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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static int ext4_mballoc_ready;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
83 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
84 static struct file_system_type ext2_fs_type = {
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
91 MODULE_ALIAS_FS("ext2");
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #define IS_EXT2_SB(sb) (0)
99 static struct file_system_type ext3_fs_type = {
100 .owner = THIS_MODULE,
103 .kill_sb = kill_block_super,
104 .fs_flags = FS_REQUIRES_DEV,
106 MODULE_ALIAS_FS("ext3");
107 MODULE_ALIAS("ext3");
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 static int ext4_verify_csum_type(struct super_block *sb,
111 struct ext4_super_block *es)
113 if (!ext4_has_feature_metadata_csum(sb))
116 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
119 static __le32 ext4_superblock_csum(struct super_block *sb,
120 struct ext4_super_block *es)
122 struct ext4_sb_info *sbi = EXT4_SB(sb);
123 int offset = offsetof(struct ext4_super_block, s_checksum);
126 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
128 return cpu_to_le32(csum);
131 static int ext4_superblock_csum_verify(struct super_block *sb,
132 struct ext4_super_block *es)
134 if (!ext4_has_metadata_csum(sb))
137 return es->s_checksum == ext4_superblock_csum(sb, es);
140 void ext4_superblock_csum_set(struct super_block *sb)
142 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
144 if (!ext4_has_metadata_csum(sb))
147 es->s_checksum = ext4_superblock_csum(sb, es);
150 void *ext4_kvmalloc(size_t size, gfp_t flags)
154 ret = kmalloc(size, flags | __GFP_NOWARN);
156 ret = __vmalloc(size, flags, PAGE_KERNEL);
160 void *ext4_kvzalloc(size_t size, gfp_t flags)
164 ret = kzalloc(size, flags | __GFP_NOWARN);
166 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
170 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
171 struct ext4_group_desc *bg)
173 return le32_to_cpu(bg->bg_block_bitmap_lo) |
174 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
175 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
178 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
181 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
186 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
187 struct ext4_group_desc *bg)
189 return le32_to_cpu(bg->bg_inode_table_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
194 __u32 ext4_free_group_clusters(struct super_block *sb,
195 struct ext4_group_desc *bg)
197 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
202 __u32 ext4_free_inodes_count(struct super_block *sb,
203 struct ext4_group_desc *bg)
205 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
210 __u32 ext4_used_dirs_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
213 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
218 __u32 ext4_itable_unused_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
221 return le16_to_cpu(bg->bg_itable_unused_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
226 void ext4_block_bitmap_set(struct super_block *sb,
227 struct ext4_group_desc *bg, ext4_fsblk_t blk)
229 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
230 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
234 void ext4_inode_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
242 void ext4_inode_table_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
250 void ext4_free_group_clusters_set(struct super_block *sb,
251 struct ext4_group_desc *bg, __u32 count)
253 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
258 void ext4_free_inodes_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
261 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
266 void ext4_used_dirs_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
269 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
274 void ext4_itable_unused_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
277 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
283 static void __save_error_info(struct super_block *sb, const char *func,
286 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
288 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
289 if (bdev_read_only(sb->s_bdev))
291 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
292 es->s_last_error_time = cpu_to_le32(get_seconds());
293 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
294 es->s_last_error_line = cpu_to_le32(line);
295 if (!es->s_first_error_time) {
296 es->s_first_error_time = es->s_last_error_time;
297 strncpy(es->s_first_error_func, func,
298 sizeof(es->s_first_error_func));
299 es->s_first_error_line = cpu_to_le32(line);
300 es->s_first_error_ino = es->s_last_error_ino;
301 es->s_first_error_block = es->s_last_error_block;
304 * Start the daily error reporting function if it hasn't been
307 if (!es->s_error_count)
308 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
309 le32_add_cpu(&es->s_error_count, 1);
312 static void save_error_info(struct super_block *sb, const char *func,
315 __save_error_info(sb, func, line);
316 ext4_commit_super(sb, 1);
320 * The del_gendisk() function uninitializes the disk-specific data
321 * structures, including the bdi structure, without telling anyone
322 * else. Once this happens, any attempt to call mark_buffer_dirty()
323 * (for example, by ext4_commit_super), will cause a kernel OOPS.
324 * This is a kludge to prevent these oops until we can put in a proper
325 * hook in del_gendisk() to inform the VFS and file system layers.
327 static int block_device_ejected(struct super_block *sb)
329 struct inode *bd_inode = sb->s_bdev->bd_inode;
330 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
332 return bdi->dev == NULL;
335 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
337 struct super_block *sb = journal->j_private;
338 struct ext4_sb_info *sbi = EXT4_SB(sb);
339 int error = is_journal_aborted(journal);
340 struct ext4_journal_cb_entry *jce;
342 BUG_ON(txn->t_state == T_FINISHED);
343 spin_lock(&sbi->s_md_lock);
344 while (!list_empty(&txn->t_private_list)) {
345 jce = list_entry(txn->t_private_list.next,
346 struct ext4_journal_cb_entry, jce_list);
347 list_del_init(&jce->jce_list);
348 spin_unlock(&sbi->s_md_lock);
349 jce->jce_func(sb, jce, error);
350 spin_lock(&sbi->s_md_lock);
352 spin_unlock(&sbi->s_md_lock);
355 /* Deal with the reporting of failure conditions on a filesystem such as
356 * inconsistencies detected or read IO failures.
358 * On ext2, we can store the error state of the filesystem in the
359 * superblock. That is not possible on ext4, because we may have other
360 * write ordering constraints on the superblock which prevent us from
361 * writing it out straight away; and given that the journal is about to
362 * be aborted, we can't rely on the current, or future, transactions to
363 * write out the superblock safely.
365 * We'll just use the jbd2_journal_abort() error code to record an error in
366 * the journal instead. On recovery, the journal will complain about
367 * that error until we've noted it down and cleared it.
370 static void ext4_handle_error(struct super_block *sb)
372 if (sb->s_flags & MS_RDONLY)
375 if (!test_opt(sb, ERRORS_CONT)) {
376 journal_t *journal = EXT4_SB(sb)->s_journal;
378 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
380 jbd2_journal_abort(journal, -EIO);
382 if (test_opt(sb, ERRORS_RO)) {
383 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
385 * Make sure updated value of ->s_mount_flags will be visible
386 * before ->s_flags update
389 sb->s_flags |= MS_RDONLY;
391 if (test_opt(sb, ERRORS_PANIC)) {
392 if (EXT4_SB(sb)->s_journal &&
393 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
395 panic("EXT4-fs (device %s): panic forced after error\n",
400 #define ext4_error_ratelimit(sb) \
401 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
404 void __ext4_error(struct super_block *sb, const char *function,
405 unsigned int line, const char *fmt, ...)
407 struct va_format vaf;
410 if (ext4_error_ratelimit(sb)) {
415 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
416 sb->s_id, function, line, current->comm, &vaf);
419 save_error_info(sb, function, line);
420 ext4_handle_error(sb);
423 void __ext4_error_inode(struct inode *inode, const char *function,
424 unsigned int line, ext4_fsblk_t block,
425 const char *fmt, ...)
428 struct va_format vaf;
429 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
431 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
432 es->s_last_error_block = cpu_to_le64(block);
433 if (ext4_error_ratelimit(inode->i_sb)) {
438 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
439 "inode #%lu: block %llu: comm %s: %pV\n",
440 inode->i_sb->s_id, function, line, inode->i_ino,
441 block, current->comm, &vaf);
443 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
444 "inode #%lu: comm %s: %pV\n",
445 inode->i_sb->s_id, function, line, inode->i_ino,
446 current->comm, &vaf);
449 save_error_info(inode->i_sb, function, line);
450 ext4_handle_error(inode->i_sb);
453 void __ext4_error_file(struct file *file, const char *function,
454 unsigned int line, ext4_fsblk_t block,
455 const char *fmt, ...)
458 struct va_format vaf;
459 struct ext4_super_block *es;
460 struct inode *inode = file_inode(file);
461 char pathname[80], *path;
463 es = EXT4_SB(inode->i_sb)->s_es;
464 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
465 if (ext4_error_ratelimit(inode->i_sb)) {
466 path = file_path(file, pathname, sizeof(pathname));
474 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
475 "block %llu: comm %s: path %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 block, current->comm, path, &vaf);
480 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
481 "comm %s: path %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, path, &vaf);
486 save_error_info(inode->i_sb, function, line);
487 ext4_handle_error(inode->i_sb);
490 const char *ext4_decode_error(struct super_block *sb, int errno,
497 errstr = "Corrupt filesystem";
500 errstr = "Filesystem failed CRC";
503 errstr = "IO failure";
506 errstr = "Out of memory";
509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 errstr = "Journal has aborted";
513 errstr = "Readonly filesystem";
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
530 /* __ext4_std_error decodes expected errors from journaling functions
531 * automatically and invokes the appropriate error response. */
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
546 if (ext4_error_ratelimit(sb)) {
547 errstr = ext4_decode_error(sb, errno, nbuf);
548 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
549 sb->s_id, function, line, errstr);
552 save_error_info(sb, function, line);
553 ext4_handle_error(sb);
557 * ext4_abort is a much stronger failure handler than ext4_error. The
558 * abort function may be used to deal with unrecoverable failures such
559 * as journal IO errors or ENOMEM at a critical moment in log management.
561 * We unconditionally force the filesystem into an ABORT|READONLY state,
562 * unless the error response on the fs has been set to panic in which
563 * case we take the easy way out and panic immediately.
566 void __ext4_abort(struct super_block *sb, const char *function,
567 unsigned int line, const char *fmt, ...)
571 save_error_info(sb, function, line);
573 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
579 if ((sb->s_flags & MS_RDONLY) == 0) {
580 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
581 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
583 * Make sure updated value of ->s_mount_flags will be visible
584 * before ->s_flags update
587 sb->s_flags |= MS_RDONLY;
588 if (EXT4_SB(sb)->s_journal)
589 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
590 save_error_info(sb, function, line);
592 if (test_opt(sb, ERRORS_PANIC)) {
593 if (EXT4_SB(sb)->s_journal &&
594 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
596 panic("EXT4-fs panic from previous error\n");
600 void __ext4_msg(struct super_block *sb,
601 const char *prefix, const char *fmt, ...)
603 struct va_format vaf;
606 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
612 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
616 #define ext4_warning_ratelimit(sb) \
617 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
620 void __ext4_warning(struct super_block *sb, const char *function,
621 unsigned int line, const char *fmt, ...)
623 struct va_format vaf;
626 if (!ext4_warning_ratelimit(sb))
632 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
633 sb->s_id, function, line, &vaf);
637 void __ext4_warning_inode(const struct inode *inode, const char *function,
638 unsigned int line, const char *fmt, ...)
640 struct va_format vaf;
643 if (!ext4_warning_ratelimit(inode->i_sb))
649 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
650 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
651 function, line, inode->i_ino, current->comm, &vaf);
655 void __ext4_grp_locked_error(const char *function, unsigned int line,
656 struct super_block *sb, ext4_group_t grp,
657 unsigned long ino, ext4_fsblk_t block,
658 const char *fmt, ...)
662 struct va_format vaf;
664 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
666 es->s_last_error_ino = cpu_to_le32(ino);
667 es->s_last_error_block = cpu_to_le64(block);
668 __save_error_info(sb, function, line);
670 if (ext4_error_ratelimit(sb)) {
674 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
675 sb->s_id, function, line, grp);
677 printk(KERN_CONT "inode %lu: ", ino);
679 printk(KERN_CONT "block %llu:",
680 (unsigned long long) block);
681 printk(KERN_CONT "%pV\n", &vaf);
685 if (test_opt(sb, ERRORS_CONT)) {
686 ext4_commit_super(sb, 0);
690 ext4_unlock_group(sb, grp);
691 ext4_handle_error(sb);
693 * We only get here in the ERRORS_RO case; relocking the group
694 * may be dangerous, but nothing bad will happen since the
695 * filesystem will have already been marked read/only and the
696 * journal has been aborted. We return 1 as a hint to callers
697 * who might what to use the return value from
698 * ext4_grp_locked_error() to distinguish between the
699 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
700 * aggressively from the ext4 function in question, with a
701 * more appropriate error code.
703 ext4_lock_group(sb, grp);
707 void ext4_update_dynamic_rev(struct super_block *sb)
709 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
711 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
715 "updating to rev %d because of new feature flag, "
716 "running e2fsck is recommended",
719 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
720 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
721 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
722 /* leave es->s_feature_*compat flags alone */
723 /* es->s_uuid will be set by e2fsck if empty */
726 * The rest of the superblock fields should be zero, and if not it
727 * means they are likely already in use, so leave them alone. We
728 * can leave it up to e2fsck to clean up any inconsistencies there.
733 * Open the external journal device
735 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
737 struct block_device *bdev;
738 char b[BDEVNAME_SIZE];
740 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
746 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
747 __bdevname(dev, b), PTR_ERR(bdev));
752 * Release the journal device
754 static void ext4_blkdev_put(struct block_device *bdev)
756 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
759 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
761 struct block_device *bdev;
762 bdev = sbi->journal_bdev;
764 ext4_blkdev_put(bdev);
765 sbi->journal_bdev = NULL;
769 static inline struct inode *orphan_list_entry(struct list_head *l)
771 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
774 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
778 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
779 le32_to_cpu(sbi->s_es->s_last_orphan));
781 printk(KERN_ERR "sb_info orphan list:\n");
782 list_for_each(l, &sbi->s_orphan) {
783 struct inode *inode = orphan_list_entry(l);
785 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
786 inode->i_sb->s_id, inode->i_ino, inode,
787 inode->i_mode, inode->i_nlink,
792 static void ext4_put_super(struct super_block *sb)
794 struct ext4_sb_info *sbi = EXT4_SB(sb);
795 struct ext4_super_block *es = sbi->s_es;
799 ext4_unregister_li_request(sb);
800 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
802 flush_workqueue(sbi->rsv_conversion_wq);
803 destroy_workqueue(sbi->rsv_conversion_wq);
805 if (sbi->s_journal) {
806 aborted = is_journal_aborted(sbi->s_journal);
807 err = jbd2_journal_destroy(sbi->s_journal);
808 sbi->s_journal = NULL;
809 if ((err < 0) && !aborted)
810 ext4_abort(sb, "Couldn't clean up the journal");
813 ext4_unregister_sysfs(sb);
814 ext4_es_unregister_shrinker(sbi);
815 del_timer_sync(&sbi->s_err_report);
816 ext4_release_system_zone(sb);
818 ext4_ext_release(sb);
820 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
821 ext4_clear_feature_journal_needs_recovery(sb);
822 es->s_state = cpu_to_le16(sbi->s_mount_state);
824 if (!(sb->s_flags & MS_RDONLY))
825 ext4_commit_super(sb, 1);
827 for (i = 0; i < sbi->s_gdb_count; i++)
828 brelse(sbi->s_group_desc[i]);
829 kvfree(sbi->s_group_desc);
830 kvfree(sbi->s_flex_groups);
831 percpu_counter_destroy(&sbi->s_freeclusters_counter);
832 percpu_counter_destroy(&sbi->s_freeinodes_counter);
833 percpu_counter_destroy(&sbi->s_dirs_counter);
834 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
837 for (i = 0; i < EXT4_MAXQUOTAS; i++)
838 kfree(sbi->s_qf_names[i]);
841 /* Debugging code just in case the in-memory inode orphan list
842 * isn't empty. The on-disk one can be non-empty if we've
843 * detected an error and taken the fs readonly, but the
844 * in-memory list had better be clean by this point. */
845 if (!list_empty(&sbi->s_orphan))
846 dump_orphan_list(sb, sbi);
847 J_ASSERT(list_empty(&sbi->s_orphan));
849 sync_blockdev(sb->s_bdev);
850 invalidate_bdev(sb->s_bdev);
851 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
853 * Invalidate the journal device's buffers. We don't want them
854 * floating about in memory - the physical journal device may
855 * hotswapped, and it breaks the `ro-after' testing code.
