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/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ratelimit_state ext4_mount_msg_ratelimit;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
77 static inline int ext2_feature_set_ok(struct super_block *sb);
78 static inline int ext3_feature_set_ok(struct super_block *sb);
79 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
80 static void ext4_destroy_lazyinit_thread(void);
81 static void ext4_unregister_li_request(struct super_block *sb);
82 static void ext4_clear_request_list(void);
83 static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
146 if (!ext4_has_feature_metadata_csum(sb))
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
152 static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
161 return cpu_to_le32(csum);
164 static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
167 if (!ext4_has_metadata_csum(sb))
170 return es->s_checksum == ext4_superblock_csum(sb, es);
173 void ext4_superblock_csum_set(struct super_block *sb)
175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
177 if (!ext4_has_metadata_csum(sb))
180 es->s_checksum = ext4_superblock_csum(sb, es);
183 void *ext4_kvmalloc(size_t size, gfp_t flags)
187 ret = kmalloc(size, flags | __GFP_NOWARN);
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
193 void *ext4_kvzalloc(size_t size, gfp_t flags)
197 ret = kzalloc(size, flags | __GFP_NOWARN);
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
203 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
211 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
219 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le32_to_cpu(bg->bg_inode_table_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
227 __u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
235 __u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
243 __u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
251 __u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
259 void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
267 void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
275 void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
283 void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
291 void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
299 void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
307 void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
316 static void __save_error_info(struct super_block *sb, const char *func,
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
322 if (bdev_read_only(sb->s_bdev))
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
337 * Start the daily error reporting function if it hasn't been
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
342 le32_add_cpu(&es->s_error_count, 1);
345 static void save_error_info(struct super_block *sb, const char *func,
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
360 static int block_device_ejected(struct super_block *sb)
362 struct inode *bd_inode = sb->s_bdev->bd_inode;
363 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
365 return bdi->dev == NULL;
368 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
370 struct super_block *sb = journal->j_private;
371 struct ext4_sb_info *sbi = EXT4_SB(sb);
372 int error = is_journal_aborted(journal);
373 struct ext4_journal_cb_entry *jce;
375 BUG_ON(txn->t_state == T_FINISHED);
377 ext4_process_freed_data(sb, txn->t_tid);
379 spin_lock(&sbi->s_md_lock);
380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
388 spin_unlock(&sbi->s_md_lock);
391 /* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
394 * On ext2, we can store the error state of the filesystem in the
395 * superblock. That is not possible on ext4, because we may have other
396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
401 * We'll just use the jbd2_journal_abort() error code to record an error in
402 * the journal instead. On recovery, the journal will complain about
403 * that error until we've noted it down and cleared it.
406 static void ext4_handle_error(struct super_block *sb)
411 if (!test_opt(sb, ERRORS_CONT)) {
412 journal_t *journal = EXT4_SB(sb)->s_journal;
414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
416 jbd2_journal_abort(journal, -EIO);
418 if (test_opt(sb, ERRORS_RO)) {
419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
425 sb->s_flags |= MS_RDONLY;
427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
431 panic("EXT4-fs (device %s): panic forced after error\n",
436 #define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
440 void __ext4_error(struct super_block *sb, const char *function,
441 unsigned int line, const char *fmt, ...)
443 struct va_format vaf;
446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
449 if (ext4_error_ratelimit(sb)) {
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb->s_id, function, line, current->comm, &vaf);
458 save_error_info(sb, function, line);
459 ext4_handle_error(sb);
462 void __ext4_error_inode(struct inode *inode, const char *function,
463 unsigned int line, ext4_fsblk_t block,
464 const char *fmt, ...)
467 struct va_format vaf;
468 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
473 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
474 es->s_last_error_block = cpu_to_le64(block);
475 if (ext4_error_ratelimit(inode->i_sb)) {
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, &vaf);
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, &vaf);
491 save_error_info(inode->i_sb, function, line);
492 ext4_handle_error(inode->i_sb);
495 void __ext4_error_file(struct file *file, const char *function,
496 unsigned int line, ext4_fsblk_t block,
497 const char *fmt, ...)
500 struct va_format vaf;
501 struct ext4_super_block *es;
502 struct inode *inode = file_inode(file);
503 char pathname[80], *path;
505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508 es = EXT4_SB(inode->i_sb)->s_es;
509 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
510 if (ext4_error_ratelimit(inode->i_sb)) {
511 path = file_path(file, pathname, sizeof(pathname));
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, path, &vaf);
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 current->comm, path, &vaf);
531 save_error_info(inode->i_sb, function, line);
532 ext4_handle_error(inode->i_sb);
535 const char *ext4_decode_error(struct super_block *sb, int errno,
542 errstr = "Corrupt filesystem";
545 errstr = "Filesystem failed CRC";
548 errstr = "IO failure";
551 errstr = "Out of memory";
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
558 errstr = "Readonly filesystem";
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
578 void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
590 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
593 if (ext4_error_ratelimit(sb)) {
594 errstr = ext4_decode_error(sb, errno, nbuf);
595 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
596 sb->s_id, function, line, errstr);
599 save_error_info(sb, function, line);
600 ext4_handle_error(sb);
604 * ext4_abort is a much stronger failure handler than ext4_error. The
605 * abort function may be used to deal with unrecoverable failures such
606 * as journal IO errors or ENOMEM at a critical moment in log management.
608 * We unconditionally force the filesystem into an ABORT|READONLY state,
609 * unless the error response on the fs has been set to panic in which
610 * case we take the easy way out and panic immediately.
613 void __ext4_abort(struct super_block *sb, const char *function,
614 unsigned int line, const char *fmt, ...)
616 struct va_format vaf;
619 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
622 save_error_info(sb, function, line);
626 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
627 sb->s_id, function, line, &vaf);
630 if (sb_rdonly(sb) == 0) {
631 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
632 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
634 * Make sure updated value of ->s_mount_flags will be visible
635 * before ->s_flags update
638 sb->s_flags |= MS_RDONLY;
639 if (EXT4_SB(sb)->s_journal)
640 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
641 save_error_info(sb, function, line);
643 if (test_opt(sb, ERRORS_PANIC)) {
644 if (EXT4_SB(sb)->s_journal &&
645 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
647 panic("EXT4-fs panic from previous error\n");
651 void __ext4_msg(struct super_block *sb,
652 const char *prefix, const char *fmt, ...)
654 struct va_format vaf;
657 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
663 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
667 #define ext4_warning_ratelimit(sb) \
668 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
671 void __ext4_warning(struct super_block *sb, const char *function,
672 unsigned int line, const char *fmt, ...)
674 struct va_format vaf;
677 if (!ext4_warning_ratelimit(sb))
683 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
684 sb->s_id, function, line, &vaf);
688 void __ext4_warning_inode(const struct inode *inode, const char *function,
689 unsigned int line, const char *fmt, ...)
691 struct va_format vaf;
694 if (!ext4_warning_ratelimit(inode->i_sb))
700 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
701 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
702 function, line, inode->i_ino, current->comm, &vaf);
706 void __ext4_grp_locked_error(const char *function, unsigned int line,
707 struct super_block *sb, ext4_group_t grp,
708 unsigned long ino, ext4_fsblk_t block,
709 const char *fmt, ...)
713 struct va_format vaf;
715 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
717 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
720 es->s_last_error_ino = cpu_to_le32(ino);
721 es->s_last_error_block = cpu_to_le64(block);
722 __save_error_info(sb, function, line);
724 if (ext4_error_ratelimit(sb)) {
728 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
729 sb->s_id, function, line, grp);
731 printk(KERN_CONT "inode %lu: ", ino);
733 printk(KERN_CONT "block %llu:",
734 (unsigned long long) block);
735 printk(KERN_CONT "%pV\n", &vaf);
739 if (test_opt(sb, ERRORS_CONT)) {
740 ext4_commit_super(sb, 0);
744 ext4_unlock_group(sb, grp);
745 ext4_handle_error(sb);
747 * We only get here in the ERRORS_RO case; relocking the group
748 * may be dangerous, but nothing bad will happen since the
749 * filesystem will have already been marked read/only and the
750 * journal has been aborted. We return 1 as a hint to callers
751 * who might what to use the return value from
752 * ext4_grp_locked_error() to distinguish between the
753 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
754 * aggressively from the ext4 function in question, with a
755 * more appropriate error code.
