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 struct ratelimit_state ext4_mount_msg_ratelimit;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
85 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
86 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
90 * page lock -> i_data_sem (rw)
92 * buffered write path:
93 * sb_start_write -> i_mutex -> mmap_sem
94 * sb_start_write -> i_mutex -> transaction start -> page lock ->
98 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
99 * i_mmap_rwsem (w) -> page lock
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * transaction start -> i_data_sem (rw)
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
106 * transaction start -> i_data_sem (rw)
109 * transaction start -> page lock(s) -> i_data_sem (rw)
112 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
113 static struct file_system_type ext2_fs_type = {
114 .owner = THIS_MODULE,
117 .kill_sb = kill_block_super,
118 .fs_flags = FS_REQUIRES_DEV,
120 MODULE_ALIAS_FS("ext2");
121 MODULE_ALIAS("ext2");
122 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
124 #define IS_EXT2_SB(sb) (0)
128 static struct file_system_type ext3_fs_type = {
129 .owner = THIS_MODULE,
132 .kill_sb = kill_block_super,
133 .fs_flags = FS_REQUIRES_DEV,
135 MODULE_ALIAS_FS("ext3");
136 MODULE_ALIAS("ext3");
137 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
139 static int ext4_verify_csum_type(struct super_block *sb,
140 struct ext4_super_block *es)
142 if (!ext4_has_feature_metadata_csum(sb))
145 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148 static __le32 ext4_superblock_csum(struct super_block *sb,
149 struct ext4_super_block *es)
151 struct ext4_sb_info *sbi = EXT4_SB(sb);
152 int offset = offsetof(struct ext4_super_block, s_checksum);
155 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
157 return cpu_to_le32(csum);
160 static int ext4_superblock_csum_verify(struct super_block *sb,
161 struct ext4_super_block *es)
163 if (!ext4_has_metadata_csum(sb))
166 return es->s_checksum == ext4_superblock_csum(sb, es);
169 void ext4_superblock_csum_set(struct super_block *sb)
171 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
173 if (!ext4_has_metadata_csum(sb))
176 es->s_checksum = ext4_superblock_csum(sb, es);
179 void *ext4_kvmalloc(size_t size, gfp_t flags)
183 ret = kmalloc(size, flags | __GFP_NOWARN);
185 ret = __vmalloc(size, flags, PAGE_KERNEL);
189 void *ext4_kvzalloc(size_t size, gfp_t flags)
193 ret = kzalloc(size, flags | __GFP_NOWARN);
195 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
199 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
202 return le32_to_cpu(bg->bg_block_bitmap_lo) |
203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
207 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
210 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
215 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
216 struct ext4_group_desc *bg)
218 return le32_to_cpu(bg->bg_inode_table_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
223 __u32 ext4_free_group_clusters(struct super_block *sb,
224 struct ext4_group_desc *bg)
226 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
231 __u32 ext4_free_inodes_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
234 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
239 __u32 ext4_used_dirs_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
242 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
247 __u32 ext4_itable_unused_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
250 return le16_to_cpu(bg->bg_itable_unused_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
255 void ext4_block_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
258 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
263 void ext4_inode_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
266 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
271 void ext4_inode_table_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
274 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
279 void ext4_free_group_clusters_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
282 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287 void ext4_free_inodes_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
290 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295 void ext4_used_dirs_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
298 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303 void ext4_itable_unused_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
306 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
312 static void __save_error_info(struct super_block *sb, const char *func,
315 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
317 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
318 if (bdev_read_only(sb->s_bdev))
320 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
321 es->s_last_error_time = cpu_to_le32(get_seconds());
322 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
323 es->s_last_error_line = cpu_to_le32(line);
324 if (!es->s_first_error_time) {
325 es->s_first_error_time = es->s_last_error_time;
326 strncpy(es->s_first_error_func, func,
327 sizeof(es->s_first_error_func));
328 es->s_first_error_line = cpu_to_le32(line);
329 es->s_first_error_ino = es->s_last_error_ino;
330 es->s_first_error_block = es->s_last_error_block;
333 * Start the daily error reporting function if it hasn't been
336 if (!es->s_error_count)
337 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
338 le32_add_cpu(&es->s_error_count, 1);
341 static void save_error_info(struct super_block *sb, const char *func,
344 __save_error_info(sb, func, line);
345 ext4_commit_super(sb, 1);
349 * The del_gendisk() function uninitializes the disk-specific data
350 * structures, including the bdi structure, without telling anyone
351 * else. Once this happens, any attempt to call mark_buffer_dirty()
352 * (for example, by ext4_commit_super), will cause a kernel OOPS.
353 * This is a kludge to prevent these oops until we can put in a proper
354 * hook in del_gendisk() to inform the VFS and file system layers.
356 static int block_device_ejected(struct super_block *sb)
358 struct inode *bd_inode = sb->s_bdev->bd_inode;
359 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
361 return bdi->dev == NULL;
364 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
366 struct super_block *sb = journal->j_private;
367 struct ext4_sb_info *sbi = EXT4_SB(sb);
368 int error = is_journal_aborted(journal);
369 struct ext4_journal_cb_entry *jce;
371 BUG_ON(txn->t_state == T_FINISHED);
372 spin_lock(&sbi->s_md_lock);
373 while (!list_empty(&txn->t_private_list)) {
374 jce = list_entry(txn->t_private_list.next,
375 struct ext4_journal_cb_entry, jce_list);
376 list_del_init(&jce->jce_list);
377 spin_unlock(&sbi->s_md_lock);
378 jce->jce_func(sb, jce, error);
379 spin_lock(&sbi->s_md_lock);
381 spin_unlock(&sbi->s_md_lock);
384 /* Deal with the reporting of failure conditions on a filesystem such as
385 * inconsistencies detected or read IO failures.
387 * On ext2, we can store the error state of the filesystem in the
388 * superblock. That is not possible on ext4, because we may have other
389 * write ordering constraints on the superblock which prevent us from
390 * writing it out straight away; and given that the journal is about to
391 * be aborted, we can't rely on the current, or future, transactions to
392 * write out the superblock safely.
394 * We'll just use the jbd2_journal_abort() error code to record an error in
395 * the journal instead. On recovery, the journal will complain about
396 * that error until we've noted it down and cleared it.
399 static void ext4_handle_error(struct super_block *sb)
401 if (sb->s_flags & MS_RDONLY)
404 if (!test_opt(sb, ERRORS_CONT)) {
405 journal_t *journal = EXT4_SB(sb)->s_journal;
407 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
409 jbd2_journal_abort(journal, -EIO);
411 if (test_opt(sb, ERRORS_RO)) {
412 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
414 * Make sure updated value of ->s_mount_flags will be visible
415 * before ->s_flags update
418 sb->s_flags |= MS_RDONLY;
420 if (test_opt(sb, ERRORS_PANIC)) {
421 if (EXT4_SB(sb)->s_journal &&
422 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
424 panic("EXT4-fs (device %s): panic forced after error\n",
429 #define ext4_error_ratelimit(sb) \
430 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 void __ext4_error(struct super_block *sb, const char *function,
434 unsigned int line, const char *fmt, ...)
436 struct va_format vaf;
439 if (ext4_error_ratelimit(sb)) {
444 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
445 sb->s_id, function, line, current->comm, &vaf);
448 save_error_info(sb, function, line);
449 ext4_handle_error(sb);
452 void __ext4_error_inode(struct inode *inode, const char *function,
453 unsigned int line, ext4_fsblk_t block,
454 const char *fmt, ...)
457 struct va_format vaf;
458 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
460 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
461 es->s_last_error_block = cpu_to_le64(block);
462 if (ext4_error_ratelimit(inode->i_sb)) {
467 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
468 "inode #%lu: block %llu: comm %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, &vaf);
472 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
473 "inode #%lu: comm %s: %pV\n",
474 inode->i_sb->s_id, function, line, inode->i_ino,
475 current->comm, &vaf);
478 save_error_info(inode->i_sb, function, line);
479 ext4_handle_error(inode->i_sb);
482 void __ext4_error_file(struct file *file, const char *function,
483 unsigned int line, ext4_fsblk_t block,
484 const char *fmt, ...)
487 struct va_format vaf;
488 struct ext4_super_block *es;
489 struct inode *inode = file_inode(file);
490 char pathname[80], *path;
492 es = EXT4_SB(inode->i_sb)->s_es;
493 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
494 if (ext4_error_ratelimit(inode->i_sb)) {
495 path = file_path(file, pathname, sizeof(pathname));
503 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
504 "block %llu: comm %s: path %s: %pV\n",
505 inode->i_sb->s_id, function, line, inode->i_ino,
506 block, current->comm, path, &vaf);
509 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
510 "comm %s: path %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, path, &vaf);
515 save_error_info(inode->i_sb, function, line);
516 ext4_handle_error(inode->i_sb);
519 const char *ext4_decode_error(struct super_block *sb, int errno,
526 errstr = "Corrupt filesystem";
529 errstr = "Filesystem failed CRC";
532 errstr = "IO failure";
535 errstr = "Out of memory";
538 if (!sb || (EXT4_SB(sb)->s_journal &&
539 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
540 errstr = "Journal has aborted";
542 errstr = "Readonly filesystem";
545 /* If the caller passed in an extra buffer for unknown
546 * errors, textualise them now. Else we just return
549 /* Check for truncated error codes... */
550 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
559 /* __ext4_std_error decodes expected errors from journaling functions
560 * automatically and invokes the appropriate error response. */
562 void __ext4_std_error(struct super_block *sb, const char *function,
563 unsigned int line, int errno)
568 /* Special case: if the error is EROFS, and we're not already
569 * inside a transaction, then there's really no point in logging
571 if (errno == -EROFS && journal_current_handle() == NULL &&
572 (sb->s_flags & MS_RDONLY))
575 if (ext4_error_ratelimit(sb)) {
576 errstr = ext4_decode_error(sb, errno, nbuf);
577 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
578 sb->s_id, function, line, errstr);
581 save_error_info(sb, function, line);
582 ext4_handle_error(sb);
586 * ext4_abort is a much stronger failure handler than ext4_error. The
587 * abort function may be used to deal with unrecoverable failures such
588 * as journal IO errors or ENOMEM at a critical moment in log management.
590 * We unconditionally force the filesystem into an ABORT|READONLY state,
591 * unless the error response on the fs has been set to panic in which
592 * case we take the easy way out and panic immediately.
595 void __ext4_abort(struct super_block *sb, const char *function,
596 unsigned int line, const char *fmt, ...)
600 save_error_info(sb, function, line);
602 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
608 if ((sb->s_flags & MS_RDONLY) == 0) {
609 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
610 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
612 * Make sure updated value of ->s_mount_flags will be visible
613 * before ->s_flags update
616 sb->s_flags |= MS_RDONLY;
617 if (EXT4_SB(sb)->s_journal)
618 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
619 save_error_info(sb, function, line);
621 if (test_opt(sb, ERRORS_PANIC)) {
622 if (EXT4_SB(sb)->s_journal &&
623 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
625 panic("EXT4-fs panic from previous error\n");
629 void __ext4_msg(struct super_block *sb,
630 const char *prefix, const char *fmt, ...)
632 struct va_format vaf;
635 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
641 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
645 #define ext4_warning_ratelimit(sb) \
646 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
649 void __ext4_warning(struct super_block *sb, const char *function,
650 unsigned int line, const char *fmt, ...)
652 struct va_format vaf;
655 if (!ext4_warning_ratelimit(sb))
661 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
662 sb->s_id, function, line, &vaf);
666 void __ext4_warning_inode(const struct inode *inode, const char *function,
667 unsigned int line, const char *fmt, ...)
669 struct va_format vaf;
672 if (!ext4_warning_ratelimit(inode->i_sb))
678 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
679 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
680 function, line, inode->i_ino, current->comm, &vaf);
684 void __ext4_grp_locked_error(const char *function, unsigned int line,
685 struct super_block *sb, ext4_group_t grp,
686 unsigned long ino, ext4_fsblk_t block,
687 const char *fmt, ...)
691 struct va_format vaf;
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
695 es->s_last_error_ino = cpu_to_le32(ino);
696 es->s_last_error_block = cpu_to_le64(block);
697 __save_error_info(sb, function, line);
699 if (ext4_error_ratelimit(sb)) {
703 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
704 sb->s_id, function, line, grp);
706 printk(KERN_CONT "inode %lu: ", ino);
708 printk(KERN_CONT "block %llu:",
709 (unsigned long long) block);
710 printk(KERN_CONT "%pV\n", &vaf);
714 if (test_opt(sb, ERRORS_CONT)) {
715 ext4_commit_super(sb, 0);
719 ext4_unlock_group(sb, grp);
720 ext4_handle_error(sb);
722 * We only get here in the ERRORS_RO case; relocking the group
723 * may be dangerous, but nothing bad will happen since the
724 * filesystem will have already been marked read/only and the
725 * journal has been aborted. We return 1 as a hint to callers
726 * who might what to use the return value from
727 * ext4_grp_locked_error() to distinguish between the
728 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
729 * aggressively from the ext4 function in question, with a
730 * more appropriate error code.
732 ext4_lock_group(sb, grp);
736 void ext4_update_dynamic_rev(struct super_block *sb)
738 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
740 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
744 "updating to rev %d because of new feature flag, "
745 "running e2fsck is recommended",
748 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
749 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
750 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
751 /* leave es->s_feature_*compat flags alone */
752 /* es->s_uuid will be set by e2fsck if empty */
755 * The rest of the superblock fields should be zero, and if not it
756 * means they are likely already in use, so leave them alone. We
757 * can leave it up to e2fsck to clean up any inconsistencies there.
762 * Open the external journal device
764 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
766 struct block_device *bdev;
767 char b[BDEVNAME_SIZE];
769 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
775 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
776 __bdevname(dev, b), PTR_ERR(bdev));
781 * Release the journal device
783 static void ext4_blkdev_put(struct block_device *bdev)
785 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
788 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
790 struct block_device *bdev;
791 bdev = sbi->journal_bdev;
793 ext4_blkdev_put(bdev);
794 sbi->journal_bdev = NULL;
798 static inline struct inode *orphan_list_entry(struct list_head *l)
800 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
803 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
807 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
808 le32_to_cpu(sbi->s_es->s_last_orphan));
810 printk(KERN_ERR "sb_info orphan list:\n");
811 list_for_each(l, &sbi->s_orphan) {
812 struct inode *inode = orphan_list_entry(l);
814 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
815 inode->i_sb->s_id, inode->i_ino, inode,
816 inode->i_mode, inode->i_nlink,
821 static void ext4_put_super(struct super_block *sb)
823 struct ext4_sb_info *sbi = EXT4_SB(sb);
824 struct ext4_super_block *es = sbi->s_es;
827 ext4_unregister_li_request(sb);
828 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
830 flush_workqueue(sbi->rsv_conversion_wq);
831 destroy_workqueue(sbi->rsv_conversion_wq);
833 if (sbi->s_journal) {
834 err = jbd2_journal_destroy(sbi->s_journal);
835 sbi->s_journal = NULL;
837 ext4_abort(sb, "Couldn't clean up the journal");
840 ext4_unregister_sysfs(sb);
841 ext4_es_unregister_shrinker(sbi);
842 del_timer_sync(&sbi->s_err_report);
843 ext4_release_system_zone(sb);
845 ext4_ext_release(sb);
847 if (!(sb->s_flags & MS_RDONLY)) {
848 ext4_clear_feature_journal_needs_recovery(sb);
849 es->s_state = cpu_to_le16(sbi->s_mount_state);
851 if (!(sb->s_flags & MS_RDONLY))
852 ext4_commit_super(sb, 1);
854 for (i = 0; i < sbi->s_gdb_count; i++)
855 brelse(sbi->s_group_desc[i]);
856 kvfree(sbi->s_group_desc);
857 kvfree(sbi->s_flex_groups);
858 percpu_counter_destroy(&sbi->s_freeclusters_counter);
859 percpu_counter_destroy(&sbi->s_freeinodes_counter);
860 percpu_counter_destroy(&sbi->s_dirs_counter);
861 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
862 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
865 for (i = 0; i < EXT4_MAXQUOTAS; i++)
866 kfree(sbi->s_qf_names[i]);
869 /* Debugging code just in case the in-memory inode orphan list
870 * isn't empty. The on-disk one can be non-empty if we've
871 * detected an error and taken the fs readonly, but the
872 * in-memory list had better be clean by this point. */
873 if (!list_empty(&sbi->s_orphan))
874 dump_orphan_list(sb, sbi);
875 J_ASSERT(list_empty(&sbi->s_orphan));
877 sync_blockdev(sb->s_bdev);
878 invalidate_bdev(sb->s_bdev);
879 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
881 * Invalidate the journal device's buffers. We don't want them
882 * floating about in memory - the physical journal device may
883 * hotswapped, and it breaks the `ro-after' testing code.