857 sync_blockdev(sbi->journal_bdev);
858 invalidate_bdev(sbi->journal_bdev);
859 ext4_blkdev_remove(sbi);
861 if (sbi->s_mb_cache) {
862 ext4_xattr_destroy_cache(sbi->s_mb_cache);
863 sbi->s_mb_cache = NULL;
866 kthread_stop(sbi->s_mmp_tsk);
867 sb->s_fs_info = NULL;
869 * Now that we are completely done shutting down the
870 * superblock, we need to actually destroy the kobject.
872 kobject_put(&sbi->s_kobj);
873 wait_for_completion(&sbi->s_kobj_unregister);
874 if (sbi->s_chksum_driver)
875 crypto_free_shash(sbi->s_chksum_driver);
876 kfree(sbi->s_blockgroup_lock);
880 static struct kmem_cache *ext4_inode_cachep;
883 * Called inside transaction, so use GFP_NOFS
885 static struct inode *ext4_alloc_inode(struct super_block *sb)
887 struct ext4_inode_info *ei;
889 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
893 ei->vfs_inode.i_version = 1;
894 spin_lock_init(&ei->i_raw_lock);
895 INIT_LIST_HEAD(&ei->i_prealloc_list);
896 spin_lock_init(&ei->i_prealloc_lock);
897 ext4_es_init_tree(&ei->i_es_tree);
898 rwlock_init(&ei->i_es_lock);
899 INIT_LIST_HEAD(&ei->i_es_list);
902 ei->i_es_shrink_lblk = 0;
903 ei->i_reserved_data_blocks = 0;
904 ei->i_reserved_meta_blocks = 0;
905 ei->i_allocated_meta_blocks = 0;
906 ei->i_da_metadata_calc_len = 0;
907 ei->i_da_metadata_calc_last_lblock = 0;
908 spin_lock_init(&(ei->i_block_reservation_lock));
910 ei->i_reserved_quota = 0;
911 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
914 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
915 spin_lock_init(&ei->i_completed_io_lock);
917 ei->i_datasync_tid = 0;
918 atomic_set(&ei->i_ioend_count, 0);
919 atomic_set(&ei->i_unwritten, 0);
920 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
921 #ifdef CONFIG_EXT4_FS_ENCRYPTION
922 ei->i_crypt_info = NULL;
924 return &ei->vfs_inode;
927 static int ext4_drop_inode(struct inode *inode)
929 int drop = generic_drop_inode(inode);
931 trace_ext4_drop_inode(inode, drop);
935 static void ext4_i_callback(struct rcu_head *head)
937 struct inode *inode = container_of(head, struct inode, i_rcu);
938 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
941 static void ext4_destroy_inode(struct inode *inode)
943 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
944 ext4_msg(inode->i_sb, KERN_ERR,
945 "Inode %lu (%p): orphan list check failed!",
946 inode->i_ino, EXT4_I(inode));
947 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
948 EXT4_I(inode), sizeof(struct ext4_inode_info),
952 call_rcu(&inode->i_rcu, ext4_i_callback);
955 static void init_once(void *foo)
957 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
959 INIT_LIST_HEAD(&ei->i_orphan);
960 init_rwsem(&ei->xattr_sem);
961 init_rwsem(&ei->i_data_sem);
962 init_rwsem(&ei->i_mmap_sem);
963 inode_init_once(&ei->vfs_inode);
966 static int __init init_inodecache(void)
968 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
969 sizeof(struct ext4_inode_info),
970 0, (SLAB_RECLAIM_ACCOUNT|
973 if (ext4_inode_cachep == NULL)
978 static void destroy_inodecache(void)
981 * Make sure all delayed rcu free inodes are flushed before we
985 kmem_cache_destroy(ext4_inode_cachep);
988 void ext4_clear_inode(struct inode *inode)
990 invalidate_inode_buffers(inode);
993 ext4_discard_preallocations(inode);
994 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
995 if (EXT4_I(inode)->jinode) {
996 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
997 EXT4_I(inode)->jinode);
998 jbd2_free_inode(EXT4_I(inode)->jinode);
999 EXT4_I(inode)->jinode = NULL;
1001 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1002 if (EXT4_I(inode)->i_crypt_info)
1003 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1007 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1008 u64 ino, u32 generation)
1010 struct inode *inode;
1012 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1013 return ERR_PTR(-ESTALE);
1014 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1015 return ERR_PTR(-ESTALE);
1017 /* iget isn't really right if the inode is currently unallocated!!
1019 * ext4_read_inode will return a bad_inode if the inode had been
1020 * deleted, so we should be safe.
1022 * Currently we don't know the generation for parent directory, so
1023 * a generation of 0 means "accept any"
1025 inode = ext4_iget_normal(sb, ino);
1027 return ERR_CAST(inode);
1028 if (generation && inode->i_generation != generation) {
1030 return ERR_PTR(-ESTALE);
1036 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1037 int fh_len, int fh_type)
1039 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1040 ext4_nfs_get_inode);
1043 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1044 int fh_len, int fh_type)
1046 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1047 ext4_nfs_get_inode);
1051 * Try to release metadata pages (indirect blocks, directories) which are
1052 * mapped via the block device. Since these pages could have journal heads
1053 * which would prevent try_to_free_buffers() from freeing them, we must use
1054 * jbd2 layer's try_to_free_buffers() function to release them.
1056 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1059 journal_t *journal = EXT4_SB(sb)->s_journal;
1061 WARN_ON(PageChecked(page));
1062 if (!page_has_buffers(page))
1065 return jbd2_journal_try_to_free_buffers(journal, page,
1066 wait & ~__GFP_DIRECT_RECLAIM);
1067 return try_to_free_buffers(page);
1071 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1072 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1074 static int ext4_write_dquot(struct dquot *dquot);
1075 static int ext4_acquire_dquot(struct dquot *dquot);
1076 static int ext4_release_dquot(struct dquot *dquot);
1077 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1078 static int ext4_write_info(struct super_block *sb, int type);
1079 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1081 static int ext4_quota_off(struct super_block *sb, int type);
1082 static int ext4_quota_on_mount(struct super_block *sb, int type);
1083 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1084 size_t len, loff_t off);
1085 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1086 const char *data, size_t len, loff_t off);
1087 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1088 unsigned int flags);
1089 static int ext4_enable_quotas(struct super_block *sb);
1091 static struct dquot **ext4_get_dquots(struct inode *inode)
1093 return EXT4_I(inode)->i_dquot;
1096 static const struct dquot_operations ext4_quota_operations = {
1097 .get_reserved_space = ext4_get_reserved_space,
1098 .write_dquot = ext4_write_dquot,
1099 .acquire_dquot = ext4_acquire_dquot,
1100 .release_dquot = ext4_release_dquot,
1101 .mark_dirty = ext4_mark_dquot_dirty,
1102 .write_info = ext4_write_info,
1103 .alloc_dquot = dquot_alloc,
1104 .destroy_dquot = dquot_destroy,
1107 static const struct quotactl_ops ext4_qctl_operations = {
1108 .quota_on = ext4_quota_on,
1109 .quota_off = ext4_quota_off,
1110 .quota_sync = dquot_quota_sync,
1111 .get_state = dquot_get_state,
1112 .set_info = dquot_set_dqinfo,
1113 .get_dqblk = dquot_get_dqblk,
1114 .set_dqblk = dquot_set_dqblk
1118 static const struct super_operations ext4_sops = {
1119 .alloc_inode = ext4_alloc_inode,
1120 .destroy_inode = ext4_destroy_inode,
1121 .write_inode = ext4_write_inode,
1122 .dirty_inode = ext4_dirty_inode,
1123 .drop_inode = ext4_drop_inode,
1124 .evict_inode = ext4_evict_inode,
1125 .put_super = ext4_put_super,
1126 .sync_fs = ext4_sync_fs,
1127 .freeze_fs = ext4_freeze,
1128 .unfreeze_fs = ext4_unfreeze,
1129 .statfs = ext4_statfs,
1130 .remount_fs = ext4_remount,
1131 .show_options = ext4_show_options,
1133 .quota_read = ext4_quota_read,
1134 .quota_write = ext4_quota_write,
1135 .get_dquots = ext4_get_dquots,
1137 .bdev_try_to_free_page = bdev_try_to_free_page,
1140 static const struct export_operations ext4_export_ops = {
1141 .fh_to_dentry = ext4_fh_to_dentry,
1142 .fh_to_parent = ext4_fh_to_parent,
1143 .get_parent = ext4_get_parent,
1147 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1148 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1149 Opt_nouid32, Opt_debug, Opt_removed,
1150 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1151 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1152 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1153 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1154 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1155 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1156 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1157 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1158 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1159 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1160 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1161 Opt_lazytime, Opt_nolazytime,
1162 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1163 Opt_inode_readahead_blks, Opt_journal_ioprio,
1164 Opt_dioread_nolock, Opt_dioread_lock,
1165 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1166 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1169 static const match_table_t tokens = {
1170 {Opt_bsd_df, "bsddf"},
1171 {Opt_minix_df, "minixdf"},
1172 {Opt_grpid, "grpid"},
1173 {Opt_grpid, "bsdgroups"},
1174 {Opt_nogrpid, "nogrpid"},
1175 {Opt_nogrpid, "sysvgroups"},
1176 {Opt_resgid, "resgid=%u"},
1177 {Opt_resuid, "resuid=%u"},
1179 {Opt_err_cont, "errors=continue"},
1180 {Opt_err_panic, "errors=panic"},
1181 {Opt_err_ro, "errors=remount-ro"},
1182 {Opt_nouid32, "nouid32"},
1183 {Opt_debug, "debug"},
1184 {Opt_removed, "oldalloc"},
1185 {Opt_removed, "orlov"},
1186 {Opt_user_xattr, "user_xattr"},
1187 {Opt_nouser_xattr, "nouser_xattr"},
1189 {Opt_noacl, "noacl"},
1190 {Opt_noload, "norecovery"},
1191 {Opt_noload, "noload"},
1192 {Opt_removed, "nobh"},
1193 {Opt_removed, "bh"},
1194 {Opt_commit, "commit=%u"},
1195 {Opt_min_batch_time, "min_batch_time=%u"},
1196 {Opt_max_batch_time, "max_batch_time=%u"},
1197 {Opt_journal_dev, "journal_dev=%u"},
1198 {Opt_journal_path, "journal_path=%s"},
1199 {Opt_journal_checksum, "journal_checksum"},
1200 {Opt_nojournal_checksum, "nojournal_checksum"},
1201 {Opt_journal_async_commit, "journal_async_commit"},
1202 {Opt_abort, "abort"},
1203 {Opt_data_journal, "data=journal"},
1204 {Opt_data_ordered, "data=ordered"},
1205 {Opt_data_writeback, "data=writeback"},
1206 {Opt_data_err_abort, "data_err=abort"},
1207 {Opt_data_err_ignore, "data_err=ignore"},
1208 {Opt_offusrjquota, "usrjquota="},
1209 {Opt_usrjquota, "usrjquota=%s"},
1210 {Opt_offgrpjquota, "grpjquota="},
1211 {Opt_grpjquota, "grpjquota=%s"},
1212 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1213 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1214 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1215 {Opt_grpquota, "grpquota"},
1216 {Opt_noquota, "noquota"},
1217 {Opt_quota, "quota"},
1218 {Opt_usrquota, "usrquota"},
1219 {Opt_barrier, "barrier=%u"},
1220 {Opt_barrier, "barrier"},
1221 {Opt_nobarrier, "nobarrier"},
1222 {Opt_i_version, "i_version"},
1224 {Opt_stripe, "stripe=%u"},
1225 {Opt_delalloc, "delalloc"},
1226 {Opt_lazytime, "lazytime"},
1227 {Opt_nolazytime, "nolazytime"},
1228 {Opt_nodelalloc, "nodelalloc"},
1229 {Opt_removed, "mblk_io_submit"},
1230 {Opt_removed, "nomblk_io_submit"},
1231 {Opt_block_validity, "block_validity"},
1232 {Opt_noblock_validity, "noblock_validity"},
1233 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1234 {Opt_journal_ioprio, "journal_ioprio=%u"},
1235 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1236 {Opt_auto_da_alloc, "auto_da_alloc"},
1237 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1238 {Opt_dioread_nolock, "dioread_nolock"},
1239 {Opt_dioread_lock, "dioread_lock"},
1240 {Opt_discard, "discard"},
1241 {Opt_nodiscard, "nodiscard"},
1242 {Opt_init_itable, "init_itable=%u"},
1243 {Opt_init_itable, "init_itable"},
1244 {Opt_noinit_itable, "noinit_itable"},
1245 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1246 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1255 static ext4_fsblk_t get_sb_block(void **data)
1257 ext4_fsblk_t sb_block;
1258 char *options = (char *) *data;
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
1264 /* TODO: use simple_strtoll with >32bit ext4 */
1265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
1267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1271 if (*options == ',')
1273 *data = (void *) options;
1278 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1279 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1283 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
1296 if (ext4_has_feature_quota(sb)) {
1297 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1298 "ignored when QUOTA feature is enabled");
1301 qname = match_strdup(args);
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
1307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1316 if (strchr(qname, '/')) {
1317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
1321 sbi->s_qf_names[qtype] = qname;
1329 static int clear_qf_name(struct super_block *sb, int qtype)
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
1340 kfree(sbi->s_qf_names[qtype]);
1341 sbi->s_qf_names[qtype] = NULL;
1346 #define MOPT_SET 0x0001
1347 #define MOPT_CLEAR 0x0002
1348 #define MOPT_NOSUPPORT 0x0004
1349 #define MOPT_EXPLICIT 0x0008
1350 #define MOPT_CLEAR_ERR 0x0010
1351 #define MOPT_GTE0 0x0020
1354 #define MOPT_QFMT 0x0040
1356 #define MOPT_Q MOPT_NOSUPPORT
1357 #define MOPT_QFMT MOPT_NOSUPPORT
1359 #define MOPT_DATAJ 0x0080
1360 #define MOPT_NO_EXT2 0x0100
1361 #define MOPT_NO_EXT3 0x0200
1362 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1363 #define MOPT_STRING 0x0400
1365 static const struct mount_opts {
1369 } ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
1380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_CLEAR},
1388 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1389 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1390 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1391 EXT4_MOUNT_JOURNAL_CHECKSUM),
1392 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1393 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1394 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1397 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_SET},
1399 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1400 MOPT_NO_EXT2 | MOPT_CLEAR},
1401 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1402 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1403 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1404 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1405 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1406 {Opt_commit, 0, MOPT_GTE0},
1407 {Opt_max_batch_time, 0, MOPT_GTE0},
1408 {Opt_min_batch_time, 0, MOPT_GTE0},
1409 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1410 {Opt_init_itable, 0, MOPT_GTE0},
1411 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1412 {Opt_stripe, 0, MOPT_GTE0},
1413 {Opt_resuid, 0, MOPT_GTE0},
1414 {Opt_resgid, 0, MOPT_GTE0},
1415 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1416 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1417 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1418 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1420 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1421 MOPT_NO_EXT2 | MOPT_DATAJ},
1422 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1423 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1424 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1425 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1426 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1428 {Opt_acl, 0, MOPT_NOSUPPORT},
1429 {Opt_noacl, 0, MOPT_NOSUPPORT},
1431 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1432 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1433 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1434 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1436 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1438 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1439 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1440 {Opt_usrjquota, 0, MOPT_Q},
1441 {Opt_grpjquota, 0, MOPT_Q},
1442 {Opt_offusrjquota, 0, MOPT_Q},
1443 {Opt_offgrpjquota, 0, MOPT_Q},
1444 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1445 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1446 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1447 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1448 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1452 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1453 substring_t *args, unsigned long *journal_devnum,
1454 unsigned int *journal_ioprio, int is_remount)
1456 struct ext4_sb_info *sbi = EXT4_SB(sb);
1457 const struct mount_opts *m;
1463 if (token == Opt_usrjquota)
1464 return set_qf_name(sb, USRQUOTA, &args[0]);
1465 else if (token == Opt_grpjquota)
1466 return set_qf_name(sb, GRPQUOTA, &args[0]);
1467 else if (token == Opt_offusrjquota)
1468 return clear_qf_name(sb, USRQUOTA);
1469 else if (token == Opt_offgrpjquota)
1470 return clear_qf_name(sb, GRPQUOTA);
1474 case Opt_nouser_xattr:
1475 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1478 return 1; /* handled by get_sb_block() */
1480 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1483 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1486 sb->s_flags |= MS_I_VERSION;
1489 sb->s_flags |= MS_LAZYTIME;
1491 case Opt_nolazytime:
1492 sb->s_flags &= ~MS_LAZYTIME;
1496 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1497 if (token == m->token)
1500 if (m->token == Opt_err) {
1501 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1502 "or missing value", opt);
1506 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1507 ext4_msg(sb, KERN_ERR,
1508 "Mount option \"%s\" incompatible with ext2", opt);
1511 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1512 ext4_msg(sb, KERN_ERR,
1513 "Mount option \"%s\" incompatible with ext3", opt);
1517 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1519 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1521 if (m->flags & MOPT_EXPLICIT) {
1522 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1523 set_opt2(sb, EXPLICIT_DELALLOC);