757 ext4_lock_group(sb, grp);
761 void ext4_update_dynamic_rev(struct super_block *sb)
763 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
765 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
769 "updating to rev %d because of new feature flag, "
770 "running e2fsck is recommended",
773 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
774 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
775 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
776 /* leave es->s_feature_*compat flags alone */
777 /* es->s_uuid will be set by e2fsck if empty */
780 * The rest of the superblock fields should be zero, and if not it
781 * means they are likely already in use, so leave them alone. We
782 * can leave it up to e2fsck to clean up any inconsistencies there.
787 * Open the external journal device
789 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
791 struct block_device *bdev;
792 char b[BDEVNAME_SIZE];
794 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
800 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
801 __bdevname(dev, b), PTR_ERR(bdev));
806 * Release the journal device
808 static void ext4_blkdev_put(struct block_device *bdev)
810 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
813 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
815 struct block_device *bdev;
816 bdev = sbi->journal_bdev;
818 ext4_blkdev_put(bdev);
819 sbi->journal_bdev = NULL;
823 static inline struct inode *orphan_list_entry(struct list_head *l)
825 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
828 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
832 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
833 le32_to_cpu(sbi->s_es->s_last_orphan));
835 printk(KERN_ERR "sb_info orphan list:\n");
836 list_for_each(l, &sbi->s_orphan) {
837 struct inode *inode = orphan_list_entry(l);
839 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
840 inode->i_sb->s_id, inode->i_ino, inode,
841 inode->i_mode, inode->i_nlink,
847 static int ext4_quota_off(struct super_block *sb, int type);
849 static inline void ext4_quota_off_umount(struct super_block *sb)
853 /* Use our quota_off function to clear inode flags etc. */
854 for (type = 0; type < EXT4_MAXQUOTAS; type++)
855 ext4_quota_off(sb, type);
858 static inline void ext4_quota_off_umount(struct super_block *sb)
863 static void ext4_put_super(struct super_block *sb)
865 struct ext4_sb_info *sbi = EXT4_SB(sb);
866 struct ext4_super_block *es = sbi->s_es;
870 ext4_unregister_li_request(sb);
871 ext4_quota_off_umount(sb);
873 flush_workqueue(sbi->rsv_conversion_wq);
874 destroy_workqueue(sbi->rsv_conversion_wq);
876 if (sbi->s_journal) {
877 aborted = is_journal_aborted(sbi->s_journal);
878 err = jbd2_journal_destroy(sbi->s_journal);
879 sbi->s_journal = NULL;
880 if ((err < 0) && !aborted)
881 ext4_abort(sb, "Couldn't clean up the journal");
884 ext4_unregister_sysfs(sb);
885 ext4_es_unregister_shrinker(sbi);
886 del_timer_sync(&sbi->s_err_report);
887 ext4_release_system_zone(sb);
889 ext4_ext_release(sb);
891 if (!sb_rdonly(sb) && !aborted) {
892 ext4_clear_feature_journal_needs_recovery(sb);
893 es->s_state = cpu_to_le16(sbi->s_mount_state);
896 ext4_commit_super(sb, 1);
898 for (i = 0; i < sbi->s_gdb_count; i++)
899 brelse(sbi->s_group_desc[i]);
900 kvfree(sbi->s_group_desc);
901 kvfree(sbi->s_flex_groups);
902 percpu_counter_destroy(&sbi->s_freeclusters_counter);
903 percpu_counter_destroy(&sbi->s_freeinodes_counter);
904 percpu_counter_destroy(&sbi->s_dirs_counter);
905 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
906 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
908 for (i = 0; i < EXT4_MAXQUOTAS; i++)
909 kfree(sbi->s_qf_names[i]);
912 /* Debugging code just in case the in-memory inode orphan list
913 * isn't empty. The on-disk one can be non-empty if we've
914 * detected an error and taken the fs readonly, but the
915 * in-memory list had better be clean by this point. */
916 if (!list_empty(&sbi->s_orphan))
917 dump_orphan_list(sb, sbi);
918 J_ASSERT(list_empty(&sbi->s_orphan));
920 sync_blockdev(sb->s_bdev);
921 invalidate_bdev(sb->s_bdev);
922 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
924 * Invalidate the journal device's buffers. We don't want them
925 * floating about in memory - the physical journal device may
926 * hotswapped, and it breaks the `ro-after' testing code.
928 sync_blockdev(sbi->journal_bdev);
929 invalidate_bdev(sbi->journal_bdev);
930 ext4_blkdev_remove(sbi);
932 if (sbi->s_ea_inode_cache) {
933 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
934 sbi->s_ea_inode_cache = NULL;
936 if (sbi->s_ea_block_cache) {
937 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
938 sbi->s_ea_block_cache = NULL;
941 kthread_stop(sbi->s_mmp_tsk);
943 sb->s_fs_info = NULL;
945 * Now that we are completely done shutting down the
946 * superblock, we need to actually destroy the kobject.
948 kobject_put(&sbi->s_kobj);
949 wait_for_completion(&sbi->s_kobj_unregister);
950 if (sbi->s_chksum_driver)
951 crypto_free_shash(sbi->s_chksum_driver);
952 kfree(sbi->s_blockgroup_lock);
953 fs_put_dax(sbi->s_daxdev);
957 static struct kmem_cache *ext4_inode_cachep;
960 * Called inside transaction, so use GFP_NOFS
962 static struct inode *ext4_alloc_inode(struct super_block *sb)
964 struct ext4_inode_info *ei;
966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
970 ei->vfs_inode.i_version = 1;
971 spin_lock_init(&ei->i_raw_lock);
972 INIT_LIST_HEAD(&ei->i_prealloc_list);
973 spin_lock_init(&ei->i_prealloc_lock);
974 ext4_es_init_tree(&ei->i_es_tree);
975 rwlock_init(&ei->i_es_lock);
976 INIT_LIST_HEAD(&ei->i_es_list);
979 ei->i_es_shrink_lblk = 0;
980 ei->i_reserved_data_blocks = 0;
981 ei->i_da_metadata_calc_len = 0;
982 ei->i_da_metadata_calc_last_lblock = 0;
983 spin_lock_init(&(ei->i_block_reservation_lock));
985 ei->i_reserved_quota = 0;
986 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
989 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
990 spin_lock_init(&ei->i_completed_io_lock);
992 ei->i_datasync_tid = 0;
993 atomic_set(&ei->i_unwritten, 0);
994 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
995 return &ei->vfs_inode;
998 static int ext4_drop_inode(struct inode *inode)
1000 int drop = generic_drop_inode(inode);
1002 trace_ext4_drop_inode(inode, drop);
1006 static void ext4_i_callback(struct rcu_head *head)
1008 struct inode *inode = container_of(head, struct inode, i_rcu);
1009 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1012 static void ext4_destroy_inode(struct inode *inode)
1014 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1015 ext4_msg(inode->i_sb, KERN_ERR,
1016 "Inode %lu (%p): orphan list check failed!",
1017 inode->i_ino, EXT4_I(inode));
1018 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1019 EXT4_I(inode), sizeof(struct ext4_inode_info),
1023 call_rcu(&inode->i_rcu, ext4_i_callback);
1026 static void init_once(void *foo)
1028 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1030 INIT_LIST_HEAD(&ei->i_orphan);
1031 init_rwsem(&ei->xattr_sem);
1032 init_rwsem(&ei->i_data_sem);
1033 init_rwsem(&ei->i_mmap_sem);
1034 inode_init_once(&ei->vfs_inode);
1037 static int __init init_inodecache(void)
1039 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1040 sizeof(struct ext4_inode_info),
1041 0, (SLAB_RECLAIM_ACCOUNT|
1042 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1044 if (ext4_inode_cachep == NULL)
1049 static void destroy_inodecache(void)
1052 * Make sure all delayed rcu free inodes are flushed before we
1056 kmem_cache_destroy(ext4_inode_cachep);
1059 void ext4_clear_inode(struct inode *inode)
1061 invalidate_inode_buffers(inode);
1064 ext4_discard_preallocations(inode);
1065 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1066 if (EXT4_I(inode)->jinode) {
1067 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1068 EXT4_I(inode)->jinode);
1069 jbd2_free_inode(EXT4_I(inode)->jinode);
1070 EXT4_I(inode)->jinode = NULL;
1072 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1073 fscrypt_put_encryption_info(inode, NULL);
1077 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1078 u64 ino, u32 generation)
1080 struct inode *inode;
1082 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1083 return ERR_PTR(-ESTALE);
1084 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1085 return ERR_PTR(-ESTALE);
1087 /* iget isn't really right if the inode is currently unallocated!!
1089 * ext4_read_inode will return a bad_inode if the inode had been
1090 * deleted, so we should be safe.
