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;
798 ext4_unregister_li_request(sb);
799 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
801 flush_workqueue(sbi->rsv_conversion_wq);
802 destroy_workqueue(sbi->rsv_conversion_wq);
804 if (sbi->s_journal) {
805 err = jbd2_journal_destroy(sbi->s_journal);
806 sbi->s_journal = NULL;
808 ext4_abort(sb, "Couldn't clean up the journal");
811 ext4_unregister_sysfs(sb);
812 ext4_es_unregister_shrinker(sbi);
813 del_timer_sync(&sbi->s_err_report);
814 ext4_release_system_zone(sb);
816 ext4_ext_release(sb);
817 ext4_xattr_put_super(sb);
819 if (!(sb->s_flags & MS_RDONLY)) {
820 ext4_clear_feature_journal_needs_recovery(sb);
821 es->s_state = cpu_to_le16(sbi->s_mount_state);
823 if (!(sb->s_flags & MS_RDONLY))
824 ext4_commit_super(sb, 1);
826 for (i = 0; i < sbi->s_gdb_count; i++)
827 brelse(sbi->s_group_desc[i]);
828 kvfree(sbi->s_group_desc);
829 kvfree(sbi->s_flex_groups);
830 percpu_counter_destroy(&sbi->s_freeclusters_counter);
831 percpu_counter_destroy(&sbi->s_freeinodes_counter);
832 percpu_counter_destroy(&sbi->s_dirs_counter);
833 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
836 for (i = 0; i < EXT4_MAXQUOTAS; i++)
837 kfree(sbi->s_qf_names[i]);
840 /* Debugging code just in case the in-memory inode orphan list
841 * isn't empty. The on-disk one can be non-empty if we've
842 * detected an error and taken the fs readonly, but the
843 * in-memory list had better be clean by this point. */
844 if (!list_empty(&sbi->s_orphan))
845 dump_orphan_list(sb, sbi);
846 J_ASSERT(list_empty(&sbi->s_orphan));
848 sync_blockdev(sb->s_bdev);
849 invalidate_bdev(sb->s_bdev);
850 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
852 * Invalidate the journal device's buffers. We don't want them
853 * floating about in memory - the physical journal device may
854 * hotswapped, and it breaks the `ro-after' testing code.
856 sync_blockdev(sbi->journal_bdev);
857 invalidate_bdev(sbi->journal_bdev);
858 ext4_blkdev_remove(sbi);
860 if (sbi->s_mb_cache) {
861 ext4_xattr_destroy_cache(sbi->s_mb_cache);
862 sbi->s_mb_cache = NULL;
865 kthread_stop(sbi->s_mmp_tsk);
866 sb->s_fs_info = NULL;
868 * Now that we are completely done shutting down the
869 * superblock, we need to actually destroy the kobject.
871 kobject_put(&sbi->s_kobj);
872 wait_for_completion(&sbi->s_kobj_unregister);
873 if (sbi->s_chksum_driver)
874 crypto_free_shash(sbi->s_chksum_driver);
875 kfree(sbi->s_blockgroup_lock);
879 static struct kmem_cache *ext4_inode_cachep;
882 * Called inside transaction, so use GFP_NOFS
884 static struct inode *ext4_alloc_inode(struct super_block *sb)
886 struct ext4_inode_info *ei;
888 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
892 ei->vfs_inode.i_version = 1;
893 spin_lock_init(&ei->i_raw_lock);
894 INIT_LIST_HEAD(&ei->i_prealloc_list);
895 spin_lock_init(&ei->i_prealloc_lock);
896 ext4_es_init_tree(&ei->i_es_tree);
897 rwlock_init(&ei->i_es_lock);
898 INIT_LIST_HEAD(&ei->i_es_list);
901 ei->i_es_shrink_lblk = 0;
902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
905 ei->i_da_metadata_calc_len = 0;
906 ei->i_da_metadata_calc_last_lblock = 0;
907 spin_lock_init(&(ei->i_block_reservation_lock));
909 ei->i_reserved_quota = 0;
910 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
913 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
914 spin_lock_init(&ei->i_completed_io_lock);
916 ei->i_datasync_tid = 0;
917 atomic_set(&ei->i_ioend_count, 0);
918 atomic_set(&ei->i_unwritten, 0);
919 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
920 #ifdef CONFIG_EXT4_FS_ENCRYPTION
921 ei->i_crypt_info = NULL;
923 return &ei->vfs_inode;
926 static int ext4_drop_inode(struct inode *inode)
928 int drop = generic_drop_inode(inode);
930 trace_ext4_drop_inode(inode, drop);
934 static void ext4_i_callback(struct rcu_head *head)
936 struct inode *inode = container_of(head, struct inode, i_rcu);
937 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
940 static void ext4_destroy_inode(struct inode *inode)
942 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
943 ext4_msg(inode->i_sb, KERN_ERR,
944 "Inode %lu (%p): orphan list check failed!",
945 inode->i_ino, EXT4_I(inode));
946 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
947 EXT4_I(inode), sizeof(struct ext4_inode_info),
951 call_rcu(&inode->i_rcu, ext4_i_callback);
954 static void init_once(void *foo)
956 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
958 INIT_LIST_HEAD(&ei->i_orphan);
959 init_rwsem(&ei->xattr_sem);
960 init_rwsem(&ei->i_data_sem);
961 init_rwsem(&ei->i_mmap_sem);
962 inode_init_once(&ei->vfs_inode);
965 static int __init init_inodecache(void)
967 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
968 sizeof(struct ext4_inode_info),
969 0, (SLAB_RECLAIM_ACCOUNT|
972 if (ext4_inode_cachep == NULL)
977 static void destroy_inodecache(void)
980 * Make sure all delayed rcu free inodes are flushed before we
984 kmem_cache_destroy(ext4_inode_cachep);
987 void ext4_clear_inode(struct inode *inode)
989 invalidate_inode_buffers(inode);
992 ext4_discard_preallocations(inode);
993 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
994 if (EXT4_I(inode)->jinode) {
995 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
996 EXT4_I(inode)->jinode);
997 jbd2_free_inode(EXT4_I(inode)->jinode);
998 EXT4_I(inode)->jinode = NULL;
1000 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1001 if (EXT4_I(inode)->i_crypt_info)
1002 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1006 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1007 u64 ino, u32 generation)
1009 struct inode *inode;
1011 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1012 return ERR_PTR(-ESTALE);
1013 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1014 return ERR_PTR(-ESTALE);
1016 /* iget isn't really right if the inode is currently unallocated!!
1018 * ext4_read_inode will return a bad_inode if the inode had been
1019 * deleted, so we should be safe.
1021 * Currently we don't know the generation for parent directory, so
1022 * a generation of 0 means "accept any"
1024 inode = ext4_iget_normal(sb, ino);
1026 return ERR_CAST(inode);
1027 if (generation && inode->i_generation != generation) {
1029 return ERR_PTR(-ESTALE);
1035 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1036 int fh_len, int fh_type)
1038 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1039 ext4_nfs_get_inode);
1042 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1043 int fh_len, int fh_type)
1045 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1046 ext4_nfs_get_inode);
1050 * Try to release metadata pages (indirect blocks, directories) which are
1051 * mapped via the block device. Since these pages could have journal heads
1052 * which would prevent try_to_free_buffers() from freeing them, we must use
1053 * jbd2 layer's try_to_free_buffers() function to release them.
1055 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1058 journal_t *journal = EXT4_SB(sb)->s_journal;
1060 WARN_ON(PageChecked(page));
1061 if (!page_has_buffers(page))
1064 return jbd2_journal_try_to_free_buffers(journal, page,
1065 wait & ~__GFP_DIRECT_RECLAIM);
1066 return try_to_free_buffers(page);
1070 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1071 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1073 static int ext4_write_dquot(struct dquot *dquot);
1074 static int ext4_acquire_dquot(struct dquot *dquot);
1075 static int ext4_release_dquot(struct dquot *dquot);
1076 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1077 static int ext4_write_info(struct super_block *sb, int type);
1078 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1080 static int ext4_quota_off(struct super_block *sb, int type);
1081 static int ext4_quota_on_mount(struct super_block *sb, int type);
1082 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1083 size_t len, loff_t off);
1084 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1085 const char *data, size_t len, loff_t off);
1086 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1087 unsigned int flags);
1088 static int ext4_enable_quotas(struct super_block *sb);
1090 static struct dquot **ext4_get_dquots(struct inode *inode)
1092 return EXT4_I(inode)->i_dquot;
1095 static const struct dquot_operations ext4_quota_operations = {
1096 .get_reserved_space = ext4_get_reserved_space,
1097 .write_dquot = ext4_write_dquot,
1098 .acquire_dquot = ext4_acquire_dquot,
1099 .release_dquot = ext4_release_dquot,
1100 .mark_dirty = ext4_mark_dquot_dirty,
1101 .write_info = ext4_write_info,
1102 .alloc_dquot = dquot_alloc,
1103 .destroy_dquot = dquot_destroy,
1106 static const struct quotactl_ops ext4_qctl_operations = {
1107 .quota_on = ext4_quota_on,
1108 .quota_off = ext4_quota_off,
1109 .quota_sync = dquot_quota_sync,
1110 .get_state = dquot_get_state,
1111 .set_info = dquot_set_dqinfo,
1112 .get_dqblk = dquot_get_dqblk,
1113 .set_dqblk = dquot_set_dqblk
1117 static const struct super_operations ext4_sops = {
1118 .alloc_inode = ext4_alloc_inode,
1119 .destroy_inode = ext4_destroy_inode,
1120 .write_inode = ext4_write_inode,
1121 .dirty_inode = ext4_dirty_inode,
1122 .drop_inode = ext4_drop_inode,
1123 .evict_inode = ext4_evict_inode,
1124 .put_super = ext4_put_super,
1125 .sync_fs = ext4_sync_fs,
1126 .freeze_fs = ext4_freeze,
1127 .unfreeze_fs = ext4_unfreeze,
1128 .statfs = ext4_statfs,
1129 .remount_fs = ext4_remount,
1130 .show_options = ext4_show_options,
1132 .quota_read = ext4_quota_read,
1133 .quota_write = ext4_quota_write,
1134 .get_dquots = ext4_get_dquots,
1136 .bdev_try_to_free_page = bdev_try_to_free_page,
1139 static const struct export_operations ext4_export_ops = {
1140 .fh_to_dentry = ext4_fh_to_dentry,
1141 .fh_to_parent = ext4_fh_to_parent,
1142 .get_parent = ext4_get_parent,
1146 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1147 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1148 Opt_nouid32, Opt_debug, Opt_removed,
1149 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1150 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1151 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1152 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1153 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1154 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1155 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1156 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1157 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1158 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1159 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1160 Opt_lazytime, Opt_nolazytime,
1161 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1162 Opt_inode_readahead_blks, Opt_journal_ioprio,
1163 Opt_dioread_nolock, Opt_dioread_lock,
1164 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1165 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1168 static const match_table_t tokens = {
1169 {Opt_bsd_df, "bsddf"},
1170 {Opt_minix_df, "minixdf"},
1171 {Opt_grpid, "grpid"},
1172 {Opt_grpid, "bsdgroups"},
1173 {Opt_nogrpid, "nogrpid"},
1174 {Opt_nogrpid, "sysvgroups"},
1175 {Opt_resgid, "resgid=%u"},
1176 {Opt_resuid, "resuid=%u"},
1178 {Opt_err_cont, "errors=continue"},
1179 {Opt_err_panic, "errors=panic"},
1180 {Opt_err_ro, "errors=remount-ro"},
1181 {Opt_nouid32, "nouid32"},
1182 {Opt_debug, "debug"},
1183 {Opt_removed, "oldalloc"},
1184 {Opt_removed, "orlov"},
1185 {Opt_user_xattr, "user_xattr"},
1186 {Opt_nouser_xattr, "nouser_xattr"},
1188 {Opt_noacl, "noacl"},
1189 {Opt_noload, "norecovery"},
1190 {Opt_noload, "noload"},
1191 {Opt_removed, "nobh"},
1192 {Opt_removed, "bh"},
1193 {Opt_commit, "commit=%u"},
1194 {Opt_min_batch_time, "min_batch_time=%u"},
1195 {Opt_max_batch_time, "max_batch_time=%u"},
1196 {Opt_journal_dev, "journal_dev=%u"},
1197 {Opt_journal_path, "journal_path=%s"},
1198 {Opt_journal_checksum, "journal_checksum"},
1199 {Opt_nojournal_checksum, "nojournal_checksum"},
1200 {Opt_journal_async_commit, "journal_async_commit"},
1201 {Opt_abort, "abort"},
1202 {Opt_data_journal, "data=journal"},
1203 {Opt_data_ordered, "data=ordered"},
1204 {Opt_data_writeback, "data=writeback"},
1205 {Opt_data_err_abort, "data_err=abort"},
1206 {Opt_data_err_ignore, "data_err=ignore"},
1207 {Opt_offusrjquota, "usrjquota="},
1208 {Opt_usrjquota, "usrjquota=%s"},
1209 {Opt_offgrpjquota, "grpjquota="},
1210 {Opt_grpjquota, "grpjquota=%s"},
1211 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1212 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1213 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1214 {Opt_grpquota, "grpquota"},
1215 {Opt_noquota, "noquota"},
1216 {Opt_quota, "quota"},
1217 {Opt_usrquota, "usrquota"},
1218 {Opt_barrier, "barrier=%u"},
1219 {Opt_barrier, "barrier"},
1220 {Opt_nobarrier, "nobarrier"},
1221 {Opt_i_version, "i_version"},
1223 {Opt_stripe, "stripe=%u"},
1224 {Opt_delalloc, "delalloc"},
1225 {Opt_lazytime, "lazytime"},
1226 {Opt_nolazytime, "nolazytime"},
1227 {Opt_nodelalloc, "nodelalloc"},
1228 {Opt_removed, "mblk_io_submit"},
1229 {Opt_removed, "nomblk_io_submit"},
1230 {Opt_block_validity, "block_validity"},
1231 {Opt_noblock_validity, "noblock_validity"},
