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 inode_init_once(&ei->vfs_inode);
964 static int __init init_inodecache(void)
966 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
967 sizeof(struct ext4_inode_info),
968 0, (SLAB_RECLAIM_ACCOUNT|
971 if (ext4_inode_cachep == NULL)
976 static void destroy_inodecache(void)
979 * Make sure all delayed rcu free inodes are flushed before we
983 kmem_cache_destroy(ext4_inode_cachep);
986 void ext4_clear_inode(struct inode *inode)
988 invalidate_inode_buffers(inode);
991 ext4_discard_preallocations(inode);
992 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
993 if (EXT4_I(inode)->jinode) {
994 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
995 EXT4_I(inode)->jinode);
996 jbd2_free_inode(EXT4_I(inode)->jinode);
997 EXT4_I(inode)->jinode = NULL;
999 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1000 if (EXT4_I(inode)->i_crypt_info)
1001 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1005 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1006 u64 ino, u32 generation)
1008 struct inode *inode;
1010 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1011 return ERR_PTR(-ESTALE);
1012 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1013 return ERR_PTR(-ESTALE);
1015 /* iget isn't really right if the inode is currently unallocated!!
1017 * ext4_read_inode will return a bad_inode if the inode had been
1018 * deleted, so we should be safe.
1020 * Currently we don't know the generation for parent directory, so
1021 * a generation of 0 means "accept any"
1023 inode = ext4_iget_normal(sb, ino);
1025 return ERR_CAST(inode);
1026 if (generation && inode->i_generation != generation) {
1028 return ERR_PTR(-ESTALE);
1034 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1035 int fh_len, int fh_type)
1037 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1038 ext4_nfs_get_inode);
1041 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1042 int fh_len, int fh_type)
1044 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1045 ext4_nfs_get_inode);
1049 * Try to release metadata pages (indirect blocks, directories) which are
1050 * mapped via the block device. Since these pages could have journal heads
1051 * which would prevent try_to_free_buffers() from freeing them, we must use
1052 * jbd2 layer's try_to_free_buffers() function to release them.
1054 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1057 journal_t *journal = EXT4_SB(sb)->s_journal;
1059 WARN_ON(PageChecked(page));
1060 if (!page_has_buffers(page))
1063 return jbd2_journal_try_to_free_buffers(journal, page,
1064 wait & ~__GFP_DIRECT_RECLAIM);
1065 return try_to_free_buffers(page);
1069 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1070 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1072 static int ext4_write_dquot(struct dquot *dquot);
1073 static int ext4_acquire_dquot(struct dquot *dquot);
1074 static int ext4_release_dquot(struct dquot *dquot);
1075 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1076 static int ext4_write_info(struct super_block *sb, int type);
1077 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1079 static int ext4_quota_off(struct super_block *sb, int type);
1080 static int ext4_quota_on_mount(struct super_block *sb, int type);
1081 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1082 size_t len, loff_t off);
1083 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1084 const char *data, size_t len, loff_t off);
1085 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1086 unsigned int flags);
1087 static int ext4_enable_quotas(struct super_block *sb);
1089 static struct dquot **ext4_get_dquots(struct inode *inode)
1091 return EXT4_I(inode)->i_dquot;
1094 static const struct dquot_operations ext4_quota_operations = {
1095 .get_reserved_space = ext4_get_reserved_space,
1096 .write_dquot = ext4_write_dquot,
1097 .acquire_dquot = ext4_acquire_dquot,
1098 .release_dquot = ext4_release_dquot,
1099 .mark_dirty = ext4_mark_dquot_dirty,
1100 .write_info = ext4_write_info,
1101 .alloc_dquot = dquot_alloc,
1102 .destroy_dquot = dquot_destroy,
1105 static const struct quotactl_ops ext4_qctl_operations = {
1106 .quota_on = ext4_quota_on,
1107 .quota_off = ext4_quota_off,
1108 .quota_sync = dquot_quota_sync,
1109 .get_state = dquot_get_state,
1110 .set_info = dquot_set_dqinfo,
1111 .get_dqblk = dquot_get_dqblk,
1112 .set_dqblk = dquot_set_dqblk
1116 static const struct super_operations ext4_sops = {
1117 .alloc_inode = ext4_alloc_inode,
1118 .destroy_inode = ext4_destroy_inode,
1119 .write_inode = ext4_write_inode,
1120 .dirty_inode = ext4_dirty_inode,
1121 .drop_inode = ext4_drop_inode,
1122 .evict_inode = ext4_evict_inode,
1123 .put_super = ext4_put_super,
1124 .sync_fs = ext4_sync_fs,
1125 .freeze_fs = ext4_freeze,
1126 .unfreeze_fs = ext4_unfreeze,
1127 .statfs = ext4_statfs,
1128 .remount_fs = ext4_remount,
1129 .show_options = ext4_show_options,
1131 .quota_read = ext4_quota_read,
1132 .quota_write = ext4_quota_write,
1133 .get_dquots = ext4_get_dquots,
1135 .bdev_try_to_free_page = bdev_try_to_free_page,
1138 static const struct export_operations ext4_export_ops = {
1139 .fh_to_dentry = ext4_fh_to_dentry,
1140 .fh_to_parent = ext4_fh_to_parent,
1141 .get_parent = ext4_get_parent,
1145 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1146 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1147 Opt_nouid32, Opt_debug, Opt_removed,
1148 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1149 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1150 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1151 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1152 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1153 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1154 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1155 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1156 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1157 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1158 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1159 Opt_lazytime, Opt_nolazytime,
1160 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1161 Opt_inode_readahead_blks, Opt_journal_ioprio,
1162 Opt_dioread_nolock, Opt_dioread_lock,
1163 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1164 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1167 static const match_table_t tokens = {
1168 {Opt_bsd_df, "bsddf"},
1169 {Opt_minix_df, "minixdf"},
1170 {Opt_grpid, "grpid"},
1171 {Opt_grpid, "bsdgroups"},
1172 {Opt_nogrpid, "nogrpid"},
1173 {Opt_nogrpid, "sysvgroups"},
1174 {Opt_resgid, "resgid=%u"},
1175 {Opt_resuid, "resuid=%u"},
1177 {Opt_err_cont, "errors=continue"},
1178 {Opt_err_panic, "errors=panic"},
1179 {Opt_err_ro, "errors=remount-ro"},
1180 {Opt_nouid32, "nouid32"},
1181 {Opt_debug, "debug"},
1182 {Opt_removed, "oldalloc"},
1183 {Opt_removed, "orlov"},
1184 {Opt_user_xattr, "user_xattr"},
1185 {Opt_nouser_xattr, "nouser_xattr"},
1187 {Opt_noacl, "noacl"},
1188 {Opt_noload, "norecovery"},
1189 {Opt_noload, "noload"},
1190 {Opt_removed, "nobh"},
1191 {Opt_removed, "bh"},
1192 {Opt_commit, "commit=%u"},
1193 {Opt_min_batch_time, "min_batch_time=%u"},
1194 {Opt_max_batch_time, "max_batch_time=%u"},
1195 {Opt_journal_dev, "journal_dev=%u"},
1196 {Opt_journal_path, "journal_path=%s"},
1197 {Opt_journal_checksum, "journal_checksum"},
1198 {Opt_nojournal_checksum, "nojournal_checksum"},
1199 {Opt_journal_async_commit, "journal_async_commit"},
1200 {Opt_abort, "abort"},
1201 {Opt_data_journal, "data=journal"},
1202 {Opt_data_ordered, "data=ordered"},
1203 {Opt_data_writeback, "data=writeback"},
1204 {Opt_data_err_abort, "data_err=abort"},
1205 {Opt_data_err_ignore, "data_err=ignore"},
1206 {Opt_offusrjquota, "usrjquota="},
1207 {Opt_usrjquota, "usrjquota=%s"},
1208 {Opt_offgrpjquota, "grpjquota="},
1209 {Opt_grpjquota, "grpjquota=%s"},
1210 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1211 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1212 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1213 {Opt_grpquota, "grpquota"},
1214 {Opt_noquota, "noquota"},
1215 {Opt_quota, "quota"},
1216 {Opt_usrquota, "usrquota"},
1217 {Opt_barrier, "barrier=%u"},
1218 {Opt_barrier, "barrier"},
1219 {Opt_nobarrier, "nobarrier"},
1220 {Opt_i_version, "i_version"},
1222 {Opt_stripe, "stripe=%u"},
1223 {Opt_delalloc, "delalloc"},
1224 {Opt_lazytime, "lazytime"},
1225 {Opt_nolazytime, "nolazytime"},
1226 {Opt_nodelalloc, "nodelalloc"},
1227 {Opt_removed, "mblk_io_submit"},
1228 {Opt_removed, "nomblk_io_submit"},
1229 {Opt_block_validity, "block_validity"},
1230 {Opt_noblock_validity, "noblock_validity"},
1231 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1232 {Opt_journal_ioprio, "journal_ioprio=%u"},
1233 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1234 {Opt_auto_da_alloc, "auto_da_alloc"},
1235 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1236 {Opt_dioread_nolock, "dioread_nolock"},
1237 {Opt_dioread_lock, "dioread_lock"},
1238 {Opt_discard, "discard"},
1239 {Opt_nodiscard, "nodiscard"},
1240 {Opt_init_itable, "init_itable=%u"},
1241 {Opt_init_itable, "init_itable"},
1242 {Opt_noinit_itable, "noinit_itable"},
1243 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1244 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1245 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1246 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1247 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1248 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1249 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1253 static ext4_fsblk_t get_sb_block(void **data)
1255 ext4_fsblk_t sb_block;
1256 char *options = (char *) *data;
1258 if (!options || strncmp(options, "sb=", 3) != 0)
1259 return 1; /* Default location */
1262 /* TODO: use simple_strtoll with >32bit ext4 */
1263 sb_block = simple_strtoul(options, &options, 0);
1264 if (*options && *options != ',') {
1265 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1269 if (*options == ',')
1271 *data = (void *) options;
1276 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1277 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1278 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1281 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1283 struct ext4_sb_info *sbi = EXT4_SB(sb);
1287 if (sb_any_quota_loaded(sb) &&
1288 !sbi->s_qf_names[qtype]) {
1289 ext4_msg(sb, KERN_ERR,
1290 "Cannot change journaled "
1291 "quota options when quota turned on");
1294 if (ext4_has_feature_quota(sb)) {
1295 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1296 "when QUOTA feature is enabled");
1299 qname = match_strdup(args);
1301 ext4_msg(sb, KERN_ERR,
1302 "Not enough memory for storing quotafile name");
1305 if (sbi->s_qf_names[qtype]) {
1306 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1309 ext4_msg(sb, KERN_ERR,
1310 "%s quota file already specified",
1314 if (strchr(qname, '/')) {
1315 ext4_msg(sb, KERN_ERR,
1316 "quotafile must be on filesystem root");
1319 sbi->s_qf_names[qtype] = qname;
1327 static int clear_qf_name(struct super_block *sb, int qtype)
1330 struct ext4_sb_info *sbi = EXT4_SB(sb);
1332 if (sb_any_quota_loaded(sb) &&
1333 sbi->s_qf_names[qtype]) {
1334 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1335 " when quota turned on");
1338 kfree(sbi->s_qf_names[qtype]);
1339 sbi->s_qf_names[qtype] = NULL;
1344 #define MOPT_SET 0x0001
1345 #define MOPT_CLEAR 0x0002
1346 #define MOPT_NOSUPPORT 0x0004
1347 #define MOPT_EXPLICIT 0x0008
1348 #define MOPT_CLEAR_ERR 0x0010
1349 #define MOPT_GTE0 0x0020
1352 #define MOPT_QFMT 0x0040
1354 #define MOPT_Q MOPT_NOSUPPORT
1355 #define MOPT_QFMT MOPT_NOSUPPORT
1357 #define MOPT_DATAJ 0x0080
1358 #define MOPT_NO_EXT2 0x0100
1359 #define MOPT_NO_EXT3 0x0200
1360 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1361 #define MOPT_STRING 0x0400
1363 static const struct mount_opts {
1367 } ext4_mount_opts[] = {
1368 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1369 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1370 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1371 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1372 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1373 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1374 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1375 MOPT_EXT4_ONLY | MOPT_SET},
1376 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_CLEAR},
1378 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1379 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1380 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1381 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1382 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_CLEAR},
1384 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1410 {Opt_stripe, 0, MOPT_GTE0},
1411 {Opt_resuid, 0, MOPT_GTE0},
1412 {Opt_resgid, 0, MOPT_GTE0},
1413 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1414 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1415 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1416 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1418 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1419 MOPT_NO_EXT2 | MOPT_DATAJ},
1420 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1421 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1423 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1424 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1426 {Opt_acl, 0, MOPT_NOSUPPORT},
1427 {Opt_noacl, 0, MOPT_NOSUPPORT},
1429 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1430 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1431 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1432 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1434 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1436 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1437 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1438 {Opt_usrjquota, 0, MOPT_Q},
