeroed)
2965 struct ext4_sb_info *sbi = EXT4_SB(sb);
2966 struct ext4_li_request *elr;
2967 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2970 if (sbi->s_li_request != NULL) {
2972 * Reset timeout so it can be computed again, because
2973 * s_li_wait_mult might have changed.
2975 sbi->s_li_request->lr_timeout = 0;
2979 if (first_not_zeroed == ngroups ||
2980 (sb->s_flags & MS_RDONLY) ||
2981 !test_opt(sb, INIT_INODE_TABLE))
2984 elr = ext4_li_request_new(sb, first_not_zeroed);
2988 mutex_lock(&ext4_li_mtx);
2990 if (NULL == ext4_li_info) {
2991 ret = ext4_li_info_new();
2996 mutex_lock(&ext4_li_info->li_list_mtx);
2997 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2998 mutex_unlock(&ext4_li_info->li_list_mtx);
3000 sbi->s_li_request = elr;
3002 * set elr to NULL here since it has been inserted to
3003 * the request_list and the removal and free of it is
3004 * handled by ext4_clear_request_list from now on.
3008 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3009 ret = ext4_run_lazyinit_thread();
3014 mutex_unlock(&ext4_li_mtx);
3021 * We do not need to lock anything since this is called on
3024 static void ext4_destroy_lazyinit_thread(void)
3027 * If thread exited earlier
3028 * there's nothing to be done.
3030 if (!ext4_li_info || !ext4_lazyinit_task)
3033 kthread_stop(ext4_lazyinit_task);
3036 static int set_journal_csum_feature_set(struct super_block *sb)
3039 int compat, incompat;
3040 struct ext4_sb_info *sbi = EXT4_SB(sb);
3042 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3043 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3044 /* journal checksum v2 */
3046 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3048 /* journal checksum v1 */
3049 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3053 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3054 ret = jbd2_journal_set_features(sbi->s_journal,
3056 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3058 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3059 ret = jbd2_journal_set_features(sbi->s_journal,
3062 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3063 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3065 jbd2_journal_clear_features(sbi->s_journal,
3066 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3067 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3068 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3074 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3076 char *orig_data = kstrdup(data, GFP_KERNEL);
3077 struct buffer_head *bh;
3078 struct ext4_super_block *es = NULL;
3079 struct ext4_sb_info *sbi;
3081 ext4_fsblk_t sb_block = get_sb_block(&data);
3082 ext4_fsblk_t logical_sb_block;
3083 unsigned long offset = 0;
3084 unsigned long journal_devnum = 0;
3085 unsigned long def_mount_opts;
3090 int blocksize, clustersize;
3091 unsigned int db_count;
3093 int needs_recovery, has_huge_files, has_bigalloc;
3096 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3097 ext4_group_t first_not_zeroed;
3099 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3103 sbi->s_blockgroup_lock =
3104 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3105 if (!sbi->s_blockgroup_lock) {
3109 sb->s_fs_info = sbi;
3111 sbi->s_mount_opt = 0;
3112 sbi->s_resuid = EXT4_DEF_RESUID;
3113 sbi->s_resgid = EXT4_DEF_RESGID;
3114 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3115 sbi->s_sb_block = sb_block;
3116 if (sb->s_bdev->bd_part)
3117 sbi->s_sectors_written_start =
3118 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3120 /* Cleanup superblock name */
3121 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3125 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3127 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3132 * The ext4 superblock will not be buffer aligned for other than 1kB
3133 * block sizes. We need to calculate the offset from buffer start.
3135 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3136 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3137 offset = do_div(logical_sb_block, blocksize);
3139 logical_sb_block = sb_block;
3142 if (!(bh = sb_bread(sb, logical_sb_block))) {
3143 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3147 * Note: s_es must be initialized as soon as possible because
3148 * some ext4 macro-instructions depend on its value
3150 es = (struct ext4_super_block *) (bh->b_data + offset);
3152 sb->s_magic = le16_to_cpu(es->s_magic);
3153 if (sb->s_magic != EXT4_SUPER_MAGIC)
3155 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3157 /* Warn if metadata_csum and gdt_csum are both set. */
3158 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3159 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3160 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3161 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3162 "redundant flags; please run fsck.");
3164 /* Check for a known checksum algorithm */
3165 if (!ext4_verify_csum_type(sb, es)) {
3166 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3167 "unknown checksum algorithm.");
3172 /* Load the checksum driver */
3173 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3174 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3175 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3176 if (IS_ERR(sbi->s_chksum_driver)) {
3177 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3178 ret = PTR_ERR(sbi->s_chksum_driver);
3179 sbi->s_chksum_driver = NULL;
3184 /* Check superblock checksum */
3185 if (!ext4_superblock_csum_verify(sb, es)) {
3186 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3187 "invalid superblock checksum. Run e2fsck?");
3192 /* Precompute checksum seed for all metadata */
3193 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3194 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3195 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3196 sizeof(es->s_uuid));
3198 /* Set defaults before we parse the mount options */
3199 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3200 set_opt(sb, INIT_INODE_TABLE);
3201 if (def_mount_opts & EXT4_DEFM_DEBUG)
3203 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3205 if (def_mount_opts & EXT4_DEFM_UID16)
3206 set_opt(sb, NO_UID32);
3207 /* xattr user namespace & acls are now defaulted on */
3208 #ifdef CONFIG_EXT4_FS_XATTR
3209 set_opt(sb, XATTR_USER);
3211 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3212 set_opt(sb, POSIX_ACL);
3214 set_opt(sb, MBLK_IO_SUBMIT);
3215 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3216 set_opt(sb, JOURNAL_DATA);
3217 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3218 set_opt(sb, ORDERED_DATA);
3219 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3220 set_opt(sb, WRITEBACK_DATA);
3222 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3223 set_opt(sb, ERRORS_PANIC);
3224 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3225 set_opt(sb, ERRORS_CONT);
3227 set_opt(sb, ERRORS_RO);
3228 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3229 set_opt(sb, BLOCK_VALIDITY);
3230 if (def_mount_opts & EXT4_DEFM_DISCARD)
3231 set_opt(sb, DISCARD);
3233 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3234 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3235 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3236 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3237 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3239 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3240 set_opt(sb, BARRIER);
3243 * enable delayed allocation by default
3244 * Use -o nodelalloc to turn it off
3246 if (!IS_EXT3_SB(sb) &&
3247 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3248 set_opt(sb, DELALLOC);
3251 * set default s_li_wait_mult for lazyinit, for the case there is
3252 * no mount option specified.