1092 * Currently we don't know the generation for parent directory, so
1093 * a generation of 0 means "accept any"
1095 inode = ext4_iget_normal(sb, ino);
1097 return ERR_CAST(inode);
1098 if (generation && inode->i_generation != generation) {
1100 return ERR_PTR(-ESTALE);
1106 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1107 int fh_len, int fh_type)
1109 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1110 ext4_nfs_get_inode);
1113 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1114 int fh_len, int fh_type)
1116 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1117 ext4_nfs_get_inode);
1121 * Try to release metadata pages (indirect blocks, directories) which are
1122 * mapped via the block device. Since these pages could have journal heads
1123 * which would prevent try_to_free_buffers() from freeing them, we must use
1124 * jbd2 layer's try_to_free_buffers() function to release them.
1126 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1129 journal_t *journal = EXT4_SB(sb)->s_journal;
1131 WARN_ON(PageChecked(page));
1132 if (!page_has_buffers(page))
1135 return jbd2_journal_try_to_free_buffers(journal, page,
1136 wait & ~__GFP_DIRECT_RECLAIM);
1137 return try_to_free_buffers(page);
1140 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1141 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1143 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1144 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1147 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1150 handle_t *handle = fs_data;
1151 int res, res2, credits, retries = 0;
1154 * Encrypting the root directory is not allowed because e2fsck expects
1155 * lost+found to exist and be unencrypted, and encrypting the root
1156 * directory would imply encrypting the lost+found directory as well as
1157 * the filename "lost+found" itself.
1159 if (inode->i_ino == EXT4_ROOT_INO)
1162 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1165 res = ext4_convert_inline_data(inode);
1170 * If a journal handle was specified, then the encryption context is
1171 * being set on a new inode via inheritance and is part of a larger
1172 * transaction to create the inode. Otherwise the encryption context is
1173 * being set on an existing inode in its own transaction. Only in the
1174 * latter case should the "retry on ENOSPC" logic be used.
1178 res = ext4_xattr_set_handle(handle, inode,
1179 EXT4_XATTR_INDEX_ENCRYPTION,
1180 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1183 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1184 ext4_clear_inode_state(inode,
1185 EXT4_STATE_MAY_INLINE_DATA);
1187 * Update inode->i_flags - e.g. S_DAX may get disabled
1189 ext4_set_inode_flags(inode);
1194 res = dquot_initialize(inode);
1198 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1203 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1205 return PTR_ERR(handle);
1207 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1208 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1211 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1212 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1213 ext4_set_inode_flags(inode);
1214 res = ext4_mark_inode_dirty(handle, inode);
1216 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1218 res2 = ext4_journal_stop(handle);
1220 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1227 static bool ext4_dummy_context(struct inode *inode)
1229 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1232 static unsigned ext4_max_namelen(struct inode *inode)
1234 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1238 static const struct fscrypt_operations ext4_cryptops = {
1239 .key_prefix = "ext4:",
1240 .get_context = ext4_get_context,
1241 .set_context = ext4_set_context,
1242 .dummy_context = ext4_dummy_context,
1243 .is_encrypted = ext4_encrypted_inode,
1244 .empty_dir = ext4_empty_dir,
1245 .max_namelen = ext4_max_namelen,
1248 static const struct fscrypt_operations ext4_cryptops = {
1249 .is_encrypted = ext4_encrypted_inode,
1254 static const char * const quotatypes[] = INITQFNAMES;
1255 #define QTYPE2NAME(t) (quotatypes[t])
1257 static int ext4_write_dquot(struct dquot *dquot);
1258 static int ext4_acquire_dquot(struct dquot *dquot);
1259 static int ext4_release_dquot(struct dquot *dquot);
1260 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1261 static int ext4_write_info(struct super_block *sb, int type);
1262 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1263 const struct path *path);
1264 static int ext4_quota_on_mount(struct super_block *sb, int type);
1265 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1266 size_t len, loff_t off);
1267 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1268 const char *data, size_t len, loff_t off);
1269 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1270 unsigned int flags);
1271 static int ext4_enable_quotas(struct super_block *sb);
1272 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1274 static struct dquot **ext4_get_dquots(struct inode *inode)
1276 return EXT4_I(inode)->i_dquot;
1279 static const struct dquot_operations ext4_quota_operations = {
1280 .get_reserved_space = ext4_get_reserved_space,
1281 .write_dquot = ext4_write_dquot,
1282 .acquire_dquot = ext4_acquire_dquot,
1283 .release_dquot = ext4_release_dquot,
1284 .mark_dirty = ext4_mark_dquot_dirty,
1285 .write_info = ext4_write_info,
1286 .alloc_dquot = dquot_alloc,
1287 .destroy_dquot = dquot_destroy,
1288 .get_projid = ext4_get_projid,
1289 .get_inode_usage = ext4_get_inode_usage,
1290 .get_next_id = ext4_get_next_id,
1293 static const struct quotactl_ops ext4_qctl_operations = {
1294 .quota_on = ext4_quota_on,
1295 .quota_off = ext4_quota_off,
1296 .quota_sync = dquot_quota_sync,
1297 .get_state = dquot_get_state,
1298 .set_info = dquot_set_dqinfo,
1299 .get_dqblk = dquot_get_dqblk,
1300 .set_dqblk = dquot_set_dqblk,
1301 .get_nextdqblk = dquot_get_next_dqblk,
1305 static const struct super_operations ext4_sops = {
1306 .alloc_inode = ext4_alloc_inode,
1307 .destroy_inode = ext4_destroy_inode,
1308 .write_inode = ext4_write_inode,
1309 .dirty_inode = ext4_dirty_inode,
1310 .drop_inode = ext4_drop_inode,
1311 .evict_inode = ext4_evict_inode,
1312 .put_super = ext4_put_super,
1313 .sync_fs = ext4_sync_fs,
1314 .freeze_fs = ext4_freeze,
1315 .unfreeze_fs = ext4_unfreeze,
1316 .statfs = ext4_statfs,
1317 .remount_fs = ext4_remount,
1318 .show_options = ext4_show_options,
1320 .quota_read = ext4_quota_read,
1321 .quota_write = ext4_quota_write,
1322 .get_dquots = ext4_get_dquots,
1324 .bdev_try_to_free_page = bdev_try_to_free_page,
1327 static const struct export_operations ext4_export_ops = {
1328 .fh_to_dentry = ext4_fh_to_dentry,
1329 .fh_to_parent = ext4_fh_to_parent,
1330 .get_parent = ext4_get_parent,
1334 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1335 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1336 Opt_nouid32, Opt_debug, Opt_removed,
1337 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1338 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1339 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1340 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1341 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1342 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1343 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1344 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1345 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1346 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1347 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1348 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1349 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1350 Opt_inode_readahead_blks, Opt_journal_ioprio,
1351 Opt_dioread_nolock, Opt_dioread_lock,
1352 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1353 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1356 static const match_table_t tokens = {
1357 {Opt_bsd_df, "bsddf"},
1358 {Opt_minix_df, "minixdf"},
1359 {Opt_grpid, "grpid"},
1360 {Opt_grpid, "bsdgroups"},
1361 {Opt_nogrpid, "nogrpid"},
1362 {Opt_nogrpid, "sysvgroups"},
1363 {Opt_resgid, "resgid=%u"},
1364 {Opt_resuid, "resuid=%u"},
1366 {Opt_err_cont, "errors=continue"},
1367 {Opt_err_panic, "errors=panic"},
1368 {Opt_err_ro, "errors=remount-ro"},
1369 {Opt_nouid32, "nouid32"},
1370 {Opt_debug, "debug"},
1371 {Opt_removed, "oldalloc"},
1372 {Opt_removed, "orlov"},
1373 {Opt_user_xattr, "user_xattr"},
1374 {Opt_nouser_xattr, "nouser_xattr"},
1376 {Opt_noacl, "noacl"},
1377 {Opt_noload, "norecovery"},
1378 {Opt_noload, "noload"},
1379 {Opt_removed, "nobh"},
1380 {Opt_removed, "bh"},
1381 {Opt_commit, "commit=%u"},
1382 {Opt_min_batch_time, "min_batch_time=%u"},
1383 {Opt_max_batch_time, "max_batch_time=%u"},
1384 {Opt_journal_dev, "journal_dev=%u"},
1385 {Opt_journal_path, "journal_path=%s"},
1386 {Opt_journal_checksum, "journal_checksum"},
1387 {Opt_nojournal_checksum, "nojournal_checksum"},
1388 {Opt_journal_async_commit, "journal_async_commit"},
1389 {Opt_abort, "abort"},
1390 {Opt_data_journal, "data=journal"},
1391 {Opt_data_ordered, "data=ordered"},
1392 {Opt_data_writeback, "data=writeback"},
1393 {Opt_data_err_abort, "data_err=abort"},
1394 {Opt_data_err_ignore, "data_err=ignore"},
1395 {Opt_offusrjquota, "usrjquota="},
1396 {Opt_usrjquota, "usrjquota=%s"},
1397 {Opt_offgrpjquota, "grpjquota="},
1398 {Opt_grpjquota, "grpjquota=%s"},
1399 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1400 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1401 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1402 {Opt_grpquota, "grpquota"},