1439 {Opt_grpjquota, 0, MOPT_Q},
1440 {Opt_offusrjquota, 0, MOPT_Q},
1441 {Opt_offgrpjquota, 0, MOPT_Q},
1442 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1444 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1445 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1446 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1450 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1451 substring_t *args, unsigned long *journal_devnum,
1452 unsigned int *journal_ioprio, int is_remount)
1454 struct ext4_sb_info *sbi = EXT4_SB(sb);
1455 const struct mount_opts *m;
1461 if (token == Opt_usrjquota)
1462 return set_qf_name(sb, USRQUOTA, &args[0]);
1463 else if (token == Opt_grpjquota)
1464 return set_qf_name(sb, GRPQUOTA, &args[0]);
1465 else if (token == Opt_offusrjquota)
1466 return clear_qf_name(sb, USRQUOTA);
1467 else if (token == Opt_offgrpjquota)
1468 return clear_qf_name(sb, GRPQUOTA);
1472 case Opt_nouser_xattr:
1473 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1476 return 1; /* handled by get_sb_block() */
1478 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1481 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1484 sb->s_flags |= MS_I_VERSION;
1487 sb->s_flags |= MS_LAZYTIME;
1489 case Opt_nolazytime:
1490 sb->s_flags &= ~MS_LAZYTIME;
1494 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1495 if (token == m->token)
1498 if (m->token == Opt_err) {
1499 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1500 "or missing value", opt);
1504 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1505 ext4_msg(sb, KERN_ERR,
1506 "Mount option \"%s\" incompatible with ext2", opt);
1509 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1510 ext4_msg(sb, KERN_ERR,
1511 "Mount option \"%s\" incompatible with ext3", opt);
1515 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1517 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1519 if (m->flags & MOPT_EXPLICIT) {
1520 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1521 set_opt2(sb, EXPLICIT_DELALLOC);
1522 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1523 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1527 if (m->flags & MOPT_CLEAR_ERR)
1528 clear_opt(sb, ERRORS_MASK);
1529 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1530 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1531 "options when quota turned on");
1535 if (m->flags & MOPT_NOSUPPORT) {
1536 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1537 } else if (token == Opt_commit) {
1539 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1540 sbi->s_commit_interval = HZ * arg;
1541 } else if (token == Opt_max_batch_time) {
1542 sbi->s_max_batch_time = arg;
1543 } else if (token == Opt_min_batch_time) {
1544 sbi->s_min_batch_time = arg;
1545 } else if (token == Opt_inode_readahead_blks) {
1546 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1547 ext4_msg(sb, KERN_ERR,
1548 "EXT4-fs: inode_readahead_blks must be "
1549 "0 or a power of 2 smaller than 2^31");
1552 sbi->s_inode_readahead_blks = arg;
1553 } else if (token == Opt_init_itable) {
1554 set_opt(sb, INIT_INODE_TABLE);
1556 arg = EXT4_DEF_LI_WAIT_MULT;
1557 sbi->s_li_wait_mult = arg;
1558 } else if (token == Opt_max_dir_size_kb) {
1559 sbi->s_max_dir_size_kb = arg;
1560 } else if (token == Opt_stripe) {
1561 sbi->s_stripe = arg;
1562 } else if (token == Opt_resuid) {
1563 uid = make_kuid(current_user_ns(), arg);
1564 if (!uid_valid(uid)) {
1565 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1568 sbi->s_resuid = uid;
1569 } else if (token == Opt_resgid) {
1570 gid = make_kgid(current_user_ns(), arg);
1571 if (!gid_valid(gid)) {
1572 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1575 sbi->s_resgid = gid;
1576 } else if (token == Opt_journal_dev) {
1578 ext4_msg(sb, KERN_ERR,
1579 "Cannot specify journal on remount");
1582 *journal_devnum = arg;
1583 } else if (token == Opt_journal_path) {
1585 struct inode *journal_inode;
1590 ext4_msg(sb, KERN_ERR,
1591 "Cannot specify journal on remount");
1594 journal_path = match_strdup(&args[0]);
1595 if (!journal_path) {
1596 ext4_msg(sb, KERN_ERR, "error: could not dup "
1597 "journal device string");
1601 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1603 ext4_msg(sb, KERN_ERR, "error: could not find "
1604 "journal device path: error %d", error);
1605 kfree(journal_path);
1609 journal_inode = d_inode(path.dentry);
1610 if (!S_ISBLK(journal_inode->i_mode)) {
1611 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1612 "is not a block device", journal_path);
1614 kfree(journal_path);
1618 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1620 kfree(journal_path);
1621 } else if (token == Opt_journal_ioprio) {
1623 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1628 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1629 } else if (token == Opt_test_dummy_encryption) {
1630 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1631 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1632 ext4_msg(sb, KERN_WARNING,
1633 "Test dummy encryption mode enabled");
1635 ext4_msg(sb, KERN_WARNING,
1636 "Test dummy encryption mount option ignored");
1638 } else if (m->flags & MOPT_DATAJ) {
1640 if (!sbi->s_journal)
1641 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1642 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1643 ext4_msg(sb, KERN_ERR,
1644 "Cannot change data mode on remount");
1648 clear_opt(sb, DATA_FLAGS);
1649 sbi->s_mount_opt |= m->mount_opt;
1652 } else if (m->flags & MOPT_QFMT) {
1653 if (sb_any_quota_loaded(sb) &&
1654 sbi->s_jquota_fmt != m->mount_opt) {
1655 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1656 "quota options when quota turned on");
1659 if (ext4_has_feature_quota(sb)) {
1660 ext4_msg(sb, KERN_ERR,
1661 "Cannot set journaled quota options "
1662 "when QUOTA feature is enabled");
1665 sbi->s_jquota_fmt = m->mount_opt;
1667 #ifndef CONFIG_FS_DAX
1668 } else if (token == Opt_dax) {
1669 ext4_msg(sb, KERN_INFO, "dax option not supported");
1675 if (m->flags & MOPT_CLEAR)
1677 else if (unlikely(!(m->flags & MOPT_SET))) {
1678 ext4_msg(sb, KERN_WARNING,
1679 "buggy handling of option %s", opt);
1684 sbi->s_mount_opt |= m->mount_opt;
1686 sbi->s_mount_opt &= ~m->mount_opt;
1691 static int parse_options(char *options, struct super_block *sb,
1692 unsigned long *journal_devnum,
1693 unsigned int *journal_ioprio,
1696 struct ext4_sb_info *sbi = EXT4_SB(sb);
1698 substring_t args[MAX_OPT_ARGS];
1704 while ((p = strsep(&options, ",")) != NULL) {
1708 * Initialize args struct so we know whether arg was
1709 * found; some options take optional arguments.
1711 args[0].to = args[0].from = NULL;
1712 token = match_token(p, tokens, args);
1713 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1714 journal_ioprio, is_remount) < 0)
1718 if (ext4_has_feature_quota(sb) &&
1719 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1720 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1721 "feature is enabled");
1724 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1725 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1726 clear_opt(sb, USRQUOTA);
1728 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1729 clear_opt(sb, GRPQUOTA);
1731 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1732 ext4_msg(sb, KERN_ERR, "old and new quota "
1737 if (!sbi->s_jquota_fmt) {
1738 ext4_msg(sb, KERN_ERR, "journaled quota format "
1744 if (test_opt(sb, DIOREAD_NOLOCK)) {
1746 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1748 if (blocksize < PAGE_CACHE_SIZE) {
1749 ext4_msg(sb, KERN_ERR, "can't mount with "
1750 "dioread_nolock if block size != PAGE_SIZE");
1754 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1755 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1756 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1757 "in data=ordered mode");
1763 static inline void ext4_show_quota_options(struct seq_file *seq,
1764 struct super_block *sb)
1766 #if defined(CONFIG_QUOTA)
1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 if (sbi->s_jquota_fmt) {
1772 switch (sbi->s_jquota_fmt) {
1783 seq_printf(seq, ",jqfmt=%s", fmtname);
1786 if (sbi->s_qf_names[USRQUOTA])
1787 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1789 if (sbi->s_qf_names[GRPQUOTA])
1790 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1794 static const char *token2str(int token)
1796 const struct match_token *t;
1798 for (t = tokens; t->token != Opt_err; t++)
1799 if (t->token == token && !strchr(t->pattern, '='))
1806 * - it's set to a non-default value OR
1807 * - if the per-sb default is different from the global default
1809 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1812 struct ext4_sb_info *sbi = EXT4_SB(sb);
1813 struct ext4_super_block *es = sbi->s_es;
1814 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1815 const struct mount_opts *m;
1816 char sep = nodefs ? '\n' : ',';
1818 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1819 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1821 if (sbi->s_sb_block != 1)
1822 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1824 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1825 int want_set = m->flags & MOPT_SET;
1826 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1827 (m->flags & MOPT_CLEAR_ERR))
1829 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1830 continue; /* skip if same as the default */
1832 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1833 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1834 continue; /* select Opt_noFoo vs Opt_Foo */
1835 SEQ_OPTS_PRINT("%s", token2str(m->token));
1838 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1839 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1840 SEQ_OPTS_PRINT("resuid=%u",
1841 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1842 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1843 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1844 SEQ_OPTS_PRINT("resgid=%u",
1845 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1846 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1847 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1848 SEQ_OPTS_PUTS("errors=remount-ro");
1849 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1850 SEQ_OPTS_PUTS("errors=continue");
1851 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1852 SEQ_OPTS_PUTS("errors=panic");
1853 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1854 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1855 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1856 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1857 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1858 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1859 if (sb->s_flags & MS_I_VERSION)
1860 SEQ_OPTS_PUTS("i_version");
1861 if (nodefs || sbi->s_stripe)
1862 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1863 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1865 SEQ_OPTS_PUTS("data=journal");
1866 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1867 SEQ_OPTS_PUTS("data=ordered");
1868 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1869 SEQ_OPTS_PUTS("data=writeback");
1872 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1873 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1874 sbi->s_inode_readahead_blks);
1876 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1877 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1878 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1879 if (nodefs || sbi->s_max_dir_size_kb)
1880 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1882 ext4_show_quota_options(seq, sb);
1886 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1888 return _ext4_show_options(seq, root->d_sb, 0);
1891 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1893 struct super_block *sb = seq->private;
1896 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1897 rc = _ext4_show_options(seq, sb, 1);
1898 seq_puts(seq, "\n");
1902 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1905 struct ext4_sb_info *sbi = EXT4_SB(sb);
1908 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1909 ext4_msg(sb, KERN_ERR, "revision level too high, "
1910 "forcing read-only mode");
1915 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1916 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1917 "running e2fsck is recommended");
1918 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1919 ext4_msg(sb, KERN_WARNING,
1920 "warning: mounting fs with errors, "
1921 "running e2fsck is recommended");
1922 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1923 le16_to_cpu(es->s_mnt_count) >=
1924 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1925 ext4_msg(sb, KERN_WARNING,
1926 "warning: maximal mount count reached, "
1927 "running e2fsck is recommended");
1928 else if (le32_to_cpu(es->s_checkinterval) &&
1929 (le32_to_cpu(es->s_lastcheck) +
1930 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1931 ext4_msg(sb, KERN_WARNING,
1932 "warning: checktime reached, "
1933 "running e2fsck is recommended");
1934 if (!sbi->s_journal)
1935 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1936 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1937 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1938 le16_add_cpu(&es->s_mnt_count, 1);
1939 es->s_mtime = cpu_to_le32(get_seconds());
1940 ext4_update_dynamic_rev(sb);
1942 ext4_set_feature_journal_needs_recovery(sb);
1944 ext4_commit_super(sb, 1);
1946 if (test_opt(sb, DEBUG))
1947 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1948 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1950 sbi->s_groups_count,
1951 EXT4_BLOCKS_PER_GROUP(sb),
1952 EXT4_INODES_PER_GROUP(sb),
1953 sbi->s_mount_opt, sbi->s_mount_opt2);
1955 cleancache_init_fs(sb);
1959 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1961 struct ext4_sb_info *sbi = EXT4_SB(sb);
1962 struct flex_groups *new_groups;
1965 if (!sbi->s_log_groups_per_flex)