3254 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3256 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3257 &journal_devnum, &journal_ioprio, 0)) {
3258 ext4_msg(sb, KERN_WARNING,
3259 "failed to parse options in superblock: %s",
3260 sbi->s_es->s_mount_opts);
3262 sbi->s_def_mount_opt = sbi->s_mount_opt;
3263 if (!parse_options((char *) data, sb, &journal_devnum,
3264 &journal_ioprio, 0))
3267 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3268 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3269 "with data=journal disables delayed "
3270 "allocation and O_DIRECT support!\n");
3271 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3272 ext4_msg(sb, KERN_ERR, "can't mount with "
3273 "both data=journal and delalloc");
3276 if (test_opt(sb, DIOREAD_NOLOCK)) {
3277 ext4_msg(sb, KERN_ERR, "can't mount with "
3278 "both data=journal and delalloc");
3281 if (test_opt(sb, DELALLOC))
3282 clear_opt(sb, DELALLOC);
3285 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3286 if (test_opt(sb, DIOREAD_NOLOCK)) {
3287 if (blocksize < PAGE_SIZE) {
3288 ext4_msg(sb, KERN_ERR, "can't mount with "
3289 "dioread_nolock if block size != PAGE_SIZE");
3294 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3295 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3297 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3298 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3299 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3300 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3301 ext4_msg(sb, KERN_WARNING,
3302 "feature flags set on rev 0 fs, "
3303 "running e2fsck is recommended");
3305 if (IS_EXT2_SB(sb)) {
3306 if (ext2_feature_set_ok(sb))
3307 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3308 "using the ext4 subsystem");
3310 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3311 "to feature incompatibilities");
3316 if (IS_EXT3_SB(sb)) {
3317 if (ext3_feature_set_ok(sb))
3318 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3319 "using the ext4 subsystem");
3321 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3322 "to feature incompatibilities");
3328 * Check feature flags regardless of the revision level, since we
3329 * previously didn't change the revision level when setting the flags,
3330 * so there is a chance incompat flags are set on a rev 0 filesystem.
3332 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3335 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3336 blocksize > EXT4_MAX_BLOCK_SIZE) {
3337 ext4_msg(sb, KERN_ERR,
3338 "Unsupported filesystem blocksize %d", blocksize);
3342 if (sb->s_blocksize != blocksize) {
3343 /* Validate the filesystem blocksize */
3344 if (!sb_set_blocksize(sb, blocksize)) {
3345 ext4_msg(sb, KERN_ERR, "bad block size %d",
3351 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3352 offset = do_div(logical_sb_block, blocksize);
3353 bh = sb_bread(sb, logical_sb_block);
3355 ext4_msg(sb, KERN_ERR,
3356 "Can't read superblock on 2nd try");
3359 es = (struct ext4_super_block *)(bh->b_data + offset);
3361 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3362 ext4_msg(sb, KERN_ERR,
3363 "Magic mismatch, very weird!");
3368 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3369 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3370 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3372 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3374 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3375 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3376 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3378 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3379 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3380 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3381 (!is_power_of_2(sbi->s_inode_size)) ||
3382 (sbi->s_inode_size > blocksize)) {
3383 ext4_msg(sb, KERN_ERR,
3384 "unsupported inode size: %d",
3388 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3389 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3392 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3393 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3394 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3395 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3396 !is_power_of_2(sbi->s_desc_size)) {
3397 ext4_msg(sb, KERN_ERR,
3398 "unsupported descriptor size %lu",
3403 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3405 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3406 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3407 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3410 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3411 if (sbi->s_inodes_per_block == 0)
3413 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3414 sbi->s_inodes_per_block;
3415 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3417 sbi->s_mount_state = le16_to_cpu(es->s_state);
3418 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3419 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3421 for (i = 0; i < 4; i++)
3422 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3423 sbi->s_def_hash_version = es->s_def_hash_version;
3424 i = le32_to_cpu(es->s_flags);
3425 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3426 sbi->s_hash_unsigned = 3;
3427 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3428 #ifdef __CHAR_UNSIGNED__
3429 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3430 sbi->s_hash_unsigned = 3;
3432 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3436 /* Handle clustersize */
3437 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3438 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3439 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3441 if (clustersize < blocksize) {
3442 ext4_msg(sb, KERN_ERR,
3443 "cluster size (%d) smaller than "
3444 "block size (%d)", clustersize, blocksize);
3447 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3448 le32_to_cpu(es->s_log_block_size);
3449 sbi->s_clusters_per_group =
3450 le32_to_cpu(es->s_clusters_per_group);
3451 if (sbi->s_clusters_per_group > blocksize * 8) {
3452 ext4_msg(sb, KERN_ERR,
3453 "#clusters per group too big: %lu",
3454 sbi->s_clusters_per_group);
3457 if (sbi->s_blocks_per_group !=
3458 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3459 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3460 "clusters per group (%lu) inconsistent",
3461 sbi->s_blocks_per_group,
3462 sbi->s_clusters_per_group);
3466 if (clustersize != blocksize) {
3467 ext4_warning(sb, "fragment/cluster size (%d) != "
3468 "block size (%d)", clustersize,
3470 clustersize = blocksize;
3472 if (sbi->s_blocks_per_group > blocksize * 8) {
3473 ext4_msg(sb, KERN_ERR,
3474 "#blocks per group too big: %lu",
3475 sbi->s_blocks_per_group);
3478 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3479 sbi->s_cluster_bits = 0;
3481 sbi->s_cluster_ratio = clustersize / blocksize;
3483 if (sbi->s_inodes_per_group > blocksize * 8) {
3484 ext4_msg(sb, KERN_ERR,
3485 "#inodes per group too big: %lu",
3486 sbi->s_inodes_per_group);
3491 * Test whether we have more sectors than will fit in sector_t,
3492 * and whether the max offset is addressable by the page cache.