1524 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1525 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1529 if (m->flags & MOPT_CLEAR_ERR)
1530 clear_opt(sb, ERRORS_MASK);
1531 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1532 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1533 "options when quota turned on");
1537 if (m->flags & MOPT_NOSUPPORT) {
1538 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1539 } else if (token == Opt_commit) {
1541 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1542 sbi->s_commit_interval = HZ * arg;
1543 } else if (token == Opt_max_batch_time) {
1544 sbi->s_max_batch_time = arg;
1545 } else if (token == Opt_min_batch_time) {
1546 sbi->s_min_batch_time = arg;
1547 } else if (token == Opt_inode_readahead_blks) {
1548 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1549 ext4_msg(sb, KERN_ERR,
1550 "EXT4-fs: inode_readahead_blks must be "
1551 "0 or a power of 2 smaller than 2^31");
1554 sbi->s_inode_readahead_blks = arg;
1555 } else if (token == Opt_init_itable) {
1556 set_opt(sb, INIT_INODE_TABLE);
1558 arg = EXT4_DEF_LI_WAIT_MULT;
1559 sbi->s_li_wait_mult = arg;
1560 } else if (token == Opt_max_dir_size_kb) {
1561 sbi->s_max_dir_size_kb = arg;
1562 } else if (token == Opt_stripe) {
1563 sbi->s_stripe = arg;
1564 } else if (token == Opt_resuid) {
1565 uid = make_kuid(current_user_ns(), arg);
1566 if (!uid_valid(uid)) {
1567 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1570 sbi->s_resuid = uid;
1571 } else if (token == Opt_resgid) {
1572 gid = make_kgid(current_user_ns(), arg);
1573 if (!gid_valid(gid)) {
1574 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1577 sbi->s_resgid = gid;
1578 } else if (token == Opt_journal_dev) {
1580 ext4_msg(sb, KERN_ERR,
1581 "Cannot specify journal on remount");
1584 *journal_devnum = arg;
1585 } else if (token == Opt_journal_path) {
1587 struct inode *journal_inode;
1592 ext4_msg(sb, KERN_ERR,
1593 "Cannot specify journal on remount");
1596 journal_path = match_strdup(&args[0]);
1597 if (!journal_path) {
1598 ext4_msg(sb, KERN_ERR, "error: could not dup "
1599 "journal device string");
1603 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1605 ext4_msg(sb, KERN_ERR, "error: could not find "
1606 "journal device path: error %d", error);
1607 kfree(journal_path);
1611 journal_inode = d_inode(path.dentry);
1612 if (!S_ISBLK(journal_inode->i_mode)) {
1613 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1614 "is not a block device", journal_path);
1616 kfree(journal_path);
1620 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1622 kfree(journal_path);
1623 } else if (token == Opt_journal_ioprio) {
1625 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1630 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1631 } else if (token == Opt_test_dummy_encryption) {
1632 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1633 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1634 ext4_msg(sb, KERN_WARNING,
1635 "Test dummy encryption mode enabled");
1637 ext4_msg(sb, KERN_WARNING,
1638 "Test dummy encryption mount option ignored");
1640 } else if (m->flags & MOPT_DATAJ) {
1642 if (!sbi->s_journal)
1643 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1644 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1645 ext4_msg(sb, KERN_ERR,
1646 "Cannot change data mode on remount");
1650 clear_opt(sb, DATA_FLAGS);
1651 sbi->s_mount_opt |= m->mount_opt;
1654 } else if (m->flags & MOPT_QFMT) {
1655 if (sb_any_quota_loaded(sb) &&
1656 sbi->s_jquota_fmt != m->mount_opt) {
1657 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1658 "quota options when quota turned on");
1661 if (ext4_has_feature_quota(sb)) {
1662 ext4_msg(sb, KERN_INFO,
1663 "Quota format mount options ignored "
1664 "when QUOTA feature is enabled");
1667 sbi->s_jquota_fmt = m->mount_opt;
1669 } else if (token == Opt_dax) {
1670 #ifdef CONFIG_FS_DAX
1671 ext4_msg(sb, KERN_WARNING,
1672 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1673 sbi->s_mount_opt |= m->mount_opt;
1675 ext4_msg(sb, KERN_INFO, "dax option not supported");
1681 if (m->flags & MOPT_CLEAR)
1683 else if (unlikely(!(m->flags & MOPT_SET))) {
1684 ext4_msg(sb, KERN_WARNING,
1685 "buggy handling of option %s", opt);
1690 sbi->s_mount_opt |= m->mount_opt;
1692 sbi->s_mount_opt &= ~m->mount_opt;
1697 static int parse_options(char *options, struct super_block *sb,
1698 unsigned long *journal_devnum,
1699 unsigned int *journal_ioprio,
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1704 substring_t args[MAX_OPT_ARGS];
1710 while ((p = strsep(&options, ",")) != NULL) {
1714 * Initialize args struct so we know whether arg was
1715 * found; some options take optional arguments.
1717 args[0].to = args[0].from = NULL;
1718 token = match_token(p, tokens, args);
1719 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1720 journal_ioprio, is_remount) < 0)
1724 if (ext4_has_feature_quota(sb) &&
1725 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1726 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1727 "mount options ignored.");
1728 clear_opt(sb, USRQUOTA);
1729 clear_opt(sb, GRPQUOTA);
1730 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1731 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1732 clear_opt(sb, USRQUOTA);
1734 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1735 clear_opt(sb, GRPQUOTA);
1737 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1738 ext4_msg(sb, KERN_ERR, "old and new quota "
1743 if (!sbi->s_jquota_fmt) {
1744 ext4_msg(sb, KERN_ERR, "journaled quota format "
1750 if (test_opt(sb, DIOREAD_NOLOCK)) {
1752 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1754 if (blocksize < PAGE_CACHE_SIZE) {
1755 ext4_msg(sb, KERN_ERR, "can't mount with "
1756 "dioread_nolock if block size != PAGE_SIZE");
1760 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1761 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1762 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1763 "in data=ordered mode");
1769 static inline void ext4_show_quota_options(struct seq_file *seq,
1770 struct super_block *sb)
1772 #if defined(CONFIG_QUOTA)
1773 struct ext4_sb_info *sbi = EXT4_SB(sb);
1775 if (sbi->s_jquota_fmt) {
1778 switch (sbi->s_jquota_fmt) {
1789 seq_printf(seq, ",jqfmt=%s", fmtname);
1792 if (sbi->s_qf_names[USRQUOTA])
1793 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1795 if (sbi->s_qf_names[GRPQUOTA])
1796 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1800 static const char *token2str(int token)
1802 const struct match_token *t;
1804 for (t = tokens; t->token != Opt_err; t++)
1805 if (t->token == token && !strchr(t->pattern, '='))
1812 * - it's set to a non-default value OR
1813 * - if the per-sb default is different from the global default
1815 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1818 struct ext4_sb_info *sbi = EXT4_SB(sb);
1819 struct ext4_super_block *es = sbi->s_es;
1820 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1821 const struct mount_opts *m;
1822 char sep = nodefs ? '\n' : ',';
1824 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1825 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1827 if (sbi->s_sb_block != 1)
1828 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1830 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1831 int want_set = m->flags & MOPT_SET;
1832 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1833 (m->flags & MOPT_CLEAR_ERR))
1835 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1836 continue; /* skip if same as the default */
1838 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1839 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1840 continue; /* select Opt_noFoo vs Opt_Foo */
1841 SEQ_OPTS_PRINT("%s", token2str(m->token));
1844 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1845 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1846 SEQ_OPTS_PRINT("resuid=%u",
1847 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1848 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1849 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1850 SEQ_OPTS_PRINT("resgid=%u",
1851 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1852 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1853 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1854 SEQ_OPTS_PUTS("errors=remount-ro");
1855 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1856 SEQ_OPTS_PUTS("errors=continue");
1857 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1858 SEQ_OPTS_PUTS("errors=panic");
1859 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1860 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1861 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1862 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1863 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1864 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1865 if (sb->s_flags & MS_I_VERSION)
1866 SEQ_OPTS_PUTS("i_version");
1867 if (nodefs || sbi->s_stripe)
1868 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1869 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1870 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1871 SEQ_OPTS_PUTS("data=journal");
1872 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1873 SEQ_OPTS_PUTS("data=ordered");
1874 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1875 SEQ_OPTS_PUTS("data=writeback");
1878 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1879 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1880 sbi->s_inode_readahead_blks);
1882 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1883 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1884 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1885 if (nodefs || sbi->s_max_dir_size_kb)
1886 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1888 ext4_show_quota_options(seq, sb);
1892 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1894 return _ext4_show_options(seq, root->d_sb, 0);
1897 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1899 struct super_block *sb = seq->private;
1902 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1903 rc = _ext4_show_options(seq, sb, 1);
1904 seq_puts(seq, "\n");
1908 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1911 struct ext4_sb_info *sbi = EXT4_SB(sb);
1914 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1915 ext4_msg(sb, KERN_ERR, "revision level too high, "
1916 "forcing read-only mode");
1921 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1922 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1923 "running e2fsck is recommended");
1924 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1925 ext4_msg(sb, KERN_WARNING,
1926 "warning: mounting fs with errors, "
1927 "running e2fsck is recommended");
1928 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1929 le16_to_cpu(es->s_mnt_count) >=
1930 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1931 ext4_msg(sb, KERN_WARNING,
1932 "warning: maximal mount count reached, "
1933 "running e2fsck is recommended");
1934 else if (le32_to_cpu(es->s_checkinterval) &&
1935 (le32_to_cpu(es->s_lastcheck) +
1936 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1937 ext4_msg(sb, KERN_WARNING,
1938 "warning: checktime reached, "
1939 "running e2fsck is recommended");
1940 if (!sbi->s_journal)
1941 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1942 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1943 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1944 le16_add_cpu(&es->s_mnt_count, 1);
1945 es->s_mtime = cpu_to_le32(get_seconds());
1946 ext4_update_dynamic_rev(sb);
1948 ext4_set_feature_journal_needs_recovery(sb);
1950 ext4_commit_super(sb, 1);
1952 if (test_opt(sb, DEBUG))
1953 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1954 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1956 sbi->s_groups_count,
1957 EXT4_BLOCKS_PER_GROUP(sb),
1958 EXT4_INODES_PER_GROUP(sb),
1959 sbi->s_mount_opt, sbi->s_mount_opt2);
1961 cleancache_init_fs(sb);
1965 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1967 struct ext4_sb_info *sbi = EXT4_SB(sb);
1968 struct flex_groups *new_groups;
1971 if (!sbi->s_log_groups_per_flex)
1974 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1975 if (size <= sbi->s_flex_groups_allocated)
1978 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1979 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1981 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1982 size / (int) sizeof(struct flex_groups));
1986 if (sbi->s_flex_groups) {
1987 memcpy(new_groups, sbi->s_flex_groups,
1988 (sbi->s_flex_groups_allocated *
1989 sizeof(struct flex_groups)));
1990 kvfree(sbi->s_flex_groups);
1992 sbi->s_flex_groups = new_groups;
1993 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1997 static int ext4_fill_flex_info(struct super_block *sb)
1999 struct ext4_sb_info *sbi = EXT4_SB(sb);
2000 struct ext4_group_desc *gdp = NULL;
2001 ext4_group_t flex_group;
2004 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2005 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2006 sbi->s_log_groups_per_flex = 0;
2010 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2014 for (i = 0; i < sbi->s_groups_count; i++) {
2015 gdp = ext4_get_group_desc(sb, i, NULL);
2017 flex_group = ext4_flex_group(sbi, i);
2018 atomic_add(ext4_free_inodes_count(sb, gdp),
2019 &sbi->s_flex_groups[flex_group].free_inodes);
2020 atomic64_add(ext4_free_group_clusters(sb, gdp),
2021 &sbi->s_flex_groups[flex_group].free_clusters);
2022 atomic_add(ext4_used_dirs_count(sb, gdp),
2023 &sbi->s_flex_groups[flex_group].used_dirs);
2031 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2032 struct ext4_group_desc *gdp)
2034 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2036 __le32 le_group = cpu_to_le32(block_group);
2037 struct ext4_sb_info *sbi = EXT4_SB(sb);
2039 if (ext4_has_metadata_csum(sbi->s_sb)) {
2040 /* Use new metadata_csum algorithm */
2042 __u16 dummy_csum = 0;
2044 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2046 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2047 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2048 sizeof(dummy_csum));
2049 offset += sizeof(dummy_csum);
2050 if (offset < sbi->s_desc_size)
2051 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2052 sbi->s_desc_size - offset);
2054 crc = csum32 & 0xFFFF;
2058 /* old crc16 code */
2059 if (!ext4_has_feature_gdt_csum(sb))
2062 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2063 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2064 crc = crc16(crc, (__u8 *)gdp, offset);
2065 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2066 /* for checksum of struct ext4_group_desc do the rest...*/
2067 if (ext4_has_feature_64bit(sb) &&
2068 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2069 crc = crc16(crc, (__u8 *)gdp + offset,
2070 le16_to_cpu(sbi->s_es->s_desc_size) -
2074 return cpu_to_le16(crc);
2077 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2078 struct ext4_group_desc *gdp)
2080 if (ext4_has_group_desc_csum(sb) &&
2081 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2087 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2088 struct ext4_group_desc *gdp)
2090 if (!ext4_has_group_desc_csum(sb))
2092 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2095 /* Called at mount-time, super-block is locked */
2096 static int ext4_check_descriptors(struct super_block *sb,
2097 ext4_fsblk_t sb_block,
2098 ext4_group_t *first_not_zeroed)
2100 struct ext4_sb_info *sbi = EXT4_SB(sb);
2101 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2102 ext4_fsblk_t last_block;
2103 ext4_fsblk_t block_bitmap;
2104 ext4_fsblk_t inode_bitmap;
2105 ext4_fsblk_t inode_table;
2106 int flexbg_flag = 0;
2107 ext4_group_t i, grp = sbi->s_groups_count;
2109 if (ext4_has_feature_flex_bg(sb))
2112 ext4_debug("Checking group descriptors");
2114 for (i = 0; i < sbi->s_groups_count; i++) {
2115 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2117 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2118 last_block = ext4_blocks_count(sbi->s_es) - 1;
2120 last_block = first_block +
2121 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2123 if ((grp == sbi->s_groups_count) &&
2124 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2127 block_bitmap = ext4_block_bitmap(sb, gdp);
2128 if (block_bitmap == sb_block) {
2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2130 "Block bitmap for group %u overlaps "
2133 if (block_bitmap < first_block || block_bitmap > last_block) {
2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2135 "Block bitmap for group %u not in group "
2136 "(block %llu)!", i, block_bitmap);
2139 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2140 if (inode_bitmap == sb_block) {
2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2142 "Inode bitmap for group %u overlaps "
2145 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2146 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2147 "Inode bitmap for group %u not in group "
2148 "(block %llu)!", i, inode_bitmap);
2151 inode_table = ext4_inode_table(sb, gdp);
2152 if (inode_table == sb_block) {
2153 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2154 "Inode table for group %u overlaps "
2157 if (inode_table < first_block ||
2158 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2159 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2160 "Inode table for group %u not in group "
2161 "(block %llu)!", i, inode_table);
2164 ext4_lock_group(sb, i);
2165 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2166 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2167 "Checksum for group %u failed (%u!=%u)",
2168 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2169 gdp)), le16_to_cpu(gdp->bg_checksum));
2170 if (!(sb->s_flags & MS_RDONLY)) {
2171 ext4_unlock_group(sb, i);
2175 ext4_unlock_group(sb, i);
2177 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2179 if (NULL != first_not_zeroed)
2180 *first_not_zeroed = grp;
2184 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2185 * the superblock) which were deleted from all directories, but held open by
2186 * a process at the time of a crash. We walk the list and try to delete these
2187 * inodes at recovery time (only with a read-write filesystem).