885 sync_blockdev(sbi->journal_bdev);
886 invalidate_bdev(sbi->journal_bdev);
887 ext4_blkdev_remove(sbi);
889 if (sbi->s_mb_cache) {
890 ext4_xattr_destroy_cache(sbi->s_mb_cache);
891 sbi->s_mb_cache = NULL;
894 kthread_stop(sbi->s_mmp_tsk);
895 sb->s_fs_info = NULL;
897 * Now that we are completely done shutting down the
898 * superblock, we need to actually destroy the kobject.
900 kobject_put(&sbi->s_kobj);
901 wait_for_completion(&sbi->s_kobj_unregister);
902 if (sbi->s_chksum_driver)
903 crypto_free_shash(sbi->s_chksum_driver);
904 kfree(sbi->s_blockgroup_lock);
908 static struct kmem_cache *ext4_inode_cachep;
911 * Called inside transaction, so use GFP_NOFS
913 static struct inode *ext4_alloc_inode(struct super_block *sb)
915 struct ext4_inode_info *ei;
917 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
921 ei->vfs_inode.i_version = 1;
922 spin_lock_init(&ei->i_raw_lock);
923 INIT_LIST_HEAD(&ei->i_prealloc_list);
924 spin_lock_init(&ei->i_prealloc_lock);
925 ext4_es_init_tree(&ei->i_es_tree);
926 rwlock_init(&ei->i_es_lock);
927 INIT_LIST_HEAD(&ei->i_es_list);
930 ei->i_es_shrink_lblk = 0;
931 ei->i_reserved_data_blocks = 0;
932 ei->i_reserved_meta_blocks = 0;
933 ei->i_allocated_meta_blocks = 0;
934 ei->i_da_metadata_calc_len = 0;
935 ei->i_da_metadata_calc_last_lblock = 0;
936 spin_lock_init(&(ei->i_block_reservation_lock));
938 ei->i_reserved_quota = 0;
939 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
942 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
943 spin_lock_init(&ei->i_completed_io_lock);
945 ei->i_datasync_tid = 0;
946 atomic_set(&ei->i_unwritten, 0);
947 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
948 #ifdef CONFIG_EXT4_FS_ENCRYPTION
949 ei->i_crypt_info = NULL;
951 return &ei->vfs_inode;
954 static int ext4_drop_inode(struct inode *inode)
956 int drop = generic_drop_inode(inode);
958 trace_ext4_drop_inode(inode, drop);
962 static void ext4_i_callback(struct rcu_head *head)
964 struct inode *inode = container_of(head, struct inode, i_rcu);
965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
968 static void ext4_destroy_inode(struct inode *inode)
970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
979 call_rcu(&inode->i_rcu, ext4_i_callback);
982 static void init_once(void *foo)
984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
986 INIT_LIST_HEAD(&ei->i_orphan);
987 init_rwsem(&ei->xattr_sem);
988 init_rwsem(&ei->i_data_sem);
989 init_rwsem(&ei->i_mmap_sem);
990 inode_init_once(&ei->vfs_inode);
993 static int __init init_inodecache(void)
995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
997 0, (SLAB_RECLAIM_ACCOUNT|
998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1000 if (ext4_inode_cachep == NULL)
1005 static void destroy_inodecache(void)
1008 * Make sure all delayed rcu free inodes are flushed before we
1012 kmem_cache_destroy(ext4_inode_cachep);
1015 void ext4_clear_inode(struct inode *inode)
1017 invalidate_inode_buffers(inode);
1020 ext4_discard_preallocations(inode);
1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1028 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1029 if (EXT4_I(inode)->i_crypt_info)
1030 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1034 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1035 u64 ino, u32 generation)
1037 struct inode *inode;
1039 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1040 return ERR_PTR(-ESTALE);
1041 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1042 return ERR_PTR(-ESTALE);
1044 /* iget isn't really right if the inode is currently unallocated!!
1046 * ext4_read_inode will return a bad_inode if the inode had been
1047 * deleted, so we should be safe.
1049 * Currently we don't know the generation for parent directory, so
1050 * a generation of 0 means "accept any"
1052 inode = ext4_iget_normal(sb, ino);
1054 return ERR_CAST(inode);
1055 if (generation && inode->i_generation != generation) {
1057 return ERR_PTR(-ESTALE);
1063 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1064 int fh_len, int fh_type)
1066 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1067 ext4_nfs_get_inode);
1070 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1071 int fh_len, int fh_type)
1073 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1074 ext4_nfs_get_inode);
1078 * Try to release metadata pages (indirect blocks, directories) which are
1079 * mapped via the block device. Since these pages could have journal heads
1080 * which would prevent try_to_free_buffers() from freeing them, we must use
1081 * jbd2 layer's try_to_free_buffers() function to release them.
1083 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1086 journal_t *journal = EXT4_SB(sb)->s_journal;
1088 WARN_ON(PageChecked(page));
1089 if (!page_has_buffers(page))
1092 return jbd2_journal_try_to_free_buffers(journal, page,
1093 wait & ~__GFP_DIRECT_RECLAIM);
1094 return try_to_free_buffers(page);
1098 static char *quotatypes[] = INITQFNAMES;
1099 #define QTYPE2NAME(t) (quotatypes[t])
1101 static int ext4_write_dquot(struct dquot *dquot);
1102 static int ext4_acquire_dquot(struct dquot *dquot);
1103 static int ext4_release_dquot(struct dquot *dquot);
1104 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1105 static int ext4_write_info(struct super_block *sb, int type);
1106 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1108 static int ext4_quota_off(struct super_block *sb, int type);
1109 static int ext4_quota_on_mount(struct super_block *sb, int type);
1110 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1111 size_t len, loff_t off);
1112 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1113 const char *data, size_t len, loff_t off);
1114 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1115 unsigned int flags);
1116 static int ext4_enable_quotas(struct super_block *sb);
1117 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1119 static struct dquot **ext4_get_dquots(struct inode *inode)
1121 return EXT4_I(inode)->i_dquot;
1124 static const struct dquot_operations ext4_quota_operations = {
1125 .get_reserved_space = ext4_get_reserved_space,
1126 .write_dquot = ext4_write_dquot,
1127 .acquire_dquot = ext4_acquire_dquot,
1128 .release_dquot = ext4_release_dquot,
1129 .mark_dirty = ext4_mark_dquot_dirty,
1130 .write_info = ext4_write_info,
1131 .alloc_dquot = dquot_alloc,
1132 .destroy_dquot = dquot_destroy,
1133 .get_projid = ext4_get_projid,
1134 .get_next_id = ext4_get_next_id,
1137 static const struct quotactl_ops ext4_qctl_operations = {
1138 .quota_on = ext4_quota_on,
1139 .quota_off = ext4_quota_off,
1140 .quota_sync = dquot_quota_sync,
1141 .get_state = dquot_get_state,
1142 .set_info = dquot_set_dqinfo,
1143 .get_dqblk = dquot_get_dqblk,
1144 .set_dqblk = dquot_set_dqblk,
1145 .get_nextdqblk = dquot_get_next_dqblk,
1149 static const struct super_operations ext4_sops = {
1150 .alloc_inode = ext4_alloc_inode,
1151 .destroy_inode = ext4_destroy_inode,
1152 .write_inode = ext4_write_inode,
1153 .dirty_inode = ext4_dirty_inode,
1154 .drop_inode = ext4_drop_inode,
1155 .evict_inode = ext4_evict_inode,
1156 .put_super = ext4_put_super,
1157 .sync_fs = ext4_sync_fs,
1158 .freeze_fs = ext4_freeze,
1159 .unfreeze_fs = ext4_unfreeze,
1160 .statfs = ext4_statfs,
1161 .remount_fs = ext4_remount,
1162 .show_options = ext4_show_options,
1164 .quota_read = ext4_quota_read,
1165 .quota_write = ext4_quota_write,
1166 .get_dquots = ext4_get_dquots,
1168 .bdev_try_to_free_page = bdev_try_to_free_page,
1171 static const struct export_operations ext4_export_ops = {
1172 .fh_to_dentry = ext4_fh_to_dentry,
1173 .fh_to_parent = ext4_fh_to_parent,
1174 .get_parent = ext4_get_parent,
1178 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1179 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1180 Opt_nouid32, Opt_debug, Opt_removed,
1181 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1182 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1183 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1184 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1185 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1186 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1187 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1188 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1189 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1190 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1191 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1192 Opt_lazytime, Opt_nolazytime,
1193 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1194 Opt_inode_readahead_blks, Opt_journal_ioprio,
1195 Opt_dioread_nolock, Opt_dioread_lock,
1196 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1197 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1200 static const match_table_t tokens = {
1201 {Opt_bsd_df, "bsddf"},
1202 {Opt_minix_df, "minixdf"},
1203 {Opt_grpid, "grpid"},
1204 {Opt_grpid, "bsdgroups"},
1205 {Opt_nogrpid, "nogrpid"},
1206 {Opt_nogrpid, "sysvgroups"},
1207 {Opt_resgid, "resgid=%u"},
1208 {Opt_resuid, "resuid=%u"},
1210 {Opt_err_cont, "errors=continue"},
1211 {Opt_err_panic, "errors=panic"},
1212 {Opt_err_ro, "errors=remount-ro"},
1213 {Opt_nouid32, "nouid32"},
1214 {Opt_debug, "debug"},
1215 {Opt_removed, "oldalloc"},
1216 {Opt_removed, "orlov"},
1217 {Opt_user_xattr, "user_xattr"},
1218 {Opt_nouser_xattr, "nouser_xattr"},
1220 {Opt_noacl, "noacl"},
1221 {Opt_noload, "norecovery"},
1222 {Opt_noload, "noload"},
1223 {Opt_removed, "nobh"},
1224 {Opt_removed, "bh"},
1225 {Opt_commit, "commit=%u"},
1226 {Opt_min_batch_time, "min_batch_time=%u"},
1227 {Opt_max_batch_time, "max_batch_time=%u"},
1228 {Opt_journal_dev, "journal_dev=%u"},
1229 {Opt_journal_path, "journal_path=%s"},
1230 {Opt_journal_checksum, "journal_checksum"},
1231 {Opt_nojournal_checksum, "nojournal_checksum"},
1232 {Opt_journal_async_commit, "journal_async_commit"},
1233 {Opt_abort, "abort"},
1234 {Opt_data_journal, "data=journal"},
1235 {Opt_data_ordered, "data=ordered"},
1236 {Opt_data_writeback, "data=writeback"},
1237 {Opt_data_err_abort, "data_err=abort"},
1238 {Opt_data_err_ignore, "data_err=ignore"},
1239 {Opt_offusrjquota, "usrjquota="},
1240 {Opt_usrjquota, "usrjquota=%s"},
1241 {Opt_offgrpjquota, "grpjquota="},
1242 {Opt_grpjquota, "grpjquota=%s"},
1243 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1244 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1245 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1246 {Opt_grpquota, "grpquota"},
1247 {Opt_noquota, "noquota"},
1248 {Opt_quota, "quota"},
1249 {Opt_usrquota, "usrquota"},
1250 {Opt_barrier, "barrier=%u"},
1251 {Opt_barrier, "barrier"},
1252 {Opt_nobarrier, "nobarrier"},
1253 {Opt_i_version, "i_version"},
1255 {Opt_stripe, "stripe=%u"},
1256 {Opt_delalloc, "delalloc"},
1257 {Opt_lazytime, "lazytime"},
1258 {Opt_nolazytime, "nolazytime"},
1259 {Opt_nodelalloc, "nodelalloc"},
1260 {Opt_removed, "mblk_io_submit"},
1261 {Opt_removed, "nomblk_io_submit"},
1262 {Opt_block_validity, "block_validity"},
1263 {Opt_noblock_validity, "noblock_validity"},
1264 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1265 {Opt_journal_ioprio, "journal_ioprio=%u"},
1266 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1267 {Opt_auto_da_alloc, "auto_da_alloc"},
1268 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1269 {Opt_dioread_nolock, "dioread_nolock"},
1270 {Opt_dioread_lock, "dioread_lock"},
1271 {Opt_discard, "discard"},
1272 {Opt_nodiscard, "nodiscard"},
1273 {Opt_init_itable, "init_itable=%u"},
1274 {Opt_init_itable, "init_itable"},
1275 {Opt_noinit_itable, "noinit_itable"},
1276 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1277 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1278 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1279 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1280 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1281 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1282 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1286 static ext4_fsblk_t get_sb_block(void **data)
1288 ext4_fsblk_t sb_block;
1289 char *options = (char *) *data;
1291 if (!options || strncmp(options, "sb=", 3) != 0)
1292 return 1; /* Default location */
1295 /* TODO: use simple_strtoll with >32bit ext4 */
1296 sb_block = simple_strtoul(options, &options, 0);
1297 if (*options && *options != ',') {
1298 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1302 if (*options == ',')
1304 *data = (void *) options;
1309 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1310 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1311 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1314 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1316 struct ext4_sb_info *sbi = EXT4_SB(sb);
1320 if (sb_any_quota_loaded(sb) &&
1321 !sbi->s_qf_names[qtype]) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot change journaled "
1324 "quota options when quota turned on");
1327 if (ext4_has_feature_quota(sb)) {
1328 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1329 "ignored when QUOTA feature is enabled");
1332 qname = match_strdup(args);
1334 ext4_msg(sb, KERN_ERR,
1335 "Not enough memory for storing quotafile name");
1338 if (sbi->s_qf_names[qtype]) {
1339 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1342 ext4_msg(sb, KERN_ERR,
1343 "%s quota file already specified",
1347 if (strchr(qname, '/')) {
1348 ext4_msg(sb, KERN_ERR,
1349 "quotafile must be on filesystem root");
1352 sbi->s_qf_names[qtype] = qname;
1360 static int clear_qf_name(struct super_block *sb, int qtype)
1363 struct ext4_sb_info *sbi = EXT4_SB(sb);
1365 if (sb_any_quota_loaded(sb) &&
1366 sbi->s_qf_names[qtype]) {
1367 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1368 " when quota turned on");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
1377 #define MOPT_SET 0x0001
1378 #define MOPT_CLEAR 0x0002
1379 #define MOPT_NOSUPPORT 0x0004
1380 #define MOPT_EXPLICIT 0x0008
1381 #define MOPT_CLEAR_ERR 0x0010
1382 #define MOPT_GTE0 0x0020
1385 #define MOPT_QFMT 0x0040
1387 #define MOPT_Q MOPT_NOSUPPORT
1388 #define MOPT_QFMT MOPT_NOSUPPORT
1390 #define MOPT_DATAJ 0x0080
1391 #define MOPT_NO_EXT2 0x0100
1392 #define MOPT_NO_EXT3 0x0200
1393 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1394 #define MOPT_STRING 0x0400
1396 static const struct mount_opts {
1400 } ext4_mount_opts[] = {
1401 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1402 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1403 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1404 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1405 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1406 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1407 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1408 MOPT_EXT4_ONLY | MOPT_SET},
1409 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1410 MOPT_EXT4_ONLY | MOPT_CLEAR},
1411 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1412 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1413 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1414 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1415 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1416 MOPT_EXT4_ONLY | MOPT_CLEAR},
1417 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1418 MOPT_EXT4_ONLY | MOPT_CLEAR},
1419 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1420 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1421 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1422 EXT4_MOUNT_JOURNAL_CHECKSUM),
1423 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1424 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1425 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1426 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1427 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1428 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1430 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1432 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1433 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1434 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1435 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1436 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1437 {Opt_commit, 0, MOPT_GTE0},
1438 {Opt_max_batch_time, 0, MOPT_GTE0},
1439 {Opt_min_batch_time, 0, MOPT_GTE0},
1440 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1441 {Opt_init_itable, 0, MOPT_GTE0},
1442 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1443 {Opt_stripe, 0, MOPT_GTE0},
1444 {Opt_resuid, 0, MOPT_GTE0},
1445 {Opt_resgid, 0, MOPT_GTE0},
1446 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1447 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1448 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1449 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1450 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1451 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1452 MOPT_NO_EXT2 | MOPT_DATAJ},
1453 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1455 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1456 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1457 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1459 {Opt_acl, 0, MOPT_NOSUPPORT},
1460 {Opt_noacl, 0, MOPT_NOSUPPORT},
1462 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1463 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1464 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1465 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1467 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1469 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1470 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1471 {Opt_usrjquota, 0, MOPT_Q},
1472 {Opt_grpjquota, 0, MOPT_Q},
1473 {Opt_offusrjquota, 0, MOPT_Q},
1474 {Opt_offgrpjquota, 0, MOPT_Q},
1475 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1476 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1477 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1478 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1479 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1483 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1484 substring_t *args, unsigned long *journal_devnum,
1485 unsigned int *journal_ioprio, int is_remount)
1487 struct ext4_sb_info *sbi = EXT4_SB(sb);
1488 const struct mount_opts *m;
1494 if (token == Opt_usrjquota)
1495 return set_qf_name(sb, USRQUOTA, &args[0]);
1496 else if (token == Opt_grpjquota)
1497 return set_qf_name(sb, GRPQUOTA, &args[0]);