1232 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1233 {Opt_journal_ioprio, "journal_ioprio=%u"},
1234 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1235 {Opt_auto_da_alloc, "auto_da_alloc"},
1236 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1237 {Opt_dioread_nolock, "dioread_nolock"},
1238 {Opt_dioread_lock, "dioread_lock"},
1239 {Opt_discard, "discard"},
1240 {Opt_nodiscard, "nodiscard"},
1241 {Opt_init_itable, "init_itable=%u"},
1242 {Opt_init_itable, "init_itable"},
1243 {Opt_noinit_itable, "noinit_itable"},
1244 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1245 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1246 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1247 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1248 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1249 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1254 static ext4_fsblk_t get_sb_block(void **data)
1256 ext4_fsblk_t sb_block;
1257 char *options = (char *) *data;
1259 if (!options || strncmp(options, "sb=", 3) != 0)
1260 return 1; /* Default location */
1263 /* TODO: use simple_strtoll with >32bit ext4 */
1264 sb_block = simple_strtoul(options, &options, 0);
1265 if (*options && *options != ',') {
1266 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1270 if (*options == ',')
1272 *data = (void *) options;
1277 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1278 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1279 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1282 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1284 struct ext4_sb_info *sbi = EXT4_SB(sb);
1288 if (sb_any_quota_loaded(sb) &&
1289 !sbi->s_qf_names[qtype]) {
1290 ext4_msg(sb, KERN_ERR,
1291 "Cannot change journaled "
1292 "quota options when quota turned on");
1295 if (ext4_has_feature_quota(sb)) {
1296 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1297 "ignored when QUOTA feature is enabled");
1300 qname = match_strdup(args);
1302 ext4_msg(sb, KERN_ERR,
1303 "Not enough memory for storing quotafile name");
1306 if (sbi->s_qf_names[qtype]) {
1307 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1310 ext4_msg(sb, KERN_ERR,
1311 "%s quota file already specified",
1315 if (strchr(qname, '/')) {
1316 ext4_msg(sb, KERN_ERR,
1317 "quotafile must be on filesystem root");
1320 sbi->s_qf_names[qtype] = qname;
1328 static int clear_qf_name(struct super_block *sb, int qtype)
1331 struct ext4_sb_info *sbi = EXT4_SB(sb);
1333 if (sb_any_quota_loaded(sb) &&
1334 sbi->s_qf_names[qtype]) {
1335 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1336 " when quota turned on");
1339 kfree(sbi->s_qf_names[qtype]);
1340 sbi->s_qf_names[qtype] = NULL;
1345 #define MOPT_SET 0x0001
1346 #define MOPT_CLEAR 0x0002
1347 #define MOPT_NOSUPPORT 0x0004
1348 #define MOPT_EXPLICIT 0x0008
1349 #define MOPT_CLEAR_ERR 0x0010
1350 #define MOPT_GTE0 0x0020
1353 #define MOPT_QFMT 0x0040
1355 #define MOPT_Q MOPT_NOSUPPORT
1356 #define MOPT_QFMT MOPT_NOSUPPORT
1358 #define MOPT_DATAJ 0x0080
1359 #define MOPT_NO_EXT2 0x0100
1360 #define MOPT_NO_EXT3 0x0200
1361 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1362 #define MOPT_STRING 0x0400
1364 static const struct mount_opts {
1368 } ext4_mount_opts[] = {
1369 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1370 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1371 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1372 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1373 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1374 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1375 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1376 MOPT_EXT4_ONLY | MOPT_SET},
1377 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1378 MOPT_EXT4_ONLY | MOPT_CLEAR},
1379 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1380 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1381 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1382 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1383 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1384 MOPT_EXT4_ONLY | MOPT_CLEAR},
1385 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1386 MOPT_EXT4_ONLY | MOPT_CLEAR},
1387 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1388 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1389 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1390 EXT4_MOUNT_JOURNAL_CHECKSUM),
1391 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1392 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1393 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1397 MOPT_NO_EXT2 | MOPT_SET},
1398 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1399 MOPT_NO_EXT2 | MOPT_CLEAR},
1400 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1401 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1402 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1403 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1404 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1405 {Opt_commit, 0, MOPT_GTE0},
1406 {Opt_max_batch_time, 0, MOPT_GTE0},
1407 {Opt_min_batch_time, 0, MOPT_GTE0},
1408 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1409 {Opt_init_itable, 0, MOPT_GTE0},
1410 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1411 {Opt_stripe, 0, MOPT_GTE0},
1412 {Opt_resuid, 0, MOPT_GTE0},
1413 {Opt_resgid, 0, MOPT_GTE0},
1414 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1415 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1416 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1417 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1418 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1420 MOPT_NO_EXT2 | MOPT_DATAJ},
1421 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1422 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1423 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1424 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1425 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1427 {Opt_acl, 0, MOPT_NOSUPPORT},
1428 {Opt_noacl, 0, MOPT_NOSUPPORT},
1430 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1431 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1432 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1433 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1435 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1437 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1438 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1439 {Opt_usrjquota, 0, MOPT_Q},
1440 {Opt_grpjquota, 0, MOPT_Q},
1441 {Opt_offusrjquota, 0, MOPT_Q},
1442 {Opt_offgrpjquota, 0, MOPT_Q},
1443 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1444 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1445 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1446 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1447 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1451 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1452 substring_t *args, unsigned long *journal_devnum,
1453 unsigned int *journal_ioprio, int is_remount)
1455 struct ext4_sb_info *sbi = EXT4_SB(sb);
1456 const struct mount_opts *m;
1462 if (token == Opt_usrjquota)
1463 return set_qf_name(sb, USRQUOTA, &args[0]);
1464 else if (token == Opt_grpjquota)
1465 return set_qf_name(sb, GRPQUOTA, &args[0]);
1466 else if (token == Opt_offusrjquota)
1467 return clear_qf_name(sb, USRQUOTA);
1468 else if (token == Opt_offgrpjquota)
1469 return clear_qf_name(sb, GRPQUOTA);
1473 case Opt_nouser_xattr:
1474 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1477 return 1; /* handled by get_sb_block() */
1479 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1482 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1485 sb->s_flags |= MS_I_VERSION;
1488 sb->s_flags |= MS_LAZYTIME;
1490 case Opt_nolazytime:
1491 sb->s_flags &= ~MS_LAZYTIME;
1495 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1496 if (token == m->token)
1499 if (m->token == Opt_err) {
1500 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1501 "or missing value", opt);
1505 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1506 ext4_msg(sb, KERN_ERR,
1507 "Mount option \"%s\" incompatible with ext2", opt);
1510 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1511 ext4_msg(sb, KERN_ERR,
1512 "Mount option \"%s\" incompatible with ext3", opt);
1516 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1518 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1520 if (m->flags & MOPT_EXPLICIT) {
1521 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1522 set_opt2(sb, EXPLICIT_DELALLOC);
1523 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1524 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1528 if (m->flags & MOPT_CLEAR_ERR)
1529 clear_opt(sb, ERRORS_MASK);
1530 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1531 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1532 "options when quota turned on");
1536 if (m->flags & MOPT_NOSUPPORT) {
1537 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1538 } else if (token == Opt_commit) {
1540 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1541 sbi->s_commit_interval = HZ * arg;
1542 } else if (token == Opt_max_batch_time) {
1543 sbi->s_max_batch_time = arg;
1544 } else if (token == Opt_min_batch_time) {
1545 sbi->s_min_batch_time = arg;
1546 } else if (token == Opt_inode_readahead_blks) {
1547 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1548 ext4_msg(sb, KERN_ERR,
1549 "EXT4-fs: inode_readahead_blks must be "
1550 "0 or a power of 2 smaller than 2^31");
1553 sbi->s_inode_readahead_blks = arg;
1554 } else if (token == Opt_init_itable) {
1555 set_opt(sb, INIT_INODE_TABLE);
1557 arg = EXT4_DEF_LI_WAIT_MULT;
1558 sbi->s_li_wait_mult = arg;
1559 } else if (token == Opt_max_dir_size_kb) {
1560 sbi->s_max_dir_size_kb = arg;
1561 } else if (token == Opt_stripe) {
1562 sbi->s_stripe = arg;
1563 } else if (token == Opt_resuid) {
1564 uid = make_kuid(current_user_ns(), arg);
1565 if (!uid_valid(uid)) {
1566 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1569 sbi->s_resuid = uid;
1570 } else if (token == Opt_resgid) {
1571 gid = make_kgid(current_user_ns(), arg);
1572 if (!gid_valid(gid)) {
1573 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1576 sbi->s_resgid = gid;
1577 } else if (token == Opt_journal_dev) {
1579 ext4_msg(sb, KERN_ERR,
1580 "Cannot specify journal on remount");
1583 *journal_devnum = arg;
1584 } else if (token == Opt_journal_path) {
1586 struct inode *journal_inode;
1591 ext4_msg(sb, KERN_ERR,
1592 "Cannot specify journal on remount");
1595 journal_path = match_strdup(&args[0]);
1596 if (!journal_path) {
1597 ext4_msg(sb, KERN_ERR, "error: could not dup "
1598 "journal device string");
1602 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1604 ext4_msg(sb, KERN_ERR, "error: could not find "
1605 "journal device path: error %d", error);
1606 kfree(journal_path);
1610 journal_inode = d_inode(path.dentry);
1611 if (!S_ISBLK(journal_inode->i_mode)) {
1612 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1613 "is not a block device", journal_path);
1615 kfree(journal_path);
1619 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1621 kfree(journal_path);
1622 } else if (token == Opt_journal_ioprio) {
1624 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1629 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1630 } else if (token == Opt_test_dummy_encryption) {
1631 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1632 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1633 ext4_msg(sb, KERN_WARNING,
1634 "Test dummy encryption mode enabled");
1636 ext4_msg(sb, KERN_WARNING,
1637 "Test dummy encryption mount option ignored");
1639 } else if (m->flags & MOPT_DATAJ) {
1641 if (!sbi->s_journal)
1642 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1643 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1644 ext4_msg(sb, KERN_ERR,
1645 "Cannot change data mode on remount");
1649 clear_opt(sb, DATA_FLAGS);
1650 sbi->s_mount_opt |= m->mount_opt;
1653 } else if (m->flags & MOPT_QFMT) {
1654 if (sb_any_quota_loaded(sb) &&
1655 sbi->s_jquota_fmt != m->mount_opt) {
1656 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1657 "quota options when quota turned on");
1660 if (ext4_has_feature_quota(sb)) {
1661 ext4_msg(sb, KERN_INFO,
1662 "Quota format mount options ignored "
1663 "when QUOTA feature is enabled");
1666 sbi->s_jquota_fmt = m->mount_opt;
1668 } else if (token == Opt_dax) {
1669 #ifdef CONFIG_FS_DAX
1670 ext4_msg(sb, KERN_WARNING,
1671 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1672 sbi->s_mount_opt |= m->mount_opt;
1674 ext4_msg(sb, KERN_INFO, "dax option not supported");
1680 if (m->flags & MOPT_CLEAR)
1682 else if (unlikely(!(m->flags & MOPT_SET))) {
1683 ext4_msg(sb, KERN_WARNING,
1684 "buggy handling of option %s", opt);
1689 sbi->s_mount_opt |= m->mount_opt;
1691 sbi->s_mount_opt &= ~m->mount_opt;
1696 static int parse_options(char *options, struct super_block *sb,
1697 unsigned long *journal_devnum,
1698 unsigned int *journal_ioprio,
1701 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703 substring_t args[MAX_OPT_ARGS];
1709 while ((p = strsep(&options, ",")) != NULL) {
1713 * Initialize args struct so we know whether arg was
1714 * found; some options take optional arguments.