2189 * In order to keep the orphan inode chain consistent during traversal (in
2190 * case of crash during recovery), we link each inode into the superblock
2191 * orphan list_head and handle it the same way as an inode deletion during
2192 * normal operation (which journals the operations for us).
2194 * We only do an iget() and an iput() on each inode, which is very safe if we
2195 * accidentally point at an in-use or already deleted inode. The worst that
2196 * can happen in this case is that we get a "bit already cleared" message from
2197 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2198 * e2fsck was run on this filesystem, and it must have already done the orphan
2199 * inode cleanup for us, so we can safely abort without any further action.
2201 static void ext4_orphan_cleanup(struct super_block *sb,
2202 struct ext4_super_block *es)
2204 unsigned int s_flags = sb->s_flags;
2205 int nr_orphans = 0, nr_truncates = 0;
2207 int quota_update = 0;
2210 if (!es->s_last_orphan) {
2211 jbd_debug(4, "no orphan inodes to clean up\n");
2215 if (bdev_read_only(sb->s_bdev)) {
2216 ext4_msg(sb, KERN_ERR, "write access "
2217 "unavailable, skipping orphan cleanup");
2221 /* Check if feature set would not allow a r/w mount */
2222 if (!ext4_feature_set_ok(sb, 0)) {
2223 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2224 "unknown ROCOMPAT features");
2228 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2229 /* don't clear list on RO mount w/ errors */
2230 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2231 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2232 "clearing orphan list.\n");
2233 es->s_last_orphan = 0;
2235 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2239 if (s_flags & MS_RDONLY) {
2240 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2241 sb->s_flags &= ~MS_RDONLY;
2244 /* Needed for iput() to work correctly and not trash data */
2245 sb->s_flags |= MS_ACTIVE;
2248 * Turn on quotas which were not enabled for read-only mounts if
2249 * filesystem has quota feature, so that they are updated correctly.
2251 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2252 int ret = ext4_enable_quotas(sb);
2257 ext4_msg(sb, KERN_ERR,
2258 "Cannot turn on quotas: error %d", ret);
2261 /* Turn on journaled quotas used for old sytle */
2262 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2263 if (EXT4_SB(sb)->s_qf_names[i]) {
2264 int ret = ext4_quota_on_mount(sb, i);
2269 ext4_msg(sb, KERN_ERR,
2270 "Cannot turn on journaled "
2271 "quota: type %d: error %d", i, ret);
2276 while (es->s_last_orphan) {
2277 struct inode *inode;
2280 * We may have encountered an error during cleanup; if
2281 * so, skip the rest.
2283 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2284 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2285 es->s_last_orphan = 0;
2289 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2290 if (IS_ERR(inode)) {
2291 es->s_last_orphan = 0;
2295 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2296 dquot_initialize(inode);
2297 if (inode->i_nlink) {
2298 if (test_opt(sb, DEBUG))
2299 ext4_msg(sb, KERN_DEBUG,
2300 "%s: truncating inode %lu to %lld bytes",
2301 __func__, inode->i_ino, inode->i_size);
2302 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2303 inode->i_ino, inode->i_size);
2304 mutex_lock(&inode->i_mutex);
2305 truncate_inode_pages(inode->i_mapping, inode->i_size);
2306 ext4_truncate(inode);
2307 mutex_unlock(&inode->i_mutex);
2310 if (test_opt(sb, DEBUG))
2311 ext4_msg(sb, KERN_DEBUG,
2312 "%s: deleting unreferenced inode %lu",
2313 __func__, inode->i_ino);
2314 jbd_debug(2, "deleting unreferenced inode %lu\n",
2318 iput(inode); /* The delete magic happens here! */
2321 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2324 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2325 PLURAL(nr_orphans));
2327 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2328 PLURAL(nr_truncates));
2330 /* Turn off quotas if they were enabled for orphan cleanup */
2332 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2333 if (sb_dqopt(sb)->files[i])
2334 dquot_quota_off(sb, i);
2338 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2342 * Maximal extent format file size.
2343 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2344 * extent format containers, within a sector_t, and within i_blocks
2345 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2346 * so that won't be a limiting factor.
2348 * However there is other limiting factor. We do store extents in the form
2349 * of starting block and length, hence the resulting length of the extent
2350 * covering maximum file size must fit into on-disk format containers as
2351 * well. Given that length is always by 1 unit bigger than max unit (because
2352 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2354 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2356 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2359 loff_t upper_limit = MAX_LFS_FILESIZE;
2361 /* small i_blocks in vfs inode? */
2362 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2364 * CONFIG_LBDAF is not enabled implies the inode
2365 * i_block represent total blocks in 512 bytes
2366 * 32 == size of vfs inode i_blocks * 8
2368 upper_limit = (1LL << 32) - 1;
2370 /* total blocks in file system block size */
2371 upper_limit >>= (blkbits - 9);
2372 upper_limit <<= blkbits;
2376 * 32-bit extent-start container, ee_block. We lower the maxbytes
2377 * by one fs block, so ee_len can cover the extent of maximum file
2380 res = (1LL << 32) - 1;
2383 /* Sanity check against vm- & vfs- imposed limits */
2384 if (res > upper_limit)
2391 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2392 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2393 * We need to be 1 filesystem block less than the 2^48 sector limit.
2395 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2397 loff_t res = EXT4_NDIR_BLOCKS;
2400 /* This is calculated to be the largest file size for a dense, block
2401 * mapped file such that the file's total number of 512-byte sectors,
2402 * including data and all indirect blocks, does not exceed (2^48 - 1).
2404 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2405 * number of 512-byte sectors of the file.
2408 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2410 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2411 * the inode i_block field represents total file blocks in
2412 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2414 upper_limit = (1LL << 32) - 1;
2416 /* total blocks in file system block size */
2417 upper_limit >>= (bits - 9);
2421 * We use 48 bit ext4_inode i_blocks
2422 * With EXT4_HUGE_FILE_FL set the i_blocks
2423 * represent total number of blocks in
2424 * file system block size
2426 upper_limit = (1LL << 48) - 1;
2430 /* indirect blocks */
2432 /* double indirect blocks */
2433 meta_blocks += 1 + (1LL << (bits-2));
2434 /* tripple indirect blocks */
2435 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2437 upper_limit -= meta_blocks;
2438 upper_limit <<= bits;
2440 res += 1LL << (bits-2);
2441 res += 1LL << (2*(bits-2));
2442 res += 1LL << (3*(bits-2));
2444 if (res > upper_limit)
2447 if (res > MAX_LFS_FILESIZE)
2448 res = MAX_LFS_FILESIZE;
2453 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2454 ext4_fsblk_t logical_sb_block, int nr)
2456 struct ext4_sb_info *sbi = EXT4_SB(sb);
2457 ext4_group_t bg, first_meta_bg;
2460 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2462 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2463 return logical_sb_block + nr + 1;
2464 bg = sbi->s_desc_per_block * nr;
2465 if (ext4_bg_has_super(sb, bg))
2469 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2470 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2471 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2474 if (sb->s_blocksize == 1024 && nr == 0 &&
2475 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2478 return (has_super + ext4_group_first_block_no(sb, bg));
2482 * ext4_get_stripe_size: Get the stripe size.
2483 * @sbi: In memory super block info
2485 * If we have specified it via mount option, then
2486 * use the mount option value. If the value specified at mount time is
2487 * greater than the blocks per group use the super block value.
2488 * If the super block value is greater than blocks per group return 0.
2489 * Allocator needs it be less than blocks per group.
2492 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2494 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2495 unsigned long stripe_width =
2496 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2499 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2500 ret = sbi->s_stripe;
2501 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2503 else if (stride && stride <= sbi->s_blocks_per_group)
2509 * If the stripe width is 1, this makes no sense and
2510 * we set it to 0 to turn off stripe handling code.
2519 * Check whether this filesystem can be mounted based on
2520 * the features present and the RDONLY/RDWR mount requested.
2521 * Returns 1 if this filesystem can be mounted as requested,
2522 * 0 if it cannot be.
2524 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2526 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2527 ext4_msg(sb, KERN_ERR,
2528 "Couldn't mount because of "
2529 "unsupported optional features (%x)",
2530 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2531 ~EXT4_FEATURE_INCOMPAT_SUPP));
2538 if (ext4_has_feature_readonly(sb)) {
2539 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2540 sb->s_flags |= MS_RDONLY;
2544 /* Check that feature set is OK for a read-write mount */
2545 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2546 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2547 "unsupported optional features (%x)",
2548 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2549 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2553 * Large file size enabled file system can only be mounted
2554 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2556 if (ext4_has_feature_huge_file(sb)) {
2557 if (sizeof(blkcnt_t) < sizeof(u64)) {
2558 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2559 "cannot be mounted RDWR without "
2564 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2565 ext4_msg(sb, KERN_ERR,
2566 "Can't support bigalloc feature without "
2567 "extents feature\n");
2571 #ifndef CONFIG_QUOTA
2572 if (ext4_has_feature_quota(sb) && !readonly) {
2573 ext4_msg(sb, KERN_ERR,
2574 "Filesystem with quota feature cannot be mounted RDWR "
2575 "without CONFIG_QUOTA");
2578 #endif /* CONFIG_QUOTA */
2583 * This function is called once a day if we have errors logged
2584 * on the file system
2586 static void print_daily_error_info(unsigned long arg)
2588 struct super_block *sb = (struct super_block *) arg;
2589 struct ext4_sb_info *sbi;
2590 struct ext4_super_block *es;
2595 if (es->s_error_count)
2596 /* fsck newer than v1.41.13 is needed to clean this condition. */
2597 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2598 le32_to_cpu(es->s_error_count));
2599 if (es->s_first_error_time) {
2600 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2601 sb->s_id, le32_to_cpu(es->s_first_error_time),
2602 (int) sizeof(es->s_first_error_func),
2603 es->s_first_error_func,
2604 le32_to_cpu(es->s_first_error_line));
2605 if (es->s_first_error_ino)
2606 printk(": inode %u",
2607 le32_to_cpu(es->s_first_error_ino));
2608 if (es->s_first_error_block)
2609 printk(": block %llu", (unsigned long long)
2610 le64_to_cpu(es->s_first_error_block));
2613 if (es->s_last_error_time) {
2614 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2615 sb->s_id, le32_to_cpu(es->s_last_error_time),
2616 (int) sizeof(es->s_last_error_func),
2617 es->s_last_error_func,
2618 le32_to_cpu(es->s_last_error_line));
2619 if (es->s_last_error_ino)
2620 printk(": inode %u",
2621 le32_to_cpu(es->s_last_error_ino));
2622 if (es->s_last_error_block)
2623 printk(": block %llu", (unsigned long long)
2624 le64_to_cpu(es->s_last_error_block));
2627 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2630 /* Find next suitable group and run ext4_init_inode_table */
2631 static int ext4_run_li_request(struct ext4_li_request *elr)
2633 struct ext4_group_desc *gdp = NULL;
2634 ext4_group_t group, ngroups;
2635 struct super_block *sb;
2636 unsigned long timeout = 0;
2640 ngroups = EXT4_SB(sb)->s_groups_count;
2643 for (group = elr->lr_next_group; group < ngroups; group++) {
2644 gdp = ext4_get_group_desc(sb, group, NULL);
2650 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2654 if (group >= ngroups)
2659 ret = ext4_init_inode_table(sb, group,
2660 elr->lr_timeout ? 0 : 1);
2661 if (elr->lr_timeout == 0) {
2662 timeout = (jiffies - timeout) *
2663 elr->lr_sbi->s_li_wait_mult;
2664 elr->lr_timeout = timeout;
2666 elr->lr_next_sched = jiffies + elr->lr_timeout;
2667 elr->lr_next_group = group + 1;
2675 * Remove lr_request from the list_request and free the
2676 * request structure. Should be called with li_list_mtx held
2678 static void ext4_remove_li_request(struct ext4_li_request *elr)
2680 struct ext4_sb_info *sbi;
2687 list_del(&elr->lr_request);
2688 sbi->s_li_request = NULL;
2692 static void ext4_unregister_li_request(struct super_block *sb)
2694 mutex_lock(&ext4_li_mtx);
2695 if (!ext4_li_info) {
2696 mutex_unlock(&ext4_li_mtx);
2700 mutex_lock(&ext4_li_info->li_list_mtx);
2701 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2702 mutex_unlock(&ext4_li_info->li_list_mtx);
2703 mutex_unlock(&ext4_li_mtx);
2706 static struct task_struct *ext4_lazyinit_task;
2709 * This is the function where ext4lazyinit thread lives. It walks
2710 * through the request list searching for next scheduled filesystem.
2711 * When such a fs is found, run the lazy initialization request
2712 * (ext4_rn_li_request) and keep track of the time spend in this
2713 * function. Based on that time we compute next schedule time of
2714 * the request. When walking through the list is complete, compute
2715 * next waking time and put itself into sleep.