1498 else if (token == Opt_offusrjquota)
1499 return clear_qf_name(sb, USRQUOTA);
1500 else if (token == Opt_offgrpjquota)
1501 return clear_qf_name(sb, GRPQUOTA);
1505 case Opt_nouser_xattr:
1506 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1509 return 1; /* handled by get_sb_block() */
1511 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1514 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1517 sb->s_flags |= MS_I_VERSION;
1520 sb->s_flags |= MS_LAZYTIME;
1522 case Opt_nolazytime:
1523 sb->s_flags &= ~MS_LAZYTIME;
1527 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1528 if (token == m->token)
1531 if (m->token == Opt_err) {
1532 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1533 "or missing value", opt);
1537 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Mount option \"%s\" incompatible with ext2", opt);
1542 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1543 ext4_msg(sb, KERN_ERR,
1544 "Mount option \"%s\" incompatible with ext3", opt);
1548 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1552 if (m->flags & MOPT_EXPLICIT) {
1553 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1554 set_opt2(sb, EXPLICIT_DELALLOC);
1555 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1556 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1560 if (m->flags & MOPT_CLEAR_ERR)
1561 clear_opt(sb, ERRORS_MASK);
1562 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564 "options when quota turned on");
1568 if (m->flags & MOPT_NOSUPPORT) {
1569 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1570 } else if (token == Opt_commit) {
1572 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1573 sbi->s_commit_interval = HZ * arg;
1574 } else if (token == Opt_max_batch_time) {
1575 sbi->s_max_batch_time = arg;
1576 } else if (token == Opt_min_batch_time) {
1577 sbi->s_min_batch_time = arg;
1578 } else if (token == Opt_inode_readahead_blks) {
1579 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1580 ext4_msg(sb, KERN_ERR,
1581 "EXT4-fs: inode_readahead_blks must be "
1582 "0 or a power of 2 smaller than 2^31");
1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
1591 } else if (token == Opt_max_dir_size_kb) {
1592 sbi->s_max_dir_size_kb = arg;
1593 } else if (token == Opt_stripe) {
1594 sbi->s_stripe = arg;
1595 } else if (token == Opt_resuid) {
1596 uid = make_kuid(current_user_ns(), arg);
1597 if (!uid_valid(uid)) {
1598 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1601 sbi->s_resuid = uid;
1602 } else if (token == Opt_resgid) {
1603 gid = make_kgid(current_user_ns(), arg);
1604 if (!gid_valid(gid)) {
1605 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1608 sbi->s_resgid = gid;
1609 } else if (token == Opt_journal_dev) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot specify journal on remount");
1615 *journal_devnum = arg;
1616 } else if (token == Opt_journal_path) {
1618 struct inode *journal_inode;
1623 ext4_msg(sb, KERN_ERR,
1624 "Cannot specify journal on remount");
1627 journal_path = match_strdup(&args[0]);
1628 if (!journal_path) {
1629 ext4_msg(sb, KERN_ERR, "error: could not dup "
1630 "journal device string");
1634 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1636 ext4_msg(sb, KERN_ERR, "error: could not find "
1637 "journal device path: error %d", error);
1638 kfree(journal_path);
1642 journal_inode = d_inode(path.dentry);
1643 if (!S_ISBLK(journal_inode->i_mode)) {
1644 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1645 "is not a block device", journal_path);
1647 kfree(journal_path);
1651 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1653 kfree(journal_path);
1654 } else if (token == Opt_journal_ioprio) {
1656 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1661 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1662 } else if (token == Opt_test_dummy_encryption) {
1663 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1664 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1665 ext4_msg(sb, KERN_WARNING,
1666 "Test dummy encryption mode enabled");
1668 ext4_msg(sb, KERN_WARNING,
1669 "Test dummy encryption mount option ignored");
1671 } else if (m->flags & MOPT_DATAJ) {
1673 if (!sbi->s_journal)
1674 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1675 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1676 ext4_msg(sb, KERN_ERR,
1677 "Cannot change data mode on remount");
1681 clear_opt(sb, DATA_FLAGS);
1682 sbi->s_mount_opt |= m->mount_opt;
1685 } else if (m->flags & MOPT_QFMT) {
1686 if (sb_any_quota_loaded(sb) &&
1687 sbi->s_jquota_fmt != m->mount_opt) {
1688 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1689 "quota options when quota turned on");
1692 if (ext4_has_feature_quota(sb)) {
1693 ext4_msg(sb, KERN_INFO,
1694 "Quota format mount options ignored "
1695 "when QUOTA feature is enabled");
1698 sbi->s_jquota_fmt = m->mount_opt;
1700 } else if (token == Opt_dax) {
1701 #ifdef CONFIG_FS_DAX
1702 ext4_msg(sb, KERN_WARNING,
1703 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1704 sbi->s_mount_opt |= m->mount_opt;
1706 ext4_msg(sb, KERN_INFO, "dax option not supported");
1709 } else if (token == Opt_data_err_abort) {
1710 sbi->s_mount_opt |= m->mount_opt;
1711 } else if (token == Opt_data_err_ignore) {
1712 sbi->s_mount_opt &= ~m->mount_opt;
1716 if (m->flags & MOPT_CLEAR)
1718 else if (unlikely(!(m->flags & MOPT_SET))) {
1719 ext4_msg(sb, KERN_WARNING,
1720 "buggy handling of option %s", opt);
1725 sbi->s_mount_opt |= m->mount_opt;
1727 sbi->s_mount_opt &= ~m->mount_opt;
1732 static int parse_options(char *options, struct super_block *sb,
1733 unsigned long *journal_devnum,
1734 unsigned int *journal_ioprio,
1737 struct ext4_sb_info *sbi = EXT4_SB(sb);
1739 substring_t args[MAX_OPT_ARGS];
1745 while ((p = strsep(&options, ",")) != NULL) {
1749 * Initialize args struct so we know whether arg was
1750 * found; some options take optional arguments.
1752 args[0].to = args[0].from = NULL;
1753 token = match_token(p, tokens, args);
1754 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1755 journal_ioprio, is_remount) < 0)
1759 if (ext4_has_feature_quota(sb) &&
1760 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1761 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1762 "mount options ignored.");
1763 clear_opt(sb, USRQUOTA);
1764 clear_opt(sb, GRPQUOTA);
1765 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1766 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1767 clear_opt(sb, USRQUOTA);
1769 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1770 clear_opt(sb, GRPQUOTA);
1772 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1773 ext4_msg(sb, KERN_ERR, "old and new quota "
1778 if (!sbi->s_jquota_fmt) {
1779 ext4_msg(sb, KERN_ERR, "journaled quota format "
1785 if (test_opt(sb, DIOREAD_NOLOCK)) {
1787 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1789 if (blocksize < PAGE_SIZE) {
1790 ext4_msg(sb, KERN_ERR, "can't mount with "
1791 "dioread_nolock if block size != PAGE_SIZE");
1795 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1796 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1797 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1798 "in data=ordered mode");
1804 static inline void ext4_show_quota_options(struct seq_file *seq,
1805 struct super_block *sb)
1807 #if defined(CONFIG_QUOTA)
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1810 if (sbi->s_jquota_fmt) {
1813 switch (sbi->s_jquota_fmt) {
1824 seq_printf(seq, ",jqfmt=%s", fmtname);
1827 if (sbi->s_qf_names[USRQUOTA])
1828 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1830 if (sbi->s_qf_names[GRPQUOTA])
1831 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1835 static const char *token2str(int token)
1837 const struct match_token *t;
1839 for (t = tokens; t->token != Opt_err; t++)
1840 if (t->token == token && !strchr(t->pattern, '='))
1847 * - it's set to a non-default value OR
1848 * - if the per-sb default is different from the global default
1850 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 struct ext4_super_block *es = sbi->s_es;
1855 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1856 const struct mount_opts *m;
1857 char sep = nodefs ? '\n' : ',';
1859 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1860 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1862 if (sbi->s_sb_block != 1)
1863 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1865 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1866 int want_set = m->flags & MOPT_SET;
1867 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1868 (m->flags & MOPT_CLEAR_ERR))
1870 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1871 continue; /* skip if same as the default */
1873 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1874 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1875 continue; /* select Opt_noFoo vs Opt_Foo */
1876 SEQ_OPTS_PRINT("%s", token2str(m->token));
1879 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1880 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1881 SEQ_OPTS_PRINT("resuid=%u",
1882 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1883 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1884 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1885 SEQ_OPTS_PRINT("resgid=%u",
1886 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1887 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1888 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1889 SEQ_OPTS_PUTS("errors=remount-ro");
1890 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1891 SEQ_OPTS_PUTS("errors=continue");
1892 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1893 SEQ_OPTS_PUTS("errors=panic");
1894 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1895 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1896 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1897 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1898 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1899 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1900 if (sb->s_flags & MS_I_VERSION)
1901 SEQ_OPTS_PUTS("i_version");
1902 if (nodefs || sbi->s_stripe)
1903 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1904 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1906 SEQ_OPTS_PUTS("data=journal");
1907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1908 SEQ_OPTS_PUTS("data=ordered");
1909 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1910 SEQ_OPTS_PUTS("data=writeback");
1913 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1914 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1915 sbi->s_inode_readahead_blks);
1917 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1918 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1919 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1920 if (nodefs || sbi->s_max_dir_size_kb)
1921 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1922 if (test_opt(sb, DATA_ERR_ABORT))
1923 SEQ_OPTS_PUTS("data_err=abort");
1925 ext4_show_quota_options(seq, sb);
1929 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1931 return _ext4_show_options(seq, root->d_sb, 0);
1934 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1936 struct super_block *sb = seq->private;
1939 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1940 rc = _ext4_show_options(seq, sb, 1);
1941 seq_puts(seq, "\n");
1945 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1952 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953 "forcing read-only mode");
1958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960 "running e2fsck is recommended");
1961 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1962 ext4_msg(sb, KERN_WARNING,
1963 "warning: mounting fs with errors, "
1964 "running e2fsck is recommended");
1965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1966 le16_to_cpu(es->s_mnt_count) >=
1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1968 ext4_msg(sb, KERN_WARNING,
1969 "warning: maximal mount count reached, "
1970 "running e2fsck is recommended");
1971 else if (le32_to_cpu(es->s_checkinterval) &&
1972 (le32_to_cpu(es->s_lastcheck) +
1973 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1974 ext4_msg(sb, KERN_WARNING,
1975 "warning: checktime reached, "
1976 "running e2fsck is recommended");
1977 if (!sbi->s_journal)
1978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1981 le16_add_cpu(&es->s_mnt_count, 1);
1982 es->s_mtime = cpu_to_le32(get_seconds());
1983 ext4_update_dynamic_rev(sb);
1985 ext4_set_feature_journal_needs_recovery(sb);
1987 ext4_commit_super(sb, 1);
1989 if (test_opt(sb, DEBUG))
1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1993 sbi->s_groups_count,
1994 EXT4_BLOCKS_PER_GROUP(sb),
1995 EXT4_INODES_PER_GROUP(sb),
1996 sbi->s_mount_opt, sbi->s_mount_opt2);
1998 cleancache_init_fs(sb);
2002 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2004 struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 struct flex_groups *new_groups;
2008 if (!sbi->s_log_groups_per_flex)
2011 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2012 if (size <= sbi->s_flex_groups_allocated)
2015 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2016 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2018 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2019 size / (int) sizeof(struct flex_groups));
2023 if (sbi->s_flex_groups) {
2024 memcpy(new_groups, sbi->s_flex_groups,
2025 (sbi->s_flex_groups_allocated *
2026 sizeof(struct flex_groups)));
2027 kvfree(sbi->s_flex_groups);
2029 sbi->s_flex_groups = new_groups;
2030 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2034 static int ext4_fill_flex_info(struct super_block *sb)
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2037 struct ext4_group_desc *gdp = NULL;
2038 ext4_group_t flex_group;
2041 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2042 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2043 sbi->s_log_groups_per_flex = 0;
2047 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2051 for (i = 0; i < sbi->s_groups_count; i++) {
2052 gdp = ext4_get_group_desc(sb, i, NULL);
2054 flex_group = ext4_flex_group(sbi, i);
2055 atomic_add(ext4_free_inodes_count(sb, gdp),
2056 &sbi->s_flex_groups[flex_group].free_inodes);
2057 atomic64_add(ext4_free_group_clusters(sb, gdp),
2058 &sbi->s_flex_groups[flex_group].free_clusters);
2059 atomic_add(ext4_used_dirs_count(sb, gdp),
2060 &sbi->s_flex_groups[flex_group].used_dirs);
2068 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2069 struct ext4_group_desc *gdp)
2073 __le32 le_group = cpu_to_le32(block_group);
2074 struct ext4_sb_info *sbi = EXT4_SB(sb);
2076 if (ext4_has_metadata_csum(sbi->s_sb)) {
2077 /* Use new metadata_csum algorithm */
2081 save_csum = gdp->bg_checksum;
2082 gdp->bg_checksum = 0;
2083 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2085 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2087 gdp->bg_checksum = save_csum;
2089 crc = csum32 & 0xFFFF;
2093 /* old crc16 code */
2094 if (!ext4_has_feature_gdt_csum(sb))
2097 offset = offsetof(struct ext4_group_desc, bg_checksum);
2099 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2100 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2101 crc = crc16(crc, (__u8 *)gdp, offset);
2102 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2103 /* for checksum of struct ext4_group_desc do the rest...*/
2104 if (ext4_has_feature_64bit(sb) &&
2105 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2106 crc = crc16(crc, (__u8 *)gdp + offset,
2107 le16_to_cpu(sbi->s_es->s_desc_size) -
2111 return cpu_to_le16(crc);
2114 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2115 struct ext4_group_desc *gdp)
2117 if (ext4_has_group_desc_csum(sb) &&
2118 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2124 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2125 struct ext4_group_desc *gdp)
2127 if (!ext4_has_group_desc_csum(sb))
2129 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2132 /* Called at mount-time, super-block is locked */
2133 static int ext4_check_descriptors(struct super_block *sb,
2134 ext4_group_t *first_not_zeroed)
2136 struct ext4_sb_info *sbi = EXT4_SB(sb);
2137 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2138 ext4_fsblk_t last_block;
2139 ext4_fsblk_t block_bitmap;
2140 ext4_fsblk_t inode_bitmap;
2141 ext4_fsblk_t inode_table;
2142 int flexbg_flag = 0;
2143 ext4_group_t i, grp = sbi->s_groups_count;
2145 if (ext4_has_feature_flex_bg(sb))
2148 ext4_debug("Checking group descriptors");
2150 for (i = 0; i < sbi->s_groups_count; i++) {
2151 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2153 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2154 last_block = ext4_blocks_count(sbi->s_es) - 1;
2156 last_block = first_block +
2157 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2159 if ((grp == sbi->s_groups_count) &&
2160 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2163 block_bitmap = ext4_block_bitmap(sb, gdp);
2164 if (block_bitmap < first_block || block_bitmap > last_block) {
2165 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2166 "Block bitmap for group %u not in group "
2167 "(block %llu)!", i, block_bitmap);
2170 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2171 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2172 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2173 "Inode bitmap for group %u not in group "
2174 "(block %llu)!", i, inode_bitmap);
2177 inode_table = ext4_inode_table(sb, gdp);
2178 if (inode_table < first_block ||
2179 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2180 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2181 "Inode table for group %u not in group "
2182 "(block %llu)!", i, inode_table);
2185 ext4_lock_group(sb, i);
2186 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2187 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2188 "Checksum for group %u failed (%u!=%u)",
2189 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2190 gdp)), le16_to_cpu(gdp->bg_checksum));
2191 if (!(sb->s_flags & MS_RDONLY)) {
2192 ext4_unlock_group(sb, i);
2196 ext4_unlock_group(sb, i);
2198 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2200 if (NULL != first_not_zeroed)
2201 *first_not_zeroed = grp;
2205 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2206 * the superblock) which were deleted from all directories, but held open by
2207 * a process at the time of a crash. We walk the list and try to delete these
2208 * inodes at recovery time (only with a read-write filesystem).