3494 err = generic_check_addressable(sb->s_blocksize_bits,
3495 ext4_blocks_count(es));
3497 ext4_msg(sb, KERN_ERR, "filesystem"
3498 " too large to mount safely on this system");
3499 if (sizeof(sector_t) < 8)
3500 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3505 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3508 /* check blocks count against device size */
3509 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3510 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3511 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3512 "exceeds size of device (%llu blocks)",
3513 ext4_blocks_count(es), blocks_count);
3518 * It makes no sense for the first data block to be beyond the end
3519 * of the filesystem.
3521 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3522 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3523 "block %u is beyond end of filesystem (%llu)",
3524 le32_to_cpu(es->s_first_data_block),
3525 ext4_blocks_count(es));
3528 blocks_count = (ext4_blocks_count(es) -
3529 le32_to_cpu(es->s_first_data_block) +
3530 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3531 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3532 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3533 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3534 "(block count %llu, first data block %u, "
3535 "blocks per group %lu)", sbi->s_groups_count,
3536 ext4_blocks_count(es),
3537 le32_to_cpu(es->s_first_data_block),
3538 EXT4_BLOCKS_PER_GROUP(sb));
3541 sbi->s_groups_count = blocks_count;
3542 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3543 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3544 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3545 EXT4_DESC_PER_BLOCK(sb);
3546 sbi->s_group_desc = ext4_kvmalloc(db_count *
3547 sizeof(struct buffer_head *),
3549 if (sbi->s_group_desc == NULL) {
3550 ext4_msg(sb, KERN_ERR, "not enough memory");
3556 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3559 proc_create_data("options", S_IRUGO, sbi->s_proc,
3560 &ext4_seq_options_fops, sb);
3562 bgl_lock_init(sbi->s_blockgroup_lock);
3564 for (i = 0; i < db_count; i++) {
3565 block = descriptor_loc(sb, logical_sb_block, i);
3566 sbi->s_group_desc[i] = sb_bread(sb, block);
3567 if (!sbi->s_group_desc[i]) {
3568 ext4_msg(sb, KERN_ERR,
3569 "can't read group descriptor %d", i);
3574 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3575 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3578 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3579 if (!ext4_fill_flex_info(sb)) {
3580 ext4_msg(sb, KERN_ERR,
3581 "unable to initialize "
3582 "flex_bg meta info!");
3586 sbi->s_gdb_count = db_count;
3587 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3588 spin_lock_init(&sbi->s_next_gen_lock);
3590 init_timer(&sbi->s_err_report);
3591 sbi->s_err_report.function = print_daily_error_info;
3592 sbi->s_err_report.data = (unsigned long) sb;
3594 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3595 ext4_count_free_clusters(sb));
3597 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3598 ext4_count_free_inodes(sb));
3601 err = percpu_counter_init(&sbi->s_dirs_counter,
3602 ext4_count_dirs(sb));
3605 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3608 ext4_msg(sb, KERN_ERR, "insufficient memory");
3613 sbi->s_stripe = ext4_get_stripe_size(sbi);
3614 sbi->s_max_writeback_mb_bump = 128;
3617 * set up enough so that it can read an inode
3619 if (!test_opt(sb, NOLOAD) &&
3620 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3621 sb->s_op = &ext4_sops;
3623 sb->s_op = &ext4_nojournal_sops;
3624 sb->s_export_op = &ext4_export_ops;
3625 sb->s_xattr = ext4_xattr_handlers;
3627 sb->s_qcop = &ext4_qctl_operations;
3628 sb->dq_op = &ext4_quota_operations;
3630 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3632 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3633 mutex_init(&sbi->s_orphan_lock);
3634 sbi->s_resize_flags = 0;
3638 needs_recovery = (es->s_last_orphan != 0 ||
3639 EXT4_HAS_INCOMPAT_FEATURE(sb,
3640 EXT4_FEATURE_INCOMPAT_RECOVER));
3642 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3643 !(sb->s_flags & MS_RDONLY))
3644 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3648 * The first inode we look at is the journal inode. Don't try
3649 * root first: it may be modified in the journal!
3651 if (!test_opt(sb, NOLOAD) &&
3652 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3653 if (ext4_load_journal(sb, es, journal_devnum))
3655 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3656 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3657 ext4_msg(sb, KERN_ERR, "required journal recovery "
3658 "suppressed and not mounted read-only");
3659 goto failed_mount_wq;
3661 clear_opt(sb, DATA_FLAGS);
3662 sbi->s_journal = NULL;
3667 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3668 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3669 JBD2_FEATURE_INCOMPAT_64BIT)) {
3670 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3671 goto failed_mount_wq;
3674 if (!set_journal_csum_feature_set(sb)) {
3675 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3677 goto failed_mount_wq;
3680 /* We have now updated the journal if required, so we can
3681 * validate the data journaling mode. */
3682 switch (test_opt(sb, DATA_FLAGS)) {
3684 /* No mode set, assume a default based on the journal
3685 * capabilities: ORDERED_DATA if the journal can
3686 * cope, else JOURNAL_DATA
3688 if (jbd2_journal_check_available_features
3689 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3690 set_opt(sb, ORDERED_DATA);
3692 set_opt(sb, JOURNAL_DATA);
3695 case EXT4_MOUNT_ORDERED_DATA:
3696 case EXT4_MOUNT_WRITEBACK_DATA:
3697 if (!jbd2_journal_check_available_features
3698 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3699 ext4_msg(sb, KERN_ERR, "Journal does not support "
3700 "requested data journaling mode");
3701 goto failed_mount_wq;
3706 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3708 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3711 * The journal may have updated the bg summary counts, so we
3712 * need to update the global counters.
3714 percpu_counter_set(&sbi->s_freeclusters_counter,
3715 ext4_count_free_clusters(sb));
3716 percpu_counter_set(&sbi->s_freeinodes_counter,
3717 ext4_count_free_inodes(sb));
3718 percpu_counter_set(&sbi->s_dirs_counter,
3719 ext4_count_dirs(sb));
3720 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3724 * The maximum number of concurrent works can be high and
3725 * concurrency isn't really necessary. Limit it to 1.
3727 EXT4_SB(sb)->dio_unwritten_wq =
3728 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3729 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3730 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3731 goto failed_mount_wq;
3735 * The jbd2_journal_load will have done any necessary log recovery,
3736 * so we can safely mount the rest of the filesystem now.