2717 static int ext4_lazyinit_thread(void *arg)
2719 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2720 struct list_head *pos, *n;
2721 struct ext4_li_request *elr;
2722 unsigned long next_wakeup, cur;
2724 BUG_ON(NULL == eli);
2728 next_wakeup = MAX_JIFFY_OFFSET;
2730 mutex_lock(&eli->li_list_mtx);
2731 if (list_empty(&eli->li_request_list)) {
2732 mutex_unlock(&eli->li_list_mtx);
2736 list_for_each_safe(pos, n, &eli->li_request_list) {
2737 elr = list_entry(pos, struct ext4_li_request,
2740 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2741 if (ext4_run_li_request(elr) != 0) {
2742 /* error, remove the lazy_init job */
2743 ext4_remove_li_request(elr);
2748 if (time_before(elr->lr_next_sched, next_wakeup))
2749 next_wakeup = elr->lr_next_sched;
2751 mutex_unlock(&eli->li_list_mtx);
2756 if ((time_after_eq(cur, next_wakeup)) ||
2757 (MAX_JIFFY_OFFSET == next_wakeup)) {
2762 schedule_timeout_interruptible(next_wakeup - cur);
2764 if (kthread_should_stop()) {
2765 ext4_clear_request_list();
2772 * It looks like the request list is empty, but we need
2773 * to check it under the li_list_mtx lock, to prevent any
2774 * additions into it, and of course we should lock ext4_li_mtx
2775 * to atomically free the list and ext4_li_info, because at
2776 * this point another ext4 filesystem could be registering
2779 mutex_lock(&ext4_li_mtx);
2780 mutex_lock(&eli->li_list_mtx);
2781 if (!list_empty(&eli->li_request_list)) {
2782 mutex_unlock(&eli->li_list_mtx);
2783 mutex_unlock(&ext4_li_mtx);
2786 mutex_unlock(&eli->li_list_mtx);
2787 kfree(ext4_li_info);
2788 ext4_li_info = NULL;
2789 mutex_unlock(&ext4_li_mtx);
2794 static void ext4_clear_request_list(void)
2796 struct list_head *pos, *n;
2797 struct ext4_li_request *elr;
2799 mutex_lock(&ext4_li_info->li_list_mtx);
2800 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2801 elr = list_entry(pos, struct ext4_li_request,
2803 ext4_remove_li_request(elr);
2805 mutex_unlock(&ext4_li_info->li_list_mtx);
2808 static int ext4_run_lazyinit_thread(void)
2810 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2811 ext4_li_info, "ext4lazyinit");
2812 if (IS_ERR(ext4_lazyinit_task)) {
2813 int err = PTR_ERR(ext4_lazyinit_task);
2814 ext4_clear_request_list();
2815 kfree(ext4_li_info);
2816 ext4_li_info = NULL;
2817 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2818 "initialization thread\n",
2822 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2827 * Check whether it make sense to run itable init. thread or not.
2828 * If there is at least one uninitialized inode table, return
2829 * corresponding group number, else the loop goes through all
2830 * groups and return total number of groups.
2832 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2834 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2835 struct ext4_group_desc *gdp = NULL;
2837 for (group = 0; group < ngroups; group++) {
2838 gdp = ext4_get_group_desc(sb, group, NULL);
2842 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2849 static int ext4_li_info_new(void)
2851 struct ext4_lazy_init *eli = NULL;
2853 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2857 INIT_LIST_HEAD(&eli->li_request_list);
2858 mutex_init(&eli->li_list_mtx);
2860 eli->li_state |= EXT4_LAZYINIT_QUIT;
2867 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2870 struct ext4_sb_info *sbi = EXT4_SB(sb);
2871 struct ext4_li_request *elr;
2873 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2879 elr->lr_next_group = start;
2882 * Randomize first schedule time of the request to
2883 * spread the inode table initialization requests
2886 elr->lr_next_sched = jiffies + (prandom_u32() %
2887 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2891 int ext4_register_li_request(struct super_block *sb,
2892 ext4_group_t first_not_zeroed)
2894 struct ext4_sb_info *sbi = EXT4_SB(sb);
2895 struct ext4_li_request *elr = NULL;
2896 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2899 mutex_lock(&ext4_li_mtx);
2900 if (sbi->s_li_request != NULL) {
2902 * Reset timeout so it can be computed again, because
2903 * s_li_wait_mult might have changed.
2905 sbi->s_li_request->lr_timeout = 0;
2909 if (first_not_zeroed == ngroups ||
2910 (sb->s_flags & MS_RDONLY) ||
2911 !test_opt(sb, INIT_INODE_TABLE))
2914 elr = ext4_li_request_new(sb, first_not_zeroed);
2920 if (NULL == ext4_li_info) {
2921 ret = ext4_li_info_new();
2926 mutex_lock(&ext4_li_info->li_list_mtx);
2927 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2928 mutex_unlock(&ext4_li_info->li_list_mtx);
2930 sbi->s_li_request = elr;
2932 * set elr to NULL here since it has been inserted to
2933 * the request_list and the removal and free of it is
2934 * handled by ext4_clear_request_list from now on.
2938 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2939 ret = ext4_run_lazyinit_thread();
2944 mutex_unlock(&ext4_li_mtx);
2951 * We do not need to lock anything since this is called on
2954 static void ext4_destroy_lazyinit_thread(void)
2957 * If thread exited earlier
2958 * there's nothing to be done.
2960 if (!ext4_li_info || !ext4_lazyinit_task)
2963 kthread_stop(ext4_lazyinit_task);
2966 static int set_journal_csum_feature_set(struct super_block *sb)
2969 int compat, incompat;
2970 struct ext4_sb_info *sbi = EXT4_SB(sb);
2972 if (ext4_has_metadata_csum(sb)) {
2973 /* journal checksum v3 */
2975 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2977 /* journal checksum v1 */
2978 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2982 jbd2_journal_clear_features(sbi->s_journal,
2983 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2984 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2985 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2986 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2987 ret = jbd2_journal_set_features(sbi->s_journal,
2989 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2991 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2992 ret = jbd2_journal_set_features(sbi->s_journal,
2995 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2996 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2998 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2999 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3006 * Note: calculating the overhead so we can be compatible with
3007 * historical BSD practice is quite difficult in the face of
3008 * clusters/bigalloc. This is because multiple metadata blocks from
3009 * different block group can end up in the same allocation cluster.
3010 * Calculating the exact overhead in the face of clustered allocation
3011 * requires either O(all block bitmaps) in memory or O(number of block
3012 * groups**2) in time. We will still calculate the superblock for
3013 * older file systems --- and if we come across with a bigalloc file
3014 * system with zero in s_overhead_clusters the estimate will be close to
3015 * correct especially for very large cluster sizes --- but for newer
3016 * file systems, it's better to calculate this figure once at mkfs
3017 * time, and store it in the superblock. If the superblock value is
3018 * present (even for non-bigalloc file systems), we will use it.
3020 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3023 struct ext4_sb_info *sbi = EXT4_SB(sb);
3024 struct ext4_group_desc *gdp;
3025 ext4_fsblk_t first_block, last_block, b;
3026 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3027 int s, j, count = 0;
3029 if (!ext4_has_feature_bigalloc(sb))
3030 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3031 sbi->s_itb_per_group + 2);
3033 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3034 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3035 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3036 for (i = 0; i < ngroups; i++) {
3037 gdp = ext4_get_group_desc(sb, i, NULL);
3038 b = ext4_block_bitmap(sb, gdp);
3039 if (b >= first_block && b <= last_block) {
3040 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3043 b = ext4_inode_bitmap(sb, gdp);
3044 if (b >= first_block && b <= last_block) {
3045 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3048 b = ext4_inode_table(sb, gdp);
3049 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3050 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3051 int c = EXT4_B2C(sbi, b - first_block);
3052 ext4_set_bit(c, buf);
3058 if (ext4_bg_has_super(sb, grp)) {
3059 ext4_set_bit(s++, buf);
3062 j = ext4_bg_num_gdb(sb, grp);
3063 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3064 ext4_error(sb, "Invalid number of block group "
3065 "descriptor blocks: %d", j);
3066 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3070 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3074 return EXT4_CLUSTERS_PER_GROUP(sb) -
3075 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3079 * Compute the overhead and stash it in sbi->s_overhead
3081 int ext4_calculate_overhead(struct super_block *sb)
3083 struct ext4_sb_info *sbi = EXT4_SB(sb);
3084 struct ext4_super_block *es = sbi->s_es;
3085 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3086 ext4_fsblk_t overhead = 0;
3087 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3093 * Compute the overhead (FS structures). This is constant
3094 * for a given filesystem unless the number of block groups
3095 * changes so we cache the previous value until it does.
3099 * All of the blocks before first_data_block are overhead
3101 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3104 * Add the overhead found in each block group
3106 for (i = 0; i < ngroups; i++) {
3109 blks = count_overhead(sb, i, buf);
3112 memset(buf, 0, PAGE_SIZE);
3115 /* Add the internal journal blocks as well */
3116 if (sbi->s_journal && !sbi->journal_bdev)
3117 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3119 sbi->s_overhead = overhead;
3121 free_page((unsigned long) buf);
3125 static void ext4_set_resv_clusters(struct super_block *sb)
3127 ext4_fsblk_t resv_clusters;
3128 struct ext4_sb_info *sbi = EXT4_SB(sb);
3131 * There's no need to reserve anything when we aren't using extents.
3132 * The space estimates are exact, there are no unwritten extents,
3133 * hole punching doesn't need new metadata... This is needed especially
3134 * to keep ext2/3 backward compatibility.
3136 if (!ext4_has_feature_extents(sb))
3139 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3140 * This should cover the situations where we can not afford to run
3141 * out of space like for example punch hole, or converting
3142 * unwritten extents in delalloc path. In most cases such
3143 * allocation would require 1, or 2 blocks, higher numbers are
3146 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3147 sbi->s_cluster_bits);
3149 do_div(resv_clusters, 50);
3150 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3152 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3155 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3157 char *orig_data = kstrdup(data, GFP_KERNEL);
3158 struct buffer_head *bh;
3159 struct ext4_super_block *es = NULL;
3160 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3162 ext4_fsblk_t sb_block = get_sb_block(&data);
3163 ext4_fsblk_t logical_sb_block;
3164 unsigned long offset = 0;
3165 unsigned long journal_devnum = 0;
3166 unsigned long def_mount_opts;
3170 int blocksize, clustersize;
3171 unsigned int db_count;
3173 int needs_recovery, has_huge_files, has_bigalloc;
3176 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3177 ext4_group_t first_not_zeroed;
3179 if ((data && !orig_data) || !sbi)
3182 sbi->s_blockgroup_lock =
3183 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3184 if (!sbi->s_blockgroup_lock)
3187 sb->s_fs_info = sbi;
3189 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3190 sbi->s_sb_block = sb_block;
3191 if (sb->s_bdev->bd_part)
3192 sbi->s_sectors_written_start =
3193 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3195 /* Cleanup superblock name */
3196 strreplace(sb->s_id, '/', '!');
3198 /* -EINVAL is default */
3200 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3202 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3207 * The ext4 superblock will not be buffer aligned for other than 1kB
3208 * block sizes. We need to calculate the offset from buffer start.
3210 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3211 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3212 offset = do_div(logical_sb_block, blocksize);
3214 logical_sb_block = sb_block;
3217 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3218 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3222 * Note: s_es must be initialized as soon as possible because
3223 * some ext4 macro-instructions depend on its value
3225 es = (struct ext4_super_block *) (bh->b_data + offset);
3227 sb->s_magic = le16_to_cpu(es->s_magic);
3228 if (sb->s_magic != EXT4_SUPER_MAGIC)
3230 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3232 /* Warn if metadata_csum and gdt_csum are both set. */
3233 if (ext4_has_feature_metadata_csum(sb) &&
3234 ext4_has_feature_gdt_csum(sb))
3235 ext4_warning(sb, "metadata_csum and uninit_bg are "
3236 "redundant flags; please run fsck.");
3238 /* Check for a known checksum algorithm */
3239 if (!ext4_verify_csum_type(sb, es)) {
3240 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3241 "unknown checksum algorithm.");
3246 /* Load the checksum driver */
3247 if (ext4_has_feature_metadata_csum(sb)) {
3248 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3249 if (IS_ERR(sbi->s_chksum_driver)) {
3250 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3251 ret = PTR_ERR(sbi->s_chksum_driver);
3252 sbi->s_chksum_driver = NULL;
3257 /* Check superblock checksum */
3258 if (!ext4_superblock_csum_verify(sb, es)) {
3259 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3260 "invalid superblock checksum. Run e2fsck?");
3266 /* Precompute checksum seed for all metadata */
3267 if (ext4_has_feature_csum_seed(sb))
3268 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3269 else if (ext4_has_metadata_csum(sb))
3270 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3271 sizeof(es->s_uuid));
3273 /* Set defaults before we parse the mount options */
3274 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3275 set_opt(sb, INIT_INODE_TABLE);
3276 if (def_mount_opts & EXT4_DEFM_DEBUG)
3278 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3280 if (def_mount_opts & EXT4_DEFM_UID16)
3281 set_opt(sb, NO_UID32);
3282 /* xattr user namespace & acls are now defaulted on */
3283 set_opt(sb, XATTR_USER);
3284 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3285 set_opt(sb, POSIX_ACL);
3287 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3288 if (ext4_has_metadata_csum(sb))
3289 set_opt(sb, JOURNAL_CHECKSUM);
3291 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3292 set_opt(sb, JOURNAL_DATA);
3293 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3294 set_opt(sb, ORDERED_DATA);
3295 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3296 set_opt(sb, WRITEBACK_DATA);
3298 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3299 set_opt(sb, ERRORS_PANIC);
3300 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3301 set_opt(sb, ERRORS_CONT);
3303 set_opt(sb, ERRORS_RO);
3304 /* block_validity enabled by default; disable with noblock_validity */
3305 set_opt(sb, BLOCK_VALIDITY);
3306 if (def_mount_opts & EXT4_DEFM_DISCARD)
3307 set_opt(sb, DISCARD);
3309 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3310 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3311 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3312 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3313 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3315 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3316 set_opt(sb, BARRIER);
3319 * enable delayed allocation by default
3320 * Use -o nodelalloc to turn it off
3322 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3323 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3324 set_opt(sb, DELALLOC);
3327 * set default s_li_wait_mult for lazyinit, for the case there is
3328 * no mount option specified.
3330 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3332 if (sbi->s_es->s_mount_opts[0]) {
3333 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3334 sizeof(sbi->s_es->s_mount_opts),
3338 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3339 &journal_ioprio, 0)) {
3340 ext4_msg(sb, KERN_WARNING,
3341 "failed to parse options in superblock: %s",
3344 kfree(s_mount_opts);
3346 sbi->s_def_mount_opt = sbi->s_mount_opt;
3347 if (!parse_options((char *) data, sb, &journal_devnum,
3348 &journal_ioprio, 0))
3351 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3352 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3353 "with data=journal disables delayed "
3354 "allocation and O_DIRECT support!\n");
3355 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3356 ext4_msg(sb, KERN_ERR, "can't mount with "
3357 "both data=journal and delalloc");
3360 if (test_opt(sb, DIOREAD_NOLOCK)) {
3361 ext4_msg(sb, KERN_ERR, "can't mount with "
3362 "both data=journal and dioread_nolock");
3365 if (test_opt(sb, DAX)) {
3366 ext4_msg(sb, KERN_ERR, "can't mount with "
3367 "both data=journal and dax");
3370 if (ext4_has_feature_encrypt(sb)) {
3371 ext4_msg(sb, KERN_WARNING,
3372 "encrypted files will use data=ordered "
3373 "instead of data journaling mode");
3375 if (test_opt(sb, DELALLOC))
3376 clear_opt(sb, DELALLOC);
3378 sb->s_iflags |= SB_I_CGROUPWB;
3381 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3382 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3384 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3385 (ext4_has_compat_features(sb) ||
3386 ext4_has_ro_compat_features(sb) ||
3387 ext4_has_incompat_features(sb)))
3388 ext4_msg(sb, KERN_WARNING,
3389 "feature flags set on rev 0 fs, "
3390 "running e2fsck is recommended");
3392 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3393 set_opt2(sb, HURD_COMPAT);
3394 if (ext4_has_feature_64bit(sb)) {
3395 ext4_msg(sb, KERN_ERR,
3396 "The Hurd can't support 64-bit file systems");
3401 if (IS_EXT2_SB(sb)) {
3402 if (ext2_feature_set_ok(sb))
3403 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3404 "using the ext4 subsystem");
3406 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3407 "to feature incompatibilities");
3412 if (IS_EXT3_SB(sb)) {
3413 if (ext3_feature_set_ok(sb))
3414 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3415 "using the ext4 subsystem");
3417 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3418 "to feature incompatibilities");
3424 * Check feature flags regardless of the revision level, since we
3425 * previously didn't change the revision level when setting the flags,
3426 * so there is a chance incompat flags are set on a rev 0 filesystem.