1968 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1969 if (size <= sbi->s_flex_groups_allocated)
1972 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1973 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1975 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1976 size / (int) sizeof(struct flex_groups));
1980 if (sbi->s_flex_groups) {
1981 memcpy(new_groups, sbi->s_flex_groups,
1982 (sbi->s_flex_groups_allocated *
1983 sizeof(struct flex_groups)));
1984 kvfree(sbi->s_flex_groups);
1986 sbi->s_flex_groups = new_groups;
1987 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1991 static int ext4_fill_flex_info(struct super_block *sb)
1993 struct ext4_sb_info *sbi = EXT4_SB(sb);
1994 struct ext4_group_desc *gdp = NULL;
1995 ext4_group_t flex_group;
1998 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1999 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2000 sbi->s_log_groups_per_flex = 0;
2004 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2008 for (i = 0; i < sbi->s_groups_count; i++) {
2009 gdp = ext4_get_group_desc(sb, i, NULL);
2011 flex_group = ext4_flex_group(sbi, i);
2012 atomic_add(ext4_free_inodes_count(sb, gdp),
2013 &sbi->s_flex_groups[flex_group].free_inodes);
2014 atomic64_add(ext4_free_group_clusters(sb, gdp),
2015 &sbi->s_flex_groups[flex_group].free_clusters);
2016 atomic_add(ext4_used_dirs_count(sb, gdp),
2017 &sbi->s_flex_groups[flex_group].used_dirs);
2025 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2026 struct ext4_group_desc *gdp)
2030 __le32 le_group = cpu_to_le32(block_group);
2031 struct ext4_sb_info *sbi = EXT4_SB(sb);
2033 if (ext4_has_metadata_csum(sbi->s_sb)) {
2034 /* Use new metadata_csum algorithm */
2038 save_csum = gdp->bg_checksum;
2039 gdp->bg_checksum = 0;
2040 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2042 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2044 gdp->bg_checksum = save_csum;
2046 crc = csum32 & 0xFFFF;
2050 /* old crc16 code */
2051 if (!ext4_has_feature_gdt_csum(sb))
2054 offset = offsetof(struct ext4_group_desc, bg_checksum);
2056 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2057 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2058 crc = crc16(crc, (__u8 *)gdp, offset);
2059 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2060 /* for checksum of struct ext4_group_desc do the rest...*/
2061 if (ext4_has_feature_64bit(sb) &&
2062 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2063 crc = crc16(crc, (__u8 *)gdp + offset,
2064 le16_to_cpu(sbi->s_es->s_desc_size) -
2068 return cpu_to_le16(crc);
2071 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2072 struct ext4_group_desc *gdp)
2074 if (ext4_has_group_desc_csum(sb) &&
2075 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2081 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2082 struct ext4_group_desc *gdp)
2084 if (!ext4_has_group_desc_csum(sb))
2086 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2089 /* Called at mount-time, super-block is locked */
2090 static int ext4_check_descriptors(struct super_block *sb,
2091 ext4_group_t *first_not_zeroed)
2093 struct ext4_sb_info *sbi = EXT4_SB(sb);
2094 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2095 ext4_fsblk_t last_block;
2096 ext4_fsblk_t block_bitmap;
2097 ext4_fsblk_t inode_bitmap;
2098 ext4_fsblk_t inode_table;
2099 int flexbg_flag = 0;
2100 ext4_group_t i, grp = sbi->s_groups_count;
2102 if (ext4_has_feature_flex_bg(sb))
2105 ext4_debug("Checking group descriptors");
2107 for (i = 0; i < sbi->s_groups_count; i++) {
2108 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2110 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2111 last_block = ext4_blocks_count(sbi->s_es) - 1;
2113 last_block = first_block +
2114 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2116 if ((grp == sbi->s_groups_count) &&
2117 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2120 block_bitmap = ext4_block_bitmap(sb, gdp);
2121 if (block_bitmap < first_block || block_bitmap > last_block) {
2122 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2123 "Block bitmap for group %u not in group "
2124 "(block %llu)!", i, block_bitmap);
2127 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2128 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2130 "Inode bitmap for group %u not in group "
2131 "(block %llu)!", i, inode_bitmap);
2134 inode_table = ext4_inode_table(sb, gdp);
2135 if (inode_table < first_block ||
2136 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2137 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2138 "Inode table for group %u not in group "
2139 "(block %llu)!", i, inode_table);
2142 ext4_lock_group(sb, i);
2143 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2144 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2145 "Checksum for group %u failed (%u!=%u)",
2146 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2147 gdp)), le16_to_cpu(gdp->bg_checksum));
2148 if (!(sb->s_flags & MS_RDONLY)) {
2149 ext4_unlock_group(sb, i);
2153 ext4_unlock_group(sb, i);
2155 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2157 if (NULL != first_not_zeroed)
2158 *first_not_zeroed = grp;
2162 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2163 * the superblock) which were deleted from all directories, but held open by
2164 * a process at the time of a crash. We walk the list and try to delete these
2165 * inodes at recovery time (only with a read-write filesystem).
2167 * In order to keep the orphan inode chain consistent during traversal (in
2168 * case of crash during recovery), we link each inode into the superblock
2169 * orphan list_head and handle it the same way as an inode deletion during
2170 * normal operation (which journals the operations for us).
2172 * We only do an iget() and an iput() on each inode, which is very safe if we
2173 * accidentally point at an in-use or already deleted inode. The worst that
2174 * can happen in this case is that we get a "bit already cleared" message from
2175 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2176 * e2fsck was run on this filesystem, and it must have already done the orphan
2177 * inode cleanup for us, so we can safely abort without any further action.
2179 static void ext4_orphan_cleanup(struct super_block *sb,
2180 struct ext4_super_block *es)
2182 unsigned int s_flags = sb->s_flags;
2183 int nr_orphans = 0, nr_truncates = 0;
2187 if (!es->s_last_orphan) {
2188 jbd_debug(4, "no orphan inodes to clean up\n");
2192 if (bdev_read_only(sb->s_bdev)) {
2193 ext4_msg(sb, KERN_ERR, "write access "
2194 "unavailable, skipping orphan cleanup");
2198 /* Check if feature set would not allow a r/w mount */
2199 if (!ext4_feature_set_ok(sb, 0)) {
2200 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2201 "unknown ROCOMPAT features");
2205 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2206 /* don't clear list on RO mount w/ errors */
2207 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2208 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2209 "clearing orphan list.\n");
2210 es->s_last_orphan = 0;
2212 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2216 if (s_flags & MS_RDONLY) {
2217 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2218 sb->s_flags &= ~MS_RDONLY;
2221 /* Needed for iput() to work correctly and not trash data */
2222 sb->s_flags |= MS_ACTIVE;
2223 /* Turn on quotas so that they are updated correctly */
2224 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2225 if (EXT4_SB(sb)->s_qf_names[i]) {
2226 int ret = ext4_quota_on_mount(sb, i);
2228 ext4_msg(sb, KERN_ERR,
2229 "Cannot turn on journaled "
2230 "quota: error %d", ret);
2235 while (es->s_last_orphan) {
2236 struct inode *inode;
2238 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2239 if (IS_ERR(inode)) {
2240 es->s_last_orphan = 0;
2244 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2245 dquot_initialize(inode);
2246 if (inode->i_nlink) {
2247 if (test_opt(sb, DEBUG))
2248 ext4_msg(sb, KERN_DEBUG,
2249 "%s: truncating inode %lu to %lld bytes",
2250 __func__, inode->i_ino, inode->i_size);
2251 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2252 inode->i_ino, inode->i_size);
2253 mutex_lock(&inode->i_mutex);
2254 truncate_inode_pages(inode->i_mapping, inode->i_size);
2255 ext4_truncate(inode);
2256 mutex_unlock(&inode->i_mutex);
2259 if (test_opt(sb, DEBUG))
2260 ext4_msg(sb, KERN_DEBUG,
2261 "%s: deleting unreferenced inode %lu",
2262 __func__, inode->i_ino);
2263 jbd_debug(2, "deleting unreferenced inode %lu\n",
2267 iput(inode); /* The delete magic happens here! */
2270 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2273 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2274 PLURAL(nr_orphans));
2276 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2277 PLURAL(nr_truncates));
2279 /* Turn quotas off */
2280 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2281 if (sb_dqopt(sb)->files[i])
2282 dquot_quota_off(sb, i);
2285 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2289 * Maximal extent format file size.
2290 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2291 * extent format containers, within a sector_t, and within i_blocks
2292 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2293 * so that won't be a limiting factor.
2295 * However there is other limiting factor. We do store extents in the form
2296 * of starting block and length, hence the resulting length of the extent
2297 * covering maximum file size must fit into on-disk format containers as
2298 * well. Given that length is always by 1 unit bigger than max unit (because
2299 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2301 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2303 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2306 loff_t upper_limit = MAX_LFS_FILESIZE;
2308 /* small i_blocks in vfs inode? */
2309 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2311 * CONFIG_LBDAF is not enabled implies the inode
2312 * i_block represent total blocks in 512 bytes
2313 * 32 == size of vfs inode i_blocks * 8
2315 upper_limit = (1LL << 32) - 1;
2317 /* total blocks in file system block size */
2318 upper_limit >>= (blkbits - 9);
2319 upper_limit <<= blkbits;
2323 * 32-bit extent-start container, ee_block. We lower the maxbytes
2324 * by one fs block, so ee_len can cover the extent of maximum file
2327 res = (1LL << 32) - 1;
2330 /* Sanity check against vm- & vfs- imposed limits */
2331 if (res > upper_limit)
2338 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2339 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2340 * We need to be 1 filesystem block less than the 2^48 sector limit.
2342 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2344 loff_t res = EXT4_NDIR_BLOCKS;
2347 /* This is calculated to be the largest file size for a dense, block
2348 * mapped file such that the file's total number of 512-byte sectors,
2349 * including data and all indirect blocks, does not exceed (2^48 - 1).
2351 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2352 * number of 512-byte sectors of the file.
2355 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2357 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2358 * the inode i_block field represents total file blocks in
2359 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2361 upper_limit = (1LL << 32) - 1;
2363 /* total blocks in file system block size */
2364 upper_limit >>= (bits - 9);
2368 * We use 48 bit ext4_inode i_blocks
2369 * With EXT4_HUGE_FILE_FL set the i_blocks
2370 * represent total number of blocks in
2371 * file system block size
2373 upper_limit = (1LL << 48) - 1;
2377 /* indirect blocks */
2379 /* double indirect blocks */
2380 meta_blocks += 1 + (1LL << (bits-2));
2381 /* tripple indirect blocks */
2382 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2384 upper_limit -= meta_blocks;
2385 upper_limit <<= bits;
2387 res += 1LL << (bits-2);
2388 res += 1LL << (2*(bits-2));
2389 res += 1LL << (3*(bits-2));
2391 if (res > upper_limit)
2394 if (res > MAX_LFS_FILESIZE)
2395 res = MAX_LFS_FILESIZE;
2400 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2401 ext4_fsblk_t logical_sb_block, int nr)
2403 struct ext4_sb_info *sbi = EXT4_SB(sb);
2404 ext4_group_t bg, first_meta_bg;
2407 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2409 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2410 return logical_sb_block + nr + 1;
2411 bg = sbi->s_desc_per_block * nr;
2412 if (ext4_bg_has_super(sb, bg))
2416 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2417 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2418 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2421 if (sb->s_blocksize == 1024 && nr == 0 &&
2422 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2425 return (has_super + ext4_group_first_block_no(sb, bg));
2429 * ext4_get_stripe_size: Get the stripe size.
2430 * @sbi: In memory super block info
2432 * If we have specified it via mount option, then
2433 * use the mount option value. If the value specified at mount time is
2434 * greater than the blocks per group use the super block value.
2435 * If the super block value is greater than blocks per group return 0.
2436 * Allocator needs it be less than blocks per group.
2439 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2441 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2442 unsigned long stripe_width =
2443 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2446 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2447 ret = sbi->s_stripe;
2448 else if (stripe_width <= sbi->s_blocks_per_group)
2450 else if (stride <= sbi->s_blocks_per_group)
2456 * If the stripe width is 1, this makes no sense and
2457 * we set it to 0 to turn off stripe handling code.
2466 * Check whether this filesystem can be mounted based on
2467 * the features present and the RDONLY/RDWR mount requested.
2468 * Returns 1 if this filesystem can be mounted as requested,
2469 * 0 if it cannot be.