1716 args[0].to = args[0].from = NULL;
1717 token = match_token(p, tokens, args);
1718 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1719 journal_ioprio, is_remount) < 0)
1723 if (ext4_has_feature_quota(sb) &&
1724 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1725 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1726 "mount options ignored.");
1727 clear_opt(sb, USRQUOTA);
1728 clear_opt(sb, GRPQUOTA);
1729 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1730 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1731 clear_opt(sb, USRQUOTA);
1733 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1734 clear_opt(sb, GRPQUOTA);
1736 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1737 ext4_msg(sb, KERN_ERR, "old and new quota "
1742 if (!sbi->s_jquota_fmt) {
1743 ext4_msg(sb, KERN_ERR, "journaled quota format "
1749 if (test_opt(sb, DIOREAD_NOLOCK)) {
1751 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1753 if (blocksize < PAGE_CACHE_SIZE) {
1754 ext4_msg(sb, KERN_ERR, "can't mount with "
1755 "dioread_nolock if block size != PAGE_SIZE");
1759 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1760 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1761 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1762 "in data=ordered mode");
1768 static inline void ext4_show_quota_options(struct seq_file *seq,
1769 struct super_block *sb)
1771 #if defined(CONFIG_QUOTA)
1772 struct ext4_sb_info *sbi = EXT4_SB(sb);
1774 if (sbi->s_jquota_fmt) {
1777 switch (sbi->s_jquota_fmt) {
1788 seq_printf(seq, ",jqfmt=%s", fmtname);
1791 if (sbi->s_qf_names[USRQUOTA])
1792 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1794 if (sbi->s_qf_names[GRPQUOTA])
1795 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1799 static const char *token2str(int token)
1801 const struct match_token *t;
1803 for (t = tokens; t->token != Opt_err; t++)
1804 if (t->token == token && !strchr(t->pattern, '='))
1811 * - it's set to a non-default value OR
1812 * - if the per-sb default is different from the global default
1814 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1817 struct ext4_sb_info *sbi = EXT4_SB(sb);
1818 struct ext4_super_block *es = sbi->s_es;
1819 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1820 const struct mount_opts *m;
1821 char sep = nodefs ? '\n' : ',';
1823 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1824 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1826 if (sbi->s_sb_block != 1)
1827 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1829 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1830 int want_set = m->flags & MOPT_SET;
1831 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1832 (m->flags & MOPT_CLEAR_ERR))
1834 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1835 continue; /* skip if same as the default */
1837 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1838 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1839 continue; /* select Opt_noFoo vs Opt_Foo */
1840 SEQ_OPTS_PRINT("%s", token2str(m->token));
1843 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1844 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1845 SEQ_OPTS_PRINT("resuid=%u",
1846 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1847 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1848 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1849 SEQ_OPTS_PRINT("resgid=%u",
1850 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1851 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1852 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1853 SEQ_OPTS_PUTS("errors=remount-ro");
1854 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1855 SEQ_OPTS_PUTS("errors=continue");
1856 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1857 SEQ_OPTS_PUTS("errors=panic");
1858 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1859 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1860 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1861 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1862 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1863 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1864 if (sb->s_flags & MS_I_VERSION)
1865 SEQ_OPTS_PUTS("i_version");
1866 if (nodefs || sbi->s_stripe)
1867 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1868 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1869 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1870 SEQ_OPTS_PUTS("data=journal");
1871 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1872 SEQ_OPTS_PUTS("data=ordered");
1873 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1874 SEQ_OPTS_PUTS("data=writeback");
1877 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1878 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1879 sbi->s_inode_readahead_blks);
1881 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1882 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1883 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1884 if (nodefs || sbi->s_max_dir_size_kb)
1885 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1887 ext4_show_quota_options(seq, sb);
1891 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1893 return _ext4_show_options(seq, root->d_sb, 0);
1896 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1898 struct super_block *sb = seq->private;
1901 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1902 rc = _ext4_show_options(seq, sb, 1);
1903 seq_puts(seq, "\n");
1907 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1910 struct ext4_sb_info *sbi = EXT4_SB(sb);
1913 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1914 ext4_msg(sb, KERN_ERR, "revision level too high, "
1915 "forcing read-only mode");
1920 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1921 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1922 "running e2fsck is recommended");
1923 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1924 ext4_msg(sb, KERN_WARNING,
1925 "warning: mounting fs with errors, "
1926 "running e2fsck is recommended");
1927 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1928 le16_to_cpu(es->s_mnt_count) >=
1929 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1930 ext4_msg(sb, KERN_WARNING,
1931 "warning: maximal mount count reached, "
1932 "running e2fsck is recommended");
1933 else if (le32_to_cpu(es->s_checkinterval) &&
1934 (le32_to_cpu(es->s_lastcheck) +
1935 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1936 ext4_msg(sb, KERN_WARNING,
1937 "warning: checktime reached, "
1938 "running e2fsck is recommended");
1939 if (!sbi->s_journal)
1940 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1941 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1942 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1943 le16_add_cpu(&es->s_mnt_count, 1);
1944 es->s_mtime = cpu_to_le32(get_seconds());
1945 ext4_update_dynamic_rev(sb);
1947 ext4_set_feature_journal_needs_recovery(sb);
1949 ext4_commit_super(sb, 1);
1951 if (test_opt(sb, DEBUG))
1952 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1953 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1955 sbi->s_groups_count,
1956 EXT4_BLOCKS_PER_GROUP(sb),
1957 EXT4_INODES_PER_GROUP(sb),
1958 sbi->s_mount_opt, sbi->s_mount_opt2);
1960 cleancache_init_fs(sb);
1964 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1966 struct ext4_sb_info *sbi = EXT4_SB(sb);
1967 struct flex_groups *new_groups;
1970 if (!sbi->s_log_groups_per_flex)
1973 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1974 if (size <= sbi->s_flex_groups_allocated)
1977 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1978 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1980 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1981 size / (int) sizeof(struct flex_groups));
1985 if (sbi->s_flex_groups) {
1986 memcpy(new_groups, sbi->s_flex_groups,
1987 (sbi->s_flex_groups_allocated *
1988 sizeof(struct flex_groups)));
1989 kvfree(sbi->s_flex_groups);
1991 sbi->s_flex_groups = new_groups;
1992 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1996 static int ext4_fill_flex_info(struct super_block *sb)
1998 struct ext4_sb_info *sbi = EXT4_SB(sb);
1999 struct ext4_group_desc *gdp = NULL;
2000 ext4_group_t flex_group;
2003 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2004 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2005 sbi->s_log_groups_per_flex = 0;
2009 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2013 for (i = 0; i < sbi->s_groups_count; i++) {
2014 gdp = ext4_get_group_desc(sb, i, NULL);
2016 flex_group = ext4_flex_group(sbi, i);
2017 atomic_add(ext4_free_inodes_count(sb, gdp),
2018 &sbi->s_flex_groups[flex_group].free_inodes);
2019 atomic64_add(ext4_free_group_clusters(sb, gdp),
2020 &sbi->s_flex_groups[flex_group].free_clusters);
2021 atomic_add(ext4_used_dirs_count(sb, gdp),
2022 &sbi->s_flex_groups[flex_group].used_dirs);
2030 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2031 struct ext4_group_desc *gdp)
2035 __le32 le_group = cpu_to_le32(block_group);
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2038 if (ext4_has_metadata_csum(sbi->s_sb)) {
2039 /* Use new metadata_csum algorithm */
2043 save_csum = gdp->bg_checksum;
2044 gdp->bg_checksum = 0;
2045 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2047 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2049 gdp->bg_checksum = save_csum;
2051 crc = csum32 & 0xFFFF;
2055 /* old crc16 code */
2056 if (!ext4_has_feature_gdt_csum(sb))
2059 offset = offsetof(struct ext4_group_desc, bg_checksum);
2061 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2062 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2063 crc = crc16(crc, (__u8 *)gdp, offset);
2064 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2065 /* for checksum of struct ext4_group_desc do the rest...*/
2066 if (ext4_has_feature_64bit(sb) &&
2067 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2068 crc = crc16(crc, (__u8 *)gdp + offset,
2069 le16_to_cpu(sbi->s_es->s_desc_size) -
2073 return cpu_to_le16(crc);
2076 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2077 struct ext4_group_desc *gdp)
2079 if (ext4_has_group_desc_csum(sb) &&
2080 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2086 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2087 struct ext4_group_desc *gdp)
2089 if (!ext4_has_group_desc_csum(sb))
2091 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2094 /* Called at mount-time, super-block is locked */
2095 static int ext4_check_descriptors(struct super_block *sb,
2096 ext4_group_t *first_not_zeroed)
2098 struct ext4_sb_info *sbi = EXT4_SB(sb);
2099 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2100 ext4_fsblk_t last_block;
2101 ext4_fsblk_t block_bitmap;
2102 ext4_fsblk_t inode_bitmap;
2103 ext4_fsblk_t inode_table;
2104 int flexbg_flag = 0;
2105 ext4_group_t i, grp = sbi->s_groups_count;
2107 if (ext4_has_feature_flex_bg(sb))
2110 ext4_debug("Checking group descriptors");
2112 for (i = 0; i < sbi->s_groups_count; i++) {
2113 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2115 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2116 last_block = ext4_blocks_count(sbi->s_es) - 1;
2118 last_block = first_block +
2119 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2121 if ((grp == sbi->s_groups_count) &&
2122 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2125 block_bitmap = ext4_block_bitmap(sb, gdp);
2126 if (block_bitmap < first_block || block_bitmap > last_block) {
2127 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2128 "Block bitmap for group %u not in group "
2129 "(block %llu)!", i, block_bitmap);
2132 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2133 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2135 "Inode bitmap for group %u not in group "
2136 "(block %llu)!", i, inode_bitmap);
2139 inode_table = ext4_inode_table(sb, gdp);
2140 if (inode_table < first_block ||
2141 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2142 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2143 "Inode table for group %u not in group "
2144 "(block %llu)!", i, inode_table);
2147 ext4_lock_group(sb, i);
2148 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2149 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2150 "Checksum for group %u failed (%u!=%u)",
2151 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2152 gdp)), le16_to_cpu(gdp->bg_checksum));
2153 if (!(sb->s_flags & MS_RDONLY)) {
2154 ext4_unlock_group(sb, i);
2158 ext4_unlock_group(sb, i);
2160 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2162 if (NULL != first_not_zeroed)
2163 *first_not_zeroed = grp;
2167 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2168 * the superblock) which were deleted from all directories, but held open by
2169 * a process at the time of a crash. We walk the list and try to delete these
2170 * inodes at recovery time (only with a read-write filesystem).
2172 * In order to keep the orphan inode chain consistent during traversal (in
2173 * case of crash during recovery), we link each inode into the superblock
2174 * orphan list_head and handle it the same way as an inode deletion during
2175 * normal operation (which journals the operations for us).
2177 * We only do an iget() and an iput() on each inode, which is very safe if we
2178 * accidentally point at an in-use or already deleted inode. The worst that
2179 * can happen in this case is that we get a "bit already cleared" message from
2180 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2181 * e2fsck was run on this filesystem, and it must have already done the orphan
2182 * inode cleanup for us, so we can safely abort without any further action.
2184 static void ext4_orphan_cleanup(struct super_block *sb,
2185 struct ext4_super_block *es)
2187 unsigned int s_flags = sb->s_flags;
2188 int nr_orphans = 0, nr_truncates = 0;
2192 if (!es->s_last_orphan) {
2193 jbd_debug(4, "no orphan inodes to clean up\n");
2197 if (bdev_read_only(sb->s_bdev)) {
2198 ext4_msg(sb, KERN_ERR, "write access "
2199 "unavailable, skipping orphan cleanup");
2203 /* Check if feature set would not allow a r/w mount */
2204 if (!ext4_feature_set_ok(sb, 0)) {
2205 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2206 "unknown ROCOMPAT features");
2210 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2211 /* don't clear list on RO mount w/ errors */
2212 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2213 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2214 "clearing orphan list.\n");
2215 es->s_last_orphan = 0;
2217 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2221 if (s_flags & MS_RDONLY) {
2222 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2223 sb->s_flags &= ~MS_RDONLY;
2226 /* Needed for iput() to work correctly and not trash data */
2227 sb->s_flags |= MS_ACTIVE;
2228 /* Turn on quotas so that they are updated correctly */
2229 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2230 if (EXT4_SB(sb)->s_qf_names[i]) {
2231 int ret = ext4_quota_on_mount(sb, i);
2233 ext4_msg(sb, KERN_ERR,
2234 "Cannot turn on journaled "
2235 "quota: error %d", ret);
2240 while (es->s_last_orphan) {
2241 struct inode *inode;
2243 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2244 if (IS_ERR(inode)) {
2245 es->s_last_orphan = 0;
2249 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2250 dquot_initialize(inode);
2251 if (inode->i_nlink) {
2252 if (test_opt(sb, DEBUG))
2253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: truncating inode %lu to %lld bytes",
2255 __func__, inode->i_ino, inode->i_size);
2256 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2257 inode->i_ino, inode->i_size);
2258 mutex_lock(&inode->i_mutex);
2259 truncate_inode_pages(inode->i_mapping, inode->i_size);
2260 ext4_truncate(inode);
2261 mutex_unlock(&inode->i_mutex);
2264 if (test_opt(sb, DEBUG))
2265 ext4_msg(sb, KERN_DEBUG,
2266 "%s: deleting unreferenced inode %lu",
2267 __func__, inode->i_ino);
2268 jbd_debug(2, "deleting unreferenced inode %lu\n",
2272 iput(inode); /* The delete magic happens here! */
2275 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2278 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2279 PLURAL(nr_orphans));
2281 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2282 PLURAL(nr_truncates));
2284 /* Turn quotas off */
2285 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2286 if (sb_dqopt(sb)->files[i])
2287 dquot_quota_off(sb, i);
2290 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2294 * Maximal extent format file size.
2295 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2296 * extent format containers, within a sector_t, and within i_blocks
2297 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2298 * so that won't be a limiting factor.
2300 * However there is other limiting factor. We do store extents in the form
2301 * of starting block and length, hence the resulting length of the extent
2302 * covering maximum file size must fit into on-disk format containers as
2303 * well. Given that length is always by 1 unit bigger than max unit (because
2304 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2306 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2308 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2311 loff_t upper_limit = MAX_LFS_FILESIZE;
2313 /* small i_blocks in vfs inode? */
2314 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2316 * CONFIG_LBDAF is not enabled implies the inode
2317 * i_block represent total blocks in 512 bytes
2318 * 32 == size of vfs inode i_blocks * 8
2320 upper_limit = (1LL << 32) - 1;
2322 /* total blocks in file system block size */
2323 upper_limit >>= (blkbits - 9);
2324 upper_limit <<= blkbits;
2328 * 32-bit extent-start container, ee_block. We lower the maxbytes
2329 * by one fs block, so ee_len can cover the extent of maximum file
2332 res = (1LL << 32) - 1;
2335 /* Sanity check against vm- & vfs- imposed limits */
2336 if (res > upper_limit)
2343 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2344 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2345 * We need to be 1 filesystem block less than the 2^48 sector limit.
2347 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2349 loff_t res = EXT4_NDIR_BLOCKS;
2352 /* This is calculated to be the largest file size for a dense, block
2353 * mapped file such that the file's total number of 512-byte sectors,
2354 * including data and all indirect blocks, does not exceed (2^48 - 1).
2356 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2357 * number of 512-byte sectors of the file.