3428 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3431 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3432 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3433 blocksize > EXT4_MAX_BLOCK_SIZE) {
3434 ext4_msg(sb, KERN_ERR,
3435 "Unsupported filesystem blocksize %d (%d log_block_size)",
3436 blocksize, le32_to_cpu(es->s_log_block_size));
3439 if (le32_to_cpu(es->s_log_block_size) >
3440 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3441 ext4_msg(sb, KERN_ERR,
3442 "Invalid log block size: %u",
3443 le32_to_cpu(es->s_log_block_size));
3447 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3448 ext4_msg(sb, KERN_ERR,
3449 "Number of reserved GDT blocks insanely large: %d",
3450 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3454 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3455 if (blocksize != PAGE_SIZE) {
3456 ext4_msg(sb, KERN_ERR,
3457 "error: unsupported blocksize for dax");
3460 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3461 ext4_msg(sb, KERN_ERR,
3462 "error: device does not support dax");
3467 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3468 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3469 es->s_encryption_level);
3473 if (sb->s_blocksize != blocksize) {
3474 /* Validate the filesystem blocksize */
3475 if (!sb_set_blocksize(sb, blocksize)) {
3476 ext4_msg(sb, KERN_ERR, "bad block size %d",
3482 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3483 offset = do_div(logical_sb_block, blocksize);
3484 bh = sb_bread_unmovable(sb, logical_sb_block);
3486 ext4_msg(sb, KERN_ERR,
3487 "Can't read superblock on 2nd try");
3490 es = (struct ext4_super_block *)(bh->b_data + offset);
3492 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3493 ext4_msg(sb, KERN_ERR,
3494 "Magic mismatch, very weird!");
3499 has_huge_files = ext4_has_feature_huge_file(sb);
3500 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3502 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3504 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3505 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3506 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3508 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3509 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3510 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3511 (!is_power_of_2(sbi->s_inode_size)) ||
3512 (sbi->s_inode_size > blocksize)) {
3513 ext4_msg(sb, KERN_ERR,
3514 "unsupported inode size: %d",
3518 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3519 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3522 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3523 if (ext4_has_feature_64bit(sb)) {
3524 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3525 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3526 !is_power_of_2(sbi->s_desc_size)) {
3527 ext4_msg(sb, KERN_ERR,
3528 "unsupported descriptor size %lu",
3533 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3535 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3536 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3538 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3539 if (sbi->s_inodes_per_block == 0)
3541 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3542 sbi->s_inodes_per_group > blocksize * 8) {
3543 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3544 sbi->s_blocks_per_group);
3547 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3548 sbi->s_inodes_per_block;
3549 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3551 sbi->s_mount_state = le16_to_cpu(es->s_state);
3552 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3553 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3555 for (i = 0; i < 4; i++)
3556 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3557 sbi->s_def_hash_version = es->s_def_hash_version;
3558 if (ext4_has_feature_dir_index(sb)) {
3559 i = le32_to_cpu(es->s_flags);
3560 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3561 sbi->s_hash_unsigned = 3;
3562 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3563 #ifdef __CHAR_UNSIGNED__
3564 if (!(sb->s_flags & MS_RDONLY))
3566 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3567 sbi->s_hash_unsigned = 3;
3569 if (!(sb->s_flags & MS_RDONLY))
3571 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3576 /* Handle clustersize */
3577 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3578 has_bigalloc = ext4_has_feature_bigalloc(sb);
3580 if (clustersize < blocksize) {
3581 ext4_msg(sb, KERN_ERR,
3582 "cluster size (%d) smaller than "
3583 "block size (%d)", clustersize, blocksize);
3586 if (le32_to_cpu(es->s_log_cluster_size) >
3587 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3588 ext4_msg(sb, KERN_ERR,
3589 "Invalid log cluster size: %u",
3590 le32_to_cpu(es->s_log_cluster_size));
3593 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3594 le32_to_cpu(es->s_log_block_size);
3595 sbi->s_clusters_per_group =
3596 le32_to_cpu(es->s_clusters_per_group);
3597 if (sbi->s_clusters_per_group > blocksize * 8) {
3598 ext4_msg(sb, KERN_ERR,
3599 "#clusters per group too big: %lu",
3600 sbi->s_clusters_per_group);
3603 if (sbi->s_blocks_per_group !=
3604 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3605 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3606 "clusters per group (%lu) inconsistent",
3607 sbi->s_blocks_per_group,
3608 sbi->s_clusters_per_group);
3612 if (clustersize != blocksize) {
3613 ext4_warning(sb, "fragment/cluster size (%d) != "
3614 "block size (%d)", clustersize,
3616 clustersize = blocksize;
3618 if (sbi->s_blocks_per_group > blocksize * 8) {
3619 ext4_msg(sb, KERN_ERR,
3620 "#blocks per group too big: %lu",
3621 sbi->s_blocks_per_group);
3624 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3625 sbi->s_cluster_bits = 0;
3627 sbi->s_cluster_ratio = clustersize / blocksize;
3629 /* Do we have standard group size of clustersize * 8 blocks ? */
3630 if (sbi->s_blocks_per_group == clustersize << 3)
3631 set_opt2(sb, STD_GROUP_SIZE);
3634 * Test whether we have more sectors than will fit in sector_t,
3635 * and whether the max offset is addressable by the page cache.
3637 err = generic_check_addressable(sb->s_blocksize_bits,
3638 ext4_blocks_count(es));
3640 ext4_msg(sb, KERN_ERR, "filesystem"
3641 " too large to mount safely on this system");
3642 if (sizeof(sector_t) < 8)
3643 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3647 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3650 /* check blocks count against device size */
3651 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3652 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3653 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3654 "exceeds size of device (%llu blocks)",
3655 ext4_blocks_count(es), blocks_count);
3660 * It makes no sense for the first data block to be beyond the end
3661 * of the filesystem.
3663 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3664 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3665 "block %u is beyond end of filesystem (%llu)",
3666 le32_to_cpu(es->s_first_data_block),
3667 ext4_blocks_count(es));
3670 blocks_count = (ext4_blocks_count(es) -
3671 le32_to_cpu(es->s_first_data_block) +
3672 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3673 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3674 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3675 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3676 "(block count %llu, first data block %u, "
3677 "blocks per group %lu)", sbi->s_groups_count,
3678 ext4_blocks_count(es),
3679 le32_to_cpu(es->s_first_data_block),
3680 EXT4_BLOCKS_PER_GROUP(sb));
3683 sbi->s_groups_count = blocks_count;
3684 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3685 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3686 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3687 EXT4_DESC_PER_BLOCK(sb);
3688 if (ext4_has_feature_meta_bg(sb)) {
3689 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3690 ext4_msg(sb, KERN_WARNING,
3691 "first meta block group too large: %u "
3692 "(group descriptor block count %u)",
3693 le32_to_cpu(es->s_first_meta_bg), db_count);
3697 sbi->s_group_desc = ext4_kvmalloc(db_count *
3698 sizeof(struct buffer_head *),
3700 if (sbi->s_group_desc == NULL) {
3701 ext4_msg(sb, KERN_ERR, "not enough memory");
3706 bgl_lock_init(sbi->s_blockgroup_lock);
3708 for (i = 0; i < db_count; i++) {
3709 block = descriptor_loc(sb, logical_sb_block, i);
3710 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3711 if (!sbi->s_group_desc[i]) {
3712 ext4_msg(sb, KERN_ERR,
3713 "can't read group descriptor %d", i);
3718 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3719 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3720 ret = -EFSCORRUPTED;
3724 sbi->s_gdb_count = db_count;
3725 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3726 spin_lock_init(&sbi->s_next_gen_lock);
3728 setup_timer(&sbi->s_err_report, print_daily_error_info,
3729 (unsigned long) sb);
3731 /* Register extent status tree shrinker */
3732 if (ext4_es_register_shrinker(sbi))
3735 sbi->s_stripe = ext4_get_stripe_size(sbi);
3736 sbi->s_extent_max_zeroout_kb = 32;
3739 * set up enough so that it can read an inode
3741 sb->s_op = &ext4_sops;
3742 sb->s_export_op = &ext4_export_ops;
3743 sb->s_xattr = ext4_xattr_handlers;
3745 sb->dq_op = &ext4_quota_operations;
3746 if (ext4_has_feature_quota(sb))
3747 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3749 sb->s_qcop = &ext4_qctl_operations;
3750 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3752 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3754 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3755 mutex_init(&sbi->s_orphan_lock);
3759 needs_recovery = (es->s_last_orphan != 0 ||
3760 ext4_has_feature_journal_needs_recovery(sb));
3762 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3763 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3764 goto failed_mount3a;
3767 * The first inode we look at is the journal inode. Don't try
3768 * root first: it may be modified in the journal!
3770 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3771 err = ext4_load_journal(sb, es, journal_devnum);
3773 goto failed_mount3a;
3774 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3775 ext4_has_feature_journal_needs_recovery(sb)) {
3776 ext4_msg(sb, KERN_ERR, "required journal recovery "
3777 "suppressed and not mounted read-only");
3778 goto failed_mount_wq;
3780 /* Nojournal mode, all journal mount options are illegal */
3781 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3782 ext4_msg(sb, KERN_ERR, "can't mount with "
3783 "journal_checksum, fs mounted w/o journal");
3784 goto failed_mount_wq;
3786 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3787 ext4_msg(sb, KERN_ERR, "can't mount with "
3788 "journal_async_commit, fs mounted w/o journal");
3789 goto failed_mount_wq;
3791 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3792 ext4_msg(sb, KERN_ERR, "can't mount with "
3793 "commit=%lu, fs mounted w/o journal",
3794 sbi->s_commit_interval / HZ);
3795 goto failed_mount_wq;
3797 if (EXT4_MOUNT_DATA_FLAGS &
3798 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3799 ext4_msg(sb, KERN_ERR, "can't mount with "
3800 "data=, fs mounted w/o journal");
3801 goto failed_mount_wq;
3803 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3804 clear_opt(sb, JOURNAL_CHECKSUM);
3805 clear_opt(sb, DATA_FLAGS);
3806 sbi->s_journal = NULL;
3811 if (ext4_has_feature_64bit(sb) &&
3812 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3813 JBD2_FEATURE_INCOMPAT_64BIT)) {
3814 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3815 goto failed_mount_wq;
3818 if (!set_journal_csum_feature_set(sb)) {
3819 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3821 goto failed_mount_wq;
3824 /* We have now updated the journal if required, so we can
3825 * validate the data journaling mode. */
3826 switch (test_opt(sb, DATA_FLAGS)) {
3828 /* No mode set, assume a default based on the journal
3829 * capabilities: ORDERED_DATA if the journal can
3830 * cope, else JOURNAL_DATA
3832 if (jbd2_journal_check_available_features
3833 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3834 set_opt(sb, ORDERED_DATA);
3836 set_opt(sb, JOURNAL_DATA);
3839 case EXT4_MOUNT_ORDERED_DATA:
3840 case EXT4_MOUNT_WRITEBACK_DATA:
3841 if (!jbd2_journal_check_available_features
3842 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3843 ext4_msg(sb, KERN_ERR, "Journal does not support "
3844 "requested data journaling mode");
3845 goto failed_mount_wq;
3850 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3852 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3855 if (ext4_mballoc_ready) {
3856 sbi->s_mb_cache = ext4_xattr_create_cache();
3857 if (!sbi->s_mb_cache) {
3858 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3859 goto failed_mount_wq;
3863 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3864 (blocksize != PAGE_CACHE_SIZE)) {
3865 ext4_msg(sb, KERN_ERR,
3866 "Unsupported blocksize for fs encryption");
3867 goto failed_mount_wq;
3870 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3871 !ext4_has_feature_encrypt(sb)) {
3872 ext4_set_feature_encrypt(sb);
3873 ext4_commit_super(sb, 1);
3877 * Get the # of file system overhead blocks from the
3878 * superblock if present.
3880 if (es->s_overhead_clusters)
3881 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3883 err = ext4_calculate_overhead(sb);
3885 goto failed_mount_wq;
3889 * The maximum number of concurrent works can be high and
3890 * concurrency isn't really necessary. Limit it to 1.
3892 EXT4_SB(sb)->rsv_conversion_wq =
3893 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3894 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3895 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3901 * The jbd2_journal_load will have done any necessary log recovery,
3902 * so we can safely mount the rest of the filesystem now.
3905 root = ext4_iget(sb, EXT4_ROOT_INO);
3907 ext4_msg(sb, KERN_ERR, "get root inode failed");
3908 ret = PTR_ERR(root);
3912 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3913 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3917 sb->s_root = d_make_root(root);
3919 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3924 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3925 sb->s_flags |= MS_RDONLY;
3927 /* determine the minimum size of new large inodes, if present */
3928 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3929 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3930 EXT4_GOOD_OLD_INODE_SIZE;
3931 if (ext4_has_feature_extra_isize(sb)) {
3932 if (sbi->s_want_extra_isize <
3933 le16_to_cpu(es->s_want_extra_isize))
3934 sbi->s_want_extra_isize =
3935 le16_to_cpu(es->s_want_extra_isize);
3936 if (sbi->s_want_extra_isize <
3937 le16_to_cpu(es->s_min_extra_isize))
3938 sbi->s_want_extra_isize =
3939 le16_to_cpu(es->s_min_extra_isize);
3942 /* Check if enough inode space is available */
3943 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3944 sbi->s_inode_size) {
3945 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3946 EXT4_GOOD_OLD_INODE_SIZE;
3947 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3951 ext4_set_resv_clusters(sb);
3953 err = ext4_setup_system_zone(sb);
3955 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3957 goto failed_mount4a;
3961 err = ext4_mb_init(sb);
3963 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3968 block = ext4_count_free_clusters(sb);
3969 ext4_free_blocks_count_set(sbi->s_es,
3970 EXT4_C2B(sbi, block));
3971 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3974 unsigned long freei = ext4_count_free_inodes(sb);
3975 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3976 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3980 err = percpu_counter_init(&sbi->s_dirs_counter,
3981 ext4_count_dirs(sb), GFP_KERNEL);
3983 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3986 ext4_msg(sb, KERN_ERR, "insufficient memory");
3990 if (ext4_has_feature_flex_bg(sb))
3991 if (!ext4_fill_flex_info(sb)) {
3992 ext4_msg(sb, KERN_ERR,
3993 "unable to initialize "
3994 "flex_bg meta info!");
3998 err = ext4_register_li_request(sb, first_not_zeroed);
4002 err = ext4_register_sysfs(sb);
4007 /* Enable quota usage during mount. */
4008 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4009 err = ext4_enable_quotas(sb);
4013 #endif /* CONFIG_QUOTA */
4015 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4016 ext4_orphan_cleanup(sb, es);
4017 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4018 if (needs_recovery) {
4019 ext4_msg(sb, KERN_INFO, "recovery complete");
4020 ext4_mark_recovery_complete(sb, es);
4022 if (EXT4_SB(sb)->s_journal) {
4023 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4024 descr = " journalled data mode";
4025 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4026 descr = " ordered data mode";
4028 descr = " writeback data mode";
4030 descr = "out journal";
4032 if (test_opt(sb, DISCARD)) {
4033 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4034 if (!blk_queue_discard(q))
4035 ext4_msg(sb, KERN_WARNING,
4036 "mounting with \"discard\" option, but "
4037 "the device does not support discard");
4040 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4041 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4042 "Opts: %.*s%s%s", descr,
4043 (int) sizeof(sbi->s_es->s_mount_opts),
4044 sbi->s_es->s_mount_opts,
4045 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4047 if (es->s_error_count)
4048 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4050 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4051 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4052 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4053 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4060 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4065 ext4_unregister_sysfs(sb);
4068 ext4_unregister_li_request(sb);
4070 ext4_mb_release(sb);
4071 if (sbi->s_flex_groups)
4072 kvfree(sbi->s_flex_groups);
4073 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4074 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4075 percpu_counter_destroy(&sbi->s_dirs_counter);
4076 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4078 ext4_ext_release(sb);
4079 ext4_release_system_zone(sb);
4084 ext4_msg(sb, KERN_ERR, "mount failed");
4085 if (EXT4_SB(sb)->rsv_conversion_wq)
4086 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4088 if (sbi->s_mb_cache) {
4089 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4090 sbi->s_mb_cache = NULL;
4092 if (sbi->s_journal) {
4093 jbd2_journal_destroy(sbi->s_journal);
4094 sbi->s_journal = NULL;
4097 ext4_es_unregister_shrinker(sbi);
4099 del_timer_sync(&sbi->s_err_report);
4101 kthread_stop(sbi->s_mmp_tsk);
4103 for (i = 0; i < db_count; i++)
4104 brelse(sbi->s_group_desc[i]);
4105 kvfree(sbi->s_group_desc);
4107 if (sbi->s_chksum_driver)
4108 crypto_free_shash(sbi->s_chksum_driver);
4110 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4111 kfree(sbi->s_qf_names[i]);
4113 ext4_blkdev_remove(sbi);
4116 sb->s_fs_info = NULL;
4117 kfree(sbi->s_blockgroup_lock);
4121 return err ? err : ret;
4125 * Setup any per-fs journal parameters now. We'll do this both on
4126 * initial mount, once the journal has been initialised but before we've
4127 * done any recovery; and again on any subsequent remount.