1403 {Opt_noquota, "noquota"},
1404 {Opt_quota, "quota"},
1405 {Opt_usrquota, "usrquota"},
1406 {Opt_prjquota, "prjquota"},
1407 {Opt_barrier, "barrier=%u"},
1408 {Opt_barrier, "barrier"},
1409 {Opt_nobarrier, "nobarrier"},
1410 {Opt_i_version, "i_version"},
1412 {Opt_stripe, "stripe=%u"},
1413 {Opt_delalloc, "delalloc"},
1414 {Opt_lazytime, "lazytime"},
1415 {Opt_nolazytime, "nolazytime"},
1416 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1417 {Opt_nodelalloc, "nodelalloc"},
1418 {Opt_removed, "mblk_io_submit"},
1419 {Opt_removed, "nomblk_io_submit"},
1420 {Opt_block_validity, "block_validity"},
1421 {Opt_noblock_validity, "noblock_validity"},
1422 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1423 {Opt_journal_ioprio, "journal_ioprio=%u"},
1424 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1425 {Opt_auto_da_alloc, "auto_da_alloc"},
1426 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1427 {Opt_dioread_nolock, "dioread_nolock"},
1428 {Opt_dioread_lock, "dioread_lock"},
1429 {Opt_discard, "discard"},
1430 {Opt_nodiscard, "nodiscard"},
1431 {Opt_init_itable, "init_itable=%u"},
1432 {Opt_init_itable, "init_itable"},
1433 {Opt_noinit_itable, "noinit_itable"},
1434 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1435 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1436 {Opt_nombcache, "nombcache"},
1437 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1438 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1439 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1440 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1441 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1442 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1446 static ext4_fsblk_t get_sb_block(void **data)
1448 ext4_fsblk_t sb_block;
1449 char *options = (char *) *data;
1451 if (!options || strncmp(options, "sb=", 3) != 0)
1452 return 1; /* Default location */
1455 /* TODO: use simple_strtoll with >32bit ext4 */
1456 sb_block = simple_strtoul(options, &options, 0);
1457 if (*options && *options != ',') {
1458 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1462 if (*options == ',')
1464 *data = (void *) options;
1469 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1470 static const char deprecated_msg[] =
1471 "Mount option \"%s\" will be removed by %s\n"
1472 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1475 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1477 struct ext4_sb_info *sbi = EXT4_SB(sb);
1481 if (sb_any_quota_loaded(sb) &&
1482 !sbi->s_qf_names[qtype]) {
1483 ext4_msg(sb, KERN_ERR,
1484 "Cannot change journaled "
1485 "quota options when quota turned on");
1488 if (ext4_has_feature_quota(sb)) {
1489 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1490 "ignored when QUOTA feature is enabled");
1493 qname = match_strdup(args);
1495 ext4_msg(sb, KERN_ERR,
1496 "Not enough memory for storing quotafile name");
1499 if (sbi->s_qf_names[qtype]) {
1500 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1503 ext4_msg(sb, KERN_ERR,
1504 "%s quota file already specified",
1508 if (strchr(qname, '/')) {
1509 ext4_msg(sb, KERN_ERR,
1510 "quotafile must be on filesystem root");
1513 sbi->s_qf_names[qtype] = qname;
1521 static int clear_qf_name(struct super_block *sb, int qtype)
1524 struct ext4_sb_info *sbi = EXT4_SB(sb);
1526 if (sb_any_quota_loaded(sb) &&
1527 sbi->s_qf_names[qtype]) {
1528 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1529 " when quota turned on");
1532 kfree(sbi->s_qf_names[qtype]);
1533 sbi->s_qf_names[qtype] = NULL;
1538 #define MOPT_SET 0x0001
1539 #define MOPT_CLEAR 0x0002
1540 #define MOPT_NOSUPPORT 0x0004
1541 #define MOPT_EXPLICIT 0x0008
1542 #define MOPT_CLEAR_ERR 0x0010
1543 #define MOPT_GTE0 0x0020
1546 #define MOPT_QFMT 0x0040
1548 #define MOPT_Q MOPT_NOSUPPORT
1549 #define MOPT_QFMT MOPT_NOSUPPORT
1551 #define MOPT_DATAJ 0x0080
1552 #define MOPT_NO_EXT2 0x0100
1553 #define MOPT_NO_EXT3 0x0200
1554 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1555 #define MOPT_STRING 0x0400
1557 static const struct mount_opts {
1561 } ext4_mount_opts[] = {
1562 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1563 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1564 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1565 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1566 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1567 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1568 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1569 MOPT_EXT4_ONLY | MOPT_SET},
1570 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1571 MOPT_EXT4_ONLY | MOPT_CLEAR},
1572 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1573 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1574 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1575 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1576 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1577 MOPT_EXT4_ONLY | MOPT_CLEAR},
1578 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1579 MOPT_EXT4_ONLY | MOPT_CLEAR},
1580 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1581 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1582 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1583 EXT4_MOUNT_JOURNAL_CHECKSUM),
1584 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1585 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1586 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1587 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1588 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1589 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1591 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1593 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1594 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1595 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1596 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1597 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1598 {Opt_commit, 0, MOPT_GTE0},
1599 {Opt_max_batch_time, 0, MOPT_GTE0},
1600 {Opt_min_batch_time, 0, MOPT_GTE0},
1601 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1602 {Opt_init_itable, 0, MOPT_GTE0},
1603 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1604 {Opt_stripe, 0, MOPT_GTE0},
1605 {Opt_resuid, 0, MOPT_GTE0},
1606 {Opt_resgid, 0, MOPT_GTE0},
1607 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1608 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1609 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1610 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1611 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1612 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1613 MOPT_NO_EXT2 | MOPT_DATAJ},
1614 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1615 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1616 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1617 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1618 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1620 {Opt_acl, 0, MOPT_NOSUPPORT},
1621 {Opt_noacl, 0, MOPT_NOSUPPORT},
1623 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1624 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1625 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1626 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1627 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1629 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1631 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1633 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1634 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1635 MOPT_CLEAR | MOPT_Q},
1636 {Opt_usrjquota, 0, MOPT_Q},
1637 {Opt_grpjquota, 0, MOPT_Q},
1638 {Opt_offusrjquota, 0, MOPT_Q},
1639 {Opt_offgrpjquota, 0, MOPT_Q},
1640 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1641 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1642 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1643 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1644 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1645 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1649 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1650 substring_t *args, unsigned long *journal_devnum,
1651 unsigned int *journal_ioprio, int is_remount)
1653 struct ext4_sb_info *sbi = EXT4_SB(sb);
1654 const struct mount_opts *m;
1660 if (token == Opt_usrjquota)
1661 return set_qf_name(sb, USRQUOTA, &args[0]);
1662 else if (token == Opt_grpjquota)
1663 return set_qf_name(sb, GRPQUOTA, &args[0]);
1664 else if (token == Opt_offusrjquota)
1665 return clear_qf_name(sb, USRQUOTA);
1666 else if (token == Opt_offgrpjquota)
1667 return clear_qf_name(sb, GRPQUOTA);
1671 case Opt_nouser_xattr:
1672 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1675 return 1; /* handled by get_sb_block() */
1677 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1680 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1683 sb->s_flags |= MS_I_VERSION;
1686 sb->s_flags |= MS_LAZYTIME;
1688 case Opt_nolazytime:
1689 sb->s_flags &= ~MS_LAZYTIME;
1693 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1694 if (token == m->token)
1697 if (m->token == Opt_err) {
1698 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1699 "or missing value", opt);
1703 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1704 ext4_msg(sb, KERN_ERR,
1705 "Mount option \"%s\" incompatible with ext2", opt);
1708 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1709 ext4_msg(sb, KERN_ERR,