2360 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2362 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2363 * the inode i_block field represents total file blocks in
2364 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2366 upper_limit = (1LL << 32) - 1;
2368 /* total blocks in file system block size */
2369 upper_limit >>= (bits - 9);
2373 * We use 48 bit ext4_inode i_blocks
2374 * With EXT4_HUGE_FILE_FL set the i_blocks
2375 * represent total number of blocks in
2376 * file system block size
2378 upper_limit = (1LL << 48) - 1;
2382 /* indirect blocks */
2384 /* double indirect blocks */
2385 meta_blocks += 1 + (1LL << (bits-2));
2386 /* tripple indirect blocks */
2387 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2389 upper_limit -= meta_blocks;
2390 upper_limit <<= bits;
2392 res += 1LL << (bits-2);
2393 res += 1LL << (2*(bits-2));
2394 res += 1LL << (3*(bits-2));
2396 if (res > upper_limit)
2399 if (res > MAX_LFS_FILESIZE)
2400 res = MAX_LFS_FILESIZE;
2405 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2406 ext4_fsblk_t logical_sb_block, int nr)
2408 struct ext4_sb_info *sbi = EXT4_SB(sb);
2409 ext4_group_t bg, first_meta_bg;
2412 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2414 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2415 return logical_sb_block + nr + 1;
2416 bg = sbi->s_desc_per_block * nr;
2417 if (ext4_bg_has_super(sb, bg))
2421 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2422 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2423 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2426 if (sb->s_blocksize == 1024 && nr == 0 &&
2427 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2430 return (has_super + ext4_group_first_block_no(sb, bg));
2434 * ext4_get_stripe_size: Get the stripe size.
2435 * @sbi: In memory super block info
2437 * If we have specified it via mount option, then
2438 * use the mount option value. If the value specified at mount time is
2439 * greater than the blocks per group use the super block value.
2440 * If the super block value is greater than blocks per group return 0.
2441 * Allocator needs it be less than blocks per group.
2444 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2446 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2447 unsigned long stripe_width =
2448 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2451 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2452 ret = sbi->s_stripe;
2453 else if (stripe_width <= sbi->s_blocks_per_group)
2455 else if (stride <= sbi->s_blocks_per_group)
2461 * If the stripe width is 1, this makes no sense and
2462 * we set it to 0 to turn off stripe handling code.
2471 * Check whether this filesystem can be mounted based on
2472 * the features present and the RDONLY/RDWR mount requested.
2473 * Returns 1 if this filesystem can be mounted as requested,
2474 * 0 if it cannot be.
2476 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2478 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2479 ext4_msg(sb, KERN_ERR,
2480 "Couldn't mount because of "
2481 "unsupported optional features (%x)",
2482 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2483 ~EXT4_FEATURE_INCOMPAT_SUPP));
2490 if (ext4_has_feature_readonly(sb)) {
2491 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2492 sb->s_flags |= MS_RDONLY;
2496 /* Check that feature set is OK for a read-write mount */
2497 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2498 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2499 "unsupported optional features (%x)",
2500 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2501 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2505 * Large file size enabled file system can only be mounted
2506 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2508 if (ext4_has_feature_huge_file(sb)) {
2509 if (sizeof(blkcnt_t) < sizeof(u64)) {
2510 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2511 "cannot be mounted RDWR without "
2516 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2517 ext4_msg(sb, KERN_ERR,
2518 "Can't support bigalloc feature without "
2519 "extents feature\n");
2523 #ifndef CONFIG_QUOTA
2524 if (ext4_has_feature_quota(sb) && !readonly) {
2525 ext4_msg(sb, KERN_ERR,
2526 "Filesystem with quota feature cannot be mounted RDWR "
2527 "without CONFIG_QUOTA");
2530 #endif /* CONFIG_QUOTA */
2535 * This function is called once a day if we have errors logged
2536 * on the file system
2538 static void print_daily_error_info(unsigned long arg)
2540 struct super_block *sb = (struct super_block *) arg;
2541 struct ext4_sb_info *sbi;
2542 struct ext4_super_block *es;
2547 if (es->s_error_count)
2548 /* fsck newer than v1.41.13 is needed to clean this condition. */
2549 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2550 le32_to_cpu(es->s_error_count));
2551 if (es->s_first_error_time) {
2552 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2553 sb->s_id, le32_to_cpu(es->s_first_error_time),
2554 (int) sizeof(es->s_first_error_func),
2555 es->s_first_error_func,
2556 le32_to_cpu(es->s_first_error_line));
2557 if (es->s_first_error_ino)
2558 printk(": inode %u",
2559 le32_to_cpu(es->s_first_error_ino));
2560 if (es->s_first_error_block)
2561 printk(": block %llu", (unsigned long long)
2562 le64_to_cpu(es->s_first_error_block));
2565 if (es->s_last_error_time) {
2566 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2567 sb->s_id, le32_to_cpu(es->s_last_error_time),
2568 (int) sizeof(es->s_last_error_func),
2569 es->s_last_error_func,
2570 le32_to_cpu(es->s_last_error_line));
2571 if (es->s_last_error_ino)
2572 printk(": inode %u",
2573 le32_to_cpu(es->s_last_error_ino));
2574 if (es->s_last_error_block)
2575 printk(": block %llu", (unsigned long long)
2576 le64_to_cpu(es->s_last_error_block));
2579 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2582 /* Find next suitable group and run ext4_init_inode_table */
2583 static int ext4_run_li_request(struct ext4_li_request *elr)
2585 struct ext4_group_desc *gdp = NULL;
2586 ext4_group_t group, ngroups;
2587 struct super_block *sb;
2588 unsigned long timeout = 0;
2592 ngroups = EXT4_SB(sb)->s_groups_count;
2595 for (group = elr->lr_next_group; group < ngroups; group++) {
2596 gdp = ext4_get_group_desc(sb, group, NULL);
2602 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2606 if (group >= ngroups)
2611 ret = ext4_init_inode_table(sb, group,
2612 elr->lr_timeout ? 0 : 1);
2613 if (elr->lr_timeout == 0) {
2614 timeout = (jiffies - timeout) *
2615 elr->lr_sbi->s_li_wait_mult;
2616 elr->lr_timeout = timeout;
2618 elr->lr_next_sched = jiffies + elr->lr_timeout;
2619 elr->lr_next_group = group + 1;
2627 * Remove lr_request from the list_request and free the
2628 * request structure. Should be called with li_list_mtx held
2630 static void ext4_remove_li_request(struct ext4_li_request *elr)
2632 struct ext4_sb_info *sbi;
2639 list_del(&elr->lr_request);
2640 sbi->s_li_request = NULL;
2644 static void ext4_unregister_li_request(struct super_block *sb)
2646 mutex_lock(&ext4_li_mtx);
2647 if (!ext4_li_info) {
2648 mutex_unlock(&ext4_li_mtx);
2652 mutex_lock(&ext4_li_info->li_list_mtx);
2653 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2654 mutex_unlock(&ext4_li_info->li_list_mtx);
2655 mutex_unlock(&ext4_li_mtx);
2658 static struct task_struct *ext4_lazyinit_task;
2661 * This is the function where ext4lazyinit thread lives. It walks
2662 * through the request list searching for next scheduled filesystem.
2663 * When such a fs is found, run the lazy initialization request
2664 * (ext4_rn_li_request) and keep track of the time spend in this
2665 * function. Based on that time we compute next schedule time of
2666 * the request. When walking through the list is complete, compute
2667 * next waking time and put itself into sleep.
2669 static int ext4_lazyinit_thread(void *arg)
2671 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2672 struct list_head *pos, *n;
2673 struct ext4_li_request *elr;
2674 unsigned long next_wakeup, cur;
2676 BUG_ON(NULL == eli);
2680 next_wakeup = MAX_JIFFY_OFFSET;
2682 mutex_lock(&eli->li_list_mtx);
2683 if (list_empty(&eli->li_request_list)) {
2684 mutex_unlock(&eli->li_list_mtx);
2688 list_for_each_safe(pos, n, &eli->li_request_list) {
2689 elr = list_entry(pos, struct ext4_li_request,
2692 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2693 if (ext4_run_li_request(elr) != 0) {
2694 /* error, remove the lazy_init job */
2695 ext4_remove_li_request(elr);
2700 if (time_before(elr->lr_next_sched, next_wakeup))
2701 next_wakeup = elr->lr_next_sched;
2703 mutex_unlock(&eli->li_list_mtx);
2708 if ((time_after_eq(cur, next_wakeup)) ||
2709 (MAX_JIFFY_OFFSET == next_wakeup)) {
2714 schedule_timeout_interruptible(next_wakeup - cur);
2716 if (kthread_should_stop()) {
2717 ext4_clear_request_list();
2724 * It looks like the request list is empty, but we need
2725 * to check it under the li_list_mtx lock, to prevent any
2726 * additions into it, and of course we should lock ext4_li_mtx
2727 * to atomically free the list and ext4_li_info, because at
2728 * this point another ext4 filesystem could be registering
2731 mutex_lock(&ext4_li_mtx);
2732 mutex_lock(&eli->li_list_mtx);
2733 if (!list_empty(&eli->li_request_list)) {
2734 mutex_unlock(&eli->li_list_mtx);
2735 mutex_unlock(&ext4_li_mtx);
2738 mutex_unlock(&eli->li_list_mtx);
2739 kfree(ext4_li_info);
2740 ext4_li_info = NULL;
2741 mutex_unlock(&ext4_li_mtx);
2746 static void ext4_clear_request_list(void)
2748 struct list_head *pos, *n;
2749 struct ext4_li_request *elr;
2751 mutex_lock(&ext4_li_info->li_list_mtx);
2752 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2753 elr = list_entry(pos, struct ext4_li_request,
2755 ext4_remove_li_request(elr);
2757 mutex_unlock(&ext4_li_info->li_list_mtx);
2760 static int ext4_run_lazyinit_thread(void)
2762 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2763 ext4_li_info, "ext4lazyinit");
2764 if (IS_ERR(ext4_lazyinit_task)) {
2765 int err = PTR_ERR(ext4_lazyinit_task);
2766 ext4_clear_request_list();
2767 kfree(ext4_li_info);
2768 ext4_li_info = NULL;
2769 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2770 "initialization thread\n",
2774 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2779 * Check whether it make sense to run itable init. thread or not.
2780 * If there is at least one uninitialized inode table, return
2781 * corresponding group number, else the loop goes through all
2782 * groups and return total number of groups.
2784 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2786 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2787 struct ext4_group_desc *gdp = NULL;
2789 for (group = 0; group < ngroups; group++) {
2790 gdp = ext4_get_group_desc(sb, group, NULL);
2794 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2801 static int ext4_li_info_new(void)
2803 struct ext4_lazy_init *eli = NULL;
2805 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2809 INIT_LIST_HEAD(&eli->li_request_list);
2810 mutex_init(&eli->li_list_mtx);
2812 eli->li_state |= EXT4_LAZYINIT_QUIT;
2819 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2822 struct ext4_sb_info *sbi = EXT4_SB(sb);
2823 struct ext4_li_request *elr;
2825 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2831 elr->lr_next_group = start;
2834 * Randomize first schedule time of the request to
2835 * spread the inode table initialization requests
2838 elr->lr_next_sched = jiffies + (prandom_u32() %
2839 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2843 int ext4_register_li_request(struct super_block *sb,
2844 ext4_group_t first_not_zeroed)
2846 struct ext4_sb_info *sbi = EXT4_SB(sb);
2847 struct ext4_li_request *elr = NULL;
2848 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2851 mutex_lock(&ext4_li_mtx);
2852 if (sbi->s_li_request != NULL) {
2854 * Reset timeout so it can be computed again, because
2855 * s_li_wait_mult might have changed.
2857 sbi->s_li_request->lr_timeout = 0;
2861 if (first_not_zeroed == ngroups ||
2862 (sb->s_flags & MS_RDONLY) ||
2863 !test_opt(sb, INIT_INODE_TABLE))
2866 elr = ext4_li_request_new(sb, first_not_zeroed);
2872 if (NULL == ext4_li_info) {
2873 ret = ext4_li_info_new();
2878 mutex_lock(&ext4_li_info->li_list_mtx);
2879 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2880 mutex_unlock(&ext4_li_info->li_list_mtx);
2882 sbi->s_li_request = elr;
2884 * set elr to NULL here since it has been inserted to
2885 * the request_list and the removal and free of it is
2886 * handled by ext4_clear_request_list from now on.
2890 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2891 ret = ext4_run_lazyinit_thread();
2896 mutex_unlock(&ext4_li_mtx);
2903 * We do not need to lock anything since this is called on
2906 static void ext4_destroy_lazyinit_thread(void)
2909 * If thread exited earlier
2910 * there's nothing to be done.
2912 if (!ext4_li_info || !ext4_lazyinit_task)
2915 kthread_stop(ext4_lazyinit_task);
2918 static int set_journal_csum_feature_set(struct super_block *sb)
2921 int compat, incompat;
2922 struct ext4_sb_info *sbi = EXT4_SB(sb);
2924 if (ext4_has_metadata_csum(sb)) {
2925 /* journal checksum v3 */
2927 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2929 /* journal checksum v1 */
2930 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2934 jbd2_journal_clear_features(sbi->s_journal,
2935 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2936 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2937 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2938 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2939 ret = jbd2_journal_set_features(sbi->s_journal,
2941 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2943 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2944 ret = jbd2_journal_set_features(sbi->s_journal,
2947 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2948 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2950 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2951 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2958 * Note: calculating the overhead so we can be compatible with
2959 * historical BSD practice is quite difficult in the face of
2960 * clusters/bigalloc. This is because multiple metadata blocks from
2961 * different block group can end up in the same allocation cluster.
2962 * Calculating the exact overhead in the face of clustered allocation
2963 * requires either O(all block bitmaps) in memory or O(number of block
2964 * groups**2) in time. We will still calculate the superblock for
2965 * older file systems --- and if we come across with a bigalloc file
2966 * system with zero in s_overhead_clusters the estimate will be close to
2967 * correct especially for very large cluster sizes --- but for newer
2968 * file systems, it's better to calculate this figure once at mkfs
2969 * time, and store it in the superblock. If the superblock value is
2970 * present (even for non-bigalloc file systems), we will use it.