2471 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2473 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2474 ext4_msg(sb, KERN_ERR,
2475 "Couldn't mount because of "
2476 "unsupported optional features (%x)",
2477 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2478 ~EXT4_FEATURE_INCOMPAT_SUPP));
2485 if (ext4_has_feature_readonly(sb)) {
2486 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2487 sb->s_flags |= MS_RDONLY;
2491 /* Check that feature set is OK for a read-write mount */
2492 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2493 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2494 "unsupported optional features (%x)",
2495 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2496 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2500 * Large file size enabled file system can only be mounted
2501 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2503 if (ext4_has_feature_huge_file(sb)) {
2504 if (sizeof(blkcnt_t) < sizeof(u64)) {
2505 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2506 "cannot be mounted RDWR without "
2511 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2512 ext4_msg(sb, KERN_ERR,
2513 "Can't support bigalloc feature without "
2514 "extents feature\n");
2518 #ifndef CONFIG_QUOTA
2519 if (ext4_has_feature_quota(sb) && !readonly) {
2520 ext4_msg(sb, KERN_ERR,
2521 "Filesystem with quota feature cannot be mounted RDWR "
2522 "without CONFIG_QUOTA");
2525 #endif /* CONFIG_QUOTA */
2530 * This function is called once a day if we have errors logged
2531 * on the file system
2533 static void print_daily_error_info(unsigned long arg)
2535 struct super_block *sb = (struct super_block *) arg;
2536 struct ext4_sb_info *sbi;
2537 struct ext4_super_block *es;
2542 if (es->s_error_count)
2543 /* fsck newer than v1.41.13 is needed to clean this condition. */
2544 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2545 le32_to_cpu(es->s_error_count));
2546 if (es->s_first_error_time) {
2547 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2548 sb->s_id, le32_to_cpu(es->s_first_error_time),
2549 (int) sizeof(es->s_first_error_func),
2550 es->s_first_error_func,
2551 le32_to_cpu(es->s_first_error_line));
2552 if (es->s_first_error_ino)
2553 printk(": inode %u",
2554 le32_to_cpu(es->s_first_error_ino));
2555 if (es->s_first_error_block)
2556 printk(": block %llu", (unsigned long long)
2557 le64_to_cpu(es->s_first_error_block));
2560 if (es->s_last_error_time) {
2561 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2562 sb->s_id, le32_to_cpu(es->s_last_error_time),
2563 (int) sizeof(es->s_last_error_func),
2564 es->s_last_error_func,
2565 le32_to_cpu(es->s_last_error_line));
2566 if (es->s_last_error_ino)
2567 printk(": inode %u",
2568 le32_to_cpu(es->s_last_error_ino));
2569 if (es->s_last_error_block)
2570 printk(": block %llu", (unsigned long long)
2571 le64_to_cpu(es->s_last_error_block));
2574 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2577 /* Find next suitable group and run ext4_init_inode_table */
2578 static int ext4_run_li_request(struct ext4_li_request *elr)
2580 struct ext4_group_desc *gdp = NULL;
2581 ext4_group_t group, ngroups;
2582 struct super_block *sb;
2583 unsigned long timeout = 0;
2587 ngroups = EXT4_SB(sb)->s_groups_count;
2590 for (group = elr->lr_next_group; group < ngroups; group++) {
2591 gdp = ext4_get_group_desc(sb, group, NULL);
2597 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2601 if (group >= ngroups)
2606 ret = ext4_init_inode_table(sb, group,
2607 elr->lr_timeout ? 0 : 1);
2608 if (elr->lr_timeout == 0) {
2609 timeout = (jiffies - timeout) *
2610 elr->lr_sbi->s_li_wait_mult;
2611 elr->lr_timeout = timeout;
2613 elr->lr_next_sched = jiffies + elr->lr_timeout;
2614 elr->lr_next_group = group + 1;
2622 * Remove lr_request from the list_request and free the
2623 * request structure. Should be called with li_list_mtx held
2625 static void ext4_remove_li_request(struct ext4_li_request *elr)
2627 struct ext4_sb_info *sbi;
2634 list_del(&elr->lr_request);
2635 sbi->s_li_request = NULL;
2639 static void ext4_unregister_li_request(struct super_block *sb)
2641 mutex_lock(&ext4_li_mtx);
2642 if (!ext4_li_info) {
2643 mutex_unlock(&ext4_li_mtx);
2647 mutex_lock(&ext4_li_info->li_list_mtx);
2648 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2649 mutex_unlock(&ext4_li_info->li_list_mtx);
2650 mutex_unlock(&ext4_li_mtx);
2653 static struct task_struct *ext4_lazyinit_task;
2656 * This is the function where ext4lazyinit thread lives. It walks
2657 * through the request list searching for next scheduled filesystem.
2658 * When such a fs is found, run the lazy initialization request
2659 * (ext4_rn_li_request) and keep track of the time spend in this
2660 * function. Based on that time we compute next schedule time of
2661 * the request. When walking through the list is complete, compute
2662 * next waking time and put itself into sleep.
2664 static int ext4_lazyinit_thread(void *arg)
2666 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2667 struct list_head *pos, *n;
2668 struct ext4_li_request *elr;
2669 unsigned long next_wakeup, cur;
2671 BUG_ON(NULL == eli);
2675 next_wakeup = MAX_JIFFY_OFFSET;
2677 mutex_lock(&eli->li_list_mtx);
2678 if (list_empty(&eli->li_request_list)) {
2679 mutex_unlock(&eli->li_list_mtx);
2683 list_for_each_safe(pos, n, &eli->li_request_list) {
2684 elr = list_entry(pos, struct ext4_li_request,
2687 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2688 if (ext4_run_li_request(elr) != 0) {
2689 /* error, remove the lazy_init job */
2690 ext4_remove_li_request(elr);
2695 if (time_before(elr->lr_next_sched, next_wakeup))
2696 next_wakeup = elr->lr_next_sched;
2698 mutex_unlock(&eli->li_list_mtx);
2703 if ((time_after_eq(cur, next_wakeup)) ||
2704 (MAX_JIFFY_OFFSET == next_wakeup)) {
2709 schedule_timeout_interruptible(next_wakeup - cur);
2711 if (kthread_should_stop()) {
2712 ext4_clear_request_list();
2719 * It looks like the request list is empty, but we need
2720 * to check it under the li_list_mtx lock, to prevent any
2721 * additions into it, and of course we should lock ext4_li_mtx
2722 * to atomically free the list and ext4_li_info, because at
2723 * this point another ext4 filesystem could be registering
2726 mutex_lock(&ext4_li_mtx);
2727 mutex_lock(&eli->li_list_mtx);
2728 if (!list_empty(&eli->li_request_list)) {
2729 mutex_unlock(&eli->li_list_mtx);
2730 mutex_unlock(&ext4_li_mtx);
2733 mutex_unlock(&eli->li_list_mtx);
2734 kfree(ext4_li_info);
2735 ext4_li_info = NULL;
2736 mutex_unlock(&ext4_li_mtx);
2741 static void ext4_clear_request_list(void)
2743 struct list_head *pos, *n;
2744 struct ext4_li_request *elr;
2746 mutex_lock(&ext4_li_info->li_list_mtx);
2747 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2748 elr = list_entry(pos, struct ext4_li_request,
2750 ext4_remove_li_request(elr);
2752 mutex_unlock(&ext4_li_info->li_list_mtx);
2755 static int ext4_run_lazyinit_thread(void)
2757 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2758 ext4_li_info, "ext4lazyinit");
2759 if (IS_ERR(ext4_lazyinit_task)) {
2760 int err = PTR_ERR(ext4_lazyinit_task);
2761 ext4_clear_request_list();
2762 kfree(ext4_li_info);
2763 ext4_li_info = NULL;
2764 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2765 "initialization thread\n",
2769 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2774 * Check whether it make sense to run itable init. thread or not.
2775 * If there is at least one uninitialized inode table, return
2776 * corresponding group number, else the loop goes through all
2777 * groups and return total number of groups.
2779 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2781 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2782 struct ext4_group_desc *gdp = NULL;
2784 for (group = 0; group < ngroups; group++) {
2785 gdp = ext4_get_group_desc(sb, group, NULL);
2789 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2796 static int ext4_li_info_new(void)
2798 struct ext4_lazy_init *eli = NULL;
2800 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2804 INIT_LIST_HEAD(&eli->li_request_list);
2805 mutex_init(&eli->li_list_mtx);
2807 eli->li_state |= EXT4_LAZYINIT_QUIT;
2814 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2817 struct ext4_sb_info *sbi = EXT4_SB(sb);
2818 struct ext4_li_request *elr;
2820 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2826 elr->lr_next_group = start;
2829 * Randomize first schedule time of the request to
2830 * spread the inode table initialization requests
2833 elr->lr_next_sched = jiffies + (prandom_u32() %
2834 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2838 int ext4_register_li_request(struct super_block *sb,
2839 ext4_group_t first_not_zeroed)
2841 struct ext4_sb_info *sbi = EXT4_SB(sb);
2842 struct ext4_li_request *elr = NULL;
2843 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2846 mutex_lock(&ext4_li_mtx);
2847 if (sbi->s_li_request != NULL) {
2849 * Reset timeout so it can be computed again, because
2850 * s_li_wait_mult might have changed.
2852 sbi->s_li_request->lr_timeout = 0;
2856 if (first_not_zeroed == ngroups ||
2857 (sb->s_flags & MS_RDONLY) ||
2858 !test_opt(sb, INIT_INODE_TABLE))
2861 elr = ext4_li_request_new(sb, first_not_zeroed);
2867 if (NULL == ext4_li_info) {
2868 ret = ext4_li_info_new();
2873 mutex_lock(&ext4_li_info->li_list_mtx);
2874 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2875 mutex_unlock(&ext4_li_info->li_list_mtx);
2877 sbi->s_li_request = elr;
2879 * set elr to NULL here since it has been inserted to
2880 * the request_list and the removal and free of it is
2881 * handled by ext4_clear_request_list from now on.
2885 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2886 ret = ext4_run_lazyinit_thread();
2891 mutex_unlock(&ext4_li_mtx);
2898 * We do not need to lock anything since this is called on
2901 static void ext4_destroy_lazyinit_thread(void)
2904 * If thread exited earlier
2905 * there's nothing to be done.
2907 if (!ext4_li_info || !ext4_lazyinit_task)
2910 kthread_stop(ext4_lazyinit_task);
2913 static int set_journal_csum_feature_set(struct super_block *sb)
2916 int compat, incompat;
2917 struct ext4_sb_info *sbi = EXT4_SB(sb);
2919 if (ext4_has_metadata_csum(sb)) {
2920 /* journal checksum v3 */
2922 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2924 /* journal checksum v1 */
2925 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2929 jbd2_journal_clear_features(sbi->s_journal,
2930 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2931 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2932 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2933 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2934 ret = jbd2_journal_set_features(sbi->s_journal,
2936 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2938 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2939 ret = jbd2_journal_set_features(sbi->s_journal,
2942 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2943 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2945 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2946 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2953 * Note: calculating the overhead so we can be compatible with
2954 * historical BSD practice is quite difficult in the face of
2955 * clusters/bigalloc. This is because multiple metadata blocks from
2956 * different block group can end up in the same allocation cluster.
2957 * Calculating the exact overhead in the face of clustered allocation
2958 * requires either O(all block bitmaps) in memory or O(number of block
2959 * groups**2) in time. We will still calculate the superblock for
2960 * older file systems --- and if we come across with a bigalloc file
2961 * system with zero in s_overhead_clusters the estimate will be close to
2962 * correct especially for very large cluster sizes --- but for newer
2963 * file systems, it's better to calculate this figure once at mkfs
2964 * time, and store it in the superblock. If the superblock value is
2965 * present (even for non-bigalloc file systems), we will use it.
2967 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2970 struct ext4_sb_info *sbi = EXT4_SB(sb);
2971 struct ext4_group_desc *gdp;
2972 ext4_fsblk_t first_block, last_block, b;
2973 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2974 int s, j, count = 0;
2976 if (!ext4_has_feature_bigalloc(sb))
2977 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2978 sbi->s_itb_per_group + 2);
2980 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2981 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2982 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2983 for (i = 0; i < ngroups; i++) {
2984 gdp = ext4_get_group_desc(sb, i, NULL);
2985 b = ext4_block_bitmap(sb, gdp);
2986 if (b >= first_block && b <= last_block) {
2987 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2990 b = ext4_inode_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_table(sb, gdp);
2996 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
2997 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
2998 int c = EXT4_B2C(sbi, b - first_block);
2999 ext4_set_bit(c, buf);
3005 if (ext4_bg_has_super(sb, grp)) {
3006 ext4_set_bit(s++, buf);
3009 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3010 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3016 return EXT4_CLUSTERS_PER_GROUP(sb) -
3017 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3021 * Compute the overhead and stash it in sbi->s_overhead
3023 int ext4_calculate_overhead(struct super_block *sb)
3025 struct ext4_sb_info *sbi = EXT4_SB(sb);
3026 struct ext4_super_block *es = sbi->s_es;
3027 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3028 ext4_fsblk_t overhead = 0;
3029 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3035 * Compute the overhead (FS structures). This is constant
3036 * for a given filesystem unless the number of block groups
3037 * changes so we cache the previous value until it does.
3041 * All of the blocks before first_data_block are overhead
3043 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3046 * Add the overhead found in each block group
3048 for (i = 0; i < ngroups; i++) {
3051 blks = count_overhead(sb, i, buf);
3054 memset(buf, 0, PAGE_SIZE);
3057 /* Add the internal journal blocks as well */
3058 if (sbi->s_journal && !sbi->journal_bdev)
3059 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3061 sbi->s_overhead = overhead;
3063 free_page((unsigned long) buf);
3067 static void ext4_set_resv_clusters(struct super_block *sb)
3069 ext4_fsblk_t resv_clusters;
3070 struct ext4_sb_info *sbi = EXT4_SB(sb);
3073 * There's no need to reserve anything when we aren't using extents.
3074 * The space estimates are exact, there are no unwritten extents,
3075 * hole punching doesn't need new metadata... This is needed especially
3076 * to keep ext2/3 backward compatibility.