1710 "Mount option \"%s\" incompatible with ext3", opt);
1714 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1716 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1718 if (m->flags & MOPT_EXPLICIT) {
1719 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1720 set_opt2(sb, EXPLICIT_DELALLOC);
1721 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1722 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1726 if (m->flags & MOPT_CLEAR_ERR)
1727 clear_opt(sb, ERRORS_MASK);
1728 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1729 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1730 "options when quota turned on");
1734 if (m->flags & MOPT_NOSUPPORT) {
1735 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1736 } else if (token == Opt_commit) {
1738 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1739 sbi->s_commit_interval = HZ * arg;
1740 } else if (token == Opt_debug_want_extra_isize) {
1741 sbi->s_want_extra_isize = arg;
1742 } else if (token == Opt_max_batch_time) {
1743 sbi->s_max_batch_time = arg;
1744 } else if (token == Opt_min_batch_time) {
1745 sbi->s_min_batch_time = arg;
1746 } else if (token == Opt_inode_readahead_blks) {
1747 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1748 ext4_msg(sb, KERN_ERR,
1749 "EXT4-fs: inode_readahead_blks must be "
1750 "0 or a power of 2 smaller than 2^31");
1753 sbi->s_inode_readahead_blks = arg;
1754 } else if (token == Opt_init_itable) {
1755 set_opt(sb, INIT_INODE_TABLE);
1757 arg = EXT4_DEF_LI_WAIT_MULT;
1758 sbi->s_li_wait_mult = arg;
1759 } else if (token == Opt_max_dir_size_kb) {
1760 sbi->s_max_dir_size_kb = arg;
1761 } else if (token == Opt_stripe) {
1762 sbi->s_stripe = arg;
1763 } else if (token == Opt_resuid) {
1764 uid = make_kuid(current_user_ns(), arg);
1765 if (!uid_valid(uid)) {
1766 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1769 sbi->s_resuid = uid;
1770 } else if (token == Opt_resgid) {
1771 gid = make_kgid(current_user_ns(), arg);
1772 if (!gid_valid(gid)) {
1773 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1776 sbi->s_resgid = gid;
1777 } else if (token == Opt_journal_dev) {
1779 ext4_msg(sb, KERN_ERR,
1780 "Cannot specify journal on remount");
1783 *journal_devnum = arg;
1784 } else if (token == Opt_journal_path) {
1786 struct inode *journal_inode;
1791 ext4_msg(sb, KERN_ERR,
1792 "Cannot specify journal on remount");
1795 journal_path = match_strdup(&args[0]);
1796 if (!journal_path) {
1797 ext4_msg(sb, KERN_ERR, "error: could not dup "
1798 "journal device string");
1802 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1804 ext4_msg(sb, KERN_ERR, "error: could not find "
1805 "journal device path: error %d", error);
1806 kfree(journal_path);
1810 journal_inode = d_inode(path.dentry);
1811 if (!S_ISBLK(journal_inode->i_mode)) {
1812 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1813 "is not a block device", journal_path);
1815 kfree(journal_path);
1819 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1821 kfree(journal_path);
1822 } else if (token == Opt_journal_ioprio) {
1824 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1829 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1830 } else if (token == Opt_test_dummy_encryption) {
1831 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1832 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1833 ext4_msg(sb, KERN_WARNING,
1834 "Test dummy encryption mode enabled");
1836 ext4_msg(sb, KERN_WARNING,
1837 "Test dummy encryption mount option ignored");
1839 } else if (m->flags & MOPT_DATAJ) {
1841 if (!sbi->s_journal)
1842 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1843 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1844 ext4_msg(sb, KERN_ERR,
1845 "Cannot change data mode on remount");
1849 clear_opt(sb, DATA_FLAGS);
1850 sbi->s_mount_opt |= m->mount_opt;
1853 } else if (m->flags & MOPT_QFMT) {
1854 if (sb_any_quota_loaded(sb) &&
1855 sbi->s_jquota_fmt != m->mount_opt) {
1856 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1857 "quota options when quota turned on");
1860 if (ext4_has_feature_quota(sb)) {
1861 ext4_msg(sb, KERN_INFO,
1862 "Quota format mount options ignored "
1863 "when QUOTA feature is enabled");
1866 sbi->s_jquota_fmt = m->mount_opt;
1868 } else if (token == Opt_dax) {
1869 #ifdef CONFIG_FS_DAX
1870 ext4_msg(sb, KERN_WARNING,
1871 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1872 sbi->s_mount_opt |= m->mount_opt;
1874 ext4_msg(sb, KERN_INFO, "dax option not supported");
1877 } else if (token == Opt_data_err_abort) {
1878 sbi->s_mount_opt |= m->mount_opt;
1879 } else if (token == Opt_data_err_ignore) {
1880 sbi->s_mount_opt &= ~m->mount_opt;
1884 if (m->flags & MOPT_CLEAR)
1886 else if (unlikely(!(m->flags & MOPT_SET))) {
1887 ext4_msg(sb, KERN_WARNING,
1888 "buggy handling of option %s", opt);
1893 sbi->s_mount_opt |= m->mount_opt;
1895 sbi->s_mount_opt &= ~m->mount_opt;
1900 static int parse_options(char *options, struct super_block *sb,
1901 unsigned long *journal_devnum,
1902 unsigned int *journal_ioprio,
1905 struct ext4_sb_info *sbi = EXT4_SB(sb);
1907 substring_t args[MAX_OPT_ARGS];
1913 while ((p = strsep(&options, ",")) != NULL) {
1917 * Initialize args struct so we know whether arg was
1918 * found; some options take optional arguments.
1920 args[0].to = args[0].from = NULL;
1921 token = match_token(p, tokens, args);
1922 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1923 journal_ioprio, is_remount) < 0)
1928 * We do the test below only for project quotas. 'usrquota' and
1929 * 'grpquota' mount options are allowed even without quota feature
1930 * to support legacy quotas in quota files.
1932 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1933 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1934 "Cannot enable project quota enforcement.");
1937 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1938 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1939 clear_opt(sb, USRQUOTA);
1941 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1942 clear_opt(sb, GRPQUOTA);
1944 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1945 ext4_msg(sb, KERN_ERR, "old and new quota "
1950 if (!sbi->s_jquota_fmt) {
1951 ext4_msg(sb, KERN_ERR, "journaled quota format "
1957 if (test_opt(sb, DIOREAD_NOLOCK)) {
1959 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1961 if (blocksize < PAGE_SIZE) {
1962 ext4_msg(sb, KERN_ERR, "can't mount with "
1963 "dioread_nolock if block size != PAGE_SIZE");
1970 static inline void ext4_show_quota_options(struct seq_file *seq,
1971 struct super_block *sb)
1973 #if defined(CONFIG_QUOTA)
1974 struct ext4_sb_info *sbi = EXT4_SB(sb);
1976 if (sbi->s_jquota_fmt) {
1979 switch (sbi->s_jquota_fmt) {
1990 seq_printf(seq, ",jqfmt=%s", fmtname);
1993 if (sbi->s_qf_names[USRQUOTA])
1994 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1996 if (sbi->s_qf_names[GRPQUOTA])
1997 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2001 static const char *token2str(int token)
2003 const struct match_token *t;
2005 for (t = tokens; t->token != Opt_err; t++)
2006 if (t->token == token && !strchr(t->pattern, '='))
2013 * - it's set to a non-default value OR
2014 * - if the per-sb default is different from the global default
2016 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2019 struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 struct ext4_super_block *es = sbi->s_es;
2021 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2022 const struct mount_opts *m;
2023 char sep = nodefs ? '\n' : ',';
2025 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2026 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2028 if (sbi->s_sb_block != 1)
2029 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2031 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2032 int want_set = m->flags & MOPT_SET;
2033 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2034 (m->flags & MOPT_CLEAR_ERR))
2036 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2037 continue; /* skip if same as the default */
2039 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2040 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2041 continue; /* select Opt_noFoo vs Opt_Foo */
2042 SEQ_OPTS_PRINT("%s", token2str(m->token));
2045 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2046 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2047 SEQ_OPTS_PRINT("resuid=%u",
2048 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2049 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2050 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2051 SEQ_OPTS_PRINT("resgid=%u",
2052 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2053 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2054 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2055 SEQ_OPTS_PUTS("errors=remount-ro");
2056 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2057 SEQ_OPTS_PUTS("errors=continue");
2058 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2059 SEQ_OPTS_PUTS("errors=panic");
2060 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2061 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2062 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2063 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2064 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2065 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2066 if (sb->s_flags & MS_I_VERSION)
2067 SEQ_OPTS_PUTS("i_version");
2068 if (nodefs || sbi->s_stripe)