2972 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2975 struct ext4_sb_info *sbi = EXT4_SB(sb);
2976 struct ext4_group_desc *gdp;
2977 ext4_fsblk_t first_block, last_block, b;
2978 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2979 int s, j, count = 0;
2981 if (!ext4_has_feature_bigalloc(sb))
2982 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2983 sbi->s_itb_per_group + 2);
2985 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2986 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2987 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2988 for (i = 0; i < ngroups; i++) {
2989 gdp = ext4_get_group_desc(sb, i, NULL);
2990 b = ext4_block_bitmap(sb, gdp);
2991 if (b >= first_block && b <= last_block) {
2992 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2995 b = ext4_inode_bitmap(sb, gdp);
2996 if (b >= first_block && b <= last_block) {
2997 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3000 b = ext4_inode_table(sb, gdp);
3001 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3002 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3003 int c = EXT4_B2C(sbi, b - first_block);
3004 ext4_set_bit(c, buf);
3010 if (ext4_bg_has_super(sb, grp)) {
3011 ext4_set_bit(s++, buf);
3014 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3015 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3021 return EXT4_CLUSTERS_PER_GROUP(sb) -
3022 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3026 * Compute the overhead and stash it in sbi->s_overhead
3028 int ext4_calculate_overhead(struct super_block *sb)
3030 struct ext4_sb_info *sbi = EXT4_SB(sb);
3031 struct ext4_super_block *es = sbi->s_es;
3032 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3033 ext4_fsblk_t overhead = 0;
3034 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3040 * Compute the overhead (FS structures). This is constant
3041 * for a given filesystem unless the number of block groups
3042 * changes so we cache the previous value until it does.
3046 * All of the blocks before first_data_block are overhead
3048 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3051 * Add the overhead found in each block group
3053 for (i = 0; i < ngroups; i++) {
3056 blks = count_overhead(sb, i, buf);
3059 memset(buf, 0, PAGE_SIZE);
3062 /* Add the internal journal blocks as well */
3063 if (sbi->s_journal && !sbi->journal_bdev)
3064 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3066 sbi->s_overhead = overhead;
3068 free_page((unsigned long) buf);
3072 static void ext4_set_resv_clusters(struct super_block *sb)
3074 ext4_fsblk_t resv_clusters;
3075 struct ext4_sb_info *sbi = EXT4_SB(sb);
3078 * There's no need to reserve anything when we aren't using extents.
3079 * The space estimates are exact, there are no unwritten extents,
3080 * hole punching doesn't need new metadata... This is needed especially
3081 * to keep ext2/3 backward compatibility.
3083 if (!ext4_has_feature_extents(sb))
3086 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3087 * This should cover the situations where we can not afford to run
3088 * out of space like for example punch hole, or converting
3089 * unwritten extents in delalloc path. In most cases such
3090 * allocation would require 1, or 2 blocks, higher numbers are
3093 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3094 sbi->s_cluster_bits);
3096 do_div(resv_clusters, 50);
3097 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3099 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3102 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3104 char *orig_data = kstrdup(data, GFP_KERNEL);
3105 struct buffer_head *bh;
3106 struct ext4_super_block *es = NULL;
3107 struct ext4_sb_info *sbi;
3109 ext4_fsblk_t sb_block = get_sb_block(&data);
3110 ext4_fsblk_t logical_sb_block;
3111 unsigned long offset = 0;
3112 unsigned long journal_devnum = 0;
3113 unsigned long def_mount_opts;
3117 int blocksize, clustersize;
3118 unsigned int db_count;
3120 int needs_recovery, has_huge_files, has_bigalloc;
3123 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3124 ext4_group_t first_not_zeroed;
3126 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3130 sbi->s_blockgroup_lock =
3131 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3132 if (!sbi->s_blockgroup_lock) {
3136 sb->s_fs_info = sbi;
3138 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3139 sbi->s_sb_block = sb_block;
3140 if (sb->s_bdev->bd_part)
3141 sbi->s_sectors_written_start =
3142 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3144 /* Cleanup superblock name */
3145 strreplace(sb->s_id, '/', '!');
3147 /* -EINVAL is default */
3149 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3151 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3156 * The ext4 superblock will not be buffer aligned for other than 1kB
3157 * block sizes. We need to calculate the offset from buffer start.
3159 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3160 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3161 offset = do_div(logical_sb_block, blocksize);
3163 logical_sb_block = sb_block;
3166 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3167 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3171 * Note: s_es must be initialized as soon as possible because
3172 * some ext4 macro-instructions depend on its value
3174 es = (struct ext4_super_block *) (bh->b_data + offset);
3176 sb->s_magic = le16_to_cpu(es->s_magic);
3177 if (sb->s_magic != EXT4_SUPER_MAGIC)
3179 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3181 /* Warn if metadata_csum and gdt_csum are both set. */
3182 if (ext4_has_feature_metadata_csum(sb) &&
3183 ext4_has_feature_gdt_csum(sb))
3184 ext4_warning(sb, "metadata_csum and uninit_bg are "
3185 "redundant flags; please run fsck.");
3187 /* Check for a known checksum algorithm */
3188 if (!ext4_verify_csum_type(sb, es)) {
3189 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3190 "unknown checksum algorithm.");
3195 /* Load the checksum driver */
3196 if (ext4_has_feature_metadata_csum(sb)) {
3197 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3198 if (IS_ERR(sbi->s_chksum_driver)) {
3199 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3200 ret = PTR_ERR(sbi->s_chksum_driver);
3201 sbi->s_chksum_driver = NULL;
3206 /* Check superblock checksum */
3207 if (!ext4_superblock_csum_verify(sb, es)) {
3208 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3209 "invalid superblock checksum. Run e2fsck?");
3215 /* Precompute checksum seed for all metadata */
3216 if (ext4_has_feature_csum_seed(sb))
3217 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3218 else if (ext4_has_metadata_csum(sb))
3219 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3220 sizeof(es->s_uuid));
3222 /* Set defaults before we parse the mount options */
3223 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3224 set_opt(sb, INIT_INODE_TABLE);
3225 if (def_mount_opts & EXT4_DEFM_DEBUG)
3227 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3229 if (def_mount_opts & EXT4_DEFM_UID16)
3230 set_opt(sb, NO_UID32);
3231 /* xattr user namespace & acls are now defaulted on */
3232 set_opt(sb, XATTR_USER);
3233 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3234 set_opt(sb, POSIX_ACL);
3236 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3237 if (ext4_has_metadata_csum(sb))
3238 set_opt(sb, JOURNAL_CHECKSUM);
3240 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3241 set_opt(sb, JOURNAL_DATA);
3242 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3243 set_opt(sb, ORDERED_DATA);
3244 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3245 set_opt(sb, WRITEBACK_DATA);
3247 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3248 set_opt(sb, ERRORS_PANIC);
3249 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3250 set_opt(sb, ERRORS_CONT);
3252 set_opt(sb, ERRORS_RO);
3253 /* block_validity enabled by default; disable with noblock_validity */
3254 set_opt(sb, BLOCK_VALIDITY);
3255 if (def_mount_opts & EXT4_DEFM_DISCARD)
3256 set_opt(sb, DISCARD);
3258 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3259 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3260 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3261 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3262 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3264 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3265 set_opt(sb, BARRIER);
3268 * enable delayed allocation by default
3269 * Use -o nodelalloc to turn it off
3271 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3272 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3273 set_opt(sb, DELALLOC);
3276 * set default s_li_wait_mult for lazyinit, for the case there is
3277 * no mount option specified.
3279 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3281 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3282 &journal_devnum, &journal_ioprio, 0)) {
3283 ext4_msg(sb, KERN_WARNING,
3284 "failed to parse options in superblock: %s",
3285 sbi->s_es->s_mount_opts);
3287 sbi->s_def_mount_opt = sbi->s_mount_opt;
3288 if (!parse_options((char *) data, sb, &journal_devnum,
3289 &journal_ioprio, 0))
3292 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3293 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3294 "with data=journal disables delayed "
3295 "allocation and O_DIRECT support!\n");
3296 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3297 ext4_msg(sb, KERN_ERR, "can't mount with "
3298 "both data=journal and delalloc");
3301 if (test_opt(sb, DIOREAD_NOLOCK)) {
3302 ext4_msg(sb, KERN_ERR, "can't mount with "
3303 "both data=journal and dioread_nolock");
3306 if (test_opt(sb, DAX)) {
3307 ext4_msg(sb, KERN_ERR, "can't mount with "
3308 "both data=journal and dax");
3311 if (test_opt(sb, DELALLOC))
3312 clear_opt(sb, DELALLOC);
3314 sb->s_iflags |= SB_I_CGROUPWB;
3317 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3318 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3320 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3321 (ext4_has_compat_features(sb) ||
3322 ext4_has_ro_compat_features(sb) ||
3323 ext4_has_incompat_features(sb)))
3324 ext4_msg(sb, KERN_WARNING,
3325 "feature flags set on rev 0 fs, "
3326 "running e2fsck is recommended");
3328 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3329 set_opt2(sb, HURD_COMPAT);
3330 if (ext4_has_feature_64bit(sb)) {
3331 ext4_msg(sb, KERN_ERR,
3332 "The Hurd can't support 64-bit file systems");
3337 if (IS_EXT2_SB(sb)) {
3338 if (ext2_feature_set_ok(sb))
3339 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3340 "using the ext4 subsystem");
3342 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3343 "to feature incompatibilities");
3348 if (IS_EXT3_SB(sb)) {
3349 if (ext3_feature_set_ok(sb))
3350 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3351 "using the ext4 subsystem");
3353 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3354 "to feature incompatibilities");
3360 * Check feature flags regardless of the revision level, since we
3361 * previously didn't change the revision level when setting the flags,
3362 * so there is a chance incompat flags are set on a rev 0 filesystem.
3364 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3367 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3368 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3369 blocksize > EXT4_MAX_BLOCK_SIZE) {
3370 ext4_msg(sb, KERN_ERR,
3371 "Unsupported filesystem blocksize %d", blocksize);
3375 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3376 if (blocksize != PAGE_SIZE) {
3377 ext4_msg(sb, KERN_ERR,
3378 "error: unsupported blocksize for dax");
3381 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3382 ext4_msg(sb, KERN_ERR,
3383 "error: device does not support dax");
3388 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3389 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3390 es->s_encryption_level);
3394 if (sb->s_blocksize != blocksize) {
3395 /* Validate the filesystem blocksize */
3396 if (!sb_set_blocksize(sb, blocksize)) {
3397 ext4_msg(sb, KERN_ERR, "bad block size %d",
3403 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3404 offset = do_div(logical_sb_block, blocksize);
3405 bh = sb_bread_unmovable(sb, logical_sb_block);
3407 ext4_msg(sb, KERN_ERR,
3408 "Can't read superblock on 2nd try");
3411 es = (struct ext4_super_block *)(bh->b_data + offset);
3413 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3414 ext4_msg(sb, KERN_ERR,
3415 "Magic mismatch, very weird!");
3420 has_huge_files = ext4_has_feature_huge_file(sb);
3421 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3423 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3425 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3426 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3427 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3429 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3430 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3431 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3432 (!is_power_of_2(sbi->s_inode_size)) ||
3433 (sbi->s_inode_size > blocksize)) {
3434 ext4_msg(sb, KERN_ERR,
3435 "unsupported inode size: %d",
3439 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3440 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3443 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3444 if (ext4_has_feature_64bit(sb)) {
3445 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3446 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3447 !is_power_of_2(sbi->s_desc_size)) {
3448 ext4_msg(sb, KERN_ERR,
3449 "unsupported descriptor size %lu",
3454 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3456 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3457 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3458 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3461 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3462 if (sbi->s_inodes_per_block == 0)
3464 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3465 sbi->s_inodes_per_block;
3466 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3468 sbi->s_mount_state = le16_to_cpu(es->s_state);
3469 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3470 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3472 for (i = 0; i < 4; i++)
3473 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3474 sbi->s_def_hash_version = es->s_def_hash_version;
3475 if (ext4_has_feature_dir_index(sb)) {
3476 i = le32_to_cpu(es->s_flags);
3477 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3478 sbi->s_hash_unsigned = 3;
3479 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3480 #ifdef __CHAR_UNSIGNED__
3481 if (!(sb->s_flags & MS_RDONLY))
3483 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3484 sbi->s_hash_unsigned = 3;
3486 if (!(sb->s_flags & MS_RDONLY))
3488 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3493 /* Handle clustersize */
3494 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3495 has_bigalloc = ext4_has_feature_bigalloc(sb);
3497 if (clustersize < blocksize) {
3498 ext4_msg(sb, KERN_ERR,
3499 "cluster size (%d) smaller than "
3500 "block size (%d)", clustersize, blocksize);
3503 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3504 le32_to_cpu(es->s_log_block_size);
3505 sbi->s_clusters_per_group =
3506 le32_to_cpu(es->s_clusters_per_group);
3507 if (sbi->s_clusters_per_group > blocksize * 8) {
3508 ext4_msg(sb, KERN_ERR,
3509 "#clusters per group too big: %lu",
3510 sbi->s_clusters_per_group);
3513 if (sbi->s_blocks_per_group !=
3514 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3515 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3516 "clusters per group (%lu) inconsistent",
3517 sbi->s_blocks_per_group,
3518 sbi->s_clusters_per_group);
3522 if (clustersize != blocksize) {
3523 ext4_warning(sb, "fragment/cluster size (%d) != "
3524 "block size (%d)", clustersize,
3526 clustersize = blocksize;
3528 if (sbi->s_blocks_per_group > blocksize * 8) {
3529 ext4_msg(sb, KERN_ERR,
3530 "#blocks per group too big: %lu",
3531 sbi->s_blocks_per_group);
3534 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3535 sbi->s_cluster_bits = 0;
3537 sbi->s_cluster_ratio = clustersize / blocksize;
3539 if (sbi->s_inodes_per_group > blocksize * 8) {
3540 ext4_msg(sb, KERN_ERR,
3541 "#inodes per group too big: %lu",
3542 sbi->s_inodes_per_group);
3546 /* Do we have standard group size of clustersize * 8 blocks ? */
3547 if (sbi->s_blocks_per_group == clustersize << 3)
3548 set_opt2(sb, STD_GROUP_SIZE);
3551 * Test whether we have more sectors than will fit in sector_t,
3552 * and whether the max offset is addressable by the page cache.