2210 * In order to keep the orphan inode chain consistent during traversal (in
2211 * case of crash during recovery), we link each inode into the superblock
2212 * orphan list_head and handle it the same way as an inode deletion during
2213 * normal operation (which journals the operations for us).
2215 * We only do an iget() and an iput() on each inode, which is very safe if we
2216 * accidentally point at an in-use or already deleted inode. The worst that
2217 * can happen in this case is that we get a "bit already cleared" message from
2218 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2219 * e2fsck was run on this filesystem, and it must have already done the orphan
2220 * inode cleanup for us, so we can safely abort without any further action.
2222 static void ext4_orphan_cleanup(struct super_block *sb,
2223 struct ext4_super_block *es)
2225 unsigned int s_flags = sb->s_flags;
2226 int nr_orphans = 0, nr_truncates = 0;
2230 if (!es->s_last_orphan) {
2231 jbd_debug(4, "no orphan inodes to clean up\n");
2235 if (bdev_read_only(sb->s_bdev)) {
2236 ext4_msg(sb, KERN_ERR, "write access "
2237 "unavailable, skipping orphan cleanup");
2241 /* Check if feature set would not allow a r/w mount */
2242 if (!ext4_feature_set_ok(sb, 0)) {
2243 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2244 "unknown ROCOMPAT features");
2248 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2249 /* don't clear list on RO mount w/ errors */
2250 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2251 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2252 "clearing orphan list.\n");
2253 es->s_last_orphan = 0;
2255 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2259 if (s_flags & MS_RDONLY) {
2260 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2261 sb->s_flags &= ~MS_RDONLY;
2264 /* Needed for iput() to work correctly and not trash data */
2265 sb->s_flags |= MS_ACTIVE;
2266 /* Turn on quotas so that they are updated correctly */
2267 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2268 if (EXT4_SB(sb)->s_qf_names[i]) {
2269 int ret = ext4_quota_on_mount(sb, i);
2271 ext4_msg(sb, KERN_ERR,
2272 "Cannot turn on journaled "
2273 "quota: error %d", ret);
2278 while (es->s_last_orphan) {
2279 struct inode *inode;
2281 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2282 if (IS_ERR(inode)) {
2283 es->s_last_orphan = 0;
2287 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2288 dquot_initialize(inode);
2289 if (inode->i_nlink) {
2290 if (test_opt(sb, DEBUG))
2291 ext4_msg(sb, KERN_DEBUG,
2292 "%s: truncating inode %lu to %lld bytes",
2293 __func__, inode->i_ino, inode->i_size);
2294 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2295 inode->i_ino, inode->i_size);
2297 truncate_inode_pages(inode->i_mapping, inode->i_size);
2298 ext4_truncate(inode);
2299 inode_unlock(inode);
2302 if (test_opt(sb, DEBUG))
2303 ext4_msg(sb, KERN_DEBUG,
2304 "%s: deleting unreferenced inode %lu",
2305 __func__, inode->i_ino);
2306 jbd_debug(2, "deleting unreferenced inode %lu\n",
2310 iput(inode); /* The delete magic happens here! */
2313 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2316 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2317 PLURAL(nr_orphans));
2319 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2320 PLURAL(nr_truncates));
2322 /* Turn quotas off */
2323 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2324 if (sb_dqopt(sb)->files[i])
2325 dquot_quota_off(sb, i);
2328 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2332 * Maximal extent format file size.
2333 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2334 * extent format containers, within a sector_t, and within i_blocks
2335 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2336 * so that won't be a limiting factor.
2338 * However there is other limiting factor. We do store extents in the form
2339 * of starting block and length, hence the resulting length of the extent
2340 * covering maximum file size must fit into on-disk format containers as
2341 * well. Given that length is always by 1 unit bigger than max unit (because
2342 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2344 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2346 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2349 loff_t upper_limit = MAX_LFS_FILESIZE;
2351 /* small i_blocks in vfs inode? */
2352 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2354 * CONFIG_LBDAF is not enabled implies the inode
2355 * i_block represent total blocks in 512 bytes
2356 * 32 == size of vfs inode i_blocks * 8
2358 upper_limit = (1LL << 32) - 1;
2360 /* total blocks in file system block size */
2361 upper_limit >>= (blkbits - 9);
2362 upper_limit <<= blkbits;
2366 * 32-bit extent-start container, ee_block. We lower the maxbytes
2367 * by one fs block, so ee_len can cover the extent of maximum file
2370 res = (1LL << 32) - 1;
2373 /* Sanity check against vm- & vfs- imposed limits */
2374 if (res > upper_limit)
2381 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2382 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2383 * We need to be 1 filesystem block less than the 2^48 sector limit.
2385 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2387 loff_t res = EXT4_NDIR_BLOCKS;
2390 /* This is calculated to be the largest file size for a dense, block
2391 * mapped file such that the file's total number of 512-byte sectors,
2392 * including data and all indirect blocks, does not exceed (2^48 - 1).
2394 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2395 * number of 512-byte sectors of the file.
2398 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2400 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2401 * the inode i_block field represents total file blocks in
2402 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2404 upper_limit = (1LL << 32) - 1;
2406 /* total blocks in file system block size */
2407 upper_limit >>= (bits - 9);
2411 * We use 48 bit ext4_inode i_blocks
2412 * With EXT4_HUGE_FILE_FL set the i_blocks
2413 * represent total number of blocks in
2414 * file system block size
2416 upper_limit = (1LL << 48) - 1;
2420 /* indirect blocks */
2422 /* double indirect blocks */
2423 meta_blocks += 1 + (1LL << (bits-2));
2424 /* tripple indirect blocks */
2425 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2427 upper_limit -= meta_blocks;
2428 upper_limit <<= bits;
2430 res += 1LL << (bits-2);
2431 res += 1LL << (2*(bits-2));
2432 res += 1LL << (3*(bits-2));
2434 if (res > upper_limit)
2437 if (res > MAX_LFS_FILESIZE)
2438 res = MAX_LFS_FILESIZE;
2443 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2444 ext4_fsblk_t logical_sb_block, int nr)
2446 struct ext4_sb_info *sbi = EXT4_SB(sb);
2447 ext4_group_t bg, first_meta_bg;
2450 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2452 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2453 return logical_sb_block + nr + 1;
2454 bg = sbi->s_desc_per_block * nr;
2455 if (ext4_bg_has_super(sb, bg))
2459 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2460 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2461 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2464 if (sb->s_blocksize == 1024 && nr == 0 &&
2465 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2468 return (has_super + ext4_group_first_block_no(sb, bg));
2472 * ext4_get_stripe_size: Get the stripe size.
2473 * @sbi: In memory super block info
2475 * If we have specified it via mount option, then
2476 * use the mount option value. If the value specified at mount time is
2477 * greater than the blocks per group use the super block value.
2478 * If the super block value is greater than blocks per group return 0.
2479 * Allocator needs it be less than blocks per group.
2482 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2484 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2485 unsigned long stripe_width =
2486 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2489 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2490 ret = sbi->s_stripe;
2491 else if (stripe_width <= sbi->s_blocks_per_group)
2493 else if (stride <= sbi->s_blocks_per_group)
2499 * If the stripe width is 1, this makes no sense and
2500 * we set it to 0 to turn off stripe handling code.
2509 * Check whether this filesystem can be mounted based on
2510 * the features present and the RDONLY/RDWR mount requested.
2511 * Returns 1 if this filesystem can be mounted as requested,
2512 * 0 if it cannot be.
2514 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2516 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2517 ext4_msg(sb, KERN_ERR,
2518 "Couldn't mount because of "
2519 "unsupported optional features (%x)",
2520 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2521 ~EXT4_FEATURE_INCOMPAT_SUPP));
2528 if (ext4_has_feature_readonly(sb)) {
2529 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2530 sb->s_flags |= MS_RDONLY;
2534 /* Check that feature set is OK for a read-write mount */
2535 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2536 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2537 "unsupported optional features (%x)",
2538 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2539 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2543 * Large file size enabled file system can only be mounted
2544 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2546 if (ext4_has_feature_huge_file(sb)) {
2547 if (sizeof(blkcnt_t) < sizeof(u64)) {
2548 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2549 "cannot be mounted RDWR without "
2554 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2555 ext4_msg(sb, KERN_ERR,
2556 "Can't support bigalloc feature without "
2557 "extents feature\n");
2561 #ifndef CONFIG_QUOTA
2562 if (ext4_has_feature_quota(sb) && !readonly) {
2563 ext4_msg(sb, KERN_ERR,
2564 "Filesystem with quota feature cannot be mounted RDWR "
2565 "without CONFIG_QUOTA");
2568 if (ext4_has_feature_project(sb) && !readonly) {
2569 ext4_msg(sb, KERN_ERR,
2570 "Filesystem with project quota feature cannot be mounted RDWR "
2571 "without CONFIG_QUOTA");
2574 #endif /* CONFIG_QUOTA */
2579 * This function is called once a day if we have errors logged
2580 * on the file system
2582 static void print_daily_error_info(unsigned long arg)
2584 struct super_block *sb = (struct super_block *) arg;
2585 struct ext4_sb_info *sbi;
2586 struct ext4_super_block *es;
2591 if (es->s_error_count)
2592 /* fsck newer than v1.41.13 is needed to clean this condition. */
2593 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2594 le32_to_cpu(es->s_error_count));
2595 if (es->s_first_error_time) {
2596 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2597 sb->s_id, le32_to_cpu(es->s_first_error_time),
2598 (int) sizeof(es->s_first_error_func),
2599 es->s_first_error_func,
2600 le32_to_cpu(es->s_first_error_line));
2601 if (es->s_first_error_ino)
2602 printk(": inode %u",
2603 le32_to_cpu(es->s_first_error_ino));
2604 if (es->s_first_error_block)
2605 printk(": block %llu", (unsigned long long)
2606 le64_to_cpu(es->s_first_error_block));
2609 if (es->s_last_error_time) {
2610 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2611 sb->s_id, le32_to_cpu(es->s_last_error_time),
2612 (int) sizeof(es->s_last_error_func),
2613 es->s_last_error_func,
2614 le32_to_cpu(es->s_last_error_line));
2615 if (es->s_last_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_last_error_ino));
2618 if (es->s_last_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_last_error_block));
2623 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2626 /* Find next suitable group and run ext4_init_inode_table */
2627 static int ext4_run_li_request(struct ext4_li_request *elr)
2629 struct ext4_group_desc *gdp = NULL;
2630 ext4_group_t group, ngroups;
2631 struct super_block *sb;
2632 unsigned long timeout = 0;
2636 ngroups = EXT4_SB(sb)->s_groups_count;
2639 for (group = elr->lr_next_group; group < ngroups; group++) {
2640 gdp = ext4_get_group_desc(sb, group, NULL);
2646 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2650 if (group >= ngroups)
2655 ret = ext4_init_inode_table(sb, group,
2656 elr->lr_timeout ? 0 : 1);
2657 if (elr->lr_timeout == 0) {
2658 timeout = (jiffies - timeout) *
2659 elr->lr_sbi->s_li_wait_mult;
2660 elr->lr_timeout = timeout;
2662 elr->lr_next_sched = jiffies + elr->lr_timeout;
2663 elr->lr_next_group = group + 1;
2671 * Remove lr_request from the list_request and free the
2672 * request structure. Should be called with li_list_mtx held
2674 static void ext4_remove_li_request(struct ext4_li_request *elr)
2676 struct ext4_sb_info *sbi;
2683 list_del(&elr->lr_request);
2684 sbi->s_li_request = NULL;
2688 static void ext4_unregister_li_request(struct super_block *sb)
2690 mutex_lock(&ext4_li_mtx);
2691 if (!ext4_li_info) {
2692 mutex_unlock(&ext4_li_mtx);
2696 mutex_lock(&ext4_li_info->li_list_mtx);
2697 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2698 mutex_unlock(&ext4_li_info->li_list_mtx);
2699 mutex_unlock(&ext4_li_mtx);
2702 static struct task_struct *ext4_lazyinit_task;
2705 * This is the function where ext4lazyinit thread lives. It walks
2706 * through the request list searching for next scheduled filesystem.
2707 * When such a fs is found, run the lazy initialization request
2708 * (ext4_rn_li_request) and keep track of the time spend in this
2709 * function. Based on that time we compute next schedule time of
2710 * the request. When walking through the list is complete, compute
2711 * next waking time and put itself into sleep.