4129 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4131 struct ext4_sb_info *sbi = EXT4_SB(sb);
4133 journal->j_commit_interval = sbi->s_commit_interval;
4134 journal->j_min_batch_time = sbi->s_min_batch_time;
4135 journal->j_max_batch_time = sbi->s_max_batch_time;
4137 write_lock(&journal->j_state_lock);
4138 if (test_opt(sb, BARRIER))
4139 journal->j_flags |= JBD2_BARRIER;
4141 journal->j_flags &= ~JBD2_BARRIER;
4142 if (test_opt(sb, DATA_ERR_ABORT))
4143 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4145 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4146 write_unlock(&journal->j_state_lock);
4149 static journal_t *ext4_get_journal(struct super_block *sb,
4150 unsigned int journal_inum)
4152 struct inode *journal_inode;
4155 BUG_ON(!ext4_has_feature_journal(sb));
4157 /* First, test for the existence of a valid inode on disk. Bad
4158 * things happen if we iget() an unused inode, as the subsequent
4159 * iput() will try to delete it. */
4161 journal_inode = ext4_iget(sb, journal_inum);
4162 if (IS_ERR(journal_inode)) {
4163 ext4_msg(sb, KERN_ERR, "no journal found");
4166 if (!journal_inode->i_nlink) {
4167 make_bad_inode(journal_inode);
4168 iput(journal_inode);
4169 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4173 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4174 journal_inode, journal_inode->i_size);
4175 if (!S_ISREG(journal_inode->i_mode)) {
4176 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4177 iput(journal_inode);
4181 journal = jbd2_journal_init_inode(journal_inode);
4183 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4184 iput(journal_inode);
4187 journal->j_private = sb;
4188 ext4_init_journal_params(sb, journal);
4192 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4195 struct buffer_head *bh;
4199 int hblock, blocksize;
4200 ext4_fsblk_t sb_block;
4201 unsigned long offset;
4202 struct ext4_super_block *es;
4203 struct block_device *bdev;
4205 BUG_ON(!ext4_has_feature_journal(sb));
4207 bdev = ext4_blkdev_get(j_dev, sb);
4211 blocksize = sb->s_blocksize;
4212 hblock = bdev_logical_block_size(bdev);
4213 if (blocksize < hblock) {
4214 ext4_msg(sb, KERN_ERR,
4215 "blocksize too small for journal device");
4219 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4220 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4221 set_blocksize(bdev, blocksize);
4222 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4223 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4224 "external journal");
4228 es = (struct ext4_super_block *) (bh->b_data + offset);
4229 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4230 !(le32_to_cpu(es->s_feature_incompat) &
4231 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4232 ext4_msg(sb, KERN_ERR, "external journal has "
4238 if ((le32_to_cpu(es->s_feature_ro_compat) &
4239 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4240 es->s_checksum != ext4_superblock_csum(sb, es)) {
4241 ext4_msg(sb, KERN_ERR, "external journal has "
4242 "corrupt superblock");
4247 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4248 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4253 len = ext4_blocks_count(es);
4254 start = sb_block + 1;
4255 brelse(bh); /* we're done with the superblock */
4257 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4258 start, len, blocksize);
4260 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4263 journal->j_private = sb;
4264 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4265 wait_on_buffer(journal->j_sb_buffer);
4266 if (!buffer_uptodate(journal->j_sb_buffer)) {
4267 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4270 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4271 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4272 "user (unsupported) - %d",
4273 be32_to_cpu(journal->j_superblock->s_nr_users));
4276 EXT4_SB(sb)->journal_bdev = bdev;
4277 ext4_init_journal_params(sb, journal);
4281 jbd2_journal_destroy(journal);
4283 ext4_blkdev_put(bdev);
4287 static int ext4_load_journal(struct super_block *sb,
4288 struct ext4_super_block *es,
4289 unsigned long journal_devnum)
4292 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4295 int really_read_only;
4297 BUG_ON(!ext4_has_feature_journal(sb));
4299 if (journal_devnum &&
4300 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4301 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4302 "numbers have changed");
4303 journal_dev = new_decode_dev(journal_devnum);
4305 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4307 really_read_only = bdev_read_only(sb->s_bdev);
4310 * Are we loading a blank journal or performing recovery after a
4311 * crash? For recovery, we need to check in advance whether we
4312 * can get read-write access to the device.
4314 if (ext4_has_feature_journal_needs_recovery(sb)) {
4315 if (sb->s_flags & MS_RDONLY) {
4316 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4317 "required on readonly filesystem");
4318 if (really_read_only) {
4319 ext4_msg(sb, KERN_ERR, "write access "
4320 "unavailable, cannot proceed");
4323 ext4_msg(sb, KERN_INFO, "write access will "
4324 "be enabled during recovery");
4328 if (journal_inum && journal_dev) {
4329 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4330 "and inode journals!");
4335 if (!(journal = ext4_get_journal(sb, journal_inum)))
4338 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4342 if (!(journal->j_flags & JBD2_BARRIER))
4343 ext4_msg(sb, KERN_INFO, "barriers disabled");
4345 if (!ext4_has_feature_journal_needs_recovery(sb))
4346 err = jbd2_journal_wipe(journal, !really_read_only);
4348 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4350 memcpy(save, ((char *) es) +
4351 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4352 err = jbd2_journal_load(journal);
4354 memcpy(((char *) es) + EXT4_S_ERR_START,
4355 save, EXT4_S_ERR_LEN);
4360 ext4_msg(sb, KERN_ERR, "error loading journal");
4361 jbd2_journal_destroy(journal);
4365 EXT4_SB(sb)->s_journal = journal;
4366 ext4_clear_journal_err(sb, es);
4368 if (!really_read_only && journal_devnum &&
4369 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4370 es->s_journal_dev = cpu_to_le32(journal_devnum);
4372 /* Make sure we flush the recovery flag to disk. */
4373 ext4_commit_super(sb, 1);
4379 static int ext4_commit_super(struct super_block *sb, int sync)
4381 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4382 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4385 if (!sbh || block_device_ejected(sb))
4387 if (buffer_write_io_error(sbh)) {
4389 * Oh, dear. A previous attempt to write the
4390 * superblock failed. This could happen because the
4391 * USB device was yanked out. Or it could happen to
4392 * be a transient write error and maybe the block will
4393 * be remapped. Nothing we can do but to retry the
4394 * write and hope for the best.
4396 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4397 "superblock detected");
4398 clear_buffer_write_io_error(sbh);
4399 set_buffer_uptodate(sbh);
4402 * If the file system is mounted read-only, don't update the
4403 * superblock write time. This avoids updating the superblock
4404 * write time when we are mounting the root file system
4405 * read/only but we need to replay the journal; at that point,
4406 * for people who are east of GMT and who make their clock
4407 * tick in localtime for Windows bug-for-bug compatibility,
4408 * the clock is set in the future, and this will cause e2fsck
4409 * to complain and force a full file system check.
4411 if (!(sb->s_flags & MS_RDONLY))
4412 es->s_wtime = cpu_to_le32(get_seconds());
4413 if (sb->s_bdev->bd_part)
4414 es->s_kbytes_written =
4415 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4416 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4417 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4419 es->s_kbytes_written =
4420 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4421 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4422 ext4_free_blocks_count_set(es,
4423 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4424 &EXT4_SB(sb)->s_freeclusters_counter)));
4425 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4426 es->s_free_inodes_count =
4427 cpu_to_le32(percpu_counter_sum_positive(
4428 &EXT4_SB(sb)->s_freeinodes_counter));
4429 BUFFER_TRACE(sbh, "marking dirty");
4430 ext4_superblock_csum_set(sb);
4431 mark_buffer_dirty(sbh);
4433 error = __sync_dirty_buffer(sbh,
4434 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4438 error = buffer_write_io_error(sbh);
4440 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4442 clear_buffer_write_io_error(sbh);
4443 set_buffer_uptodate(sbh);
4450 * Have we just finished recovery? If so, and if we are mounting (or
4451 * remounting) the filesystem readonly, then we will end up with a
4452 * consistent fs on disk. Record that fact.
4454 static void ext4_mark_recovery_complete(struct super_block *sb,
4455 struct ext4_super_block *es)
4457 journal_t *journal = EXT4_SB(sb)->s_journal;
4459 if (!ext4_has_feature_journal(sb)) {
4460 BUG_ON(journal != NULL);
4463 jbd2_journal_lock_updates(journal);
4464 if (jbd2_journal_flush(journal) < 0)
4467 if (ext4_has_feature_journal_needs_recovery(sb) &&
4468 sb->s_flags & MS_RDONLY) {
4469 ext4_clear_feature_journal_needs_recovery(sb);
4470 ext4_commit_super(sb, 1);
4474 jbd2_journal_unlock_updates(journal);
4478 * If we are mounting (or read-write remounting) a filesystem whose journal
4479 * has recorded an error from a previous lifetime, move that error to the
4480 * main filesystem now.
4482 static void ext4_clear_journal_err(struct super_block *sb,
4483 struct ext4_super_block *es)
4489 BUG_ON(!ext4_has_feature_journal(sb));
4491 journal = EXT4_SB(sb)->s_journal;
4494 * Now check for any error status which may have been recorded in the
4495 * journal by a prior ext4_error() or ext4_abort()
4498 j_errno = jbd2_journal_errno(journal);
4502 errstr = ext4_decode_error(sb, j_errno, nbuf);
4503 ext4_warning(sb, "Filesystem error recorded "
4504 "from previous mount: %s", errstr);
4505 ext4_warning(sb, "Marking fs in need of filesystem check.");
4507 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4508 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4509 ext4_commit_super(sb, 1);
4511 jbd2_journal_clear_err(journal);
4512 jbd2_journal_update_sb_errno(journal);
4517 * Force the running and committing transactions to commit,
4518 * and wait on the commit.
4520 int ext4_force_commit(struct super_block *sb)
4524 if (sb->s_flags & MS_RDONLY)
4527 journal = EXT4_SB(sb)->s_journal;
4528 return ext4_journal_force_commit(journal);
4531 static int ext4_sync_fs(struct super_block *sb, int wait)
4535 bool needs_barrier = false;
4536 struct ext4_sb_info *sbi = EXT4_SB(sb);
4538 trace_ext4_sync_fs(sb, wait);
4539 flush_workqueue(sbi->rsv_conversion_wq);
4541 * Writeback quota in non-journalled quota case - journalled quota has
4544 dquot_writeback_dquots(sb, -1);
4546 * Data writeback is possible w/o journal transaction, so barrier must
4547 * being sent at the end of the function. But we can skip it if
4548 * transaction_commit will do it for us.
4550 if (sbi->s_journal) {
4551 target = jbd2_get_latest_transaction(sbi->s_journal);
4552 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4553 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4554 needs_barrier = true;
4556 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4558 ret = jbd2_log_wait_commit(sbi->s_journal,
4561 } else if (wait && test_opt(sb, BARRIER))
4562 needs_barrier = true;
4563 if (needs_barrier) {
4565 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4574 * LVM calls this function before a (read-only) snapshot is created. This
4575 * gives us a chance to flush the journal completely and mark the fs clean.
4577 * Note that only this function cannot bring a filesystem to be in a clean
4578 * state independently. It relies on upper layer to stop all data & metadata
4581 static int ext4_freeze(struct super_block *sb)
4586 if (sb->s_flags & MS_RDONLY)
4589 journal = EXT4_SB(sb)->s_journal;
4592 /* Now we set up the journal barrier. */
4593 jbd2_journal_lock_updates(journal);
4596 * Don't clear the needs_recovery flag if we failed to
4597 * flush the journal.
4599 error = jbd2_journal_flush(journal);
4603 /* Journal blocked and flushed, clear needs_recovery flag. */
4604 ext4_clear_feature_journal_needs_recovery(sb);
4607 error = ext4_commit_super(sb, 1);
4610 /* we rely on upper layer to stop further updates */
4611 jbd2_journal_unlock_updates(journal);
4616 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4617 * flag here, even though the filesystem is not technically dirty yet.
4619 static int ext4_unfreeze(struct super_block *sb)
4621 if (sb->s_flags & MS_RDONLY)
4624 if (EXT4_SB(sb)->s_journal) {
4625 /* Reset the needs_recovery flag before the fs is unlocked. */
4626 ext4_set_feature_journal_needs_recovery(sb);
4629 ext4_commit_super(sb, 1);
4634 * Structure to save mount options for ext4_remount's benefit
4636 struct ext4_mount_options {
4637 unsigned long s_mount_opt;
4638 unsigned long s_mount_opt2;
4641 unsigned long s_commit_interval;
4642 u32 s_min_batch_time, s_max_batch_time;
4645 char *s_qf_names[EXT4_MAXQUOTAS];
4649 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4651 struct ext4_super_block *es;
4652 struct ext4_sb_info *sbi = EXT4_SB(sb);
4653 unsigned long old_sb_flags;
4654 struct ext4_mount_options old_opts;
4655 int enable_quota = 0;
4657 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4662 char *orig_data = kstrdup(data, GFP_KERNEL);
4664 /* Store the original options */
4665 old_sb_flags = sb->s_flags;
4666 old_opts.s_mount_opt = sbi->s_mount_opt;
4667 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4668 old_opts.s_resuid = sbi->s_resuid;
4669 old_opts.s_resgid = sbi->s_resgid;
4670 old_opts.s_commit_interval = sbi->s_commit_interval;
4671 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4672 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4674 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4675 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4676 if (sbi->s_qf_names[i]) {
4677 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4679 if (!old_opts.s_qf_names[i]) {
4680 for (j = 0; j < i; j++)
4681 kfree(old_opts.s_qf_names[j]);
4686 old_opts.s_qf_names[i] = NULL;
4688 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4689 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4691 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4696 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4697 test_opt(sb, JOURNAL_CHECKSUM)) {
4698 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4699 "during remount not supported; ignoring");
4700 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4703 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4704 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4705 ext4_msg(sb, KERN_ERR, "can't mount with "
4706 "both data=journal and delalloc");
4710 if (test_opt(sb, DIOREAD_NOLOCK)) {
4711 ext4_msg(sb, KERN_ERR, "can't mount with "
4712 "both data=journal and dioread_nolock");
4716 if (test_opt(sb, DAX)) {
4717 ext4_msg(sb, KERN_ERR, "can't mount with "
4718 "both data=journal and dax");
4724 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4725 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4726 "dax flag with busy inodes while remounting");
4727 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4730 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4731 ext4_abort(sb, "Abort forced by user");
4733 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4734 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4738 if (sbi->s_journal) {
4739 ext4_init_journal_params(sb, sbi->s_journal);
4740 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4743 if (*flags & MS_LAZYTIME)
4744 sb->s_flags |= MS_LAZYTIME;
4746 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4747 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4752 if (*flags & MS_RDONLY) {
4753 err = sync_filesystem(sb);
4756 err = dquot_suspend(sb, -1);
4761 * First of all, the unconditional stuff we have to do
4762 * to disable replay of the journal when we next remount
4764 sb->s_flags |= MS_RDONLY;
4767 * OK, test if we are remounting a valid rw partition
4768 * readonly, and if so set the rdonly flag and then
4769 * mark the partition as valid again.