3554 err = generic_check_addressable(sb->s_blocksize_bits,
3555 ext4_blocks_count(es));
3557 ext4_msg(sb, KERN_ERR, "filesystem"
3558 " too large to mount safely on this system");
3559 if (sizeof(sector_t) < 8)
3560 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3564 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3567 /* check blocks count against device size */
3568 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3569 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3570 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3571 "exceeds size of device (%llu blocks)",
3572 ext4_blocks_count(es), blocks_count);
3577 * It makes no sense for the first data block to be beyond the end
3578 * of the filesystem.
3580 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3581 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3582 "block %u is beyond end of filesystem (%llu)",
3583 le32_to_cpu(es->s_first_data_block),
3584 ext4_blocks_count(es));
3587 blocks_count = (ext4_blocks_count(es) -
3588 le32_to_cpu(es->s_first_data_block) +
3589 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3590 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3591 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3592 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3593 "(block count %llu, first data block %u, "
3594 "blocks per group %lu)", sbi->s_groups_count,
3595 ext4_blocks_count(es),
3596 le32_to_cpu(es->s_first_data_block),
3597 EXT4_BLOCKS_PER_GROUP(sb));
3600 sbi->s_groups_count = blocks_count;
3601 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3602 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3603 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3604 EXT4_DESC_PER_BLOCK(sb);
3605 sbi->s_group_desc = ext4_kvmalloc(db_count *
3606 sizeof(struct buffer_head *),
3608 if (sbi->s_group_desc == NULL) {
3609 ext4_msg(sb, KERN_ERR, "not enough memory");
3614 bgl_lock_init(sbi->s_blockgroup_lock);
3616 for (i = 0; i < db_count; i++) {
3617 block = descriptor_loc(sb, logical_sb_block, i);
3618 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3619 if (!sbi->s_group_desc[i]) {
3620 ext4_msg(sb, KERN_ERR,
3621 "can't read group descriptor %d", i);
3626 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3627 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3628 ret = -EFSCORRUPTED;
3632 sbi->s_gdb_count = db_count;
3633 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3634 spin_lock_init(&sbi->s_next_gen_lock);
3636 setup_timer(&sbi->s_err_report, print_daily_error_info,
3637 (unsigned long) sb);
3639 /* Register extent status tree shrinker */
3640 if (ext4_es_register_shrinker(sbi))
3643 sbi->s_stripe = ext4_get_stripe_size(sbi);
3644 sbi->s_extent_max_zeroout_kb = 32;
3647 * set up enough so that it can read an inode
3649 sb->s_op = &ext4_sops;
3650 sb->s_export_op = &ext4_export_ops;
3651 sb->s_xattr = ext4_xattr_handlers;
3653 sb->dq_op = &ext4_quota_operations;
3654 if (ext4_has_feature_quota(sb))
3655 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3657 sb->s_qcop = &ext4_qctl_operations;
3658 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3660 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3662 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3663 mutex_init(&sbi->s_orphan_lock);
3667 needs_recovery = (es->s_last_orphan != 0 ||
3668 ext4_has_feature_journal_needs_recovery(sb));
3670 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3671 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3672 goto failed_mount3a;
3675 * The first inode we look at is the journal inode. Don't try
3676 * root first: it may be modified in the journal!
3678 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3679 if (ext4_load_journal(sb, es, journal_devnum))
3680 goto failed_mount3a;
3681 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3682 ext4_has_feature_journal_needs_recovery(sb)) {
3683 ext4_msg(sb, KERN_ERR, "required journal recovery "
3684 "suppressed and not mounted read-only");
3685 goto failed_mount_wq;
3687 /* Nojournal mode, all journal mount options are illegal */
3688 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3689 ext4_msg(sb, KERN_ERR, "can't mount with "
3690 "journal_checksum, fs mounted w/o journal");
3691 goto failed_mount_wq;
3693 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3694 ext4_msg(sb, KERN_ERR, "can't mount with "
3695 "journal_async_commit, fs mounted w/o journal");
3696 goto failed_mount_wq;
3698 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3699 ext4_msg(sb, KERN_ERR, "can't mount with "
3700 "commit=%lu, fs mounted w/o journal",
3701 sbi->s_commit_interval / HZ);
3702 goto failed_mount_wq;
3704 if (EXT4_MOUNT_DATA_FLAGS &
3705 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3706 ext4_msg(sb, KERN_ERR, "can't mount with "
3707 "data=, fs mounted w/o journal");
3708 goto failed_mount_wq;
3710 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3711 clear_opt(sb, JOURNAL_CHECKSUM);
3712 clear_opt(sb, DATA_FLAGS);
3713 sbi->s_journal = NULL;
3718 if (ext4_has_feature_64bit(sb) &&
3719 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3720 JBD2_FEATURE_INCOMPAT_64BIT)) {
3721 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3722 goto failed_mount_wq;
3725 if (!set_journal_csum_feature_set(sb)) {
3726 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3728 goto failed_mount_wq;
3731 /* We have now updated the journal if required, so we can
3732 * validate the data journaling mode. */
3733 switch (test_opt(sb, DATA_FLAGS)) {
3735 /* No mode set, assume a default based on the journal
3736 * capabilities: ORDERED_DATA if the journal can
3737 * cope, else JOURNAL_DATA
3739 if (jbd2_journal_check_available_features
3740 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3741 set_opt(sb, ORDERED_DATA);
3743 set_opt(sb, JOURNAL_DATA);
3746 case EXT4_MOUNT_ORDERED_DATA:
3747 case EXT4_MOUNT_WRITEBACK_DATA:
3748 if (!jbd2_journal_check_available_features
3749 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3750 ext4_msg(sb, KERN_ERR, "Journal does not support "
3751 "requested data journaling mode");
3752 goto failed_mount_wq;
3757 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3759 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3762 if (ext4_mballoc_ready) {
3763 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3764 if (!sbi->s_mb_cache) {
3765 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3766 goto failed_mount_wq;
3770 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3771 (blocksize != PAGE_CACHE_SIZE)) {
3772 ext4_msg(sb, KERN_ERR,
3773 "Unsupported blocksize for fs encryption");
3774 goto failed_mount_wq;
3777 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3778 !ext4_has_feature_encrypt(sb)) {
3779 ext4_set_feature_encrypt(sb);
3780 ext4_commit_super(sb, 1);
3784 * Get the # of file system overhead blocks from the
3785 * superblock if present.
3787 if (es->s_overhead_clusters)
3788 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3790 err = ext4_calculate_overhead(sb);
3792 goto failed_mount_wq;
3796 * The maximum number of concurrent works can be high and
3797 * concurrency isn't really necessary. Limit it to 1.
3799 EXT4_SB(sb)->rsv_conversion_wq =
3800 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3801 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3802 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3808 * The jbd2_journal_load will have done any necessary log recovery,
3809 * so we can safely mount the rest of the filesystem now.
3812 root = ext4_iget(sb, EXT4_ROOT_INO);
3814 ext4_msg(sb, KERN_ERR, "get root inode failed");
3815 ret = PTR_ERR(root);
3819 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3820 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3824 sb->s_root = d_make_root(root);
3826 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3831 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3832 sb->s_flags |= MS_RDONLY;
3834 /* determine the minimum size of new large inodes, if present */
3835 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3836 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3837 EXT4_GOOD_OLD_INODE_SIZE;
3838 if (ext4_has_feature_extra_isize(sb)) {
3839 if (sbi->s_want_extra_isize <
3840 le16_to_cpu(es->s_want_extra_isize))
3841 sbi->s_want_extra_isize =
3842 le16_to_cpu(es->s_want_extra_isize);
3843 if (sbi->s_want_extra_isize <
3844 le16_to_cpu(es->s_min_extra_isize))
3845 sbi->s_want_extra_isize =
3846 le16_to_cpu(es->s_min_extra_isize);
3849 /* Check if enough inode space is available */
3850 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3851 sbi->s_inode_size) {
3852 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3853 EXT4_GOOD_OLD_INODE_SIZE;
3854 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3858 ext4_set_resv_clusters(sb);
3860 err = ext4_setup_system_zone(sb);
3862 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3864 goto failed_mount4a;
3868 err = ext4_mb_init(sb);
3870 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3875 block = ext4_count_free_clusters(sb);
3876 ext4_free_blocks_count_set(sbi->s_es,
3877 EXT4_C2B(sbi, block));
3878 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3881 unsigned long freei = ext4_count_free_inodes(sb);
3882 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3883 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3887 err = percpu_counter_init(&sbi->s_dirs_counter,
3888 ext4_count_dirs(sb), GFP_KERNEL);
3890 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3893 ext4_msg(sb, KERN_ERR, "insufficient memory");
3897 if (ext4_has_feature_flex_bg(sb))
3898 if (!ext4_fill_flex_info(sb)) {
3899 ext4_msg(sb, KERN_ERR,
3900 "unable to initialize "
3901 "flex_bg meta info!");
3905 err = ext4_register_li_request(sb, first_not_zeroed);
3909 err = ext4_register_sysfs(sb);
3914 /* Enable quota usage during mount. */
3915 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3916 err = ext4_enable_quotas(sb);
3920 #endif /* CONFIG_QUOTA */
3922 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3923 ext4_orphan_cleanup(sb, es);
3924 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3925 if (needs_recovery) {
3926 ext4_msg(sb, KERN_INFO, "recovery complete");
3927 ext4_mark_recovery_complete(sb, es);
3929 if (EXT4_SB(sb)->s_journal) {
3930 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3931 descr = " journalled data mode";
3932 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3933 descr = " ordered data mode";
3935 descr = " writeback data mode";
3937 descr = "out journal";
3939 if (test_opt(sb, DISCARD)) {
3940 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3941 if (!blk_queue_discard(q))
3942 ext4_msg(sb, KERN_WARNING,
3943 "mounting with \"discard\" option, but "
3944 "the device does not support discard");
3947 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3948 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3949 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3950 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3952 if (es->s_error_count)
3953 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3955 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3956 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3957 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3958 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3965 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3970 ext4_unregister_sysfs(sb);
3973 ext4_unregister_li_request(sb);
3975 ext4_mb_release(sb);
3976 if (sbi->s_flex_groups)
3977 kvfree(sbi->s_flex_groups);
3978 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3979 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3980 percpu_counter_destroy(&sbi->s_dirs_counter);
3981 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3983 ext4_ext_release(sb);
3984 ext4_release_system_zone(sb);
3989 ext4_msg(sb, KERN_ERR, "mount failed");
3990 if (EXT4_SB(sb)->rsv_conversion_wq)
3991 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
3993 if (sbi->s_journal) {
3994 jbd2_journal_destroy(sbi->s_journal);
3995 sbi->s_journal = NULL;
3998 ext4_es_unregister_shrinker(sbi);
4000 del_timer_sync(&sbi->s_err_report);
4002 kthread_stop(sbi->s_mmp_tsk);
4004 for (i = 0; i < db_count; i++)
4005 brelse(sbi->s_group_desc[i]);
4006 kvfree(sbi->s_group_desc);
4008 if (sbi->s_chksum_driver)
4009 crypto_free_shash(sbi->s_chksum_driver);
4011 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4012 kfree(sbi->s_qf_names[i]);
4014 ext4_blkdev_remove(sbi);
4017 sb->s_fs_info = NULL;
4018 kfree(sbi->s_blockgroup_lock);
4022 return err ? err : ret;
4026 * Setup any per-fs journal parameters now. We'll do this both on
4027 * initial mount, once the journal has been initialised but before we've
4028 * done any recovery; and again on any subsequent remount.