3078 if (!ext4_has_feature_extents(sb))
3081 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3082 * This should cover the situations where we can not afford to run
3083 * out of space like for example punch hole, or converting
3084 * unwritten extents in delalloc path. In most cases such
3085 * allocation would require 1, or 2 blocks, higher numbers are
3088 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3089 sbi->s_cluster_bits);
3091 do_div(resv_clusters, 50);
3092 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3094 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3097 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3099 char *orig_data = kstrdup(data, GFP_KERNEL);
3100 struct buffer_head *bh;
3101 struct ext4_super_block *es = NULL;
3102 struct ext4_sb_info *sbi;
3104 ext4_fsblk_t sb_block = get_sb_block(&data);
3105 ext4_fsblk_t logical_sb_block;
3106 unsigned long offset = 0;
3107 unsigned long journal_devnum = 0;
3108 unsigned long def_mount_opts;
3112 int blocksize, clustersize;
3113 unsigned int db_count;
3115 int needs_recovery, has_huge_files, has_bigalloc;
3118 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3119 ext4_group_t first_not_zeroed;
3121 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3125 sbi->s_blockgroup_lock =
3126 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3127 if (!sbi->s_blockgroup_lock) {
3131 sb->s_fs_info = sbi;
3133 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3134 sbi->s_sb_block = sb_block;
3135 if (sb->s_bdev->bd_part)
3136 sbi->s_sectors_written_start =
3137 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3139 /* Cleanup superblock name */
3140 strreplace(sb->s_id, '/', '!');
3142 /* -EINVAL is default */
3144 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3146 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3151 * The ext4 superblock will not be buffer aligned for other than 1kB
3152 * block sizes. We need to calculate the offset from buffer start.
3154 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3155 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3156 offset = do_div(logical_sb_block, blocksize);
3158 logical_sb_block = sb_block;
3161 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3162 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3166 * Note: s_es must be initialized as soon as possible because
3167 * some ext4 macro-instructions depend on its value
3169 es = (struct ext4_super_block *) (bh->b_data + offset);
3171 sb->s_magic = le16_to_cpu(es->s_magic);
3172 if (sb->s_magic != EXT4_SUPER_MAGIC)
3174 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3176 /* Warn if metadata_csum and gdt_csum are both set. */
3177 if (ext4_has_feature_metadata_csum(sb) &&
3178 ext4_has_feature_gdt_csum(sb))
3179 ext4_warning(sb, "metadata_csum and uninit_bg are "
3180 "redundant flags; please run fsck.");
3182 /* Check for a known checksum algorithm */
3183 if (!ext4_verify_csum_type(sb, es)) {
3184 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3185 "unknown checksum algorithm.");
3190 /* Load the checksum driver */
3191 if (ext4_has_feature_metadata_csum(sb)) {
3192 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3193 if (IS_ERR(sbi->s_chksum_driver)) {
3194 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3195 ret = PTR_ERR(sbi->s_chksum_driver);
3196 sbi->s_chksum_driver = NULL;
3201 /* Check superblock checksum */
3202 if (!ext4_superblock_csum_verify(sb, es)) {
3203 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3204 "invalid superblock checksum. Run e2fsck?");
3210 /* Precompute checksum seed for all metadata */
3211 if (ext4_has_feature_csum_seed(sb))
3212 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3213 else if (ext4_has_metadata_csum(sb))
3214 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3215 sizeof(es->s_uuid));
3217 /* Set defaults before we parse the mount options */
3218 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3219 set_opt(sb, INIT_INODE_TABLE);
3220 if (def_mount_opts & EXT4_DEFM_DEBUG)
3222 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3224 if (def_mount_opts & EXT4_DEFM_UID16)
3225 set_opt(sb, NO_UID32);
3226 /* xattr user namespace & acls are now defaulted on */
3227 set_opt(sb, XATTR_USER);
3228 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3229 set_opt(sb, POSIX_ACL);
3231 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3232 if (ext4_has_metadata_csum(sb))
3233 set_opt(sb, JOURNAL_CHECKSUM);
3235 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3236 set_opt(sb, JOURNAL_DATA);
3237 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3238 set_opt(sb, ORDERED_DATA);
3239 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3240 set_opt(sb, WRITEBACK_DATA);
3242 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3243 set_opt(sb, ERRORS_PANIC);
3244 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3245 set_opt(sb, ERRORS_CONT);
3247 set_opt(sb, ERRORS_RO);
3248 /* block_validity enabled by default; disable with noblock_validity */
3249 set_opt(sb, BLOCK_VALIDITY);
3250 if (def_mount_opts & EXT4_DEFM_DISCARD)
3251 set_opt(sb, DISCARD);
3253 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3254 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3255 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3256 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3257 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3259 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3260 set_opt(sb, BARRIER);
3263 * enable delayed allocation by default
3264 * Use -o nodelalloc to turn it off
3266 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3267 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3268 set_opt(sb, DELALLOC);
3271 * set default s_li_wait_mult for lazyinit, for the case there is
3272 * no mount option specified.
3274 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3276 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3277 &journal_devnum, &journal_ioprio, 0)) {
3278 ext4_msg(sb, KERN_WARNING,
3279 "failed to parse options in superblock: %s",
3280 sbi->s_es->s_mount_opts);
3282 sbi->s_def_mount_opt = sbi->s_mount_opt;
3283 if (!parse_options((char *) data, sb, &journal_devnum,
3284 &journal_ioprio, 0))
3287 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3288 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3289 "with data=journal disables delayed "
3290 "allocation and O_DIRECT support!\n");
3291 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3292 ext4_msg(sb, KERN_ERR, "can't mount with "
3293 "both data=journal and delalloc");
3296 if (test_opt(sb, DIOREAD_NOLOCK)) {
3297 ext4_msg(sb, KERN_ERR, "can't mount with "
3298 "both data=journal and dioread_nolock");
3301 if (test_opt(sb, DAX)) {
3302 ext4_msg(sb, KERN_ERR, "can't mount with "
3303 "both data=journal and dax");
3306 if (test_opt(sb, DELALLOC))
3307 clear_opt(sb, DELALLOC);
3309 sb->s_iflags |= SB_I_CGROUPWB;
3312 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3313 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3315 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3316 (ext4_has_compat_features(sb) ||
3317 ext4_has_ro_compat_features(sb) ||
3318 ext4_has_incompat_features(sb)))
3319 ext4_msg(sb, KERN_WARNING,
3320 "feature flags set on rev 0 fs, "
3321 "running e2fsck is recommended");
3323 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3324 set_opt2(sb, HURD_COMPAT);
3325 if (ext4_has_feature_64bit(sb)) {
3326 ext4_msg(sb, KERN_ERR,
3327 "The Hurd can't support 64-bit file systems");
3332 if (IS_EXT2_SB(sb)) {
3333 if (ext2_feature_set_ok(sb))
3334 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3335 "using the ext4 subsystem");
3337 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3338 "to feature incompatibilities");
3343 if (IS_EXT3_SB(sb)) {
3344 if (ext3_feature_set_ok(sb))
3345 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3346 "using the ext4 subsystem");
3348 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3349 "to feature incompatibilities");
3355 * Check feature flags regardless of the revision level, since we
3356 * previously didn't change the revision level when setting the flags,
3357 * so there is a chance incompat flags are set on a rev 0 filesystem.
3359 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3362 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3363 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3364 blocksize > EXT4_MAX_BLOCK_SIZE) {
3365 ext4_msg(sb, KERN_ERR,
3366 "Unsupported filesystem blocksize %d", blocksize);
3370 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3371 if (blocksize != PAGE_SIZE) {
3372 ext4_msg(sb, KERN_ERR,
3373 "error: unsupported blocksize for dax");
3376 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3377 ext4_msg(sb, KERN_ERR,
3378 "error: device does not support dax");
3383 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3384 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3385 es->s_encryption_level);
3389 if (sb->s_blocksize != blocksize) {
3390 /* Validate the filesystem blocksize */
3391 if (!sb_set_blocksize(sb, blocksize)) {
3392 ext4_msg(sb, KERN_ERR, "bad block size %d",
3398 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3399 offset = do_div(logical_sb_block, blocksize);
3400 bh = sb_bread_unmovable(sb, logical_sb_block);
3402 ext4_msg(sb, KERN_ERR,
3403 "Can't read superblock on 2nd try");
3406 es = (struct ext4_super_block *)(bh->b_data + offset);
3408 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3409 ext4_msg(sb, KERN_ERR,
3410 "Magic mismatch, very weird!");
3415 has_huge_files = ext4_has_feature_huge_file(sb);
3416 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3418 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3420 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3421 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3422 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3424 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3425 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3426 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3427 (!is_power_of_2(sbi->s_inode_size)) ||
3428 (sbi->s_inode_size > blocksize)) {
3429 ext4_msg(sb, KERN_ERR,
3430 "unsupported inode size: %d",
3434 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3435 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3438 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3439 if (ext4_has_feature_64bit(sb)) {
3440 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3441 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3442 !is_power_of_2(sbi->s_desc_size)) {
3443 ext4_msg(sb, KERN_ERR,
3444 "unsupported descriptor size %lu",
3449 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3451 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3452 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3453 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3456 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3457 if (sbi->s_inodes_per_block == 0)
3459 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3460 sbi->s_inodes_per_block;
3461 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3463 sbi->s_mount_state = le16_to_cpu(es->s_state);
3464 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3465 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3467 for (i = 0; i < 4; i++)
3468 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3469 sbi->s_def_hash_version = es->s_def_hash_version;
3470 if (ext4_has_feature_dir_index(sb)) {
3471 i = le32_to_cpu(es->s_flags);
3472 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3473 sbi->s_hash_unsigned = 3;
3474 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3475 #ifdef __CHAR_UNSIGNED__
3476 if (!(sb->s_flags & MS_RDONLY))
3478 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3479 sbi->s_hash_unsigned = 3;
3481 if (!(sb->s_flags & MS_RDONLY))
3483 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3488 /* Handle clustersize */
3489 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3490 has_bigalloc = ext4_has_feature_bigalloc(sb);
3492 if (clustersize < blocksize) {
3493 ext4_msg(sb, KERN_ERR,
3494 "cluster size (%d) smaller than "
3495 "block size (%d)", clustersize, blocksize);
3498 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3499 le32_to_cpu(es->s_log_block_size);
3500 sbi->s_clusters_per_group =
3501 le32_to_cpu(es->s_clusters_per_group);
3502 if (sbi->s_clusters_per_group > blocksize * 8) {
3503 ext4_msg(sb, KERN_ERR,
3504 "#clusters per group too big: %lu",
3505 sbi->s_clusters_per_group);
3508 if (sbi->s_blocks_per_group !=
3509 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3510 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3511 "clusters per group (%lu) inconsistent",
3512 sbi->s_blocks_per_group,
3513 sbi->s_clusters_per_group);
3517 if (clustersize != blocksize) {
3518 ext4_warning(sb, "fragment/cluster size (%d) != "
3519 "block size (%d)", clustersize,
3521 clustersize = blocksize;
3523 if (sbi->s_blocks_per_group > blocksize * 8) {
3524 ext4_msg(sb, KERN_ERR,
3525 "#blocks per group too big: %lu",
3526 sbi->s_blocks_per_group);
3529 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3530 sbi->s_cluster_bits = 0;
3532 sbi->s_cluster_ratio = clustersize / blocksize;
3534 if (sbi->s_inodes_per_group > blocksize * 8) {
3535 ext4_msg(sb, KERN_ERR,
3536 "#inodes per group too big: %lu",
3537 sbi->s_inodes_per_group);
3541 /* Do we have standard group size of clustersize * 8 blocks ? */
3542 if (sbi->s_blocks_per_group == clustersize << 3)
3543 set_opt2(sb, STD_GROUP_SIZE);
3546 * Test whether we have more sectors than will fit in sector_t,
3547 * and whether the max offset is addressable by the page cache.
3549 err = generic_check_addressable(sb->s_blocksize_bits,
3550 ext4_blocks_count(es));
3552 ext4_msg(sb, KERN_ERR, "filesystem"
3553 " too large to mount safely on this system");
3554 if (sizeof(sector_t) < 8)
3555 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3559 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3562 /* check blocks count against device size */
3563 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3564 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3565 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3566 "exceeds size of device (%llu blocks)",
3567 ext4_blocks_count(es), blocks_count);
3572 * It makes no sense for the first data block to be beyond the end
3573 * of the filesystem.