2713 static int ext4_lazyinit_thread(void *arg)
2715 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2716 struct list_head *pos, *n;
2717 struct ext4_li_request *elr;
2718 unsigned long next_wakeup, cur;
2720 BUG_ON(NULL == eli);
2724 next_wakeup = MAX_JIFFY_OFFSET;
2726 mutex_lock(&eli->li_list_mtx);
2727 if (list_empty(&eli->li_request_list)) {
2728 mutex_unlock(&eli->li_list_mtx);
2732 list_for_each_safe(pos, n, &eli->li_request_list) {
2733 elr = list_entry(pos, struct ext4_li_request,
2736 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2737 if (ext4_run_li_request(elr) != 0) {
2738 /* error, remove the lazy_init job */
2739 ext4_remove_li_request(elr);
2744 if (time_before(elr->lr_next_sched, next_wakeup))
2745 next_wakeup = elr->lr_next_sched;
2747 mutex_unlock(&eli->li_list_mtx);
2752 if ((time_after_eq(cur, next_wakeup)) ||
2753 (MAX_JIFFY_OFFSET == next_wakeup)) {
2758 schedule_timeout_interruptible(next_wakeup - cur);
2760 if (kthread_should_stop()) {
2761 ext4_clear_request_list();
2768 * It looks like the request list is empty, but we need
2769 * to check it under the li_list_mtx lock, to prevent any
2770 * additions into it, and of course we should lock ext4_li_mtx
2771 * to atomically free the list and ext4_li_info, because at
2772 * this point another ext4 filesystem could be registering
2775 mutex_lock(&ext4_li_mtx);
2776 mutex_lock(&eli->li_list_mtx);
2777 if (!list_empty(&eli->li_request_list)) {
2778 mutex_unlock(&eli->li_list_mtx);
2779 mutex_unlock(&ext4_li_mtx);
2782 mutex_unlock(&eli->li_list_mtx);
2783 kfree(ext4_li_info);
2784 ext4_li_info = NULL;
2785 mutex_unlock(&ext4_li_mtx);
2790 static void ext4_clear_request_list(void)
2792 struct list_head *pos, *n;
2793 struct ext4_li_request *elr;
2795 mutex_lock(&ext4_li_info->li_list_mtx);
2796 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2797 elr = list_entry(pos, struct ext4_li_request,
2799 ext4_remove_li_request(elr);
2801 mutex_unlock(&ext4_li_info->li_list_mtx);
2804 static int ext4_run_lazyinit_thread(void)
2806 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2807 ext4_li_info, "ext4lazyinit");
2808 if (IS_ERR(ext4_lazyinit_task)) {
2809 int err = PTR_ERR(ext4_lazyinit_task);
2810 ext4_clear_request_list();
2811 kfree(ext4_li_info);
2812 ext4_li_info = NULL;
2813 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2814 "initialization thread\n",
2818 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2823 * Check whether it make sense to run itable init. thread or not.
2824 * If there is at least one uninitialized inode table, return
2825 * corresponding group number, else the loop goes through all
2826 * groups and return total number of groups.
2828 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2830 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2831 struct ext4_group_desc *gdp = NULL;
2833 for (group = 0; group < ngroups; group++) {
2834 gdp = ext4_get_group_desc(sb, group, NULL);
2838 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2845 static int ext4_li_info_new(void)
2847 struct ext4_lazy_init *eli = NULL;
2849 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2853 INIT_LIST_HEAD(&eli->li_request_list);
2854 mutex_init(&eli->li_list_mtx);
2856 eli->li_state |= EXT4_LAZYINIT_QUIT;
2863 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2866 struct ext4_sb_info *sbi = EXT4_SB(sb);
2867 struct ext4_li_request *elr;
2869 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2875 elr->lr_next_group = start;
2878 * Randomize first schedule time of the request to
2879 * spread the inode table initialization requests
2882 elr->lr_next_sched = jiffies + (prandom_u32() %
2883 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2887 int ext4_register_li_request(struct super_block *sb,
2888 ext4_group_t first_not_zeroed)
2890 struct ext4_sb_info *sbi = EXT4_SB(sb);
2891 struct ext4_li_request *elr = NULL;
2892 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2895 mutex_lock(&ext4_li_mtx);
2896 if (sbi->s_li_request != NULL) {
2898 * Reset timeout so it can be computed again, because
2899 * s_li_wait_mult might have changed.
2901 sbi->s_li_request->lr_timeout = 0;
2905 if (first_not_zeroed == ngroups ||
2906 (sb->s_flags & MS_RDONLY) ||
2907 !test_opt(sb, INIT_INODE_TABLE))
2910 elr = ext4_li_request_new(sb, first_not_zeroed);
2916 if (NULL == ext4_li_info) {
2917 ret = ext4_li_info_new();
2922 mutex_lock(&ext4_li_info->li_list_mtx);
2923 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2924 mutex_unlock(&ext4_li_info->li_list_mtx);
2926 sbi->s_li_request = elr;
2928 * set elr to NULL here since it has been inserted to
2929 * the request_list and the removal and free of it is
2930 * handled by ext4_clear_request_list from now on.
2934 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2935 ret = ext4_run_lazyinit_thread();
2940 mutex_unlock(&ext4_li_mtx);
2947 * We do not need to lock anything since this is called on
2950 static void ext4_destroy_lazyinit_thread(void)
2953 * If thread exited earlier
2954 * there's nothing to be done.
2956 if (!ext4_li_info || !ext4_lazyinit_task)
2959 kthread_stop(ext4_lazyinit_task);
2962 static int set_journal_csum_feature_set(struct super_block *sb)
2965 int compat, incompat;
2966 struct ext4_sb_info *sbi = EXT4_SB(sb);
2968 if (ext4_has_metadata_csum(sb)) {
2969 /* journal checksum v3 */
2971 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2973 /* journal checksum v1 */
2974 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2978 jbd2_journal_clear_features(sbi->s_journal,
2979 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2980 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2981 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2982 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2983 ret = jbd2_journal_set_features(sbi->s_journal,
2985 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2987 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2988 ret = jbd2_journal_set_features(sbi->s_journal,
2991 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2992 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2994 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2995 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3002 * Note: calculating the overhead so we can be compatible with
3003 * historical BSD practice is quite difficult in the face of
3004 * clusters/bigalloc. This is because multiple metadata blocks from
3005 * different block group can end up in the same allocation cluster.
3006 * Calculating the exact overhead in the face of clustered allocation
3007 * requires either O(all block bitmaps) in memory or O(number of block
3008 * groups**2) in time. We will still calculate the superblock for
3009 * older file systems --- and if we come across with a bigalloc file
3010 * system with zero in s_overhead_clusters the estimate will be close to
3011 * correct especially for very large cluster sizes --- but for newer
3012 * file systems, it's better to calculate this figure once at mkfs
3013 * time, and store it in the superblock. If the superblock value is
3014 * present (even for non-bigalloc file systems), we will use it.
3016 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3019 struct ext4_sb_info *sbi = EXT4_SB(sb);
3020 struct ext4_group_desc *gdp;
3021 ext4_fsblk_t first_block, last_block, b;
3022 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3023 int s, j, count = 0;
3025 if (!ext4_has_feature_bigalloc(sb))
3026 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3027 sbi->s_itb_per_group + 2);
3029 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3030 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3031 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3032 for (i = 0; i < ngroups; i++) {
3033 gdp = ext4_get_group_desc(sb, i, NULL);
3034 b = ext4_block_bitmap(sb, gdp);
3035 if (b >= first_block && b <= last_block) {
3036 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3039 b = ext4_inode_bitmap(sb, gdp);
3040 if (b >= first_block && b <= last_block) {
3041 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3044 b = ext4_inode_table(sb, gdp);
3045 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3046 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3047 int c = EXT4_B2C(sbi, b - first_block);
3048 ext4_set_bit(c, buf);
3054 if (ext4_bg_has_super(sb, grp)) {
3055 ext4_set_bit(s++, buf);
3058 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3059 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3065 return EXT4_CLUSTERS_PER_GROUP(sb) -
3066 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3070 * Compute the overhead and stash it in sbi->s_overhead
3072 int ext4_calculate_overhead(struct super_block *sb)
3074 struct ext4_sb_info *sbi = EXT4_SB(sb);
3075 struct ext4_super_block *es = sbi->s_es;
3076 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3077 ext4_fsblk_t overhead = 0;
3078 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3084 * Compute the overhead (FS structures). This is constant
3085 * for a given filesystem unless the number of block groups
3086 * changes so we cache the previous value until it does.
3090 * All of the blocks before first_data_block are overhead
3092 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3095 * Add the overhead found in each block group
3097 for (i = 0; i < ngroups; i++) {
3100 blks = count_overhead(sb, i, buf);
3103 memset(buf, 0, PAGE_SIZE);
3106 /* Add the internal journal blocks as well */
3107 if (sbi->s_journal && !sbi->journal_bdev)
3108 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3110 sbi->s_overhead = overhead;
3112 free_page((unsigned long) buf);
3116 static void ext4_set_resv_clusters(struct super_block *sb)
3118 ext4_fsblk_t resv_clusters;
3119 struct ext4_sb_info *sbi = EXT4_SB(sb);
3122 * There's no need to reserve anything when we aren't using extents.
3123 * The space estimates are exact, there are no unwritten extents,
3124 * hole punching doesn't need new metadata... This is needed especially
3125 * to keep ext2/3 backward compatibility.
3127 if (!ext4_has_feature_extents(sb))
3130 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3131 * This should cover the situations where we can not afford to run
3132 * out of space like for example punch hole, or converting
3133 * unwritten extents in delalloc path. In most cases such
3134 * allocation would require 1, or 2 blocks, higher numbers are
3137 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3138 sbi->s_cluster_bits);
3140 do_div(resv_clusters, 50);
3141 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3143 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3146 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3148 char *orig_data = kstrdup(data, GFP_KERNEL);
3149 struct buffer_head *bh;
3150 struct ext4_super_block *es = NULL;
3151 struct ext4_sb_info *sbi;
3153 ext4_fsblk_t sb_block = get_sb_block(&data);
3154 ext4_fsblk_t logical_sb_block;
3155 unsigned long offset = 0;
3156 unsigned long journal_devnum = 0;
3157 unsigned long def_mount_opts;
3161 int blocksize, clustersize;
3162 unsigned int db_count;
3164 int needs_recovery, has_huge_files, has_bigalloc;
3167 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3168 ext4_group_t first_not_zeroed;
3170 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3174 sbi->s_blockgroup_lock =
3175 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3176 if (!sbi->s_blockgroup_lock) {
3180 sb->s_fs_info = sbi;
3182 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3183 sbi->s_sb_block = sb_block;
3184 if (sb->s_bdev->bd_part)
3185 sbi->s_sectors_written_start =
3186 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3188 /* Cleanup superblock name */
3189 strreplace(sb->s_id, '/', '!');
3191 /* -EINVAL is default */
3193 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3195 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3200 * The ext4 superblock will not be buffer aligned for other than 1kB
3201 * block sizes. We need to calculate the offset from buffer start.
3203 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3204 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3205 offset = do_div(logical_sb_block, blocksize);
3207 logical_sb_block = sb_block;
3210 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3211 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3215 * Note: s_es must be initialized as soon as possible because
3216 * some ext4 macro-instructions depend on its value
3218 es = (struct ext4_super_block *) (bh->b_data + offset);
3220 sb->s_magic = le16_to_cpu(es->s_magic);
3221 if (sb->s_magic != EXT4_SUPER_MAGIC)
3223 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3225 /* Warn if metadata_csum and gdt_csum are both set. */
3226 if (ext4_has_feature_metadata_csum(sb) &&
3227 ext4_has_feature_gdt_csum(sb))
3228 ext4_warning(sb, "metadata_csum and uninit_bg are "
3229 "redundant flags; please run fsck.");
3231 /* Check for a known checksum algorithm */
3232 if (!ext4_verify_csum_type(sb, es)) {
3233 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3234 "unknown checksum algorithm.");
3239 /* Load the checksum driver */
3240 if (ext4_has_feature_metadata_csum(sb)) {
3241 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3242 if (IS_ERR(sbi->s_chksum_driver)) {
3243 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3244 ret = PTR_ERR(sbi->s_chksum_driver);
3245 sbi->s_chksum_driver = NULL;
3250 /* Check superblock checksum */
3251 if (!ext4_superblock_csum_verify(sb, es)) {
3252 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3253 "invalid superblock checksum. Run e2fsck?");
3259 /* Precompute checksum seed for all metadata */
3260 if (ext4_has_feature_csum_seed(sb))
3261 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3262 else if (ext4_has_metadata_csum(sb))
3263 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3264 sizeof(es->s_uuid));
3266 /* Set defaults before we parse the mount options */
3267 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3268 set_opt(sb, INIT_INODE_TABLE);
3269 if (def_mount_opts & EXT4_DEFM_DEBUG)
3271 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3273 if (def_mount_opts & EXT4_DEFM_UID16)
3274 set_opt(sb, NO_UID32);
3275 /* xattr user namespace & acls are now defaulted on */
3276 set_opt(sb, XATTR_USER);
3277 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3278 set_opt(sb, POSIX_ACL);
3280 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3281 if (ext4_has_metadata_csum(sb))
3282 set_opt(sb, JOURNAL_CHECKSUM);
3284 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3285 set_opt(sb, JOURNAL_DATA);
3286 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3287 set_opt(sb, ORDERED_DATA);
3288 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3289 set_opt(sb, WRITEBACK_DATA);
3291 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3292 set_opt(sb, ERRORS_PANIC);
3293 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3294 set_opt(sb, ERRORS_CONT);
3296 set_opt(sb, ERRORS_RO);
3297 /* block_validity enabled by default; disable with noblock_validity */
3298 set_opt(sb, BLOCK_VALIDITY);
3299 if (def_mount_opts & EXT4_DEFM_DISCARD)
3300 set_opt(sb, DISCARD);
3302 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3303 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3304 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3305 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3306 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3308 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3309 set_opt(sb, BARRIER);
3312 * enable delayed allocation by default
3313 * Use -o nodelalloc to turn it off
3315 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3316 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3317 set_opt(sb, DELALLOC);
3320 * set default s_li_wait_mult for lazyinit, for the case there is
3321 * no mount option specified.
3323 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3325 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3326 &journal_devnum, &journal_ioprio, 0)) {
3327 ext4_msg(sb, KERN_WARNING,
3328 "failed to parse options in superblock: %s",
3329 sbi->s_es->s_mount_opts);
3331 sbi->s_def_mount_opt = sbi->s_mount_opt;
3332 if (!parse_options((char *) data, sb, &journal_devnum,
3333 &journal_ioprio, 0))
3336 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3337 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3338 "with data=journal disables delayed "
3339 "allocation and O_DIRECT support!\n");
3340 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3341 ext4_msg(sb, KERN_ERR, "can't mount with "
3342 "both data=journal and delalloc");
3345 if (test_opt(sb, DIOREAD_NOLOCK)) {
3346 ext4_msg(sb, KERN_ERR, "can't mount with "
3347 "both data=journal and dioread_nolock");
3350 if (test_opt(sb, DAX)) {
3351 ext4_msg(sb, KERN_ERR, "can't mount with "
3352 "both data=journal and dax");
3355 if (test_opt(sb, DELALLOC))
3356 clear_opt(sb, DELALLOC);
3358 sb->s_iflags |= SB_I_CGROUPWB;
3361 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3362 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3364 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3365 (ext4_has_compat_features(sb) ||
3366 ext4_has_ro_compat_features(sb) ||
3367 ext4_has_incompat_features(sb)))
3368 ext4_msg(sb, KERN_WARNING,
3369 "feature flags set on rev 0 fs, "
3370 "running e2fsck is recommended");
3372 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3373 set_opt2(sb, HURD_COMPAT);
3374 if (ext4_has_feature_64bit(sb)) {
3375 ext4_msg(sb, KERN_ERR,
3376 "The Hurd can't support 64-bit file systems");
3381 if (IS_EXT2_SB(sb)) {
3382 if (ext2_feature_set_ok(sb))
3383 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3384 "using the ext4 subsystem");
3386 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3387 "to feature incompatibilities");
3392 if (IS_EXT3_SB(sb)) {
3393 if (ext3_feature_set_ok(sb))
3394 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3395 "using the ext4 subsystem");
3397 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3398 "to feature incompatibilities");
3404 * Check feature flags regardless of the revision level, since we
3405 * previously didn't change the revision level when setting the flags,
3406 * so there is a chance incompat flags are set on a rev 0 filesystem.