4030 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4032 struct ext4_sb_info *sbi = EXT4_SB(sb);
4034 journal->j_commit_interval = sbi->s_commit_interval;
4035 journal->j_min_batch_time = sbi->s_min_batch_time;
4036 journal->j_max_batch_time = sbi->s_max_batch_time;
4038 write_lock(&journal->j_state_lock);
4039 if (test_opt(sb, BARRIER))
4040 journal->j_flags |= JBD2_BARRIER;
4042 journal->j_flags &= ~JBD2_BARRIER;
4043 if (test_opt(sb, DATA_ERR_ABORT))
4044 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4046 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4047 write_unlock(&journal->j_state_lock);
4050 static journal_t *ext4_get_journal(struct super_block *sb,
4051 unsigned int journal_inum)
4053 struct inode *journal_inode;
4056 BUG_ON(!ext4_has_feature_journal(sb));
4058 /* First, test for the existence of a valid inode on disk. Bad
4059 * things happen if we iget() an unused inode, as the subsequent
4060 * iput() will try to delete it. */
4062 journal_inode = ext4_iget(sb, journal_inum);
4063 if (IS_ERR(journal_inode)) {
4064 ext4_msg(sb, KERN_ERR, "no journal found");
4067 if (!journal_inode->i_nlink) {
4068 make_bad_inode(journal_inode);
4069 iput(journal_inode);
4070 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4074 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4075 journal_inode, journal_inode->i_size);
4076 if (!S_ISREG(journal_inode->i_mode)) {
4077 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4078 iput(journal_inode);
4082 journal = jbd2_journal_init_inode(journal_inode);
4084 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4085 iput(journal_inode);
4088 journal->j_private = sb;
4089 ext4_init_journal_params(sb, journal);
4093 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4096 struct buffer_head *bh;
4100 int hblock, blocksize;
4101 ext4_fsblk_t sb_block;
4102 unsigned long offset;
4103 struct ext4_super_block *es;
4104 struct block_device *bdev;
4106 BUG_ON(!ext4_has_feature_journal(sb));
4108 bdev = ext4_blkdev_get(j_dev, sb);
4112 blocksize = sb->s_blocksize;
4113 hblock = bdev_logical_block_size(bdev);
4114 if (blocksize < hblock) {
4115 ext4_msg(sb, KERN_ERR,
4116 "blocksize too small for journal device");
4120 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4121 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4122 set_blocksize(bdev, blocksize);
4123 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4124 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4125 "external journal");
4129 es = (struct ext4_super_block *) (bh->b_data + offset);
4130 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4131 !(le32_to_cpu(es->s_feature_incompat) &
4132 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4133 ext4_msg(sb, KERN_ERR, "external journal has "
4139 if ((le32_to_cpu(es->s_feature_ro_compat) &
4140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4141 es->s_checksum != ext4_superblock_csum(sb, es)) {
4142 ext4_msg(sb, KERN_ERR, "external journal has "
4143 "corrupt superblock");
4148 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4149 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4154 len = ext4_blocks_count(es);
4155 start = sb_block + 1;
4156 brelse(bh); /* we're done with the superblock */
4158 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4159 start, len, blocksize);
4161 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4164 journal->j_private = sb;
4165 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4166 wait_on_buffer(journal->j_sb_buffer);
4167 if (!buffer_uptodate(journal->j_sb_buffer)) {
4168 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4171 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4172 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4173 "user (unsupported) - %d",
4174 be32_to_cpu(journal->j_superblock->s_nr_users));
4177 EXT4_SB(sb)->journal_bdev = bdev;
4178 ext4_init_journal_params(sb, journal);
4182 jbd2_journal_destroy(journal);
4184 ext4_blkdev_put(bdev);
4188 static int ext4_load_journal(struct super_block *sb,
4189 struct ext4_super_block *es,
4190 unsigned long journal_devnum)
4193 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4196 int really_read_only;
4198 BUG_ON(!ext4_has_feature_journal(sb));
4200 if (journal_devnum &&
4201 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4202 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4203 "numbers have changed");
4204 journal_dev = new_decode_dev(journal_devnum);
4206 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4208 really_read_only = bdev_read_only(sb->s_bdev);
4211 * Are we loading a blank journal or performing recovery after a
4212 * crash? For recovery, we need to check in advance whether we
4213 * can get read-write access to the device.
4215 if (ext4_has_feature_journal_needs_recovery(sb)) {
4216 if (sb->s_flags & MS_RDONLY) {
4217 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4218 "required on readonly filesystem");
4219 if (really_read_only) {
4220 ext4_msg(sb, KERN_ERR, "write access "
4221 "unavailable, cannot proceed");
4224 ext4_msg(sb, KERN_INFO, "write access will "
4225 "be enabled during recovery");
4229 if (journal_inum && journal_dev) {
4230 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4231 "and inode journals!");
4236 if (!(journal = ext4_get_journal(sb, journal_inum)))
4239 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4243 if (!(journal->j_flags & JBD2_BARRIER))
4244 ext4_msg(sb, KERN_INFO, "barriers disabled");
4246 if (!ext4_has_feature_journal_needs_recovery(sb))
4247 err = jbd2_journal_wipe(journal, !really_read_only);
4249 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4251 memcpy(save, ((char *) es) +
4252 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4253 err = jbd2_journal_load(journal);
4255 memcpy(((char *) es) + EXT4_S_ERR_START,
4256 save, EXT4_S_ERR_LEN);
4261 ext4_msg(sb, KERN_ERR, "error loading journal");
4262 jbd2_journal_destroy(journal);
4266 EXT4_SB(sb)->s_journal = journal;
4267 ext4_clear_journal_err(sb, es);
4269 if (!really_read_only && journal_devnum &&
4270 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4271 es->s_journal_dev = cpu_to_le32(journal_devnum);
4273 /* Make sure we flush the recovery flag to disk. */
4274 ext4_commit_super(sb, 1);
4280 static int ext4_commit_super(struct super_block *sb, int sync)
4282 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4283 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4286 if (!sbh || block_device_ejected(sb))
4288 if (buffer_write_io_error(sbh)) {
4290 * Oh, dear. A previous attempt to write the
4291 * superblock failed. This could happen because the
4292 * USB device was yanked out. Or it could happen to
4293 * be a transient write error and maybe the block will
4294 * be remapped. Nothing we can do but to retry the
4295 * write and hope for the best.
4297 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4298 "superblock detected");
4299 clear_buffer_write_io_error(sbh);
4300 set_buffer_uptodate(sbh);
4303 * If the file system is mounted read-only, don't update the
4304 * superblock write time. This avoids updating the superblock
4305 * write time when we are mounting the root file system
4306 * read/only but we need to replay the journal; at that point,
4307 * for people who are east of GMT and who make their clock
4308 * tick in localtime for Windows bug-for-bug compatibility,
4309 * the clock is set in the future, and this will cause e2fsck
4310 * to complain and force a full file system check.
4312 if (!(sb->s_flags & MS_RDONLY))
4313 es->s_wtime = cpu_to_le32(get_seconds());
4314 if (sb->s_bdev->bd_part)
4315 es->s_kbytes_written =
4316 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4317 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4318 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4320 es->s_kbytes_written =
4321 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4322 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4323 ext4_free_blocks_count_set(es,
4324 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4325 &EXT4_SB(sb)->s_freeclusters_counter)));
4326 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4327 es->s_free_inodes_count =
4328 cpu_to_le32(percpu_counter_sum_positive(
4329 &EXT4_SB(sb)->s_freeinodes_counter));
4330 BUFFER_TRACE(sbh, "marking dirty");
4331 ext4_superblock_csum_set(sb);
4332 mark_buffer_dirty(sbh);
4334 error = __sync_dirty_buffer(sbh,
4335 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4339 error = buffer_write_io_error(sbh);
4341 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4343 clear_buffer_write_io_error(sbh);
4344 set_buffer_uptodate(sbh);
4351 * Have we just finished recovery? If so, and if we are mounting (or
4352 * remounting) the filesystem readonly, then we will end up with a
4353 * consistent fs on disk. Record that fact.
4355 static void ext4_mark_recovery_complete(struct super_block *sb,
4356 struct ext4_super_block *es)
4358 journal_t *journal = EXT4_SB(sb)->s_journal;
4360 if (!ext4_has_feature_journal(sb)) {
4361 BUG_ON(journal != NULL);
4364 jbd2_journal_lock_updates(journal);
4365 if (jbd2_journal_flush(journal) < 0)
4368 if (ext4_has_feature_journal_needs_recovery(sb) &&
4369 sb->s_flags & MS_RDONLY) {
4370 ext4_clear_feature_journal_needs_recovery(sb);
4371 ext4_commit_super(sb, 1);
4375 jbd2_journal_unlock_updates(journal);
4379 * If we are mounting (or read-write remounting) a filesystem whose journal
4380 * has recorded an error from a previous lifetime, move that error to the
4381 * main filesystem now.
4383 static void ext4_clear_journal_err(struct super_block *sb,
4384 struct ext4_super_block *es)
4390 BUG_ON(!ext4_has_feature_journal(sb));
4392 journal = EXT4_SB(sb)->s_journal;
4395 * Now check for any error status which may have been recorded in the
4396 * journal by a prior ext4_error() or ext4_abort()
4399 j_errno = jbd2_journal_errno(journal);
4403 errstr = ext4_decode_error(sb, j_errno, nbuf);
4404 ext4_warning(sb, "Filesystem error recorded "
4405 "from previous mount: %s", errstr);
4406 ext4_warning(sb, "Marking fs in need of filesystem check.");
4408 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4409 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4410 ext4_commit_super(sb, 1);
4412 jbd2_journal_clear_err(journal);
4413 jbd2_journal_update_sb_errno(journal);
4418 * Force the running and committing transactions to commit,
4419 * and wait on the commit.
4421 int ext4_force_commit(struct super_block *sb)
4425 if (sb->s_flags & MS_RDONLY)
4428 journal = EXT4_SB(sb)->s_journal;
4429 return ext4_journal_force_commit(journal);
4432 static int ext4_sync_fs(struct super_block *sb, int wait)
4436 bool needs_barrier = false;
4437 struct ext4_sb_info *sbi = EXT4_SB(sb);
4439 trace_ext4_sync_fs(sb, wait);
4440 flush_workqueue(sbi->rsv_conversion_wq);
4442 * Writeback quota in non-journalled quota case - journalled quota has
4445 dquot_writeback_dquots(sb, -1);
4447 * Data writeback is possible w/o journal transaction, so barrier must
4448 * being sent at the end of the function. But we can skip it if
4449 * transaction_commit will do it for us.
4451 if (sbi->s_journal) {
4452 target = jbd2_get_latest_transaction(sbi->s_journal);
4453 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4454 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4455 needs_barrier = true;
4457 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4459 ret = jbd2_log_wait_commit(sbi->s_journal,
4462 } else if (wait && test_opt(sb, BARRIER))
4463 needs_barrier = true;
4464 if (needs_barrier) {
4466 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4475 * LVM calls this function before a (read-only) snapshot is created. This
4476 * gives us a chance to flush the journal completely and mark the fs clean.
4478 * Note that only this function cannot bring a filesystem to be in a clean
4479 * state independently. It relies on upper layer to stop all data & metadata
4482 static int ext4_freeze(struct super_block *sb)
4487 if (sb->s_flags & MS_RDONLY)
4490 journal = EXT4_SB(sb)->s_journal;
4493 /* Now we set up the journal barrier. */
4494 jbd2_journal_lock_updates(journal);
4497 * Don't clear the needs_recovery flag if we failed to
4498 * flush the journal.
4500 error = jbd2_journal_flush(journal);
4504 /* Journal blocked and flushed, clear needs_recovery flag. */
4505 ext4_clear_feature_journal_needs_recovery(sb);
4508 error = ext4_commit_super(sb, 1);
4511 /* we rely on upper layer to stop further updates */
4512 jbd2_journal_unlock_updates(journal);
4517 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4518 * flag here, even though the filesystem is not technically dirty yet.
4520 static int ext4_unfreeze(struct super_block *sb)
4522 if (sb->s_flags & MS_RDONLY)
4525 if (EXT4_SB(sb)->s_journal) {
4526 /* Reset the needs_recovery flag before the fs is unlocked. */
4527 ext4_set_feature_journal_needs_recovery(sb);
4530 ext4_commit_super(sb, 1);
4535 * Structure to save mount options for ext4_remount's benefit
4537 struct ext4_mount_options {
4538 unsigned long s_mount_opt;
4539 unsigned long s_mount_opt2;
4542 unsigned long s_commit_interval;
4543 u32 s_min_batch_time, s_max_batch_time;
4546 char *s_qf_names[EXT4_MAXQUOTAS];
4550 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4552 struct ext4_super_block *es;
4553 struct ext4_sb_info *sbi = EXT4_SB(sb);
4554 unsigned long old_sb_flags;
4555 struct ext4_mount_options old_opts;
4556 int enable_quota = 0;
4558 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4563 char *orig_data = kstrdup(data, GFP_KERNEL);
4565 /* Store the original options */
4566 old_sb_flags = sb->s_flags;
4567 old_opts.s_mount_opt = sbi->s_mount_opt;
4568 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4569 old_opts.s_resuid = sbi->s_resuid;
4570 old_opts.s_resgid = sbi->s_resgid;
4571 old_opts.s_commit_interval = sbi->s_commit_interval;
4572 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4573 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4575 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4576 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4577 if (sbi->s_qf_names[i]) {
4578 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4580 if (!old_opts.s_qf_names[i]) {
4581 for (j = 0; j < i; j++)
4582 kfree(old_opts.s_qf_names[j]);
4587 old_opts.s_qf_names[i] = NULL;
4589 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4590 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4592 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4597 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4598 test_opt(sb, JOURNAL_CHECKSUM)) {
4599 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4600 "during remount not supported; ignoring");
4601 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4604 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4605 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4606 ext4_msg(sb, KERN_ERR, "can't mount with "
4607 "both data=journal and delalloc");
4611 if (test_opt(sb, DIOREAD_NOLOCK)) {
4612 ext4_msg(sb, KERN_ERR, "can't mount with "
4613 "both data=journal and dioread_nolock");
4617 if (test_opt(sb, DAX)) {
4618 ext4_msg(sb, KERN_ERR, "can't mount with "
4619 "both data=journal and dax");
4625 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4626 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4627 "dax flag with busy inodes while remounting");
4628 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4631 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4632 ext4_abort(sb, "Abort forced by user");
4634 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4635 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4639 if (sbi->s_journal) {
4640 ext4_init_journal_params(sb, sbi->s_journal);
4641 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4644 if (*flags & MS_LAZYTIME)
4645 sb->s_flags |= MS_LAZYTIME;
4647 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4648 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4653 if (*flags & MS_RDONLY) {
4654 err = sync_filesystem(sb);
4657 err = dquot_suspend(sb, -1);
4662 * First of all, the unconditional stuff we have to do
4663 * to disable replay of the journal when we next remount
4665 sb->s_flags |= MS_RDONLY;
4668 * OK, test if we are remounting a valid rw partition
4669 * readonly, and if so set the rdonly flag and then
4670 * mark the partition as valid again.
4672 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4673 (sbi->s_mount_state & EXT4_VALID_FS))
4674 es->s_state = cpu_to_le16(sbi->s_mount_state);
4677 ext4_mark_recovery_complete(sb, es);
4679 /* Make sure we can mount this feature set readwrite */
4680 if (ext4_has_feature_readonly(sb) ||
4681 !ext4_feature_set_ok(sb, 0)) {
4686 * Make sure the group descriptor checksums
4687 * are sane. If they aren't, refuse to remount r/w.