2069 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2070 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2071 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2072 SEQ_OPTS_PUTS("data=journal");
2073 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2074 SEQ_OPTS_PUTS("data=ordered");
2075 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2076 SEQ_OPTS_PUTS("data=writeback");
2079 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2080 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2081 sbi->s_inode_readahead_blks);
2083 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2084 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2085 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2086 if (nodefs || sbi->s_max_dir_size_kb)
2087 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2088 if (test_opt(sb, DATA_ERR_ABORT))
2089 SEQ_OPTS_PUTS("data_err=abort");
2091 ext4_show_quota_options(seq, sb);
2095 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2097 return _ext4_show_options(seq, root->d_sb, 0);
2100 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2102 struct super_block *sb = seq->private;
2105 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2106 rc = _ext4_show_options(seq, sb, 1);
2107 seq_puts(seq, "\n");
2111 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2114 struct ext4_sb_info *sbi = EXT4_SB(sb);
2117 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2118 ext4_msg(sb, KERN_ERR, "revision level too high, "
2119 "forcing read-only mode");
2124 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2125 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2126 "running e2fsck is recommended");
2127 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2128 ext4_msg(sb, KERN_WARNING,
2129 "warning: mounting fs with errors, "
2130 "running e2fsck is recommended");
2131 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2132 le16_to_cpu(es->s_mnt_count) >=
2133 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2134 ext4_msg(sb, KERN_WARNING,
2135 "warning: maximal mount count reached, "
2136 "running e2fsck is recommended");
2137 else if (le32_to_cpu(es->s_checkinterval) &&
2138 (le32_to_cpu(es->s_lastcheck) +
2139 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2140 ext4_msg(sb, KERN_WARNING,
2141 "warning: checktime reached, "
2142 "running e2fsck is recommended");
2143 if (!sbi->s_journal)
2144 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2145 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2146 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2147 le16_add_cpu(&es->s_mnt_count, 1);
2148 es->s_mtime = cpu_to_le32(get_seconds());
2149 ext4_update_dynamic_rev(sb);
2151 ext4_set_feature_journal_needs_recovery(sb);
2153 ext4_commit_super(sb, 1);
2155 if (test_opt(sb, DEBUG))
2156 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2157 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2159 sbi->s_groups_count,
2160 EXT4_BLOCKS_PER_GROUP(sb),
2161 EXT4_INODES_PER_GROUP(sb),
2162 sbi->s_mount_opt, sbi->s_mount_opt2);
2164 cleancache_init_fs(sb);
2168 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2170 struct ext4_sb_info *sbi = EXT4_SB(sb);
2171 struct flex_groups *new_groups;
2174 if (!sbi->s_log_groups_per_flex)
2177 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2178 if (size <= sbi->s_flex_groups_allocated)
2181 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2182 new_groups = kvzalloc(size, GFP_KERNEL);
2184 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2185 size / (int) sizeof(struct flex_groups));
2189 if (sbi->s_flex_groups) {
2190 memcpy(new_groups, sbi->s_flex_groups,
2191 (sbi->s_flex_groups_allocated *
2192 sizeof(struct flex_groups)));
2193 kvfree(sbi->s_flex_groups);
2195 sbi->s_flex_groups = new_groups;
2196 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2200 static int ext4_fill_flex_info(struct super_block *sb)
2202 struct ext4_sb_info *sbi = EXT4_SB(sb);
2203 struct ext4_group_desc *gdp = NULL;
2204 ext4_group_t flex_group;
2207 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2208 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2209 sbi->s_log_groups_per_flex = 0;
2213 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2217 for (i = 0; i < sbi->s_groups_count; i++) {
2218 gdp = ext4_get_group_desc(sb, i, NULL);
2220 flex_group = ext4_flex_group(sbi, i);
2221 atomic_add(ext4_free_inodes_count(sb, gdp),
2222 &sbi->s_flex_groups[flex_group].free_inodes);
2223 atomic64_add(ext4_free_group_clusters(sb, gdp),
2224 &sbi->s_flex_groups[flex_group].free_clusters);
2225 atomic_add(ext4_used_dirs_count(sb, gdp),
2226 &sbi->s_flex_groups[flex_group].used_dirs);
2234 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2235 struct ext4_group_desc *gdp)
2237 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2239 __le32 le_group = cpu_to_le32(block_group);
2240 struct ext4_sb_info *sbi = EXT4_SB(sb);
2242 if (ext4_has_metadata_csum(sbi->s_sb)) {
2243 /* Use new metadata_csum algorithm */
2245 __u16 dummy_csum = 0;
2247 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2249 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2250 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2251 sizeof(dummy_csum));
2252 offset += sizeof(dummy_csum);
2253 if (offset < sbi->s_desc_size)
2254 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2255 sbi->s_desc_size - offset);
2257 crc = csum32 & 0xFFFF;
2261 /* old crc16 code */
2262 if (!ext4_has_feature_gdt_csum(sb))
2265 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2266 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2267 crc = crc16(crc, (__u8 *)gdp, offset);
2268 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2269 /* for checksum of struct ext4_group_desc do the rest...*/
2270 if (ext4_has_feature_64bit(sb) &&
2271 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2272 crc = crc16(crc, (__u8 *)gdp + offset,
2273 le16_to_cpu(sbi->s_es->s_desc_size) -
2277 return cpu_to_le16(crc);
2280 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2281 struct ext4_group_desc *gdp)
2283 if (ext4_has_group_desc_csum(sb) &&
2284 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2290 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2291 struct ext4_group_desc *gdp)
2293 if (!ext4_has_group_desc_csum(sb))
2295 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2298 /* Called at mount-time, super-block is locked */
2299 static int ext4_check_descriptors(struct super_block *sb,
2300 ext4_fsblk_t sb_block,
2301 ext4_group_t *first_not_zeroed)
2303 struct ext4_sb_info *sbi = EXT4_SB(sb);
2304 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2305 ext4_fsblk_t last_block;
2306 ext4_fsblk_t block_bitmap;
2307 ext4_fsblk_t inode_bitmap;
2308 ext4_fsblk_t inode_table;
2309 int flexbg_flag = 0;
2310 ext4_group_t i, grp = sbi->s_groups_count;
2312 if (ext4_has_feature_flex_bg(sb))
2315 ext4_debug("Checking group descriptors");
2317 for (i = 0; i < sbi->s_groups_count; i++) {
2318 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2320 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2321 last_block = ext4_blocks_count(sbi->s_es) - 1;
2323 last_block = first_block +
2324 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2326 if ((grp == sbi->s_groups_count) &&
2327 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2330 block_bitmap = ext4_block_bitmap(sb, gdp);
2331 if (block_bitmap == sb_block) {
2332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2333 "Block bitmap for group %u overlaps "
2336 if (block_bitmap < first_block || block_bitmap > last_block) {
2337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2338 "Block bitmap for group %u not in group "
2339 "(block %llu)!", i, block_bitmap);
2342 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2343 if (inode_bitmap == sb_block) {
2344 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2345 "Inode bitmap for group %u overlaps "
2348 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2349 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2350 "Inode bitmap for group %u not in group "
2351 "(block %llu)!", i, inode_bitmap);
2354 inode_table = ext4_inode_table(sb, gdp);
2355 if (inode_table == sb_block) {
2356 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2357 "Inode table for group %u overlaps "
2360 if (inode_table < first_block ||
2361 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2362 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2363 "Inode table for group %u not in group "
2364 "(block %llu)!", i, inode_table);
2367 ext4_lock_group(sb, i);
2368 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2369 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2370 "Checksum for group %u failed (%u!=%u)",
2371 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2372 gdp)), le16_to_cpu(gdp->bg_checksum));
2373 if (!sb_rdonly(sb)) {
2374 ext4_unlock_group(sb, i);
2378 ext4_unlock_group(sb, i);
2380 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2382 if (NULL != first_not_zeroed)
2383 *first_not_zeroed = grp;
2387 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2388 * the superblock) which were deleted from all directories, but held open by
2389 * a process at the time of a crash. We walk the list and try to delete these
2390 * inodes at recovery time (only with a read-write filesystem).
2392 * In order to keep the orphan inode chain consistent during traversal (in
2393 * case of crash during recovery), we link each inode into the superblock
2394 * orphan list_head and handle it the same way as an inode deletion during
2395 * normal operation (which journals the operations for us).