3575 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3576 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3577 "block %u is beyond end of filesystem (%llu)",
3578 le32_to_cpu(es->s_first_data_block),
3579 ext4_blocks_count(es));
3582 blocks_count = (ext4_blocks_count(es) -
3583 le32_to_cpu(es->s_first_data_block) +
3584 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3585 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3586 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3587 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3588 "(block count %llu, first data block %u, "
3589 "blocks per group %lu)", sbi->s_groups_count,
3590 ext4_blocks_count(es),
3591 le32_to_cpu(es->s_first_data_block),
3592 EXT4_BLOCKS_PER_GROUP(sb));
3595 sbi->s_groups_count = blocks_count;
3596 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3597 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3598 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3599 EXT4_DESC_PER_BLOCK(sb);
3600 sbi->s_group_desc = ext4_kvmalloc(db_count *
3601 sizeof(struct buffer_head *),
3603 if (sbi->s_group_desc == NULL) {
3604 ext4_msg(sb, KERN_ERR, "not enough memory");
3609 bgl_lock_init(sbi->s_blockgroup_lock);
3611 for (i = 0; i < db_count; i++) {
3612 block = descriptor_loc(sb, logical_sb_block, i);
3613 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3614 if (!sbi->s_group_desc[i]) {
3615 ext4_msg(sb, KERN_ERR,
3616 "can't read group descriptor %d", i);
3621 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3622 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3623 ret = -EFSCORRUPTED;
3627 sbi->s_gdb_count = db_count;
3628 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3629 spin_lock_init(&sbi->s_next_gen_lock);
3631 setup_timer(&sbi->s_err_report, print_daily_error_info,
3632 (unsigned long) sb);
3634 /* Register extent status tree shrinker */
3635 if (ext4_es_register_shrinker(sbi))
3638 sbi->s_stripe = ext4_get_stripe_size(sbi);
3639 sbi->s_extent_max_zeroout_kb = 32;
3642 * set up enough so that it can read an inode
3644 sb->s_op = &ext4_sops;
3645 sb->s_export_op = &ext4_export_ops;
3646 sb->s_xattr = ext4_xattr_handlers;
3648 sb->dq_op = &ext4_quota_operations;
3649 if (ext4_has_feature_quota(sb))
3650 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3652 sb->s_qcop = &ext4_qctl_operations;
3653 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3655 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3657 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3658 mutex_init(&sbi->s_orphan_lock);
3662 needs_recovery = (es->s_last_orphan != 0 ||
3663 ext4_has_feature_journal_needs_recovery(sb));
3665 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3666 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3667 goto failed_mount3a;
3670 * The first inode we look at is the journal inode. Don't try
3671 * root first: it may be modified in the journal!
3673 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3674 if (ext4_load_journal(sb, es, journal_devnum))
3675 goto failed_mount3a;
3676 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3677 ext4_has_feature_journal_needs_recovery(sb)) {
3678 ext4_msg(sb, KERN_ERR, "required journal recovery "
3679 "suppressed and not mounted read-only");
3680 goto failed_mount_wq;
3682 /* Nojournal mode, all journal mount options are illegal */
3683 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3684 ext4_msg(sb, KERN_ERR, "can't mount with "
3685 "journal_checksum, fs mounted w/o journal");
3686 goto failed_mount_wq;
3688 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3689 ext4_msg(sb, KERN_ERR, "can't mount with "
3690 "journal_async_commit, fs mounted w/o journal");
3691 goto failed_mount_wq;
3693 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3694 ext4_msg(sb, KERN_ERR, "can't mount with "
3695 "commit=%lu, fs mounted w/o journal",
3696 sbi->s_commit_interval / HZ);
3697 goto failed_mount_wq;
3699 if (EXT4_MOUNT_DATA_FLAGS &
3700 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3701 ext4_msg(sb, KERN_ERR, "can't mount with "
3702 "data=, fs mounted w/o journal");
3703 goto failed_mount_wq;
3705 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3706 clear_opt(sb, JOURNAL_CHECKSUM);
3707 clear_opt(sb, DATA_FLAGS);
3708 sbi->s_journal = NULL;
3713 if (ext4_has_feature_64bit(sb) &&
3714 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3715 JBD2_FEATURE_INCOMPAT_64BIT)) {
3716 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3717 goto failed_mount_wq;
3720 if (!set_journal_csum_feature_set(sb)) {
3721 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3723 goto failed_mount_wq;
3726 /* We have now updated the journal if required, so we can
3727 * validate the data journaling mode. */
3728 switch (test_opt(sb, DATA_FLAGS)) {
3730 /* No mode set, assume a default based on the journal
3731 * capabilities: ORDERED_DATA if the journal can
3732 * cope, else JOURNAL_DATA
3734 if (jbd2_journal_check_available_features
3735 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3736 set_opt(sb, ORDERED_DATA);
3738 set_opt(sb, JOURNAL_DATA);
3741 case EXT4_MOUNT_ORDERED_DATA:
3742 case EXT4_MOUNT_WRITEBACK_DATA:
3743 if (!jbd2_journal_check_available_features
3744 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3745 ext4_msg(sb, KERN_ERR, "Journal does not support "
3746 "requested data journaling mode");
3747 goto failed_mount_wq;
3752 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3754 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3757 if (ext4_mballoc_ready) {
3758 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3759 if (!sbi->s_mb_cache) {
3760 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3761 goto failed_mount_wq;
3765 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3766 (blocksize != PAGE_CACHE_SIZE)) {
3767 ext4_msg(sb, KERN_ERR,
3768 "Unsupported blocksize for fs encryption");
3769 goto failed_mount_wq;
3772 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3773 !ext4_has_feature_encrypt(sb)) {
3774 ext4_set_feature_encrypt(sb);
3775 ext4_commit_super(sb, 1);
3779 * Get the # of file system overhead blocks from the
3780 * superblock if present.
3782 if (es->s_overhead_clusters)
3783 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3785 err = ext4_calculate_overhead(sb);
3787 goto failed_mount_wq;
3791 * The maximum number of concurrent works can be high and
3792 * concurrency isn't really necessary. Limit it to 1.
3794 EXT4_SB(sb)->rsv_conversion_wq =
3795 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3796 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3797 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3803 * The jbd2_journal_load will have done any necessary log recovery,
3804 * so we can safely mount the rest of the filesystem now.
3807 root = ext4_iget(sb, EXT4_ROOT_INO);
3809 ext4_msg(sb, KERN_ERR, "get root inode failed");
3810 ret = PTR_ERR(root);
3814 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3815 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3819 sb->s_root = d_make_root(root);
3821 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3826 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3827 sb->s_flags |= MS_RDONLY;
3829 /* determine the minimum size of new large inodes, if present */
3830 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3831 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3832 EXT4_GOOD_OLD_INODE_SIZE;
3833 if (ext4_has_feature_extra_isize(sb)) {
3834 if (sbi->s_want_extra_isize <
3835 le16_to_cpu(es->s_want_extra_isize))
3836 sbi->s_want_extra_isize =
3837 le16_to_cpu(es->s_want_extra_isize);
3838 if (sbi->s_want_extra_isize <
3839 le16_to_cpu(es->s_min_extra_isize))
3840 sbi->s_want_extra_isize =
3841 le16_to_cpu(es->s_min_extra_isize);
3844 /* Check if enough inode space is available */
3845 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3846 sbi->s_inode_size) {
3847 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3848 EXT4_GOOD_OLD_INODE_SIZE;
3849 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3853 ext4_set_resv_clusters(sb);
3855 err = ext4_setup_system_zone(sb);
3857 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3859 goto failed_mount4a;
3863 err = ext4_mb_init(sb);
3865 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3870 block = ext4_count_free_clusters(sb);
3871 ext4_free_blocks_count_set(sbi->s_es,
3872 EXT4_C2B(sbi, block));
3873 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3876 unsigned long freei = ext4_count_free_inodes(sb);
3877 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3878 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3882 err = percpu_counter_init(&sbi->s_dirs_counter,
3883 ext4_count_dirs(sb), GFP_KERNEL);
3885 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3888 ext4_msg(sb, KERN_ERR, "insufficient memory");
3892 if (ext4_has_feature_flex_bg(sb))
3893 if (!ext4_fill_flex_info(sb)) {
3894 ext4_msg(sb, KERN_ERR,
3895 "unable to initialize "
3896 "flex_bg meta info!");
3900 err = ext4_register_li_request(sb, first_not_zeroed);
3904 err = ext4_register_sysfs(sb);
3909 /* Enable quota usage during mount. */
3910 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3911 err = ext4_enable_quotas(sb);
3915 #endif /* CONFIG_QUOTA */
3917 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3918 ext4_orphan_cleanup(sb, es);
3919 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3920 if (needs_recovery) {
3921 ext4_msg(sb, KERN_INFO, "recovery complete");
3922 ext4_mark_recovery_complete(sb, es);
3924 if (EXT4_SB(sb)->s_journal) {
3925 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3926 descr = " journalled data mode";
3927 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3928 descr = " ordered data mode";
3930 descr = " writeback data mode";
3932 descr = "out journal";
3934 if (test_opt(sb, DISCARD)) {
3935 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3936 if (!blk_queue_discard(q))
3937 ext4_msg(sb, KERN_WARNING,
3938 "mounting with \"discard\" option, but "
3939 "the device does not support discard");
3942 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3943 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3944 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3945 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3947 if (es->s_error_count)
3948 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3950 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3951 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3952 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3953 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3960 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3965 ext4_unregister_sysfs(sb);
3968 ext4_unregister_li_request(sb);
3970 ext4_mb_release(sb);
3971 if (sbi->s_flex_groups)
3972 kvfree(sbi->s_flex_groups);
3973 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3974 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3975 percpu_counter_destroy(&sbi->s_dirs_counter);
3976 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3978 ext4_ext_release(sb);
3979 ext4_release_system_zone(sb);
3984 ext4_msg(sb, KERN_ERR, "mount failed");
3985 if (EXT4_SB(sb)->rsv_conversion_wq)
3986 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
3988 if (sbi->s_journal) {
3989 jbd2_journal_destroy(sbi->s_journal);
3990 sbi->s_journal = NULL;
3993 ext4_es_unregister_shrinker(sbi);
3995 del_timer_sync(&sbi->s_err_report);
3997 kthread_stop(sbi->s_mmp_tsk);
3999 for (i = 0; i < db_count; i++)
4000 brelse(sbi->s_group_desc[i]);
4001 kvfree(sbi->s_group_desc);
4003 if (sbi->s_chksum_driver)
4004 crypto_free_shash(sbi->s_chksum_driver);
4006 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4007 kfree(sbi->s_qf_names[i]);
4009 ext4_blkdev_remove(sbi);
4012 sb->s_fs_info = NULL;
4013 kfree(sbi->s_blockgroup_lock);
4017 return err ? err : ret;
4021 * Setup any per-fs journal parameters now. We'll do this both on
4022 * initial mount, once the journal has been initialised but before we've
4023 * done any recovery; and again on any subsequent remount.
4025 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4027 struct ext4_sb_info *sbi = EXT4_SB(sb);
4029 journal->j_commit_interval = sbi->s_commit_interval;
4030 journal->j_min_batch_time = sbi->s_min_batch_time;
4031 journal->j_max_batch_time = sbi->s_max_batch_time;
4033 write_lock(&journal->j_state_lock);
4034 if (test_opt(sb, BARRIER))
4035 journal->j_flags |= JBD2_BARRIER;
4037 journal->j_flags &= ~JBD2_BARRIER;
4038 if (test_opt(sb, DATA_ERR_ABORT))
4039 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4041 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4042 write_unlock(&journal->j_state_lock);
4045 static journal_t *ext4_get_journal(struct super_block *sb,
4046 unsigned int journal_inum)
4048 struct inode *journal_inode;
4051 BUG_ON(!ext4_has_feature_journal(sb));
4053 /* First, test for the existence of a valid inode on disk. Bad
4054 * things happen if we iget() an unused inode, as the subsequent
4055 * iput() will try to delete it. */
4057 journal_inode = ext4_iget(sb, journal_inum);
4058 if (IS_ERR(journal_inode)) {
4059 ext4_msg(sb, KERN_ERR, "no journal found");
4062 if (!journal_inode->i_nlink) {
4063 make_bad_inode(journal_inode);
4064 iput(journal_inode);
4065 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4069 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4070 journal_inode, journal_inode->i_size);
4071 if (!S_ISREG(journal_inode->i_mode)) {
4072 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4073 iput(journal_inode);
4077 journal = jbd2_journal_init_inode(journal_inode);
4079 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4080 iput(journal_inode);
4083 journal->j_private = sb;
4084 ext4_init_journal_params(sb, journal);
4088 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4091 struct buffer_head *bh;
4095 int hblock, blocksize;
4096 ext4_fsblk_t sb_block;
4097 unsigned long offset;
4098 struct ext4_super_block *es;
4099 struct block_device *bdev;
4101 BUG_ON(!ext4_has_feature_journal(sb));
4103 bdev = ext4_blkdev_get(j_dev, sb);
4107 blocksize = sb->s_blocksize;
4108 hblock = bdev_logical_block_size(bdev);
4109 if (blocksize < hblock) {
4110 ext4_msg(sb, KERN_ERR,
4111 "blocksize too small for journal device");
4115 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4116 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4117 set_blocksize(bdev, blocksize);
4118 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4119 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4120 "external journal");
4124 es = (struct ext4_super_block *) (bh->b_data + offset);
4125 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4126 !(le32_to_cpu(es->s_feature_incompat) &
4127 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4128 ext4_msg(sb, KERN_ERR, "external journal has "
4134 if ((le32_to_cpu(es->s_feature_ro_compat) &
4135 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4136 es->s_checksum != ext4_superblock_csum(sb, es)) {
4137 ext4_msg(sb, KERN_ERR, "external journal has "
4138 "corrupt superblock");
4143 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4144 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4149 len = ext4_blocks_count(es);
4150 start = sb_block + 1;
4151 brelse(bh); /* we're done with the superblock */
4153 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4154 start, len, blocksize);
4156 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4159 journal->j_private = sb;
4160 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4161 wait_on_buffer(journal->j_sb_buffer);
4162 if (!buffer_uptodate(journal->j_sb_buffer)) {
4163 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4166 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4167 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4168 "user (unsupported) - %d",
4169 be32_to_cpu(journal->j_superblock->s_nr_users));
4172 EXT4_SB(sb)->journal_bdev = bdev;
4173 ext4_init_journal_params(sb, journal);
4177 jbd2_journal_destroy(journal);
4179 ext4_blkdev_put(bdev);
4183 static int ext4_load_journal(struct super_block *sb,
4184 struct ext4_super_block *es,
4185 unsigned long journal_devnum)
4188 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4191 int really_read_only;
4193 BUG_ON(!ext4_has_feature_journal(sb));
4195 if (journal_devnum &&
4196 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4197 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4198 "numbers have changed");
4199 journal_dev = new_decode_dev(journal_devnum);
4201 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4203 really_read_only = bdev_read_only(sb->s_bdev);
4206 * Are we loading a blank journal or performing recovery after a
4207 * crash? For recovery, we need to check in advance whether we
4208 * can get read-write access to the device.