3408 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3411 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3412 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3413 blocksize > EXT4_MAX_BLOCK_SIZE) {
3414 ext4_msg(sb, KERN_ERR,
3415 "Unsupported filesystem blocksize %d", blocksize);
3419 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3420 if (blocksize != PAGE_SIZE) {
3421 ext4_msg(sb, KERN_ERR,
3422 "error: unsupported blocksize for dax");
3425 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3426 ext4_msg(sb, KERN_ERR,
3427 "error: device does not support dax");
3432 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3433 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3434 es->s_encryption_level);
3438 if (sb->s_blocksize != blocksize) {
3439 /* Validate the filesystem blocksize */
3440 if (!sb_set_blocksize(sb, blocksize)) {
3441 ext4_msg(sb, KERN_ERR, "bad block size %d",
3447 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3448 offset = do_div(logical_sb_block, blocksize);
3449 bh = sb_bread_unmovable(sb, logical_sb_block);
3451 ext4_msg(sb, KERN_ERR,
3452 "Can't read superblock on 2nd try");
3455 es = (struct ext4_super_block *)(bh->b_data + offset);
3457 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3458 ext4_msg(sb, KERN_ERR,
3459 "Magic mismatch, very weird!");
3464 has_huge_files = ext4_has_feature_huge_file(sb);
3465 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3467 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3469 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3470 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3471 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3473 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3474 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3475 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3476 (!is_power_of_2(sbi->s_inode_size)) ||
3477 (sbi->s_inode_size > blocksize)) {
3478 ext4_msg(sb, KERN_ERR,
3479 "unsupported inode size: %d",
3483 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3484 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3487 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3488 if (ext4_has_feature_64bit(sb)) {
3489 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3490 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3491 !is_power_of_2(sbi->s_desc_size)) {
3492 ext4_msg(sb, KERN_ERR,
3493 "unsupported descriptor size %lu",
3498 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3500 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3501 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3502 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3505 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3506 if (sbi->s_inodes_per_block == 0)
3508 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3509 sbi->s_inodes_per_block;
3510 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3512 sbi->s_mount_state = le16_to_cpu(es->s_state);
3513 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3514 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3516 for (i = 0; i < 4; i++)
3517 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3518 sbi->s_def_hash_version = es->s_def_hash_version;
3519 if (ext4_has_feature_dir_index(sb)) {
3520 i = le32_to_cpu(es->s_flags);
3521 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3522 sbi->s_hash_unsigned = 3;
3523 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3524 #ifdef __CHAR_UNSIGNED__
3525 if (!(sb->s_flags & MS_RDONLY))
3527 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3528 sbi->s_hash_unsigned = 3;
3530 if (!(sb->s_flags & MS_RDONLY))
3532 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3537 /* Handle clustersize */
3538 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3539 has_bigalloc = ext4_has_feature_bigalloc(sb);
3541 if (clustersize < blocksize) {
3542 ext4_msg(sb, KERN_ERR,
3543 "cluster size (%d) smaller than "
3544 "block size (%d)", clustersize, blocksize);
3547 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3548 le32_to_cpu(es->s_log_block_size);
3549 sbi->s_clusters_per_group =
3550 le32_to_cpu(es->s_clusters_per_group);
3551 if (sbi->s_clusters_per_group > blocksize * 8) {
3552 ext4_msg(sb, KERN_ERR,
3553 "#clusters per group too big: %lu",
3554 sbi->s_clusters_per_group);
3557 if (sbi->s_blocks_per_group !=
3558 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3559 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3560 "clusters per group (%lu) inconsistent",
3561 sbi->s_blocks_per_group,
3562 sbi->s_clusters_per_group);
3566 if (clustersize != blocksize) {
3567 ext4_warning(sb, "fragment/cluster size (%d) != "
3568 "block size (%d)", clustersize,
3570 clustersize = blocksize;
3572 if (sbi->s_blocks_per_group > blocksize * 8) {
3573 ext4_msg(sb, KERN_ERR,
3574 "#blocks per group too big: %lu",
3575 sbi->s_blocks_per_group);
3578 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3579 sbi->s_cluster_bits = 0;
3581 sbi->s_cluster_ratio = clustersize / blocksize;
3583 if (sbi->s_inodes_per_group > blocksize * 8) {
3584 ext4_msg(sb, KERN_ERR,
3585 "#inodes per group too big: %lu",
3586 sbi->s_inodes_per_group);
3590 /* Do we have standard group size of clustersize * 8 blocks ? */
3591 if (sbi->s_blocks_per_group == clustersize << 3)
3592 set_opt2(sb, STD_GROUP_SIZE);
3595 * Test whether we have more sectors than will fit in sector_t,
3596 * and whether the max offset is addressable by the page cache.
3598 err = generic_check_addressable(sb->s_blocksize_bits,
3599 ext4_blocks_count(es));
3601 ext4_msg(sb, KERN_ERR, "filesystem"
3602 " too large to mount safely on this system");
3603 if (sizeof(sector_t) < 8)
3604 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3608 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3611 /* check blocks count against device size */
3612 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3613 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3614 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3615 "exceeds size of device (%llu blocks)",
3616 ext4_blocks_count(es), blocks_count);
3621 * It makes no sense for the first data block to be beyond the end
3622 * of the filesystem.
3624 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3625 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3626 "block %u is beyond end of filesystem (%llu)",
3627 le32_to_cpu(es->s_first_data_block),
3628 ext4_blocks_count(es));
3631 blocks_count = (ext4_blocks_count(es) -
3632 le32_to_cpu(es->s_first_data_block) +
3633 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3634 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3635 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3636 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3637 "(block count %llu, first data block %u, "
3638 "blocks per group %lu)", sbi->s_groups_count,
3639 ext4_blocks_count(es),
3640 le32_to_cpu(es->s_first_data_block),
3641 EXT4_BLOCKS_PER_GROUP(sb));
3644 sbi->s_groups_count = blocks_count;
3645 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3646 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3647 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3648 EXT4_DESC_PER_BLOCK(sb);
3649 sbi->s_group_desc = ext4_kvmalloc(db_count *
3650 sizeof(struct buffer_head *),
3652 if (sbi->s_group_desc == NULL) {
3653 ext4_msg(sb, KERN_ERR, "not enough memory");
3658 bgl_lock_init(sbi->s_blockgroup_lock);
3660 for (i = 0; i < db_count; i++) {
3661 block = descriptor_loc(sb, logical_sb_block, i);
3662 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3663 if (!sbi->s_group_desc[i]) {
3664 ext4_msg(sb, KERN_ERR,
3665 "can't read group descriptor %d", i);
3670 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3671 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3672 ret = -EFSCORRUPTED;
3676 sbi->s_gdb_count = db_count;
3677 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3678 spin_lock_init(&sbi->s_next_gen_lock);
3680 setup_timer(&sbi->s_err_report, print_daily_error_info,
3681 (unsigned long) sb);
3683 /* Register extent status tree shrinker */
3684 if (ext4_es_register_shrinker(sbi))
3687 sbi->s_stripe = ext4_get_stripe_size(sbi);
3688 sbi->s_extent_max_zeroout_kb = 32;
3691 * set up enough so that it can read an inode
3693 sb->s_op = &ext4_sops;
3694 sb->s_export_op = &ext4_export_ops;
3695 sb->s_xattr = ext4_xattr_handlers;
3697 sb->dq_op = &ext4_quota_operations;
3698 if (ext4_has_feature_quota(sb))
3699 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3701 sb->s_qcop = &ext4_qctl_operations;
3702 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3704 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3706 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3707 mutex_init(&sbi->s_orphan_lock);
3711 needs_recovery = (es->s_last_orphan != 0 ||
3712 ext4_has_feature_journal_needs_recovery(sb));
3714 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3715 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3716 goto failed_mount3a;
3719 * The first inode we look at is the journal inode. Don't try
3720 * root first: it may be modified in the journal!
3722 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3723 if (ext4_load_journal(sb, es, journal_devnum))
3724 goto failed_mount3a;
3725 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3726 ext4_has_feature_journal_needs_recovery(sb)) {
3727 ext4_msg(sb, KERN_ERR, "required journal recovery "
3728 "suppressed and not mounted read-only");
3729 goto failed_mount_wq;
3731 /* Nojournal mode, all journal mount options are illegal */
3732 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3733 ext4_msg(sb, KERN_ERR, "can't mount with "
3734 "journal_checksum, fs mounted w/o journal");
3735 goto failed_mount_wq;
3737 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3738 ext4_msg(sb, KERN_ERR, "can't mount with "
3739 "journal_async_commit, fs mounted w/o journal");
3740 goto failed_mount_wq;
3742 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3743 ext4_msg(sb, KERN_ERR, "can't mount with "
3744 "commit=%lu, fs mounted w/o journal",
3745 sbi->s_commit_interval / HZ);
3746 goto failed_mount_wq;
3748 if (EXT4_MOUNT_DATA_FLAGS &
3749 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3750 ext4_msg(sb, KERN_ERR, "can't mount with "
3751 "data=, fs mounted w/o journal");
3752 goto failed_mount_wq;
3754 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3755 clear_opt(sb, JOURNAL_CHECKSUM);
3756 clear_opt(sb, DATA_FLAGS);
3757 sbi->s_journal = NULL;
3762 if (ext4_has_feature_64bit(sb) &&
3763 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3764 JBD2_FEATURE_INCOMPAT_64BIT)) {
3765 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3766 goto failed_mount_wq;
3769 if (!set_journal_csum_feature_set(sb)) {
3770 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3772 goto failed_mount_wq;
3775 /* We have now updated the journal if required, so we can
3776 * validate the data journaling mode. */
3777 switch (test_opt(sb, DATA_FLAGS)) {
3779 /* No mode set, assume a default based on the journal
3780 * capabilities: ORDERED_DATA if the journal can
3781 * cope, else JOURNAL_DATA
3783 if (jbd2_journal_check_available_features
3784 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3785 set_opt(sb, ORDERED_DATA);
3787 set_opt(sb, JOURNAL_DATA);
3790 case EXT4_MOUNT_ORDERED_DATA:
3791 case EXT4_MOUNT_WRITEBACK_DATA:
3792 if (!jbd2_journal_check_available_features
3793 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3794 ext4_msg(sb, KERN_ERR, "Journal does not support "
3795 "requested data journaling mode");
3796 goto failed_mount_wq;
3801 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3803 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3806 sbi->s_mb_cache = ext4_xattr_create_cache();
3807 if (!sbi->s_mb_cache) {
3808 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3809 goto failed_mount_wq;
3812 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3813 (blocksize != PAGE_SIZE)) {
3814 ext4_msg(sb, KERN_ERR,
3815 "Unsupported blocksize for fs encryption");
3816 goto failed_mount_wq;
3819 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3820 !ext4_has_feature_encrypt(sb)) {
3821 ext4_set_feature_encrypt(sb);
3822 ext4_commit_super(sb, 1);
3826 * Get the # of file system overhead blocks from the
3827 * superblock if present.
3829 if (es->s_overhead_clusters)
3830 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3832 err = ext4_calculate_overhead(sb);
3834 goto failed_mount_wq;
3838 * The maximum number of concurrent works can be high and
3839 * concurrency isn't really necessary. Limit it to 1.
3841 EXT4_SB(sb)->rsv_conversion_wq =
3842 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3843 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3844 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3850 * The jbd2_journal_load will have done any necessary log recovery,
3851 * so we can safely mount the rest of the filesystem now.
3854 root = ext4_iget(sb, EXT4_ROOT_INO);
3856 ext4_msg(sb, KERN_ERR, "get root inode failed");
3857 ret = PTR_ERR(root);
3861 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3862 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3866 sb->s_root = d_make_root(root);
3868 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3873 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3874 sb->s_flags |= MS_RDONLY;
3876 /* determine the minimum size of new large inodes, if present */
3877 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3878 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3879 EXT4_GOOD_OLD_INODE_SIZE;
3880 if (ext4_has_feature_extra_isize(sb)) {
3881 if (sbi->s_want_extra_isize <
3882 le16_to_cpu(es->s_want_extra_isize))
3883 sbi->s_want_extra_isize =
3884 le16_to_cpu(es->s_want_extra_isize);
3885 if (sbi->s_want_extra_isize <
3886 le16_to_cpu(es->s_min_extra_isize))
3887 sbi->s_want_extra_isize =
3888 le16_to_cpu(es->s_min_extra_isize);
3891 /* Check if enough inode space is available */
3892 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3893 sbi->s_inode_size) {
3894 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3895 EXT4_GOOD_OLD_INODE_SIZE;
3896 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3900 ext4_set_resv_clusters(sb);
3902 err = ext4_setup_system_zone(sb);
3904 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3906 goto failed_mount4a;
3910 err = ext4_mb_init(sb);
3912 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3917 block = ext4_count_free_clusters(sb);
3918 ext4_free_blocks_count_set(sbi->s_es,
3919 EXT4_C2B(sbi, block));
3920 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3923 unsigned long freei = ext4_count_free_inodes(sb);
3924 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3925 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3929 err = percpu_counter_init(&sbi->s_dirs_counter,
3930 ext4_count_dirs(sb), GFP_KERNEL);
3932 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3935 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
3938 ext4_msg(sb, KERN_ERR, "insufficient memory");
3942 if (ext4_has_feature_flex_bg(sb))
3943 if (!ext4_fill_flex_info(sb)) {
3944 ext4_msg(sb, KERN_ERR,
3945 "unable to initialize "
3946 "flex_bg meta info!");
3950 err = ext4_register_li_request(sb, first_not_zeroed);
3954 err = ext4_register_sysfs(sb);
3959 /* Enable quota usage during mount. */
3960 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3961 err = ext4_enable_quotas(sb);
3965 #endif /* CONFIG_QUOTA */
3967 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3968 ext4_orphan_cleanup(sb, es);
3969 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3970 if (needs_recovery) {
3971 ext4_msg(sb, KERN_INFO, "recovery complete");
3972 ext4_mark_recovery_complete(sb, es);
3974 if (EXT4_SB(sb)->s_journal) {
3975 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3976 descr = " journalled data mode";
3977 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3978 descr = " ordered data mode";
3980 descr = " writeback data mode";
3982 descr = "out journal";
3984 if (test_opt(sb, DISCARD)) {
3985 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3986 if (!blk_queue_discard(q))
3987 ext4_msg(sb, KERN_WARNING,
3988 "mounting with \"discard\" option, but "
3989 "the device does not support discard");
3992 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3993 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3994 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3995 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3997 if (es->s_error_count)
3998 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4000 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4001 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4002 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4003 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4010 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4015 ext4_unregister_sysfs(sb);
4018 ext4_unregister_li_request(sb);
4020 ext4_mb_release(sb);
4021 if (sbi->s_flex_groups)
4022 kvfree(sbi->s_flex_groups);
4023 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4024 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4025 percpu_counter_destroy(&sbi->s_dirs_counter);
4026 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4028 ext4_ext_release(sb);
4029 ext4_release_system_zone(sb);
4034 ext4_msg(sb, KERN_ERR, "mount failed");
4035 if (EXT4_SB(sb)->rsv_conversion_wq)
4036 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4038 if (sbi->s_mb_cache) {
4039 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4040 sbi->s_mb_cache = NULL;
4042 if (sbi->s_journal) {
4043 jbd2_journal_destroy(sbi->s_journal);
4044 sbi->s_journal = NULL;
4047 ext4_es_unregister_shrinker(sbi);
4049 del_timer_sync(&sbi->s_err_report);
4051 kthread_stop(sbi->s_mmp_tsk);
4053 for (i = 0; i < db_count; i++)
4054 brelse(sbi->s_group_desc[i]);
4055 kvfree(sbi->s_group_desc);
4057 if (sbi->s_chksum_driver)
4058 crypto_free_shash(sbi->s_chksum_driver);
4060 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4061 kfree(sbi->s_qf_names[i]);
4063 ext4_blkdev_remove(sbi);
4066 sb->s_fs_info = NULL;
4067 kfree(sbi->s_blockgroup_lock);
4071 return err ? err : ret;
4075 * Setup any per-fs journal parameters now. We'll do this both on
4076 * initial mount, once the journal has been initialised but before we've
4077 * done any recovery; and again on any subsequent remount.