4689 for (g = 0; g < sbi->s_groups_count; g++) {
4690 struct ext4_group_desc *gdp =
4691 ext4_get_group_desc(sb, g, NULL);
4693 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4694 ext4_msg(sb, KERN_ERR,
4695 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4696 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4697 le16_to_cpu(gdp->bg_checksum));
4704 * If we have an unprocessed orphan list hanging
4705 * around from a previously readonly bdev mount,
4706 * require a full umount/remount for now.
4708 if (es->s_last_orphan) {
4709 ext4_msg(sb, KERN_WARNING, "Couldn't "
4710 "remount RDWR because of unprocessed "
4711 "orphan inode list. Please "
4712 "umount/remount instead");
4718 * Mounting a RDONLY partition read-write, so reread
4719 * and store the current valid flag. (It may have
4720 * been changed by e2fsck since we originally mounted
4724 ext4_clear_journal_err(sb, es);
4725 sbi->s_mount_state = le16_to_cpu(es->s_state);
4726 if (!ext4_setup_super(sb, es, 0))
4727 sb->s_flags &= ~MS_RDONLY;
4728 if (ext4_has_feature_mmp(sb))
4729 if (ext4_multi_mount_protect(sb,
4730 le64_to_cpu(es->s_mmp_block))) {
4739 * Reinitialize lazy itable initialization thread based on
4742 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4743 ext4_unregister_li_request(sb);
4745 ext4_group_t first_not_zeroed;
4746 first_not_zeroed = ext4_has_uninit_itable(sb);
4747 ext4_register_li_request(sb, first_not_zeroed);
4750 ext4_setup_system_zone(sb);
4751 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4752 ext4_commit_super(sb, 1);
4755 /* Release old quota file names */
4756 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4757 kfree(old_opts.s_qf_names[i]);
4759 if (sb_any_quota_suspended(sb))
4760 dquot_resume(sb, -1);
4761 else if (ext4_has_feature_quota(sb)) {
4762 err = ext4_enable_quotas(sb);
4769 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4770 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4775 sb->s_flags = old_sb_flags;
4776 sbi->s_mount_opt = old_opts.s_mount_opt;
4777 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4778 sbi->s_resuid = old_opts.s_resuid;
4779 sbi->s_resgid = old_opts.s_resgid;
4780 sbi->s_commit_interval = old_opts.s_commit_interval;
4781 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4782 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4784 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4785 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4786 kfree(sbi->s_qf_names[i]);
4787 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4794 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4796 struct super_block *sb = dentry->d_sb;
4797 struct ext4_sb_info *sbi = EXT4_SB(sb);
4798 struct ext4_super_block *es = sbi->s_es;
4799 ext4_fsblk_t overhead = 0, resv_blocks;
4802 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4804 if (!test_opt(sb, MINIX_DF))
4805 overhead = sbi->s_overhead;
4807 buf->f_type = EXT4_SUPER_MAGIC;
4808 buf->f_bsize = sb->s_blocksize;
4809 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4810 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4811 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4812 /* prevent underflow in case that few free space is available */
4813 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4814 buf->f_bavail = buf->f_bfree -
4815 (ext4_r_blocks_count(es) + resv_blocks);
4816 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4818 buf->f_files = le32_to_cpu(es->s_inodes_count);
4819 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4820 buf->f_namelen = EXT4_NAME_LEN;
4821 fsid = le64_to_cpup((void *)es->s_uuid) ^
4822 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4823 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4824 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4829 /* Helper function for writing quotas on sync - we need to start transaction
4830 * before quota file is locked for write. Otherwise the are possible deadlocks:
4831 * Process 1 Process 2
4832 * ext4_create() quota_sync()
4833 * jbd2_journal_start() write_dquot()
4834 * dquot_initialize() down(dqio_mutex)
4835 * down(dqio_mutex) jbd2_journal_start()
4841 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4843 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4846 static int ext4_write_dquot(struct dquot *dquot)
4850 struct inode *inode;
4852 inode = dquot_to_inode(dquot);
4853 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4854 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4856 return PTR_ERR(handle);
4857 ret = dquot_commit(dquot);
4858 err = ext4_journal_stop(handle);
4864 static int ext4_acquire_dquot(struct dquot *dquot)
4869 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4870 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4872 return PTR_ERR(handle);
4873 ret = dquot_acquire(dquot);
4874 err = ext4_journal_stop(handle);
4880 static int ext4_release_dquot(struct dquot *dquot)
4885 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4886 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4887 if (IS_ERR(handle)) {
4888 /* Release dquot anyway to avoid endless cycle in dqput() */
4889 dquot_release(dquot);
4890 return PTR_ERR(handle);
4892 ret = dquot_release(dquot);
4893 err = ext4_journal_stop(handle);
4899 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4901 struct super_block *sb = dquot->dq_sb;
4902 struct ext4_sb_info *sbi = EXT4_SB(sb);
4904 /* Are we journaling quotas? */
4905 if (ext4_has_feature_quota(sb) ||
4906 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4907 dquot_mark_dquot_dirty(dquot);
4908 return ext4_write_dquot(dquot);
4910 return dquot_mark_dquot_dirty(dquot);
4914 static int ext4_write_info(struct super_block *sb, int type)
4919 /* Data block + inode block */
4920 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
4922 return PTR_ERR(handle);
4923 ret = dquot_commit_info(sb, type);
4924 err = ext4_journal_stop(handle);
4931 * Turn on quotas during mount time - we need to find
4932 * the quota file and such...
4934 static int ext4_quota_on_mount(struct super_block *sb, int type)
4936 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4937 EXT4_SB(sb)->s_jquota_fmt, type);
4940 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
4942 struct ext4_inode_info *ei = EXT4_I(inode);
4944 /* The first argument of lockdep_set_subclass has to be
4945 * *exactly* the same as the argument to init_rwsem() --- in
4946 * this case, in init_once() --- or lockdep gets unhappy
4947 * because the name of the lock is set using the
4948 * stringification of the argument to init_rwsem().
4950 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
4951 lockdep_set_subclass(&ei->i_data_sem, subclass);
4955 * Standard function to be called on quota_on
4957 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4962 if (!test_opt(sb, QUOTA))
4965 /* Quotafile not on the same filesystem? */
4966 if (path->dentry->d_sb != sb)
4968 /* Journaling quota? */
4969 if (EXT4_SB(sb)->s_qf_names[type]) {
4970 /* Quotafile not in fs root? */
4971 if (path->dentry->d_parent != sb->s_root)
4972 ext4_msg(sb, KERN_WARNING,
4973 "Quota file not on filesystem root. "
4974 "Journaled quota will not work");
4978 * When we journal data on quota file, we have to flush journal to see
4979 * all updates to the file when we bypass pagecache...
4981 if (EXT4_SB(sb)->s_journal &&
4982 ext4_should_journal_data(d_inode(path->dentry))) {
4984 * We don't need to lock updates but journal_flush() could
4985 * otherwise be livelocked...
4987 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4988 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4989 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4993 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
4994 err = dquot_quota_on(sb, type, format_id, path);
4996 lockdep_set_quota_inode(path->dentry->d_inode,
5001 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5005 struct inode *qf_inode;
5006 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5007 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5008 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5011 BUG_ON(!ext4_has_feature_quota(sb));
5013 if (!qf_inums[type])
5016 qf_inode = ext4_iget(sb, qf_inums[type]);
5017 if (IS_ERR(qf_inode)) {
5018 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5019 return PTR_ERR(qf_inode);
5022 /* Don't account quota for quota files to avoid recursion */
5023 qf_inode->i_flags |= S_NOQUOTA;
5024 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5025 err = dquot_enable(qf_inode, type, format_id, flags);
5028 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5033 /* Enable usage tracking for all quota types. */
5034 static int ext4_enable_quotas(struct super_block *sb)
5037 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5038 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5039 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5042 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5043 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5044 if (qf_inums[type]) {
5045 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5046 DQUOT_USAGE_ENABLED);
5049 "Failed to enable quota tracking "
5050 "(type=%d, err=%d). Please run "
5051 "e2fsck to fix.", type, err);
5059 static int ext4_quota_off(struct super_block *sb, int type)
5061 struct inode *inode = sb_dqopt(sb)->files[type];
5064 /* Force all delayed allocation blocks to be allocated.
5065 * Caller already holds s_umount sem */
5066 if (test_opt(sb, DELALLOC))
5067 sync_filesystem(sb);
5072 /* Update modification times of quota files when userspace can
5073 * start looking at them */
5074 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5077 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5078 ext4_mark_inode_dirty(handle, inode);
5079 ext4_journal_stop(handle);
5082 return dquot_quota_off(sb, type);
5085 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5086 * acquiring the locks... As quota files are never truncated and quota code
5087 * itself serializes the operations (and no one else should touch the files)
5088 * we don't have to be afraid of races */
5089 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5090 size_t len, loff_t off)
5092 struct inode *inode = sb_dqopt(sb)->files[type];
5093 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5094 int offset = off & (sb->s_blocksize - 1);
5097 struct buffer_head *bh;
5098 loff_t i_size = i_size_read(inode);
5102 if (off+len > i_size)
5105 while (toread > 0) {
5106 tocopy = sb->s_blocksize - offset < toread ?
5107 sb->s_blocksize - offset : toread;
5108 bh = ext4_bread(NULL, inode, blk, 0);
5111 if (!bh) /* A hole? */
5112 memset(data, 0, tocopy);
5114 memcpy(data, bh->b_data+offset, tocopy);
5124 /* Write to quotafile (we know the transaction is already started and has
5125 * enough credits) */
5126 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5127 const char *data, size_t len, loff_t off)
5129 struct inode *inode = sb_dqopt(sb)->files[type];
5130 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5131 int err, offset = off & (sb->s_blocksize - 1);
5133 struct buffer_head *bh;
5134 handle_t *handle = journal_current_handle();
5136 if (EXT4_SB(sb)->s_journal && !handle) {
5137 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5138 " cancelled because transaction is not started",
5139 (unsigned long long)off, (unsigned long long)len);
5143 * Since we account only one data block in transaction credits,
5144 * then it is impossible to cross a block boundary.
5146 if (sb->s_blocksize - offset < len) {
5147 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5148 " cancelled because not block aligned",
5149 (unsigned long long)off, (unsigned long long)len);
5154 bh = ext4_bread(handle, inode, blk,
5155 EXT4_GET_BLOCKS_CREATE |
5156 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5157 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5158 ext4_should_retry_alloc(inode->i_sb, &retries));
5163 BUFFER_TRACE(bh, "get write access");
5164 err = ext4_journal_get_write_access(handle, bh);
5170 memcpy(bh->b_data+offset, data, len);
5171 flush_dcache_page(bh->b_page);
5173 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5176 if (inode->i_size < off + len) {
5177 i_size_write(inode, off + len);
5178 EXT4_I(inode)->i_disksize = inode->i_size;
5179 ext4_mark_inode_dirty(handle, inode);
5186 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5187 const char *dev_name, void *data)
5189 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5192 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5193 static inline void register_as_ext2(void)
5195 int err = register_filesystem(&ext2_fs_type);
5198 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5201 static inline void unregister_as_ext2(void)
5203 unregister_filesystem(&ext2_fs_type);
5206 static inline int ext2_feature_set_ok(struct super_block *sb)
5208 if (ext4_has_unknown_ext2_incompat_features(sb))
5210 if (sb->s_flags & MS_RDONLY)
5212 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5217 static inline void register_as_ext2(void) { }
5218 static inline void unregister_as_ext2(void) { }
5219 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5222 static inline void register_as_ext3(void)
5224 int err = register_filesystem(&ext3_fs_type);
5227 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5230 static inline void unregister_as_ext3(void)
5232 unregister_filesystem(&ext3_fs_type);
5235 static inline int ext3_feature_set_ok(struct super_block *sb)
5237 if (ext4_has_unknown_ext3_incompat_features(sb))
5239 if (!ext4_has_feature_journal(sb))
5241 if (sb->s_flags & MS_RDONLY)
5243 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5248 static struct file_system_type ext4_fs_type = {
5249 .owner = THIS_MODULE,
5251 .mount = ext4_mount,
5252 .kill_sb = kill_block_super,
5253 .fs_flags = FS_REQUIRES_DEV,
5255 MODULE_ALIAS_FS("ext4");
5257 /* Shared across all ext4 file systems */
5258 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5259 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5261 static int __init ext4_init_fs(void)
5265 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5266 ext4_li_info = NULL;
5267 mutex_init(&ext4_li_mtx);
5269 /* Build-time check for flags consistency */
5270 ext4_check_flag_values();
5272 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5273 mutex_init(&ext4__aio_mutex[i]);
5274 init_waitqueue_head(&ext4__ioend_wq[i]);
5277 err = ext4_init_es();
5281 err = ext4_init_pageio();
5285 err = ext4_init_system_zone();
5289 err = ext4_init_sysfs();
5293 err = ext4_init_mballoc();
5297 ext4_mballoc_ready = 1;
5298 err = init_inodecache();
5303 err = register_filesystem(&ext4_fs_type);
5309 unregister_as_ext2();
5310 unregister_as_ext3();
5311 destroy_inodecache();
5313 ext4_mballoc_ready = 0;
5314 ext4_exit_mballoc();
5318 ext4_exit_system_zone();
5327 static void __exit ext4_exit_fs(void)
5330 ext4_destroy_lazyinit_thread();
5331 unregister_as_ext2();
5332 unregister_as_ext3();
5333 unregister_filesystem(&ext4_fs_type);
5334 destroy_inodecache();
5335 ext4_exit_mballoc();
5337 ext4_exit_system_zone();
5342 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5343 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5344 MODULE_LICENSE("GPL");
5345 module_init(ext4_init_fs)
5346 module_exit(ext4_exit_fs)