3739 root = ext4_iget(sb, EXT4_ROOT_INO);
3741 ext4_msg(sb, KERN_ERR, "get root inode failed");
3742 ret = PTR_ERR(root);
3746 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3747 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3751 sb->s_root = d_make_root(root);
3753 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3758 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3759 sb->s_flags |= MS_RDONLY;
3761 /* determine the minimum size of new large inodes, if present */
3762 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3763 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3764 EXT4_GOOD_OLD_INODE_SIZE;
3765 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3766 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3767 if (sbi->s_want_extra_isize <
3768 le16_to_cpu(es->s_want_extra_isize))
3769 sbi->s_want_extra_isize =
3770 le16_to_cpu(es->s_want_extra_isize);
3771 if (sbi->s_want_extra_isize <
3772 le16_to_cpu(es->s_min_extra_isize))
3773 sbi->s_want_extra_isize =
3774 le16_to_cpu(es->s_min_extra_isize);
3777 /* Check if enough inode space is available */
3778 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3779 sbi->s_inode_size) {
3780 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3781 EXT4_GOOD_OLD_INODE_SIZE;
3782 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3786 err = ext4_setup_system_zone(sb);
3788 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3790 goto failed_mount4a;
3794 err = ext4_mb_init(sb);
3796 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3801 err = ext4_register_li_request(sb, first_not_zeroed);
3805 sbi->s_kobj.kset = ext4_kset;
3806 init_completion(&sbi->s_kobj_unregister);
3807 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3812 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3813 ext4_orphan_cleanup(sb, es);
3814 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3815 if (needs_recovery) {
3816 ext4_msg(sb, KERN_INFO, "recovery complete");
3817 ext4_mark_recovery_complete(sb, es);
3819 if (EXT4_SB(sb)->s_journal) {
3820 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3821 descr = " journalled data mode";
3822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3823 descr = " ordered data mode";
3825 descr = " writeback data mode";
3827 descr = "out journal";
3829 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3830 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3831 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3833 if (es->s_error_count)
3834 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3841 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3845 ext4_unregister_li_request(sb);
3847 ext4_mb_release(sb);
3849 ext4_ext_release(sb);
3850 ext4_release_system_zone(sb);
3855 ext4_msg(sb, KERN_ERR, "mount failed");
3856 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3858 if (sbi->s_journal) {
3859 jbd2_journal_destroy(sbi->s_journal);
3860 sbi->s_journal = NULL;
3863 del_timer(&sbi->s_err_report);
3864 if (sbi->s_flex_groups)
3865 ext4_kvfree(sbi->s_flex_groups);
3866 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3867 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3868 percpu_counter_destroy(&sbi->s_dirs_counter);
3869 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3871 kthread_stop(sbi->s_mmp_tsk);
3873 for (i = 0; i < db_count; i++)
3874 brelse(sbi->s_group_desc[i]);
3875 ext4_kvfree(sbi->s_group_desc);
3877 if (sbi->s_chksum_driver)
3878 crypto_free_shash(sbi->s_chksum_driver);
3880 remove_proc_entry("options", sbi->s_proc);
3881 remove_proc_entry(sb->s_id, ext4_proc_root);
3884 for (i = 0; i < MAXQUOTAS; i++)
3885 kfree(sbi->s_qf_names[i]);
3887 ext4_blkdev_remove(sbi);
3890 sb->s_fs_info = NULL;
3891 kfree(sbi->s_blockgroup_lock);
3899 * Setup any per-fs journal parameters now. We'll do this both on
3900 * initial mount, once the journal has been initialised but before we've
3901 * done any recovery; and again on any subsequent remount.
3903 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3905 struct ext4_sb_info *sbi = EXT4_SB(sb);
3907 journal->j_commit_interval = sbi->s_commit_interval;
3908 journal->j_min_batch_time = sbi->s_min_batch_time;
3909 journal->j_max_batch_time = sbi->s_max_batch_time;
3911 write_lock(&journal->j_state_lock);
3912 if (test_opt(sb, BARRIER))
3913 journal->j_flags |= JBD2_BARRIER;
3915 journal->j_flags &= ~JBD2_BARRIER;
3916 if (test_opt(sb, DATA_ERR_ABORT))
3917 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3919 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3920 write_unlock(&journal->j_state_lock);
3923 static journal_t *ext4_get_journal(struct super_block *sb,
3924 unsigned int journal_inum)
3926 struct inode *journal_inode;
3929 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3931 /* First, test for the existence of a valid inode on disk. Bad
3932 * things happen if we iget() an unused inode, as the subsequent
3933 * iput() will try to delete it. */
3935 journal_inode = ext4_iget(sb, journal_inum);
3936 if (IS_ERR(journal_inode)) {
3937 ext4_msg(sb, KERN_ERR, "no journal found");
3940 if (!journal_inode->i_nlink) {
3941 make_bad_inode(journal_inode);
3942 iput(journal_inode);
3943 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3947 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3948 journal_inode, journal_inode->i_size);
3949 if (!S_ISREG(journal_inode->i_mode)) {
3950 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3951 iput(journal_inode);
3955 journal = jbd2_journal_init_inode(journal_inode);
3957 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3958 iput(journal_inode);
3961 journal->j_private = sb;
3962 ext4_init_journal_params(sb, journal);
3966 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3969 struct buffer_head *bh;
3973 int hblock, blocksize;
3974 ext4_fsblk_t sb_block;
3975 unsigned long offset;
3976 struct ext4_super_block *es;
3977 struct block_device *bdev;
3979 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3981 bdev = ext4_blkdev_get(j_dev, sb);
3985 blocksize = sb->s_blocksize;
3986 hblock = bdev_logical_block_size(bdev);
3987 if (blocksize < hblock) {
3988 ext4_msg(sb, KERN_ERR,
3989 "blocksize too small for journal device");
3993 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3994 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3995 set_blocksize(bdev, blocksize);
3996 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3997 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3998 "external journal");
4002 es = (struct ext4_super_block *) (bh->b_data + offset);
4003 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4004 !(le32_to_cpu(es->s_feature_incompat) &
4005 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4006 ext4_msg(sb, KERN_ERR, "external journal has "
4012 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4013 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4018 len = ext4_blocks_count(es);
4019 start = sb_block + 1;
4020 brelse(bh); /* we're done with the superblock */
4022 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4023 start, len, blocksize);
4025 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4028 journal->j_private = sb;
4029 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4030 wait_on_buffer(journal->j_sb_buffer);
4031 if (!buffer_uptodate(journal->j_sb_buffer)) {
4032 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4035 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4036 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4037 "user (unsupported) - %d",
4038 be32_to_cpu(journal->j_superblock->s_nr_users));
4041 EXT4_SB(sb)->journal_bdev = bdev;
4042 ext4_init_journal_params(sb, journal);
4046 jbd2_journal_destroy(journal);
4048 ext4_blkdev_put(bdev);
4052 static int ext4_load_journal(struct super_block *sb,
4053 struct ext4_super_block *es,
4054 unsigned long journal_devnum)
4057 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4060 int really_read_only;
4062 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4064 if (journal_devnum &&
4065 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4066 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4067 "numbers have changed");
4068 journal_dev = new_decode_dev(journal_devnum);
4070 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4072 really_read_only = bdev_read_only(sb->s_bdev);
4075 * Are we loading a blank journal or performing recovery after a
4076 * crash? For recovery, we need to check in advance whether we
4077 * can get read-write access to the device.