4079 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4081 struct ext4_sb_info *sbi = EXT4_SB(sb);
4083 journal->j_commit_interval = sbi->s_commit_interval;
4084 journal->j_min_batch_time = sbi->s_min_batch_time;
4085 journal->j_max_batch_time = sbi->s_max_batch_time;
4087 write_lock(&journal->j_state_lock);
4088 if (test_opt(sb, BARRIER))
4089 journal->j_flags |= JBD2_BARRIER;
4091 journal->j_flags &= ~JBD2_BARRIER;
4092 if (test_opt(sb, DATA_ERR_ABORT))
4093 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4095 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4096 write_unlock(&journal->j_state_lock);
4099 static journal_t *ext4_get_journal(struct super_block *sb,
4100 unsigned int journal_inum)
4102 struct inode *journal_inode;
4105 BUG_ON(!ext4_has_feature_journal(sb));
4107 /* First, test for the existence of a valid inode on disk. Bad
4108 * things happen if we iget() an unused inode, as the subsequent
4109 * iput() will try to delete it. */
4111 journal_inode = ext4_iget(sb, journal_inum);
4112 if (IS_ERR(journal_inode)) {
4113 ext4_msg(sb, KERN_ERR, "no journal found");
4116 if (!journal_inode->i_nlink) {
4117 make_bad_inode(journal_inode);
4118 iput(journal_inode);
4119 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4123 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4124 journal_inode, journal_inode->i_size);
4125 if (!S_ISREG(journal_inode->i_mode)) {
4126 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4127 iput(journal_inode);
4131 journal = jbd2_journal_init_inode(journal_inode);
4133 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4134 iput(journal_inode);
4137 journal->j_private = sb;
4138 ext4_init_journal_params(sb, journal);
4142 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4145 struct buffer_head *bh;
4149 int hblock, blocksize;
4150 ext4_fsblk_t sb_block;
4151 unsigned long offset;
4152 struct ext4_super_block *es;
4153 struct block_device *bdev;
4155 BUG_ON(!ext4_has_feature_journal(sb));
4157 bdev = ext4_blkdev_get(j_dev, sb);
4161 blocksize = sb->s_blocksize;
4162 hblock = bdev_logical_block_size(bdev);
4163 if (blocksize < hblock) {
4164 ext4_msg(sb, KERN_ERR,
4165 "blocksize too small for journal device");
4169 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4170 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4171 set_blocksize(bdev, blocksize);
4172 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4173 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4174 "external journal");
4178 es = (struct ext4_super_block *) (bh->b_data + offset);
4179 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4180 !(le32_to_cpu(es->s_feature_incompat) &
4181 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4182 ext4_msg(sb, KERN_ERR, "external journal has "
4188 if ((le32_to_cpu(es->s_feature_ro_compat) &
4189 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4190 es->s_checksum != ext4_superblock_csum(sb, es)) {
4191 ext4_msg(sb, KERN_ERR, "external journal has "
4192 "corrupt superblock");
4197 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4198 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4203 len = ext4_blocks_count(es);
4204 start = sb_block + 1;
4205 brelse(bh); /* we're done with the superblock */
4207 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4208 start, len, blocksize);
4210 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4213 journal->j_private = sb;
4214 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4215 wait_on_buffer(journal->j_sb_buffer);
4216 if (!buffer_uptodate(journal->j_sb_buffer)) {
4217 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4220 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4221 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4222 "user (unsupported) - %d",
4223 be32_to_cpu(journal->j_superblock->s_nr_users));
4226 EXT4_SB(sb)->journal_bdev = bdev;
4227 ext4_init_journal_params(sb, journal);
4231 jbd2_journal_destroy(journal);
4233 ext4_blkdev_put(bdev);
4237 static int ext4_load_journal(struct super_block *sb,
4238 struct ext4_super_block *es,
4239 unsigned long journal_devnum)
4242 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4245 int really_read_only;
4247 BUG_ON(!ext4_has_feature_journal(sb));
4249 if (journal_devnum &&
4250 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4251 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4252 "numbers have changed");
4253 journal_dev = new_decode_dev(journal_devnum);
4255 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4257 really_read_only = bdev_read_only(sb->s_bdev);
4260 * Are we loading a blank journal or performing recovery after a
4261 * crash? For recovery, we need to check in advance whether we
4262 * can get read-write access to the device.
4264 if (ext4_has_feature_journal_needs_recovery(sb)) {
4265 if (sb->s_flags & MS_RDONLY) {
4266 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4267 "required on readonly filesystem");
4268 if (really_read_only) {
4269 ext4_msg(sb, KERN_ERR, "write access "
4270 "unavailable, cannot proceed");
4273 ext4_msg(sb, KERN_INFO, "write access will "
4274 "be enabled during recovery");
4278 if (journal_inum && journal_dev) {
4279 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4280 "and inode journals!");
4285 if (!(journal = ext4_get_journal(sb, journal_inum)))
4288 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4292 if (!(journal->j_flags & JBD2_BARRIER))
4293 ext4_msg(sb, KERN_INFO, "barriers disabled");
4295 if (!ext4_has_feature_journal_needs_recovery(sb))
4296 err = jbd2_journal_wipe(journal, !really_read_only);
4298 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4300 memcpy(save, ((char *) es) +
4301 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4302 err = jbd2_journal_load(journal);
4304 memcpy(((char *) es) + EXT4_S_ERR_START,
4305 save, EXT4_S_ERR_LEN);
4310 ext4_msg(sb, KERN_ERR, "error loading journal");
4311 jbd2_journal_destroy(journal);
4315 EXT4_SB(sb)->s_journal = journal;
4316 ext4_clear_journal_err(sb, es);
4318 if (!really_read_only && journal_devnum &&
4319 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4320 es->s_journal_dev = cpu_to_le32(journal_devnum);
4322 /* Make sure we flush the recovery flag to disk. */
4323 ext4_commit_super(sb, 1);
4329 static int ext4_commit_super(struct super_block *sb, int sync)
4331 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4332 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4335 if (!sbh || block_device_ejected(sb))
4337 if (buffer_write_io_error(sbh)) {
4339 * Oh, dear. A previous attempt to write the
4340 * superblock failed. This could happen because the
4341 * USB device was yanked out. Or it could happen to
4342 * be a transient write error and maybe the block will
4343 * be remapped. Nothing we can do but to retry the
4344 * write and hope for the best.
4346 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4347 "superblock detected");
4348 clear_buffer_write_io_error(sbh);
4349 set_buffer_uptodate(sbh);
4352 * If the file system is mounted read-only, don't update the
4353 * superblock write time. This avoids updating the superblock
4354 * write time when we are mounting the root file system
4355 * read/only but we need to replay the journal; at that point,
4356 * for people who are east of GMT and who make their clock
4357 * tick in localtime for Windows bug-for-bug compatibility,
4358 * the clock is set in the future, and this will cause e2fsck
4359 * to complain and force a full file system check.
4361 if (!(sb->s_flags & MS_RDONLY))
4362 es->s_wtime = cpu_to_le32(get_seconds());
4363 if (sb->s_bdev->bd_part)
4364 es->s_kbytes_written =
4365 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4366 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4367 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4369 es->s_kbytes_written =
4370 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4371 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4372 ext4_free_blocks_count_set(es,
4373 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4374 &EXT4_SB(sb)->s_freeclusters_counter)));
4375 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4376 es->s_free_inodes_count =
4377 cpu_to_le32(percpu_counter_sum_positive(
4378 &EXT4_SB(sb)->s_freeinodes_counter));
4379 BUFFER_TRACE(sbh, "marking dirty");
4380 ext4_superblock_csum_set(sb);
4381 mark_buffer_dirty(sbh);
4383 error = __sync_dirty_buffer(sbh,
4384 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4388 error = buffer_write_io_error(sbh);
4390 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4392 clear_buffer_write_io_error(sbh);
4393 set_buffer_uptodate(sbh);
4400 * Have we just finished recovery? If so, and if we are mounting (or
4401 * remounting) the filesystem readonly, then we will end up with a
4402 * consistent fs on disk. Record that fact.
4404 static void ext4_mark_recovery_complete(struct super_block *sb,
4405 struct ext4_super_block *es)
4407 journal_t *journal = EXT4_SB(sb)->s_journal;
4409 if (!ext4_has_feature_journal(sb)) {
4410 BUG_ON(journal != NULL);
4413 jbd2_journal_lock_updates(journal);
4414 if (jbd2_journal_flush(journal) < 0)
4417 if (ext4_has_feature_journal_needs_recovery(sb) &&
4418 sb->s_flags & MS_RDONLY) {
4419 ext4_clear_feature_journal_needs_recovery(sb);
4420 ext4_commit_super(sb, 1);
4424 jbd2_journal_unlock_updates(journal);
4428 * If we are mounting (or read-write remounting) a filesystem whose journal
4429 * has recorded an error from a previous lifetime, move that error to the
4430 * main filesystem now.
4432 static void ext4_clear_journal_err(struct super_block *sb,
4433 struct ext4_super_block *es)
4439 BUG_ON(!ext4_has_feature_journal(sb));
4441 journal = EXT4_SB(sb)->s_journal;
4444 * Now check for any error status which may have been recorded in the
4445 * journal by a prior ext4_error() or ext4_abort()
4448 j_errno = jbd2_journal_errno(journal);
4452 errstr = ext4_decode_error(sb, j_errno, nbuf);
4453 ext4_warning(sb, "Filesystem error recorded "
4454 "from previous mount: %s", errstr);
4455 ext4_warning(sb, "Marking fs in need of filesystem check.");
4457 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4458 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4459 ext4_commit_super(sb, 1);
4461 jbd2_journal_clear_err(journal);
4462 jbd2_journal_update_sb_errno(journal);
4467 * Force the running and committing transactions to commit,
4468 * and wait on the commit.
4470 int ext4_force_commit(struct super_block *sb)
4474 if (sb->s_flags & MS_RDONLY)
4477 journal = EXT4_SB(sb)->s_journal;
4478 return ext4_journal_force_commit(journal);
4481 static int ext4_sync_fs(struct super_block *sb, int wait)
4485 bool needs_barrier = false;
4486 struct ext4_sb_info *sbi = EXT4_SB(sb);
4488 trace_ext4_sync_fs(sb, wait);
4489 flush_workqueue(sbi->rsv_conversion_wq);
4491 * Writeback quota in non-journalled quota case - journalled quota has
4494 dquot_writeback_dquots(sb, -1);
4496 * Data writeback is possible w/o journal transaction, so barrier must
4497 * being sent at the end of the function. But we can skip it if
4498 * transaction_commit will do it for us.
4500 if (sbi->s_journal) {
4501 target = jbd2_get_latest_transaction(sbi->s_journal);
4502 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4503 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4504 needs_barrier = true;
4506 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4508 ret = jbd2_log_wait_commit(sbi->s_journal,
4511 } else if (wait && test_opt(sb, BARRIER))
4512 needs_barrier = true;
4513 if (needs_barrier) {
4515 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4524 * LVM calls this function before a (read-only) snapshot is created. This
4525 * gives us a chance to flush the journal completely and mark the fs clean.
4527 * Note that only this function cannot bring a filesystem to be in a clean
4528 * state independently. It relies on upper layer to stop all data & metadata
4531 static int ext4_freeze(struct super_block *sb)
4536 if (sb->s_flags & MS_RDONLY)
4539 journal = EXT4_SB(sb)->s_journal;
4542 /* Now we set up the journal barrier. */
4543 jbd2_journal_lock_updates(journal);
4546 * Don't clear the needs_recovery flag if we failed to
4547 * flush the journal.
4549 error = jbd2_journal_flush(journal);
4553 /* Journal blocked and flushed, clear needs_recovery flag. */
4554 ext4_clear_feature_journal_needs_recovery(sb);
4557 error = ext4_commit_super(sb, 1);
4560 /* we rely on upper layer to stop further updates */
4561 jbd2_journal_unlock_updates(journal);
4566 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4567 * flag here, even though the filesystem is not technically dirty yet.
4569 static int ext4_unfreeze(struct super_block *sb)
4571 if (sb->s_flags & MS_RDONLY)
4574 if (EXT4_SB(sb)->s_journal) {
4575 /* Reset the needs_recovery flag before the fs is unlocked. */
4576 ext4_set_feature_journal_needs_recovery(sb);
4579 ext4_commit_super(sb, 1);
4584 * Structure to save mount options for ext4_remount's benefit
4586 struct ext4_mount_options {
4587 unsigned long s_mount_opt;
4588 unsigned long s_mount_opt2;
4591 unsigned long s_commit_interval;
4592 u32 s_min_batch_time, s_max_batch_time;
4595 char *s_qf_names[EXT4_MAXQUOTAS];
4599 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4601 struct ext4_super_block *es;
4602 struct ext4_sb_info *sbi = EXT4_SB(sb);
4603 unsigned long old_sb_flags;
4604 struct ext4_mount_options old_opts;
4605 int enable_quota = 0;
4607 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4612 char *orig_data = kstrdup(data, GFP_KERNEL);
4614 /* Store the original options */
4615 old_sb_flags = sb->s_flags;
4616 old_opts.s_mount_opt = sbi->s_mount_opt;
4617 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4618 old_opts.s_resuid = sbi->s_resuid;
4619 old_opts.s_resgid = sbi->s_resgid;
4620 old_opts.s_commit_interval = sbi->s_commit_interval;
4621 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4622 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4624 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4625 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4626 if (sbi->s_qf_names[i]) {
4627 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4629 if (!old_opts.s_qf_names[i]) {
4630 for (j = 0; j < i; j++)
4631 kfree(old_opts.s_qf_names[j]);
4636 old_opts.s_qf_names[i] = NULL;
4638 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4639 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4641 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4646 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4647 test_opt(sb, JOURNAL_CHECKSUM)) {
4648 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4649 "during remount not supported; ignoring");
4650 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4653 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4654 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4655 ext4_msg(sb, KERN_ERR, "can't mount with "
4656 "both data=journal and delalloc");
4660 if (test_opt(sb, DIOREAD_NOLOCK)) {
4661 ext4_msg(sb, KERN_ERR, "can't mount with "
4662 "both data=journal and dioread_nolock");
4666 if (test_opt(sb, DAX)) {
4667 ext4_msg(sb, KERN_ERR, "can't mount with "
4668 "both data=journal and dax");
4674 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4675 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4676 "dax flag with busy inodes while remounting");
4677 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4680 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4681 ext4_abort(sb, "Abort forced by user");
4683 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4684 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4688 if (sbi->s_journal) {
4689 ext4_init_journal_params(sb, sbi->s_journal);
4690 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4693 if (*flags & MS_LAZYTIME)
4694 sb->s_flags |= MS_LAZYTIME;
4696 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4697 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4702 if (*flags & MS_RDONLY) {
4703 err = sync_filesystem(sb);
4706 err = dquot_suspend(sb, -1);
4711 * First of all, the unconditional stuff we have to do
4712 * to disable replay of the journal when we next remount
4714 sb->s_flags |= MS_RDONLY;
4717 * OK, test if we are remounting a valid rw partition
4718 * readonly, and if so set the rdonly flag and then
4719 * mark the partition as valid again.