2397 * We only do an iget() and an iput() on each inode, which is very safe if we
2398 * accidentally point at an in-use or already deleted inode. The worst that
2399 * can happen in this case is that we get a "bit already cleared" message from
2400 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2401 * e2fsck was run on this filesystem, and it must have already done the orphan
2402 * inode cleanup for us, so we can safely abort without any further action.
2404 static void ext4_orphan_cleanup(struct super_block *sb,
2405 struct ext4_super_block *es)
2407 unsigned int s_flags = sb->s_flags;
2408 int ret, nr_orphans = 0, nr_truncates = 0;
2410 int quota_update = 0;
2413 if (!es->s_last_orphan) {
2414 jbd_debug(4, "no orphan inodes to clean up\n");
2418 if (bdev_read_only(sb->s_bdev)) {
2419 ext4_msg(sb, KERN_ERR, "write access "
2420 "unavailable, skipping orphan cleanup");
2424 /* Check if feature set would not allow a r/w mount */
2425 if (!ext4_feature_set_ok(sb, 0)) {
2426 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2427 "unknown ROCOMPAT features");
2431 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2432 /* don't clear list on RO mount w/ errors */
2433 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2434 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2435 "clearing orphan list.\n");
2436 es->s_last_orphan = 0;
2438 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2442 if (s_flags & MS_RDONLY) {
2443 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2444 sb->s_flags &= ~MS_RDONLY;
2447 /* Needed for iput() to work correctly and not trash data */
2448 sb->s_flags |= MS_ACTIVE;
2451 * Turn on quotas which were not enabled for read-only mounts if
2452 * filesystem has quota feature, so that they are updated correctly.
2454 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2455 int ret = ext4_enable_quotas(sb);
2460 ext4_msg(sb, KERN_ERR,
2461 "Cannot turn on quotas: error %d", ret);
2464 /* Turn on journaled quotas used for old sytle */
2465 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2466 if (EXT4_SB(sb)->s_qf_names[i]) {
2467 int ret = ext4_quota_on_mount(sb, i);
2472 ext4_msg(sb, KERN_ERR,
2473 "Cannot turn on journaled "
2474 "quota: type %d: error %d", i, ret);
2479 while (es->s_last_orphan) {
2480 struct inode *inode;
2483 * We may have encountered an error during cleanup; if
2484 * so, skip the rest.
2486 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2487 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2488 es->s_last_orphan = 0;
2492 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2493 if (IS_ERR(inode)) {
2494 es->s_last_orphan = 0;
2498 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2499 dquot_initialize(inode);
2500 if (inode->i_nlink) {
2501 if (test_opt(sb, DEBUG))
2502 ext4_msg(sb, KERN_DEBUG,
2503 "%s: truncating inode %lu to %lld bytes",
2504 __func__, inode->i_ino, inode->i_size);
2505 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2506 inode->i_ino, inode->i_size);
2508 truncate_inode_pages(inode->i_mapping, inode->i_size);
2509 ret = ext4_truncate(inode);
2511 ext4_std_error(inode->i_sb, ret);
2512 inode_unlock(inode);
2515 if (test_opt(sb, DEBUG))
2516 ext4_msg(sb, KERN_DEBUG,
2517 "%s: deleting unreferenced inode %lu",
2518 __func__, inode->i_ino);
2519 jbd_debug(2, "deleting unreferenced inode %lu\n",
2523 iput(inode); /* The delete magic happens here! */
2526 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2529 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2530 PLURAL(nr_orphans));
2532 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2533 PLURAL(nr_truncates));
2535 /* Turn off quotas if they were enabled for orphan cleanup */
2537 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2538 if (sb_dqopt(sb)->files[i])
2539 dquot_quota_off(sb, i);
2543 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2547 * Maximal extent format file size.
2548 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2549 * extent format containers, within a sector_t, and within i_blocks
2550 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2551 * so that won't be a limiting factor.
2553 * However there is other limiting factor. We do store extents in the form
2554 * of starting block and length, hence the resulting length of the extent
2555 * covering maximum file size must fit into on-disk format containers as
2556 * well. Given that length is always by 1 unit bigger than max unit (because
2557 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2559 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2561 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2564 loff_t upper_limit = MAX_LFS_FILESIZE;
2566 /* small i_blocks in vfs inode? */
2567 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2569 * CONFIG_LBDAF is not enabled implies the inode
2570 * i_block represent total blocks in 512 bytes
2571 * 32 == size of vfs inode i_blocks * 8
2573 upper_limit = (1LL << 32) - 1;
2575 /* total blocks in file system block size */
2576 upper_limit >>= (blkbits - 9);
2577 upper_limit <<= blkbits;
2581 * 32-bit extent-start container, ee_block. We lower the maxbytes
2582 * by one fs block, so ee_len can cover the extent of maximum file
2585 res = (1LL << 32) - 1;
2588 /* Sanity check against vm- & vfs- imposed limits */
2589 if (res > upper_limit)
2596 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2597 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2598 * We need to be 1 filesystem block less than the 2^48 sector limit.
2600 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2602 loff_t res = EXT4_NDIR_BLOCKS;
2605 /* This is calculated to be the largest file size for a dense, block
2606 * mapped file such that the file's total number of 512-byte sectors,
2607 * including data and all indirect blocks, does not exceed (2^48 - 1).
2609 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2610 * number of 512-byte sectors of the file.
2613 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2615 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2616 * the inode i_block field represents total file blocks in
2617 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2619 upper_limit = (1LL << 32) - 1;
2621 /* total blocks in file system block size */
2622 upper_limit >>= (bits - 9);
2626 * We use 48 bit ext4_inode i_blocks
2627 * With EXT4_HUGE_FILE_FL set the i_blocks
2628 * represent total number of blocks in
2629 * file system block size
2631 upper_limit = (1LL << 48) - 1;
2635 /* indirect blocks */
2637 /* double indirect blocks */
2638 meta_blocks += 1 + (1LL << (bits-2));
2639 /* tripple indirect blocks */
2640 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2642 upper_limit -= meta_blocks;
2643 upper_limit <<= bits;
2645 res += 1LL << (bits-2);
2646 res += 1LL << (2*(bits-2));
2647 res += 1LL << (3*(bits-2));
2649 if (res > upper_limit)
2652 if (res > MAX_LFS_FILESIZE)
2653 res = MAX_LFS_FILESIZE;
2658 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2659 ext4_fsblk_t logical_sb_block, int nr)
2661 struct ext4_sb_info *sbi = EXT4_SB(sb);
2662 ext4_group_t bg, first_meta_bg;
2665 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2667 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2668 return logical_sb_block + nr + 1;
2669 bg = sbi->s_desc_per_block * nr;
2670 if (ext4_bg_has_super(sb, bg))
2674 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2675 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2676 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2679 if (sb->s_blocksize == 1024 && nr == 0 &&
2680 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2683 return (has_super + ext4_group_first_block_no(sb, bg));
2687 * ext4_get_stripe_size: Get the stripe size.
2688 * @sbi: In memory super block info
2690 * If we have specified it via mount option, then
2691 * use the mount option value. If the value specified at mount time is
2692 * greater than the blocks per group use the super block value.
2693 * If the super block value is greater than blocks per group return 0.
2694 * Allocator needs it be less than blocks per group.
2697 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2699 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2700 unsigned long stripe_width =
2701 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2704 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2705 ret = sbi->s_stripe;
2706 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2708 else if (stride && stride <= sbi->s_blocks_per_group)
2714 * If the stripe width is 1, this makes no sense and
2715 * we set it to 0 to turn off stripe handling code.
2724 * Check whether this filesystem can be mounted based on
2725 * the features present and the RDONLY/RDWR mount requested.
2726 * Returns 1 if this filesystem can be mounted as requested,
2727 * 0 if it cannot be.