4210 if (ext4_has_feature_journal_needs_recovery(sb)) {
4211 if (sb->s_flags & MS_RDONLY) {
4212 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4213 "required on readonly filesystem");
4214 if (really_read_only) {
4215 ext4_msg(sb, KERN_ERR, "write access "
4216 "unavailable, cannot proceed");
4219 ext4_msg(sb, KERN_INFO, "write access will "
4220 "be enabled during recovery");
4224 if (journal_inum && journal_dev) {
4225 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4226 "and inode journals!");
4231 if (!(journal = ext4_get_journal(sb, journal_inum)))
4234 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4238 if (!(journal->j_flags & JBD2_BARRIER))
4239 ext4_msg(sb, KERN_INFO, "barriers disabled");
4241 if (!ext4_has_feature_journal_needs_recovery(sb))
4242 err = jbd2_journal_wipe(journal, !really_read_only);
4244 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4246 memcpy(save, ((char *) es) +
4247 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4248 err = jbd2_journal_load(journal);
4250 memcpy(((char *) es) + EXT4_S_ERR_START,
4251 save, EXT4_S_ERR_LEN);
4256 ext4_msg(sb, KERN_ERR, "error loading journal");
4257 jbd2_journal_destroy(journal);
4261 EXT4_SB(sb)->s_journal = journal;
4262 ext4_clear_journal_err(sb, es);
4264 if (!really_read_only && journal_devnum &&
4265 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4266 es->s_journal_dev = cpu_to_le32(journal_devnum);
4268 /* Make sure we flush the recovery flag to disk. */
4269 ext4_commit_super(sb, 1);
4275 static int ext4_commit_super(struct super_block *sb, int sync)
4277 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4278 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4281 if (!sbh || block_device_ejected(sb))
4283 if (buffer_write_io_error(sbh)) {
4285 * Oh, dear. A previous attempt to write the
4286 * superblock failed. This could happen because the
4287 * USB device was yanked out. Or it could happen to
4288 * be a transient write error and maybe the block will
4289 * be remapped. Nothing we can do but to retry the
4290 * write and hope for the best.
4292 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4293 "superblock detected");
4294 clear_buffer_write_io_error(sbh);
4295 set_buffer_uptodate(sbh);
4298 * If the file system is mounted read-only, don't update the
4299 * superblock write time. This avoids updating the superblock
4300 * write time when we are mounting the root file system
4301 * read/only but we need to replay the journal; at that point,
4302 * for people who are east of GMT and who make their clock
4303 * tick in localtime for Windows bug-for-bug compatibility,
4304 * the clock is set in the future, and this will cause e2fsck
4305 * to complain and force a full file system check.
4307 if (!(sb->s_flags & MS_RDONLY))
4308 es->s_wtime = cpu_to_le32(get_seconds());
4309 if (sb->s_bdev->bd_part)
4310 es->s_kbytes_written =
4311 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4312 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4313 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4315 es->s_kbytes_written =
4316 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4317 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4318 ext4_free_blocks_count_set(es,
4319 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4320 &EXT4_SB(sb)->s_freeclusters_counter)));
4321 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4322 es->s_free_inodes_count =
4323 cpu_to_le32(percpu_counter_sum_positive(
4324 &EXT4_SB(sb)->s_freeinodes_counter));
4325 BUFFER_TRACE(sbh, "marking dirty");
4326 ext4_superblock_csum_set(sb);
4327 mark_buffer_dirty(sbh);
4329 error = __sync_dirty_buffer(sbh,
4330 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4334 error = buffer_write_io_error(sbh);
4336 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4338 clear_buffer_write_io_error(sbh);
4339 set_buffer_uptodate(sbh);
4346 * Have we just finished recovery? If so, and if we are mounting (or
4347 * remounting) the filesystem readonly, then we will end up with a
4348 * consistent fs on disk. Record that fact.
4350 static void ext4_mark_recovery_complete(struct super_block *sb,
4351 struct ext4_super_block *es)
4353 journal_t *journal = EXT4_SB(sb)->s_journal;
4355 if (!ext4_has_feature_journal(sb)) {
4356 BUG_ON(journal != NULL);
4359 jbd2_journal_lock_updates(journal);
4360 if (jbd2_journal_flush(journal) < 0)
4363 if (ext4_has_feature_journal_needs_recovery(sb) &&
4364 sb->s_flags & MS_RDONLY) {
4365 ext4_clear_feature_journal_needs_recovery(sb);
4366 ext4_commit_super(sb, 1);
4370 jbd2_journal_unlock_updates(journal);
4374 * If we are mounting (or read-write remounting) a filesystem whose journal
4375 * has recorded an error from a previous lifetime, move that error to the
4376 * main filesystem now.
4378 static void ext4_clear_journal_err(struct super_block *sb,
4379 struct ext4_super_block *es)
4385 BUG_ON(!ext4_has_feature_journal(sb));
4387 journal = EXT4_SB(sb)->s_journal;
4390 * Now check for any error status which may have been recorded in the
4391 * journal by a prior ext4_error() or ext4_abort()
4394 j_errno = jbd2_journal_errno(journal);
4398 errstr = ext4_decode_error(sb, j_errno, nbuf);
4399 ext4_warning(sb, "Filesystem error recorded "
4400 "from previous mount: %s", errstr);
4401 ext4_warning(sb, "Marking fs in need of filesystem check.");
4403 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4404 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4405 ext4_commit_super(sb, 1);
4407 jbd2_journal_clear_err(journal);
4408 jbd2_journal_update_sb_errno(journal);
4413 * Force the running and committing transactions to commit,
4414 * and wait on the commit.
4416 int ext4_force_commit(struct super_block *sb)
4420 if (sb->s_flags & MS_RDONLY)
4423 journal = EXT4_SB(sb)->s_journal;
4424 return ext4_journal_force_commit(journal);
4427 static int ext4_sync_fs(struct super_block *sb, int wait)
4431 bool needs_barrier = false;
4432 struct ext4_sb_info *sbi = EXT4_SB(sb);
4434 trace_ext4_sync_fs(sb, wait);
4435 flush_workqueue(sbi->rsv_conversion_wq);
4437 * Writeback quota in non-journalled quota case - journalled quota has
4440 dquot_writeback_dquots(sb, -1);
4442 * Data writeback is possible w/o journal transaction, so barrier must
4443 * being sent at the end of the function. But we can skip it if
4444 * transaction_commit will do it for us.
4446 if (sbi->s_journal) {
4447 target = jbd2_get_latest_transaction(sbi->s_journal);
4448 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4449 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4450 needs_barrier = true;
4452 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4454 ret = jbd2_log_wait_commit(sbi->s_journal,
4457 } else if (wait && test_opt(sb, BARRIER))
4458 needs_barrier = true;
4459 if (needs_barrier) {
4461 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4470 * LVM calls this function before a (read-only) snapshot is created. This
4471 * gives us a chance to flush the journal completely and mark the fs clean.
4473 * Note that only this function cannot bring a filesystem to be in a clean
4474 * state independently. It relies on upper layer to stop all data & metadata
4477 static int ext4_freeze(struct super_block *sb)
4482 if (sb->s_flags & MS_RDONLY)
4485 journal = EXT4_SB(sb)->s_journal;
4488 /* Now we set up the journal barrier. */
4489 jbd2_journal_lock_updates(journal);
4492 * Don't clear the needs_recovery flag if we failed to
4493 * flush the journal.
4495 error = jbd2_journal_flush(journal);
4499 /* Journal blocked and flushed, clear needs_recovery flag. */
4500 ext4_clear_feature_journal_needs_recovery(sb);
4503 error = ext4_commit_super(sb, 1);
4506 /* we rely on upper layer to stop further updates */
4507 jbd2_journal_unlock_updates(journal);
4512 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4513 * flag here, even though the filesystem is not technically dirty yet.
4515 static int ext4_unfreeze(struct super_block *sb)
4517 if (sb->s_flags & MS_RDONLY)
4520 if (EXT4_SB(sb)->s_journal) {
4521 /* Reset the needs_recovery flag before the fs is unlocked. */
4522 ext4_set_feature_journal_needs_recovery(sb);
4525 ext4_commit_super(sb, 1);
4530 * Structure to save mount options for ext4_remount's benefit
4532 struct ext4_mount_options {
4533 unsigned long s_mount_opt;
4534 unsigned long s_mount_opt2;
4537 unsigned long s_commit_interval;
4538 u32 s_min_batch_time, s_max_batch_time;
4541 char *s_qf_names[EXT4_MAXQUOTAS];
4545 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4547 struct ext4_super_block *es;
4548 struct ext4_sb_info *sbi = EXT4_SB(sb);
4549 unsigned long old_sb_flags;
4550 struct ext4_mount_options old_opts;
4551 int enable_quota = 0;
4553 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4558 char *orig_data = kstrdup(data, GFP_KERNEL);
4560 /* Store the original options */
4561 old_sb_flags = sb->s_flags;
4562 old_opts.s_mount_opt = sbi->s_mount_opt;
4563 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4564 old_opts.s_resuid = sbi->s_resuid;
4565 old_opts.s_resgid = sbi->s_resgid;
4566 old_opts.s_commit_interval = sbi->s_commit_interval;
4567 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4568 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4570 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4571 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4572 if (sbi->s_qf_names[i]) {
4573 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4575 if (!old_opts.s_qf_names[i]) {
4576 for (j = 0; j < i; j++)
4577 kfree(old_opts.s_qf_names[j]);
4582 old_opts.s_qf_names[i] = NULL;
4584 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4585 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4587 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4592 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4593 test_opt(sb, JOURNAL_CHECKSUM)) {
4594 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4595 "during remount not supported; ignoring");
4596 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4599 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4600 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4601 ext4_msg(sb, KERN_ERR, "can't mount with "
4602 "both data=journal and delalloc");
4606 if (test_opt(sb, DIOREAD_NOLOCK)) {
4607 ext4_msg(sb, KERN_ERR, "can't mount with "
4608 "both data=journal and dioread_nolock");
4612 if (test_opt(sb, DAX)) {
4613 ext4_msg(sb, KERN_ERR, "can't mount with "
4614 "both data=journal and dax");
4620 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4621 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4622 "dax flag with busy inodes while remounting");
4623 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4626 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4627 ext4_abort(sb, "Abort forced by user");
4629 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4630 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4634 if (sbi->s_journal) {
4635 ext4_init_journal_params(sb, sbi->s_journal);
4636 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4639 if (*flags & MS_LAZYTIME)
4640 sb->s_flags |= MS_LAZYTIME;
4642 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4643 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4648 if (*flags & MS_RDONLY) {
4649 err = sync_filesystem(sb);
4652 err = dquot_suspend(sb, -1);
4657 * First of all, the unconditional stuff we have to do
4658 * to disable replay of the journal when we next remount
4660 sb->s_flags |= MS_RDONLY;
4663 * OK, test if we are remounting a valid rw partition
4664 * readonly, and if so set the rdonly flag and then
4665 * mark the partition as valid again.
4667 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4668 (sbi->s_mount_state & EXT4_VALID_FS))
4669 es->s_state = cpu_to_le16(sbi->s_mount_state);
4672 ext4_mark_recovery_complete(sb, es);
4674 /* Make sure we can mount this feature set readwrite */
4675 if (ext4_has_feature_readonly(sb) ||
4676 !ext4_feature_set_ok(sb, 0)) {
4681 * Make sure the group descriptor checksums
4682 * are sane. If they aren't, refuse to remount r/w.