4079 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4080 if (sb->s_flags & MS_RDONLY) {
4081 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4082 "required on readonly filesystem");
4083 if (really_read_only) {
4084 ext4_msg(sb, KERN_ERR, "write access "
4085 "unavailable, cannot proceed");
4088 ext4_msg(sb, KERN_INFO, "write access will "
4089 "be enabled during recovery");
4093 if (journal_inum && journal_dev) {
4094 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4095 "and inode journals!");
4100 if (!(journal = ext4_get_journal(sb, journal_inum)))
4103 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4107 if (!(journal->j_flags & JBD2_BARRIER))
4108 ext4_msg(sb, KERN_INFO, "barriers disabled");
4110 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4111 err = jbd2_journal_wipe(journal, !really_read_only);
4113 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4115 memcpy(save, ((char *) es) +
4116 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4117 err = jbd2_journal_load(journal);
4119 memcpy(((char *) es) + EXT4_S_ERR_START,
4120 save, EXT4_S_ERR_LEN);
4125 ext4_msg(sb, KERN_ERR, "error loading journal");
4126 jbd2_journal_destroy(journal);
4130 EXT4_SB(sb)->s_journal = journal;
4131 ext4_clear_journal_err(sb, es);
4133 if (!really_read_only && journal_devnum &&
4134 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4135 es->s_journal_dev = cpu_to_le32(journal_devnum);
4137 /* Make sure we flush the recovery flag to disk. */
4138 ext4_commit_super(sb, 1);
4144 static int ext4_commit_super(struct super_block *sb, int sync)
4146 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4147 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4150 if (!sbh || block_device_ejected(sb))
4152 if (buffer_write_io_error(sbh)) {
4154 * Oh, dear. A previous attempt to write the
4155 * superblock failed. This could happen because the
4156 * USB device was yanked out. Or it could happen to
4157 * be a transient write error and maybe the block will
4158 * be remapped. Nothing we can do but to retry the
4159 * write and hope for the best.
4161 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4162 "superblock detected");
4163 clear_buffer_write_io_error(sbh);
4164 set_buffer_uptodate(sbh);
4167 * If the file system is mounted read-only, don't update the
4168 * superblock write time. This avoids updating the superblock
4169 * write time when we are mounting the root file system
4170 * read/only but we need to replay the journal; at that point,
4171 * for people who are east of GMT and who make their clock
4172 * tick in localtime for Windows bug-for-bug compatibility,
4173 * the clock is set in the future, and this will cause e2fsck
4174 * to complain and force a full file system check.
4176 if (!(sb->s_flags & MS_RDONLY))
4177 es->s_wtime = cpu_to_le32(get_seconds());
4178 if (sb->s_bdev->bd_part)
4179 es->s_kbytes_written =
4180 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4181 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4182 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4184 es->s_kbytes_written =
4185 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4186 ext4_free_blocks_count_set(es,
4187 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4188 &EXT4_SB(sb)->s_freeclusters_counter)));
4189 es->s_free_inodes_count =
4190 cpu_to_le32(percpu_counter_sum_positive(
4191 &EXT4_SB(sb)->s_freeinodes_counter));
4193 BUFFER_TRACE(sbh, "marking dirty");
4194 ext4_superblock_csum_set(sb, es);
4195 mark_buffer_dirty(sbh);
4197 error = sync_dirty_buffer(sbh);
4201 error = buffer_write_io_error(sbh);
4203 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4205 clear_buffer_write_io_error(sbh);
4206 set_buffer_uptodate(sbh);
4213 * Have we just finished recovery? If so, and if we are mounting (or
4214 * remounting) the filesystem readonly, then we will end up with a
4215 * consistent fs on disk. Record that fact.
4217 static void ext4_mark_recovery_complete(struct super_block *sb,
4218 struct ext4_super_block *es)
4220 journal_t *journal = EXT4_SB(sb)->s_journal;
4222 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4223 BUG_ON(journal != NULL);
4226 jbd2_journal_lock_updates(journal);
4227 if (jbd2_journal_flush(journal) < 0)
4230 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4231 sb->s_flags & MS_RDONLY) {
4232 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4233 ext4_commit_super(sb, 1);
4237 jbd2_journal_unlock_updates(journal);
4241 * If we are mounting (or read-write remounting) a filesystem whose journal
4242 * has recorded an error from a previous lifetime, move that error to the
4243 * main filesystem now.
4245 static void ext4_clear_journal_err(struct super_block *sb,
4246 struct ext4_super_block *es)
4252 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4254 journal = EXT4_SB(sb)->s_journal;
4257 * Now check for any error status which may have been recorded in the
4258 * journal by a prior ext4_error() or ext4_abort()
4261 j_errno = jbd2_journal_errno(journal);
4265 errstr = ext4_decode_error(sb, j_errno, nbuf);
4266 ext4_warning(sb, "Filesystem error recorded "
4267 "from previous mount: %s", errstr);
4268 ext4_warning(sb, "Marking fs in need of filesystem check.");
4270 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4271 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4272 ext4_commit_super(sb, 1);
4274 jbd2_journal_clear_err(journal);
4279 * Force the running and committing transactions to commit,
4280 * and wait on the commit.