4721 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4722 (sbi->s_mount_state & EXT4_VALID_FS))
4723 es->s_state = cpu_to_le16(sbi->s_mount_state);
4726 ext4_mark_recovery_complete(sb, es);
4728 /* Make sure we can mount this feature set readwrite */
4729 if (ext4_has_feature_readonly(sb) ||
4730 !ext4_feature_set_ok(sb, 0)) {
4735 * Make sure the group descriptor checksums
4736 * are sane. If they aren't, refuse to remount r/w.
4738 for (g = 0; g < sbi->s_groups_count; g++) {
4739 struct ext4_group_desc *gdp =
4740 ext4_get_group_desc(sb, g, NULL);
4742 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4743 ext4_msg(sb, KERN_ERR,
4744 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4745 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4746 le16_to_cpu(gdp->bg_checksum));
4753 * If we have an unprocessed orphan list hanging
4754 * around from a previously readonly bdev mount,
4755 * require a full umount/remount for now.
4757 if (es->s_last_orphan) {
4758 ext4_msg(sb, KERN_WARNING, "Couldn't "
4759 "remount RDWR because of unprocessed "
4760 "orphan inode list. Please "
4761 "umount/remount instead");
4767 * Mounting a RDONLY partition read-write, so reread
4768 * and store the current valid flag. (It may have
4769 * been changed by e2fsck since we originally mounted
4773 ext4_clear_journal_err(sb, es);
4774 sbi->s_mount_state = le16_to_cpu(es->s_state);
4775 if (!ext4_setup_super(sb, es, 0))
4776 sb->s_flags &= ~MS_RDONLY;
4777 if (ext4_has_feature_mmp(sb))
4778 if (ext4_multi_mount_protect(sb,
4779 le64_to_cpu(es->s_mmp_block))) {
4788 * Reinitialize lazy itable initialization thread based on
4791 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4792 ext4_unregister_li_request(sb);
4794 ext4_group_t first_not_zeroed;
4795 first_not_zeroed = ext4_has_uninit_itable(sb);
4796 ext4_register_li_request(sb, first_not_zeroed);
4799 ext4_setup_system_zone(sb);
4800 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4801 ext4_commit_super(sb, 1);
4804 /* Release old quota file names */
4805 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4806 kfree(old_opts.s_qf_names[i]);
4808 if (sb_any_quota_suspended(sb))
4809 dquot_resume(sb, -1);
4810 else if (ext4_has_feature_quota(sb)) {
4811 err = ext4_enable_quotas(sb);
4818 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4819 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4824 sb->s_flags = old_sb_flags;
4825 sbi->s_mount_opt = old_opts.s_mount_opt;
4826 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4827 sbi->s_resuid = old_opts.s_resuid;
4828 sbi->s_resgid = old_opts.s_resgid;
4829 sbi->s_commit_interval = old_opts.s_commit_interval;
4830 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4831 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4833 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4834 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4835 kfree(sbi->s_qf_names[i]);
4836 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4844 static int ext4_statfs_project(struct super_block *sb,
4845 kprojid_t projid, struct kstatfs *buf)
4848 struct dquot *dquot;
4852 qid = make_kqid_projid(projid);
4853 dquot = dqget(sb, qid);
4855 return PTR_ERR(dquot);
4856 spin_lock(&dq_data_lock);
4858 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4859 dquot->dq_dqb.dqb_bsoftlimit :
4860 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4861 if (limit && buf->f_blocks > limit) {
4862 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4863 buf->f_blocks = limit;
4864 buf->f_bfree = buf->f_bavail =
4865 (buf->f_blocks > curblock) ?
4866 (buf->f_blocks - curblock) : 0;
4869 limit = dquot->dq_dqb.dqb_isoftlimit ?
4870 dquot->dq_dqb.dqb_isoftlimit :
4871 dquot->dq_dqb.dqb_ihardlimit;
4872 if (limit && buf->f_files > limit) {
4873 buf->f_files = limit;
4875 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4876 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4879 spin_unlock(&dq_data_lock);
4885 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4887 struct super_block *sb = dentry->d_sb;
4888 struct ext4_sb_info *sbi = EXT4_SB(sb);
4889 struct ext4_super_block *es = sbi->s_es;
4890 ext4_fsblk_t overhead = 0, resv_blocks;
4893 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4895 if (!test_opt(sb, MINIX_DF))
4896 overhead = sbi->s_overhead;
4898 buf->f_type = EXT4_SUPER_MAGIC;
4899 buf->f_bsize = sb->s_blocksize;
4900 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4901 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4902 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4903 /* prevent underflow in case that few free space is available */
4904 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4905 buf->f_bavail = buf->f_bfree -
4906 (ext4_r_blocks_count(es) + resv_blocks);
4907 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4909 buf->f_files = le32_to_cpu(es->s_inodes_count);
4910 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4911 buf->f_namelen = EXT4_NAME_LEN;
4912 fsid = le64_to_cpup((void *)es->s_uuid) ^
4913 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4914 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4915 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4918 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4919 sb_has_quota_limits_enabled(sb, PRJQUOTA))
4920 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4925 /* Helper function for writing quotas on sync - we need to start transaction
4926 * before quota file is locked for write. Otherwise the are possible deadlocks:
4927 * Process 1 Process 2
4928 * ext4_create() quota_sync()
4929 * jbd2_journal_start() write_dquot()
4930 * dquot_initialize() down(dqio_mutex)
4931 * down(dqio_mutex) jbd2_journal_start()
4937 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4939 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4942 static int ext4_write_dquot(struct dquot *dquot)
4946 struct inode *inode;
4948 inode = dquot_to_inode(dquot);
4949 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4950 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4952 return PTR_ERR(handle);
4953 ret = dquot_commit(dquot);
4954 err = ext4_journal_stop(handle);
4960 static int ext4_acquire_dquot(struct dquot *dquot)
4965 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4966 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4968 return PTR_ERR(handle);
4969 ret = dquot_acquire(dquot);
4970 err = ext4_journal_stop(handle);
4976 static int ext4_release_dquot(struct dquot *dquot)
4981 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4982 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4983 if (IS_ERR(handle)) {
4984 /* Release dquot anyway to avoid endless cycle in dqput() */
4985 dquot_release(dquot);
4986 return PTR_ERR(handle);
4988 ret = dquot_release(dquot);
4989 err = ext4_journal_stop(handle);
4995 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4997 struct super_block *sb = dquot->dq_sb;
4998 struct ext4_sb_info *sbi = EXT4_SB(sb);
5000 /* Are we journaling quotas? */
5001 if (ext4_has_feature_quota(sb) ||
5002 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5003 dquot_mark_dquot_dirty(dquot);
5004 return ext4_write_dquot(dquot);
5006 return dquot_mark_dquot_dirty(dquot);
5010 static int ext4_write_info(struct super_block *sb, int type)
5015 /* Data block + inode block */
5016 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5018 return PTR_ERR(handle);
5019 ret = dquot_commit_info(sb, type);
5020 err = ext4_journal_stop(handle);
5027 * Turn on quotas during mount time - we need to find
5028 * the quota file and such...
5030 static int ext4_quota_on_mount(struct super_block *sb, int type)
5032 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5033 EXT4_SB(sb)->s_jquota_fmt, type);
5036 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5038 struct ext4_inode_info *ei = EXT4_I(inode);
5040 /* The first argument of lockdep_set_subclass has to be
5041 * *exactly* the same as the argument to init_rwsem() --- in
5042 * this case, in init_once() --- or lockdep gets unhappy
5043 * because the name of the lock is set using the
5044 * stringification of the argument to init_rwsem().
5046 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5047 lockdep_set_subclass(&ei->i_data_sem, subclass);
5051 * Standard function to be called on quota_on
5053 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5058 if (!test_opt(sb, QUOTA))
5061 /* Quotafile not on the same filesystem? */
5062 if (path->dentry->d_sb != sb)
5064 /* Journaling quota? */
5065 if (EXT4_SB(sb)->s_qf_names[type]) {
5066 /* Quotafile not in fs root? */
5067 if (path->dentry->d_parent != sb->s_root)
5068 ext4_msg(sb, KERN_WARNING,
5069 "Quota file not on filesystem root. "
5070 "Journaled quota will not work");
5074 * When we journal data on quota file, we have to flush journal to see
5075 * all updates to the file when we bypass pagecache...
5077 if (EXT4_SB(sb)->s_journal &&
5078 ext4_should_journal_data(d_inode(path->dentry))) {
5080 * We don't need to lock updates but journal_flush() could
5081 * otherwise be livelocked...
5083 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5084 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5085 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5089 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5090 err = dquot_quota_on(sb, type, format_id, path);
5092 lockdep_set_quota_inode(path->dentry->d_inode,
5097 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5101 struct inode *qf_inode;
5102 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5103 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5104 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5105 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5108 BUG_ON(!ext4_has_feature_quota(sb));
5110 if (!qf_inums[type])
5113 qf_inode = ext4_iget(sb, qf_inums[type]);
5114 if (IS_ERR(qf_inode)) {
5115 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5116 return PTR_ERR(qf_inode);
5119 /* Don't account quota for quota files to avoid recursion */
5120 qf_inode->i_flags |= S_NOQUOTA;
5121 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5122 err = dquot_enable(qf_inode, type, format_id, flags);
5125 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5130 /* Enable usage tracking for all quota types. */
5131 static int ext4_enable_quotas(struct super_block *sb)
5134 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5135 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5136 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5137 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5140 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5141 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5142 if (qf_inums[type]) {
5143 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5144 DQUOT_USAGE_ENABLED);
5147 "Failed to enable quota tracking "
5148 "(type=%d, err=%d). Please run "
5149 "e2fsck to fix.", type, err);
5157 static int ext4_quota_off(struct super_block *sb, int type)
5159 struct inode *inode = sb_dqopt(sb)->files[type];
5162 /* Force all delayed allocation blocks to be allocated.
5163 * Caller already holds s_umount sem */
5164 if (test_opt(sb, DELALLOC))
5165 sync_filesystem(sb);
5170 /* Update modification times of quota files when userspace can
5171 * start looking at them */
5172 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5175 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5176 ext4_mark_inode_dirty(handle, inode);
5177 ext4_journal_stop(handle);
5180 return dquot_quota_off(sb, type);
5183 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5184 * acquiring the locks... As quota files are never truncated and quota code
5185 * itself serializes the operations (and no one else should touch the files)
5186 * we don't have to be afraid of races */
5187 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5188 size_t len, loff_t off)
5190 struct inode *inode = sb_dqopt(sb)->files[type];
5191 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5192 int offset = off & (sb->s_blocksize - 1);
5195 struct buffer_head *bh;
5196 loff_t i_size = i_size_read(inode);
5200 if (off+len > i_size)
5203 while (toread > 0) {
5204 tocopy = sb->s_blocksize - offset < toread ?
5205 sb->s_blocksize - offset : toread;
5206 bh = ext4_bread(NULL, inode, blk, 0);
5209 if (!bh) /* A hole? */
5210 memset(data, 0, tocopy);
5212 memcpy(data, bh->b_data+offset, tocopy);
5222 /* Write to quotafile (we know the transaction is already started and has
5223 * enough credits) */
5224 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5225 const char *data, size_t len, loff_t off)
5227 struct inode *inode = sb_dqopt(sb)->files[type];
5228 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5229 int err, offset = off & (sb->s_blocksize - 1);
5231 struct buffer_head *bh;
5232 handle_t *handle = journal_current_handle();
5234 if (EXT4_SB(sb)->s_journal && !handle) {
5235 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5236 " cancelled because transaction is not started",
5237 (unsigned long long)off, (unsigned long long)len);
5241 * Since we account only one data block in transaction credits,
5242 * then it is impossible to cross a block boundary.
5244 if (sb->s_blocksize - offset < len) {
5245 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5246 " cancelled because not block aligned",
5247 (unsigned long long)off, (unsigned long long)len);
5252 bh = ext4_bread(handle, inode, blk,
5253 EXT4_GET_BLOCKS_CREATE |
5254 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5255 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5256 ext4_should_retry_alloc(inode->i_sb, &retries));
5261 BUFFER_TRACE(bh, "get write access");
5262 err = ext4_journal_get_write_access(handle, bh);
5268 memcpy(bh->b_data+offset, data, len);
5269 flush_dcache_page(bh->b_page);
5271 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5274 if (inode->i_size < off + len) {
5275 i_size_write(inode, off + len);
5276 EXT4_I(inode)->i_disksize = inode->i_size;
5277 ext4_mark_inode_dirty(handle, inode);
5282 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5284 const struct quota_format_ops *ops;
5286 if (!sb_has_quota_loaded(sb, qid->type))
5288 ops = sb_dqopt(sb)->ops[qid->type];
5289 if (!ops || !ops->get_next_id)
5291 return dquot_get_next_id(sb, qid);
5295 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5296 const char *dev_name, void *data)
5298 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5301 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5302 static inline void register_as_ext2(void)
5304 int err = register_filesystem(&ext2_fs_type);
5307 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5310 static inline void unregister_as_ext2(void)
5312 unregister_filesystem(&ext2_fs_type);
5315 static inline int ext2_feature_set_ok(struct super_block *sb)
5317 if (ext4_has_unknown_ext2_incompat_features(sb))
5319 if (sb->s_flags & MS_RDONLY)
5321 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5326 static inline void register_as_ext2(void) { }
5327 static inline void unregister_as_ext2(void) { }
5328 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5331 static inline void register_as_ext3(void)
5333 int err = register_filesystem(&ext3_fs_type);
5336 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5339 static inline void unregister_as_ext3(void)
5341 unregister_filesystem(&ext3_fs_type);
5344 static inline int ext3_feature_set_ok(struct super_block *sb)
5346 if (ext4_has_unknown_ext3_incompat_features(sb))
5348 if (!ext4_has_feature_journal(sb))
5350 if (sb->s_flags & MS_RDONLY)
5352 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5357 static struct file_system_type ext4_fs_type = {
5358 .owner = THIS_MODULE,
5360 .mount = ext4_mount,
5361 .kill_sb = kill_block_super,
5362 .fs_flags = FS_REQUIRES_DEV,
5364 MODULE_ALIAS_FS("ext4");
5366 /* Shared across all ext4 file systems */
5367 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5369 static int __init ext4_init_fs(void)
5373 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5374 ext4_li_info = NULL;
5375 mutex_init(&ext4_li_mtx);
5377 /* Build-time check for flags consistency */
5378 ext4_check_flag_values();
5380 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5381 init_waitqueue_head(&ext4__ioend_wq[i]);
5383 err = ext4_init_es();
5387 err = ext4_init_pageio();
5391 err = ext4_init_system_zone();
5395 err = ext4_init_sysfs();
5399 err = ext4_init_mballoc();
5402 err = init_inodecache();
5407 err = register_filesystem(&ext4_fs_type);
5413 unregister_as_ext2();
5414 unregister_as_ext3();
5415 destroy_inodecache();
5417 ext4_exit_mballoc();
5421 ext4_exit_system_zone();
5430 static void __exit ext4_exit_fs(void)
5433 ext4_destroy_lazyinit_thread();
5434 unregister_as_ext2();
5435 unregister_as_ext3();
5436 unregister_filesystem(&ext4_fs_type);
5437 destroy_inodecache();
5438 ext4_exit_mballoc();
5440 ext4_exit_system_zone();
5445 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5446 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5447 MODULE_LICENSE("GPL");
5448 module_init(ext4_init_fs)
5449 module_exit(ext4_exit_fs)