4684 for (g = 0; g < sbi->s_groups_count; g++) {
4685 struct ext4_group_desc *gdp =
4686 ext4_get_group_desc(sb, g, NULL);
4688 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4689 ext4_msg(sb, KERN_ERR,
4690 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4691 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4692 le16_to_cpu(gdp->bg_checksum));
4699 * If we have an unprocessed orphan list hanging
4700 * around from a previously readonly bdev mount,
4701 * require a full umount/remount for now.
4703 if (es->s_last_orphan) {
4704 ext4_msg(sb, KERN_WARNING, "Couldn't "
4705 "remount RDWR because of unprocessed "
4706 "orphan inode list. Please "
4707 "umount/remount instead");
4713 * Mounting a RDONLY partition read-write, so reread
4714 * and store the current valid flag. (It may have
4715 * been changed by e2fsck since we originally mounted
4719 ext4_clear_journal_err(sb, es);
4720 sbi->s_mount_state = le16_to_cpu(es->s_state);
4721 if (!ext4_setup_super(sb, es, 0))
4722 sb->s_flags &= ~MS_RDONLY;
4723 if (ext4_has_feature_mmp(sb))
4724 if (ext4_multi_mount_protect(sb,
4725 le64_to_cpu(es->s_mmp_block))) {
4734 * Reinitialize lazy itable initialization thread based on
4737 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4738 ext4_unregister_li_request(sb);
4740 ext4_group_t first_not_zeroed;
4741 first_not_zeroed = ext4_has_uninit_itable(sb);
4742 ext4_register_li_request(sb, first_not_zeroed);
4745 ext4_setup_system_zone(sb);
4746 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4747 ext4_commit_super(sb, 1);
4750 /* Release old quota file names */
4751 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4752 kfree(old_opts.s_qf_names[i]);
4754 if (sb_any_quota_suspended(sb))
4755 dquot_resume(sb, -1);
4756 else if (ext4_has_feature_quota(sb)) {
4757 err = ext4_enable_quotas(sb);
4764 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4765 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4770 sb->s_flags = old_sb_flags;
4771 sbi->s_mount_opt = old_opts.s_mount_opt;
4772 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4773 sbi->s_resuid = old_opts.s_resuid;
4774 sbi->s_resgid = old_opts.s_resgid;
4775 sbi->s_commit_interval = old_opts.s_commit_interval;
4776 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4777 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4779 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4780 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4781 kfree(sbi->s_qf_names[i]);
4782 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4789 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4791 struct super_block *sb = dentry->d_sb;
4792 struct ext4_sb_info *sbi = EXT4_SB(sb);
4793 struct ext4_super_block *es = sbi->s_es;
4794 ext4_fsblk_t overhead = 0, resv_blocks;
4797 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4799 if (!test_opt(sb, MINIX_DF))
4800 overhead = sbi->s_overhead;
4802 buf->f_type = EXT4_SUPER_MAGIC;
4803 buf->f_bsize = sb->s_blocksize;
4804 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4805 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4806 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4807 /* prevent underflow in case that few free space is available */
4808 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4809 buf->f_bavail = buf->f_bfree -
4810 (ext4_r_blocks_count(es) + resv_blocks);
4811 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4813 buf->f_files = le32_to_cpu(es->s_inodes_count);
4814 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4815 buf->f_namelen = EXT4_NAME_LEN;
4816 fsid = le64_to_cpup((void *)es->s_uuid) ^
4817 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4818 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4819 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4824 /* Helper function for writing quotas on sync - we need to start transaction
4825 * before quota file is locked for write. Otherwise the are possible deadlocks:
4826 * Process 1 Process 2
4827 * ext4_create() quota_sync()
4828 * jbd2_journal_start() write_dquot()
4829 * dquot_initialize() down(dqio_mutex)
4830 * down(dqio_mutex) jbd2_journal_start()
4836 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4838 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4841 static int ext4_write_dquot(struct dquot *dquot)
4845 struct inode *inode;
4847 inode = dquot_to_inode(dquot);
4848 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4849 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4851 return PTR_ERR(handle);
4852 ret = dquot_commit(dquot);
4853 err = ext4_journal_stop(handle);
4859 static int ext4_acquire_dquot(struct dquot *dquot)
4864 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4865 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4867 return PTR_ERR(handle);
4868 ret = dquot_acquire(dquot);
4869 err = ext4_journal_stop(handle);
4875 static int ext4_release_dquot(struct dquot *dquot)
4880 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4881 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4882 if (IS_ERR(handle)) {
4883 /* Release dquot anyway to avoid endless cycle in dqput() */
4884 dquot_release(dquot);
4885 return PTR_ERR(handle);
4887 ret = dquot_release(dquot);
4888 err = ext4_journal_stop(handle);
4894 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4896 struct super_block *sb = dquot->dq_sb;
4897 struct ext4_sb_info *sbi = EXT4_SB(sb);
4899 /* Are we journaling quotas? */
4900 if (ext4_has_feature_quota(sb) ||
4901 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4902 dquot_mark_dquot_dirty(dquot);
4903 return ext4_write_dquot(dquot);
4905 return dquot_mark_dquot_dirty(dquot);
4909 static int ext4_write_info(struct super_block *sb, int type)
4914 /* Data block + inode block */
4915 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
4917 return PTR_ERR(handle);
4918 ret = dquot_commit_info(sb, type);
4919 err = ext4_journal_stop(handle);
4926 * Turn on quotas during mount time - we need to find
4927 * the quota file and such...
4929 static int ext4_quota_on_mount(struct super_block *sb, int type)
4931 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4932 EXT4_SB(sb)->s_jquota_fmt, type);
4936 * Standard function to be called on quota_on
4938 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4943 if (!test_opt(sb, QUOTA))
4946 /* Quotafile not on the same filesystem? */
4947 if (path->dentry->d_sb != sb)
4949 /* Journaling quota? */
4950 if (EXT4_SB(sb)->s_qf_names[type]) {
4951 /* Quotafile not in fs root? */
4952 if (path->dentry->d_parent != sb->s_root)
4953 ext4_msg(sb, KERN_WARNING,
4954 "Quota file not on filesystem root. "
4955 "Journaled quota will not work");
4959 * When we journal data on quota file, we have to flush journal to see
4960 * all updates to the file when we bypass pagecache...
4962 if (EXT4_SB(sb)->s_journal &&
4963 ext4_should_journal_data(d_inode(path->dentry))) {
4965 * We don't need to lock updates but journal_flush() could
4966 * otherwise be livelocked...
4968 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4969 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4970 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4975 return dquot_quota_on(sb, type, format_id, path);
4978 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4982 struct inode *qf_inode;
4983 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4984 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4985 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4988 BUG_ON(!ext4_has_feature_quota(sb));
4990 if (!qf_inums[type])
4993 qf_inode = ext4_iget(sb, qf_inums[type]);
4994 if (IS_ERR(qf_inode)) {
4995 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4996 return PTR_ERR(qf_inode);
4999 /* Don't account quota for quota files to avoid recursion */
5000 qf_inode->i_flags |= S_NOQUOTA;
5001 err = dquot_enable(qf_inode, type, format_id, flags);
5007 /* Enable usage tracking for all quota types. */
5008 static int ext4_enable_quotas(struct super_block *sb)
5011 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5012 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5013 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5016 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5017 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5018 if (qf_inums[type]) {
5019 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5020 DQUOT_USAGE_ENABLED);
5023 "Failed to enable quota tracking "
5024 "(type=%d, err=%d). Please run "
5025 "e2fsck to fix.", type, err);
5033 static int ext4_quota_off(struct super_block *sb, int type)
5035 struct inode *inode = sb_dqopt(sb)->files[type];
5038 /* Force all delayed allocation blocks to be allocated.
5039 * Caller already holds s_umount sem */
5040 if (test_opt(sb, DELALLOC))
5041 sync_filesystem(sb);
5046 /* Update modification times of quota files when userspace can
5047 * start looking at them */
5048 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5051 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5052 ext4_mark_inode_dirty(handle, inode);
5053 ext4_journal_stop(handle);
5056 return dquot_quota_off(sb, type);
5059 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5060 * acquiring the locks... As quota files are never truncated and quota code
5061 * itself serializes the operations (and no one else should touch the files)
5062 * we don't have to be afraid of races */
5063 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5064 size_t len, loff_t off)
5066 struct inode *inode = sb_dqopt(sb)->files[type];
5067 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5068 int offset = off & (sb->s_blocksize - 1);
5071 struct buffer_head *bh;
5072 loff_t i_size = i_size_read(inode);
5076 if (off+len > i_size)
5079 while (toread > 0) {
5080 tocopy = sb->s_blocksize - offset < toread ?
5081 sb->s_blocksize - offset : toread;
5082 bh = ext4_bread(NULL, inode, blk, 0);
5085 if (!bh) /* A hole? */
5086 memset(data, 0, tocopy);
5088 memcpy(data, bh->b_data+offset, tocopy);
5098 /* Write to quotafile (we know the transaction is already started and has
5099 * enough credits) */
5100 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5101 const char *data, size_t len, loff_t off)
5103 struct inode *inode = sb_dqopt(sb)->files[type];
5104 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5105 int err, offset = off & (sb->s_blocksize - 1);
5107 struct buffer_head *bh;
5108 handle_t *handle = journal_current_handle();
5110 if (EXT4_SB(sb)->s_journal && !handle) {
5111 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5112 " cancelled because transaction is not started",
5113 (unsigned long long)off, (unsigned long long)len);
5117 * Since we account only one data block in transaction credits,
5118 * then it is impossible to cross a block boundary.
5120 if (sb->s_blocksize - offset < len) {
5121 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5122 " cancelled because not block aligned",
5123 (unsigned long long)off, (unsigned long long)len);
5128 bh = ext4_bread(handle, inode, blk,
5129 EXT4_GET_BLOCKS_CREATE |
5130 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5131 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5132 ext4_should_retry_alloc(inode->i_sb, &retries));
5137 BUFFER_TRACE(bh, "get write access");
5138 err = ext4_journal_get_write_access(handle, bh);
5144 memcpy(bh->b_data+offset, data, len);
5145 flush_dcache_page(bh->b_page);
5147 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5150 if (inode->i_size < off + len) {
5151 i_size_write(inode, off + len);
5152 EXT4_I(inode)->i_disksize = inode->i_size;
5153 ext4_mark_inode_dirty(handle, inode);
5160 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5161 const char *dev_name, void *data)
5163 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5166 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5167 static inline void register_as_ext2(void)
5169 int err = register_filesystem(&ext2_fs_type);
5172 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5175 static inline void unregister_as_ext2(void)
5177 unregister_filesystem(&ext2_fs_type);
5180 static inline int ext2_feature_set_ok(struct super_block *sb)
5182 if (ext4_has_unknown_ext2_incompat_features(sb))
5184 if (sb->s_flags & MS_RDONLY)
5186 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5191 static inline void register_as_ext2(void) { }
5192 static inline void unregister_as_ext2(void) { }
5193 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5196 static inline void register_as_ext3(void)
5198 int err = register_filesystem(&ext3_fs_type);
5201 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5204 static inline void unregister_as_ext3(void)
5206 unregister_filesystem(&ext3_fs_type);
5209 static inline int ext3_feature_set_ok(struct super_block *sb)
5211 if (ext4_has_unknown_ext3_incompat_features(sb))
5213 if (!ext4_has_feature_journal(sb))
5215 if (sb->s_flags & MS_RDONLY)
5217 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5222 static struct file_system_type ext4_fs_type = {
5223 .owner = THIS_MODULE,
5225 .mount = ext4_mount,
5226 .kill_sb = kill_block_super,
5227 .fs_flags = FS_REQUIRES_DEV,
5229 MODULE_ALIAS_FS("ext4");
5231 /* Shared across all ext4 file systems */
5232 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5233 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5235 static int __init ext4_init_fs(void)
5239 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5240 ext4_li_info = NULL;
5241 mutex_init(&ext4_li_mtx);
5243 /* Build-time check for flags consistency */
5244 ext4_check_flag_values();
5246 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5247 mutex_init(&ext4__aio_mutex[i]);
5248 init_waitqueue_head(&ext4__ioend_wq[i]);
5251 err = ext4_init_es();
5255 err = ext4_init_pageio();
5259 err = ext4_init_system_zone();
5263 err = ext4_init_sysfs();
5267 err = ext4_init_mballoc();
5271 ext4_mballoc_ready = 1;
5272 err = init_inodecache();
5277 err = register_filesystem(&ext4_fs_type);
5283 unregister_as_ext2();
5284 unregister_as_ext3();
5285 destroy_inodecache();
5287 ext4_mballoc_ready = 0;
5288 ext4_exit_mballoc();
5292 ext4_exit_system_zone();
5301 static void __exit ext4_exit_fs(void)
5304 ext4_destroy_lazyinit_thread();
5305 unregister_as_ext2();
5306 unregister_as_ext3();
5307 unregister_filesystem(&ext4_fs_type);
5308 destroy_inodecache();
5309 ext4_exit_mballoc();
5311 ext4_exit_system_zone();
5316 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5317 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5318 MODULE_LICENSE("GPL");
5319 module_init(ext4_init_fs)
5320 module_exit(ext4_exit_fs)