4282 int ext4_force_commit(struct super_block *sb)
4287 if (sb->s_flags & MS_RDONLY)
4290 journal = EXT4_SB(sb)->s_journal;
4292 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4293 ret = ext4_journal_force_commit(journal);
4299 static void ext4_write_super(struct super_block *sb)
4302 ext4_commit_super(sb, 1);
4306 static int ext4_sync_fs(struct super_block *sb, int wait)
4310 struct ext4_sb_info *sbi = EXT4_SB(sb);
4312 trace_ext4_sync_fs(sb, wait);
4313 flush_workqueue(sbi->dio_unwritten_wq);
4314 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4316 jbd2_log_wait_commit(sbi->s_journal, target);
4322 * LVM calls this function before a (read-only) snapshot is created. This
4323 * gives us a chance to flush the journal completely and mark the fs clean.
4325 * Note that only this function cannot bring a filesystem to be in a clean
4326 * state independently, because ext4 prevents a new handle from being started
4327 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4330 static int ext4_freeze(struct super_block *sb)
4335 if (sb->s_flags & MS_RDONLY)
4338 journal = EXT4_SB(sb)->s_journal;
4340 /* Now we set up the journal barrier. */
4341 jbd2_journal_lock_updates(journal);
4344 * Don't clear the needs_recovery flag if we failed to flush
4347 error = jbd2_journal_flush(journal);
4351 /* Journal blocked and flushed, clear needs_recovery flag. */
4352 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4353 error = ext4_commit_super(sb, 1);
4355 /* we rely on s_frozen to stop further updates */
4356 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4361 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4362 * flag here, even though the filesystem is not technically dirty yet.
4364 static int ext4_unfreeze(struct super_block *sb)
4366 if (sb->s_flags & MS_RDONLY)
4370 /* Reset the needs_recovery flag before the fs is unlocked. */
4371 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4372 ext4_commit_super(sb, 1);
4378 * Structure to save mount options for ext4_remount's benefit
4380 struct ext4_mount_options {
4381 unsigned long s_mount_opt;
4382 unsigned long s_mount_opt2;
4385 unsigned long s_commit_interval;
4386 u32 s_min_batch_time, s_max_batch_time;
4389 char *s_qf_names[MAXQUOTAS];
4393 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4395 struct ext4_super_block *es;
4396 struct ext4_sb_info *sbi = EXT4_SB(sb);
4397 unsigned long old_sb_flags;
4398 struct ext4_mount_options old_opts;
4399 int enable_quota = 0;
4401 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4406 char *orig_data = kstrdup(data, GFP_KERNEL);
4408 /* Store the original options */
4410 old_sb_flags = sb->s_flags;
4411 old_opts.s_mount_opt = sbi->s_mount_opt;
4412 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4413 old_opts.s_resuid = sbi->s_resuid;
4414 old_opts.s_resgid = sbi->s_resgid;
4415 old_opts.s_commit_interval = sbi->s_commit_interval;
4416 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4417 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4419 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4420 for (i = 0; i < MAXQUOTAS; i++)
4421 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4423 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4424 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4427 * Allow the "check" option to be passed as a remount option.
4429 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4434 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4435 ext4_abort(sb, "Abort forced by user");
4437 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4438 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4442 if (sbi->s_journal) {
4443 ext4_init_journal_params(sb, sbi->s_journal);
4444 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4447 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4448 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4453 if (*flags & MS_RDONLY) {
4454 err = dquot_suspend(sb, -1);
4459 * First of all, the unconditional stuff we have to do
4460 * to disable replay of the journal when we next remount
4462 sb->s_flags |= MS_RDONLY;
4465 * OK, test if we are remounting a valid rw partition
4466 * readonly, and if so set the rdonly flag and then
4467 * mark the partition as valid again.
4469 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4470 (sbi->s_mount_state & EXT4_VALID_FS))
4471 es->s_state = cpu_to_le16(sbi->s_mount_state);
4474 ext4_mark_recovery_complete(sb, es);
4476 /* Make sure we can mount this feature set readwrite */
4477 if (!ext4_feature_set_ok(sb, 0)) {
4482 * Make sure the group descriptor checksums
4483 * are sane. If they aren't, refuse to remount r/w.
4485 for (g = 0; g < sbi->s_groups_count; g++) {
4486 struct ext4_group_desc *gdp =
4487 ext4_get_group_desc(sb, g, NULL);
4489 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4490 ext4_msg(sb, KERN_ERR,
4491 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4492 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4493 le16_to_cpu(gdp->bg_checksum));
4500 * If we have an unprocessed orphan list hanging
4501 * around from a previously readonly bdev mount,
4502 * require a full umount/remount for now.
4504 if (es->s_last_orphan) {
4505 ext4_msg(sb, KERN_WARNING, "Couldn't "
4506 "remount RDWR because of unprocessed "
4507 "orphan inode list. Please "
4508 "umount/remount instead");
4514 * Mounting a RDONLY partition read-write, so reread
4515 * and store the current valid flag. (It may have
4516 * been changed by e2fsck since we originally mounted
4520 ext4_clear_journal_err(sb, es);
4521 sbi->s_mount_state = le16_to_cpu(es->s_state);
4522 if (!ext4_setup_super(sb, es, 0))
4523 sb->s_flags &= ~MS_RDONLY;
4524 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4525 EXT4_FEATURE_INCOMPAT_MMP))
4526 if (ext4_multi_mount_protect(sb,
4527 le64_to_cpu(es->s_mmp_block))) {
4536 * Reinitialize lazy itable initialization thread based on
4539 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4540 ext4_unregister_li_request(sb);
4542 ext4_group_t first_not_zeroed;
4543 first_not_zeroed = ext4_has_uninit_itable(sb);
4544 ext4_register_li_request(sb, first_not_zeroed);
4547 ext4_setup_system_zone(sb);
4548 if (sbi->s_journal == NULL)
4549 ext4_commit_super(sb, 1);
4552 /* Release old quota file names */
4553 for (i = 0; i < MAXQUOTAS; i++)
4554 if (old_opts.s_qf_names[i] &&
4555 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4556 kfree(old_opts.s_qf_names[i]);
4560 dquot_resume(sb, -1);
4562 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4567 sb->s_flags = old_sb_flags;
4568 sbi->s_mount_opt = old_opts.s_mount_opt;
4569 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4570 sbi->s_resuid = old_opts.s_resuid;
4571 sbi->s_resgid = old_opts.s_resgid;
4572 sbi->s_commit_interval = old_opts.s_commit_interval;
4573 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4574 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4576 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4577 for (i = 0; i < MAXQUOTAS; i++) {
4578 if (sbi->s_qf_names[i] &&
4579 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4580 kfree(sbi->s_qf_names[i]);
4581 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4590 * Note: calculating the overhead so we can be compatible with
4591 * historical BSD practice is quite difficult in the face of
4592 * clusters/bigalloc. This is because multiple metadata blocks from
4593 * different block group can end up in the same allocation cluster.