4771 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4772 (sbi->s_mount_state & EXT4_VALID_FS))
4773 es->s_state = cpu_to_le16(sbi->s_mount_state);
4776 ext4_mark_recovery_complete(sb, es);
4778 /* Make sure we can mount this feature set readwrite */
4779 if (ext4_has_feature_readonly(sb) ||
4780 !ext4_feature_set_ok(sb, 0)) {
4785 * Make sure the group descriptor checksums
4786 * are sane. If they aren't, refuse to remount r/w.
4788 for (g = 0; g < sbi->s_groups_count; g++) {
4789 struct ext4_group_desc *gdp =
4790 ext4_get_group_desc(sb, g, NULL);
4792 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4793 ext4_msg(sb, KERN_ERR,
4794 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4795 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4796 le16_to_cpu(gdp->bg_checksum));
4803 * If we have an unprocessed orphan list hanging
4804 * around from a previously readonly bdev mount,
4805 * require a full umount/remount for now.
4807 if (es->s_last_orphan) {
4808 ext4_msg(sb, KERN_WARNING, "Couldn't "
4809 "remount RDWR because of unprocessed "
4810 "orphan inode list. Please "
4811 "umount/remount instead");
4817 * Mounting a RDONLY partition read-write, so reread
4818 * and store the current valid flag. (It may have
4819 * been changed by e2fsck since we originally mounted
4823 ext4_clear_journal_err(sb, es);
4824 sbi->s_mount_state = le16_to_cpu(es->s_state);
4825 if (!ext4_setup_super(sb, es, 0))
4826 sb->s_flags &= ~MS_RDONLY;
4827 if (ext4_has_feature_mmp(sb))
4828 if (ext4_multi_mount_protect(sb,
4829 le64_to_cpu(es->s_mmp_block))) {
4838 * Reinitialize lazy itable initialization thread based on
4841 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4842 ext4_unregister_li_request(sb);
4844 ext4_group_t first_not_zeroed;
4845 first_not_zeroed = ext4_has_uninit_itable(sb);
4846 ext4_register_li_request(sb, first_not_zeroed);
4849 ext4_setup_system_zone(sb);
4850 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4851 ext4_commit_super(sb, 1);
4854 /* Release old quota file names */
4855 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4856 kfree(old_opts.s_qf_names[i]);
4858 if (sb_any_quota_suspended(sb))
4859 dquot_resume(sb, -1);
4860 else if (ext4_has_feature_quota(sb)) {
4861 err = ext4_enable_quotas(sb);
4868 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4869 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4874 sb->s_flags = old_sb_flags;
4875 sbi->s_mount_opt = old_opts.s_mount_opt;
4876 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4877 sbi->s_resuid = old_opts.s_resuid;
4878 sbi->s_resgid = old_opts.s_resgid;
4879 sbi->s_commit_interval = old_opts.s_commit_interval;
4880 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4881 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4883 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4884 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4885 kfree(sbi->s_qf_names[i]);
4886 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4893 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4895 struct super_block *sb = dentry->d_sb;
4896 struct ext4_sb_info *sbi = EXT4_SB(sb);
4897 struct ext4_super_block *es = sbi->s_es;
4898 ext4_fsblk_t overhead = 0, resv_blocks;
4901 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4903 if (!test_opt(sb, MINIX_DF))
4904 overhead = sbi->s_overhead;
4906 buf->f_type = EXT4_SUPER_MAGIC;
4907 buf->f_bsize = sb->s_blocksize;
4908 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4909 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4910 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4911 /* prevent underflow in case that few free space is available */
4912 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4913 buf->f_bavail = buf->f_bfree -
4914 (ext4_r_blocks_count(es) + resv_blocks);
4915 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4917 buf->f_files = le32_to_cpu(es->s_inodes_count);
4918 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4919 buf->f_namelen = EXT4_NAME_LEN;
4920 fsid = le64_to_cpup((void *)es->s_uuid) ^
4921 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4922 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4923 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4928 /* Helper function for writing quotas on sync - we need to start transaction
4929 * before quota file is locked for write. Otherwise the are possible deadlocks:
4930 * Process 1 Process 2
4931 * ext4_create() quota_sync()
4932 * jbd2_journal_start() write_dquot()
4933 * dquot_initialize() down(dqio_mutex)
4934 * down(dqio_mutex) jbd2_journal_start()
4940 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4942 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4945 static int ext4_write_dquot(struct dquot *dquot)
4949 struct inode *inode;
4951 inode = dquot_to_inode(dquot);
4952 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4953 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4955 return PTR_ERR(handle);
4956 ret = dquot_commit(dquot);
4957 err = ext4_journal_stop(handle);
4963 static int ext4_acquire_dquot(struct dquot *dquot)
4968 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4969 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4971 return PTR_ERR(handle);
4972 ret = dquot_acquire(dquot);
4973 err = ext4_journal_stop(handle);
4979 static int ext4_release_dquot(struct dquot *dquot)
4984 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4985 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4986 if (IS_ERR(handle)) {
4987 /* Release dquot anyway to avoid endless cycle in dqput() */
4988 dquot_release(dquot);
4989 return PTR_ERR(handle);
4991 ret = dquot_release(dquot);
4992 err = ext4_journal_stop(handle);
4998 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5000 struct super_block *sb = dquot->dq_sb;
5001 struct ext4_sb_info *sbi = EXT4_SB(sb);
5003 /* Are we journaling quotas? */
5004 if (ext4_has_feature_quota(sb) ||
5005 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5006 dquot_mark_dquot_dirty(dquot);
5007 return ext4_write_dquot(dquot);
5009 return dquot_mark_dquot_dirty(dquot);
5013 static int ext4_write_info(struct super_block *sb, int type)
5018 /* Data block + inode block */
5019 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5021 return PTR_ERR(handle);
5022 ret = dquot_commit_info(sb, type);
5023 err = ext4_journal_stop(handle);
5030 * Turn on quotas during mount time - we need to find
5031 * the quota file and such...
5033 static int ext4_quota_on_mount(struct super_block *sb, int type)
5035 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5036 EXT4_SB(sb)->s_jquota_fmt, type);
5039 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5041 struct ext4_inode_info *ei = EXT4_I(inode);
5043 /* The first argument of lockdep_set_subclass has to be
5044 * *exactly* the same as the argument to init_rwsem() --- in
5045 * this case, in init_once() --- or lockdep gets unhappy
5046 * because the name of the lock is set using the
5047 * stringification of the argument to init_rwsem().
5049 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5050 lockdep_set_subclass(&ei->i_data_sem, subclass);
5054 * Standard function to be called on quota_on
5056 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5061 if (!test_opt(sb, QUOTA))
5064 /* Quotafile not on the same filesystem? */
5065 if (path->dentry->d_sb != sb)
5067 /* Journaling quota? */
5068 if (EXT4_SB(sb)->s_qf_names[type]) {
5069 /* Quotafile not in fs root? */
5070 if (path->dentry->d_parent != sb->s_root)
5071 ext4_msg(sb, KERN_WARNING,
5072 "Quota file not on filesystem root. "
5073 "Journaled quota will not work");
5077 * When we journal data on quota file, we have to flush journal to see
5078 * all updates to the file when we bypass pagecache...
5080 if (EXT4_SB(sb)->s_journal &&
5081 ext4_should_journal_data(d_inode(path->dentry))) {
5083 * We don't need to lock updates but journal_flush() could
5084 * otherwise be livelocked...
5086 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5087 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5088 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5092 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5093 err = dquot_quota_on(sb, type, format_id, path);
5095 lockdep_set_quota_inode(path->dentry->d_inode,
5100 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5104 struct inode *qf_inode;
5105 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5106 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5107 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5110 BUG_ON(!ext4_has_feature_quota(sb));
5112 if (!qf_inums[type])
5115 qf_inode = ext4_iget(sb, qf_inums[type]);
5116 if (IS_ERR(qf_inode)) {
5117 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5118 return PTR_ERR(qf_inode);
5121 /* Don't account quota for quota files to avoid recursion */
5122 qf_inode->i_flags |= S_NOQUOTA;
5123 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5124 err = dquot_enable(qf_inode, type, format_id, flags);
5127 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5132 /* Enable usage tracking for all quota types. */
5133 static int ext4_enable_quotas(struct super_block *sb)
5136 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5137 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5138 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5141 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5142 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5143 if (qf_inums[type]) {
5144 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5145 DQUOT_USAGE_ENABLED);
5147 for (type--; type >= 0; type--)
5148 dquot_quota_off(sb, type);
5151 "Failed to enable quota tracking "
5152 "(type=%d, err=%d). Please run "
5153 "e2fsck to fix.", type, err);
5161 static int ext4_quota_off(struct super_block *sb, int type)
5163 struct inode *inode = sb_dqopt(sb)->files[type];
5166 /* Force all delayed allocation blocks to be allocated.
5167 * Caller already holds s_umount sem */
5168 if (test_opt(sb, DELALLOC))
5169 sync_filesystem(sb);
5174 /* Update modification times of quota files when userspace can
5175 * start looking at them */
5176 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5179 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5180 ext4_mark_inode_dirty(handle, inode);
5181 ext4_journal_stop(handle);
5184 return dquot_quota_off(sb, type);
5187 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5188 * acquiring the locks... As quota files are never truncated and quota code
5189 * itself serializes the operations (and no one else should touch the files)
5190 * we don't have to be afraid of races */
5191 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5192 size_t len, loff_t off)
5194 struct inode *inode = sb_dqopt(sb)->files[type];
5195 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5196 int offset = off & (sb->s_blocksize - 1);
5199 struct buffer_head *bh;
5200 loff_t i_size = i_size_read(inode);
5204 if (off+len > i_size)
5207 while (toread > 0) {
5208 tocopy = sb->s_blocksize - offset < toread ?
5209 sb->s_blocksize - offset : toread;
5210 bh = ext4_bread(NULL, inode, blk, 0);
5213 if (!bh) /* A hole? */
5214 memset(data, 0, tocopy);
5216 memcpy(data, bh->b_data+offset, tocopy);
5226 /* Write to quotafile (we know the transaction is already started and has
5227 * enough credits) */
5228 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5229 const char *data, size_t len, loff_t off)
5231 struct inode *inode = sb_dqopt(sb)->files[type];
5232 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5233 int err, offset = off & (sb->s_blocksize - 1);
5235 struct buffer_head *bh;
5236 handle_t *handle = journal_current_handle();
5238 if (EXT4_SB(sb)->s_journal && !handle) {
5239 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5240 " cancelled because transaction is not started",
5241 (unsigned long long)off, (unsigned long long)len);
5245 * Since we account only one data block in transaction credits,
5246 * then it is impossible to cross a block boundary.
5248 if (sb->s_blocksize - offset < len) {
5249 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5250 " cancelled because not block aligned",
5251 (unsigned long long)off, (unsigned long long)len);
5256 bh = ext4_bread(handle, inode, blk,
5257 EXT4_GET_BLOCKS_CREATE |
5258 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5259 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5260 ext4_should_retry_alloc(inode->i_sb, &retries));
5265 BUFFER_TRACE(bh, "get write access");
5266 err = ext4_journal_get_write_access(handle, bh);
5272 memcpy(bh->b_data+offset, data, len);
5273 flush_dcache_page(bh->b_page);
5275 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5278 if (inode->i_size < off + len) {
5279 i_size_write(inode, off + len);
5280 EXT4_I(inode)->i_disksize = inode->i_size;
5281 ext4_mark_inode_dirty(handle, inode);
5288 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5289 const char *dev_name, void *data)
5291 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5294 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5295 static inline void register_as_ext2(void)
5297 int err = register_filesystem(&ext2_fs_type);
5300 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5303 static inline void unregister_as_ext2(void)
5305 unregister_filesystem(&ext2_fs_type);
5308 static inline int ext2_feature_set_ok(struct super_block *sb)
5310 if (ext4_has_unknown_ext2_incompat_features(sb))
5312 if (sb->s_flags & MS_RDONLY)
5314 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5319 static inline void register_as_ext2(void) { }
5320 static inline void unregister_as_ext2(void) { }
5321 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5324 static inline void register_as_ext3(void)
5326 int err = register_filesystem(&ext3_fs_type);
5329 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5332 static inline void unregister_as_ext3(void)
5334 unregister_filesystem(&ext3_fs_type);
5337 static inline int ext3_feature_set_ok(struct super_block *sb)
5339 if (ext4_has_unknown_ext3_incompat_features(sb))
5341 if (!ext4_has_feature_journal(sb))
5343 if (sb->s_flags & MS_RDONLY)
5345 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5350 static struct file_system_type ext4_fs_type = {
5351 .owner = THIS_MODULE,
5353 .mount = ext4_mount,
5354 .kill_sb = kill_block_super,
5355 .fs_flags = FS_REQUIRES_DEV,
5357 MODULE_ALIAS_FS("ext4");
5359 /* Shared across all ext4 file systems */
5360 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5361 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5363 static int __init ext4_init_fs(void)
5367 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5368 ext4_li_info = NULL;
5369 mutex_init(&ext4_li_mtx);
5371 /* Build-time check for flags consistency */
5372 ext4_check_flag_values();
5374 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5375 mutex_init(&ext4__aio_mutex[i]);
5376 init_waitqueue_head(&ext4__ioend_wq[i]);
5379 err = ext4_init_es();
5383 err = ext4_init_pageio();
5387 err = ext4_init_system_zone();
5391 err = ext4_init_sysfs();
5395 err = ext4_init_mballoc();
5399 ext4_mballoc_ready = 1;
5400 err = init_inodecache();
5405 err = register_filesystem(&ext4_fs_type);
5411 unregister_as_ext2();
5412 unregister_as_ext3();
5413 destroy_inodecache();
5415 ext4_mballoc_ready = 0;
5416 ext4_exit_mballoc();
5420 ext4_exit_system_zone();
5429 static void __exit ext4_exit_fs(void)
5432 ext4_destroy_lazyinit_thread();
5433 unregister_as_ext2();
5434 unregister_as_ext3();
5435 unregister_filesystem(&ext4_fs_type);
5436 destroy_inodecache();
5437 ext4_exit_mballoc();
5439 ext4_exit_system_zone();
5444 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5445 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5446 MODULE_LICENSE("GPL");
5447 module_init(ext4_init_fs)
5448 module_exit(ext4_exit_fs)