4594 * Calculating the exact overhead in the face of clustered allocation
4595 * requires either O(all block bitmaps) in memory or O(number of block
4596 * groups**2) in time. We will still calculate the superblock for
4597 * older file systems --- and if we come across with a bigalloc file
4598 * system with zero in s_overhead_clusters the estimate will be close to
4599 * correct especially for very large cluster sizes --- but for newer
4600 * file systems, it's better to calculate this figure once at mkfs
4601 * time, and store it in the superblock. If the superblock value is
4602 * present (even for non-bigalloc file systems), we will use it.
4604 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4606 struct super_block *sb = dentry->d_sb;
4607 struct ext4_sb_info *sbi = EXT4_SB(sb);
4608 struct ext4_super_block *es = sbi->s_es;
4609 struct ext4_group_desc *gdp;
4613 if (test_opt(sb, MINIX_DF)) {
4614 sbi->s_overhead_last = 0;
4615 } else if (es->s_overhead_clusters) {
4616 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4617 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4618 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4619 ext4_fsblk_t overhead = 0;
4622 * Compute the overhead (FS structures). This is constant
4623 * for a given filesystem unless the number of block groups
4624 * changes so we cache the previous value until it does.
4628 * All of the blocks before first_data_block are
4631 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4634 * Add the overhead found in each block group
4636 for (i = 0; i < ngroups; i++) {
4637 gdp = ext4_get_group_desc(sb, i, NULL);
4638 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4641 sbi->s_overhead_last = overhead;
4643 sbi->s_blocks_last = ext4_blocks_count(es);
4646 buf->f_type = EXT4_SUPER_MAGIC;
4647 buf->f_bsize = sb->s_blocksize;
4648 buf->f_blocks = (ext4_blocks_count(es) -
4649 EXT4_C2B(sbi, sbi->s_overhead_last));
4650 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4651 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4652 /* prevent underflow in case that few free space is available */
4653 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4654 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4655 if (buf->f_bfree < ext4_r_blocks_count(es))
4657 buf->f_files = le32_to_cpu(es->s_inodes_count);
4658 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4659 buf->f_namelen = EXT4_NAME_LEN;
4660 fsid = le64_to_cpup((void *)es->s_uuid) ^
4661 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4662 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4663 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4668 /* Helper function for writing quotas on sync - we need to start transaction
4669 * before quota file is locked for write. Otherwise the are possible deadlocks:
4670 * Process 1 Process 2
4671 * ext4_create() quota_sync()
4672 * jbd2_journal_start() write_dquot()
4673 * dquot_initialize() down(dqio_mutex)
4674 * down(dqio_mutex) jbd2_journal_start()
4680 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4682 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4685 static int ext4_write_dquot(struct dquot *dquot)
4689 struct inode *inode;
4691 inode = dquot_to_inode(dquot);
4692 handle = ext4_journal_start(inode,
4693 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4695 return PTR_ERR(handle);
4696 ret = dquot_commit(dquot);
4697 err = ext4_journal_stop(handle);
4703 static int ext4_acquire_dquot(struct dquot *dquot)
4708 handle = ext4_journal_start(dquot_to_inode(dquot),
4709 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4711 return PTR_ERR(handle);
4712 ret = dquot_acquire(dquot);
4713 err = ext4_journal_stop(handle);
4719 static int ext4_release_dquot(struct dquot *dquot)
4724 handle = ext4_journal_start(dquot_to_inode(dquot),
4725 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4726 if (IS_ERR(handle)) {
4727 /* Release dquot anyway to avoid endless cycle in dqput() */
4728 dquot_release(dquot);
4729 return PTR_ERR(handle);
4731 ret = dquot_release(dquot);
4732 err = ext4_journal_stop(handle);
4738 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4740 /* Are we journaling quotas? */
4741 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4742 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4743 dquot_mark_dquot_dirty(dquot);
4744 return ext4_write_dquot(dquot);
4746 return dquot_mark_dquot_dirty(dquot);
4750 static int ext4_write_info(struct super_block *sb, int type)
4755 /* Data block + inode block */
4756 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4758 return PTR_ERR(handle);
4759 ret = dquot_commit_info(sb, type);
4760 err = ext4_journal_stop(handle);
4767 * Turn on quotas during mount time - we need to find
4768 * the quota file and such...
4770 static int ext4_quota_on_mount(struct super_block *sb, int type)
4772 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4773 EXT4_SB(sb)->s_jquota_fmt, type);
4777 * Standard function to be called on quota_on
4779 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4784 if (!test_opt(sb, QUOTA))
4787 /* Quotafile not on the same filesystem? */
4788 if (path->dentry->d_sb != sb)
4790 /* Journaling quota? */
4791 if (EXT4_SB(sb)->s_qf_names[type]) {
4792 /* Quotafile not in fs root? */
4793 if (path->dentry->d_parent != sb->s_root)
4794 ext4_msg(sb, KERN_WARNING,
4795 "Quota file not on filesystem root. "
4796 "Journaled quota will not work");
4800 * When we journal data on quota file, we have to flush journal to see
4801 * all updates to the file when we bypass pagecache...
4803 if (EXT4_SB(sb)->s_journal &&
4804 ext4_should_journal_data(path->dentry->d_inode)) {
4806 * We don't need to lock updates but journal_flush() could
4807 * otherwise be livelocked...
4809 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4810 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4811 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4816 return dquot_quota_on(sb, type, format_id, path);
4819 static int ext4_quota_off(struct super_block *sb, int type)
4821 struct inode *inode = sb_dqopt(sb)->files[type];
4824 /* Force all delayed allocation blocks to be allocated.
4825 * Caller already holds s_umount sem */
4826 if (test_opt(sb, DELALLOC))
4827 sync_filesystem(sb);
4832 /* Update modification times of quota files when userspace can
4833 * start looking at them */
4834 handle = ext4_journal_start(inode, 1);
4837 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4838 ext4_mark_inode_dirty(handle, inode);
4839 ext4_journal_stop(handle);
4842 return dquot_quota_off(sb, type);
4845 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4846 * acquiring the locks... As quota files are never truncated and quota code
4847 * itself serializes the operations (and no one else should touch the files)
4848 * we don't have to be afraid of races */
4849 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4850 size_t len, loff_t off)
4852 struct inode *inode = sb_dqopt(sb)->files[type];
4853 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4855 int offset = off & (sb->s_blocksize - 1);
4858 struct buffer_head *bh;
4859 loff_t i_size = i_size_read(inode);
4863 if (off+len > i_size)
4866 while (toread > 0) {
4867 tocopy = sb->s_blocksize - offset < toread ?
4868 sb->s_blocksize - offset : toread;
4869 bh = ext4_bread(NULL, inode, blk, 0, &err);
4872 if (!bh) /* A hole? */
4873 memset(data, 0, tocopy);
4875 memcpy(data, bh->b_data+offset, tocopy);
4885 /* Write to quotafile (we know the transaction is already started and has
4886 * enough credits) */
4887 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4888 const char *data, size_t len, loff_t off)
4890 struct inode *inode = sb_dqopt(sb)->files[type];
4891 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4893 int offset = off & (sb->s_blocksize - 1);
4894 struct buffer_head *bh;
4895 handle_t *handle = journal_current_handle();
4897 if (EXT4_SB(sb)->s_journal && !handle) {
4898 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4899 " cancelled because transaction is not started",
4900 (unsigned long long)off, (unsigned long long)len);
4904 * Since we account only one data block in transaction credits,
4905 * then it is impossible to cross a block boundary.
4907 if (sb->s_blocksize - offset < len) {
4908 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4909 " cancelled because not block aligned",
4910 (unsigned long long)off, (unsigned long long)len);
4914 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4915 bh = ext4_bread(handle, inode, blk, 1, &err);
4918 err = ext4_journal_get_write_access(handle, bh);
4924 memcpy(bh->b_data+offset, data, len);
4925 flush_dcache_page(bh->b_page);
4927 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4931 mutex_unlock(&inode->i_mutex);
4934 if (inode->i_size < off + len) {
4935 i_size_write(inode, off + len);
4936 EXT4_I(inode)->i_disksize = inode->i_size;
4937 ext4_mark_inode_dirty(handle, inode);
4939 mutex_unlock(&inode->i_mutex);
4945 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4946 const char *dev_name, void *data)
4948 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4951 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4952 static inline void register_as_ext2(void)
4954 int err = register_filesystem(&ext2_fs_type);
4957 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4960 static inline void unregister_as_ext2(void)
4962 unregister_filesystem(&ext2_fs_type);
4965 static inline int ext2_feature_set_ok(struct super_block *sb)
4967 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4969 if (sb->s_flags & MS_RDONLY)
4971 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4975 MODULE_ALIAS("ext2");
4977 static inline void register_as_ext2(void) { }
4978 static inline void unregister_as_ext2(void) { }
4979 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4982 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4983 static inline void register_as_ext3(void)
4985 int err = register_filesystem(&ext3_fs_type);
4988 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4991 static inline void unregister_as_ext3(void)
4993 unregister_filesystem(&ext3_fs_type);
4996 static inline int ext3_feature_set_ok(struct super_block *sb)
4998 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5000 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5002 if (sb->s_flags & MS_RDONLY)
5004 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5008 MODULE_ALIAS("ext3");
5010 static inline void register_as_ext3(void) { }
5011 static inline void unregister_as_ext3(void) { }
5012 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5015 static struct file_system_type ext4_fs_type = {
5016 .owner = THIS_MODULE,
5018 .mount = ext4_mount,
5019 .kill_sb = kill_block_super,
5020 .fs_flags = FS_REQUIRES_DEV,
5023 static int __init ext4_init_feat_adverts(void)
5025 struct ext4_features *ef;
5028 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5032 ef->f_kobj.kset = ext4_kset;
5033 init_completion(&ef->f_kobj_unregister);
5034 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5047 static void ext4_exit_feat_adverts(void)
5049 kobject_put(&ext4_feat->f_kobj);
5050 wait_for_completion(&ext4_feat->f_kobj_unregister);
5054 /* Shared across all ext4 file systems */
5055 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5056 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5058 static int __init ext4_init_fs(void)
5062 ext4_li_info = NULL;
5063 mutex_init(&ext4_li_mtx);
5065 ext4_check_flag_values();
5067 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5068 mutex_init(&ext4__aio_mutex[i]);
5069 init_waitqueue_head(&ext4__ioend_wq[i]);
5072 err = ext4_init_pageio();
5075 err = ext4_init_system_zone();
5078 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5081 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5083 err = ext4_init_feat_adverts();
5087 err = ext4_init_mballoc();
5091 err = ext4_init_xattr();
5094 err = init_inodecache();
5099 err = register_filesystem(&ext4_fs_type);
5105 unregister_as_ext2();
5106 unregister_as_ext3();
5107 destroy_inodecache();
5111 ext4_exit_mballoc();
5113 ext4_exit_feat_adverts();
5116 remove_proc_entry("fs/ext4", NULL);
5117 kset_unregister(ext4_kset);
5119 ext4_exit_system_zone();
5125 static void __exit ext4_exit_fs(void)
5127 ext4_destroy_lazyinit_thread();
5128 unregister_as_ext2();
5129 unregister_as_ext3();
5130 unregister_filesystem(&ext4_fs_type);
5131 destroy_inodecache();
5133 ext4_exit_mballoc();
5134 ext4_exit_feat_adverts();
5135 remove_proc_entry("fs/ext4", NULL);
5136 kset_unregister(ext4_kset);
5137 ext4_exit_system_zone();
5141 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5142 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5143 MODULE_LICENSE("GPL");
5144 module_init(ext4_init_fs)
5145 module_exit(ext4_exit_fs)