2729 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2731 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2732 ext4_msg(sb, KERN_ERR,
2733 "Couldn't mount because of "
2734 "unsupported optional features (%x)",
2735 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2736 ~EXT4_FEATURE_INCOMPAT_SUPP));
2743 if (ext4_has_feature_readonly(sb)) {
2744 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2745 sb->s_flags |= MS_RDONLY;
2749 /* Check that feature set is OK for a read-write mount */
2750 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2751 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2752 "unsupported optional features (%x)",
2753 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2754 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2758 * Large file size enabled file system can only be mounted
2759 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2761 if (ext4_has_feature_huge_file(sb)) {
2762 if (sizeof(blkcnt_t) < sizeof(u64)) {
2763 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2764 "cannot be mounted RDWR without "
2769 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2770 ext4_msg(sb, KERN_ERR,
2771 "Can't support bigalloc feature without "
2772 "extents feature\n");
2776 #ifndef CONFIG_QUOTA
2777 if (ext4_has_feature_quota(sb) && !readonly) {
2778 ext4_msg(sb, KERN_ERR,
2779 "Filesystem with quota feature cannot be mounted RDWR "
2780 "without CONFIG_QUOTA");
2783 if (ext4_has_feature_project(sb) && !readonly) {
2784 ext4_msg(sb, KERN_ERR,
2785 "Filesystem with project quota feature cannot be mounted RDWR "
2786 "without CONFIG_QUOTA");
2789 #endif /* CONFIG_QUOTA */
2794 * This function is called once a day if we have errors logged
2795 * on the file system
2797 static void print_daily_error_info(unsigned long arg)
2799 struct super_block *sb = (struct super_block *) arg;
2800 struct ext4_sb_info *sbi;
2801 struct ext4_super_block *es;
2806 if (es->s_error_count)
2807 /* fsck newer than v1.41.13 is needed to clean this condition. */
2808 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2809 le32_to_cpu(es->s_error_count));
2810 if (es->s_first_error_time) {
2811 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2812 sb->s_id, le32_to_cpu(es->s_first_error_time),
2813 (int) sizeof(es->s_first_error_func),
2814 es->s_first_error_func,
2815 le32_to_cpu(es->s_first_error_line));
2816 if (es->s_first_error_ino)
2817 printk(KERN_CONT ": inode %u",
2818 le32_to_cpu(es->s_first_error_ino));
2819 if (es->s_first_error_block)
2820 printk(KERN_CONT ": block %llu", (unsigned long long)
2821 le64_to_cpu(es->s_first_error_block));
2822 printk(KERN_CONT "\n");
2824 if (es->s_last_error_time) {
2825 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2826 sb->s_id, le32_to_cpu(es->s_last_error_time),
2827 (int) sizeof(es->s_last_error_func),
2828 es->s_last_error_func,
2829 le32_to_cpu(es->s_last_error_line));
2830 if (es->s_last_error_ino)
2831 printk(KERN_CONT ": inode %u",
2832 le32_to_cpu(es->s_last_error_ino));
2833 if (es->s_last_error_block)
2834 printk(KERN_CONT ": block %llu", (unsigned long long)
2835 le64_to_cpu(es->s_last_error_block));
2836 printk(KERN_CONT "\n");
2838 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2841 /* Find next suitable group and run ext4_init_inode_table */
2842 static int ext4_run_li_request(struct ext4_li_request *elr)
2844 struct ext4_group_desc *gdp = NULL;
2845 ext4_group_t group, ngroups;
2846 struct super_block *sb;
2847 unsigned long timeout = 0;
2851 ngroups = EXT4_SB(sb)->s_groups_count;
2853 for (group = elr->lr_next_group; group < ngroups; group++) {
2854 gdp = ext4_get_group_desc(sb, group, NULL);
2860 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2864 if (group >= ngroups)
2869 ret = ext4_init_inode_table(sb, group,
2870 elr->lr_timeout ? 0 : 1);
2871 if (elr->lr_timeout == 0) {
2872 timeout = (jiffies - timeout) *
2873 elr->lr_sbi->s_li_wait_mult;
2874 elr->lr_timeout = timeout;
2876 elr->lr_next_sched = jiffies + elr->lr_timeout;
2877 elr->lr_next_group = group + 1;
2883 * Remove lr_request from the list_request and free the
2884 * request structure. Should be called with li_list_mtx held
2886 static void ext4_remove_li_request(struct ext4_li_request *elr)
2888 struct ext4_sb_info *sbi;
2895 list_del(&elr->lr_request);
2896 sbi->s_li_request = NULL;
2900 static void ext4_unregister_li_request(struct super_block *sb)
2902 mutex_lock(&ext4_li_mtx);
2903 if (!ext4_li_info) {
2904 mutex_unlock(&ext4_li_mtx);
2908 mutex_lock(&ext4_li_info->li_list_mtx);
2909 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2910 mutex_unlock(&ext4_li_info->li_list_mtx);
2911 mutex_unlock(&ext4_li_mtx);
2914 static struct task_struct *ext4_lazyinit_task;
2917 * This is the function where ext4lazyinit thread lives. It walks
2918 * through the request list searching for next scheduled filesystem.
2919 * When such a fs is found, run the lazy initialization request
2920 * (ext4_rn_li_request) and keep track of the time spend in this
2921 * function. Based on that time we compute next schedule time of
2922 * the request. When walking through the list is complete, compute
2923 * next waking time and put itself into sleep.
2925 static int ext4_lazyinit_thread(void *arg)
2927 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2928 struct list_head *pos, *n;
2929 struct ext4_li_request *elr;
2930 unsigned long next_wakeup, cur;
2932 BUG_ON(NULL == eli);
2936 next_wakeup = MAX_JIFFY_OFFSET;
2938 mutex_lock(&eli->li_list_mtx);
2939 if (list_empty(&eli->li_request_list)) {
2940 mutex_unlock(&eli->li_list_mtx);
2943 list_for_each_safe(pos, n, &eli->li_request_list) {
2946 elr = list_entry(pos, struct ext4_li_request,
2949 if (time_before(jiffies, elr->lr_next_sched)) {
2950 if (time_before(elr->lr_next_sched, next_wakeup))
2951 next_wakeup = elr->lr_next_sched;
2954 if (down_read_trylock(&elr->lr_super->s_umount)) {
2955 if (sb_start_write_trylock(elr->lr_super)) {
2958 * We hold sb->s_umount, sb can not
2959 * be removed from the list, it is
2960 * now safe to drop li_list_mtx
2962 mutex_unlock(&eli->li_list_mtx);
2963 err = ext4_run_li_request(elr);
2964 sb_end_write(elr->lr_super);
2965 mutex_lock(&eli->li_list_mtx);
2968 up_read((&elr->lr_super->s_umount));
2970 /* error, remove the lazy_init job */
2972 ext4_remove_li_request(elr);
2976 elr->lr_next_sched = jiffies +
2978 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2980 if (time_before(elr->lr_next_sched, next_wakeup))
2981 next_wakeup = elr->lr_next_sched;
2983 mutex_unlock(&eli->li_list_mtx);
2988 if ((time_after_eq(cur, next_wakeup)) ||
2989 (MAX_JIFFY_OFFSET == next_wakeup)) {
2994 schedule_timeout_interruptible(next_wakeup - cur);
2996 if (kthread_should_stop()) {
2997 ext4_clear_request_list();
3004 * It looks like the request list is empty, but we need
3005 * to check it under the li_list_mtx lock, to prevent any
3006 * additions into it, and of course we should lock ext4_li_mtx
3007 * to atomically free the list and ext4_li_info, because at
3008 * this point another ext4 filesystem could be registering
3011 mutex_lock(&ext4_li_mtx);
3012 mutex_lock(&eli->li_list_mtx);
3013 if (!list_empty(&eli->li_request_list)) {
3014 mutex_unlock(&eli->li_list_mtx);
3015 mutex_unlock(&ext4_li_mtx);
3018 mutex_unlock(&eli->li_list_mtx);
3019 kfree(ext4_li_info);
3020 ext4_li_info = NULL;
3021 mutex_unlock(&ext4_li_mtx);
3026 static void ext4_clear_request_list(void)
3028 struct list_head *pos, *n;
3029 struct ext4_li_request *elr;
3031 mutex_lock(&ext4_li_info->li_list_mtx);
3032 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3033 elr = list_entry(pos, struct ext4_li_request,
3035 ext4_remove_li_request(elr);
3037 mutex_unlock(&ext4_li_info->li_list_mtx);
3040 static int ext4_run_lazyinit_thread(void)
3042 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3043 ext4_li_info, "ext4lazyinit");
3044 if (IS_ERR(ext4_lazyinit_task)) {
3045 int err = PTR_ERR(ext4_lazyinit_task);
3046 ext4_clear_request_list();
3047 kfree(ext4_li_info);
3048 ext4_li_info = NULL;
3049 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3050 "initialization thread\n",
3054 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3059 * Check whether it make sense to run itable init. thread or not.
3060 * If there is at least one uninitialized inode table, return
3061 * corresponding group number, else the loop goes through all
3062 * groups and return total number of groups.
3064 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3066 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3067 struct ext4_group_desc *gdp = NULL;
3069 for (group = 0; group < ngroups; group++) {
3070 gdp = ext4_get_group_desc(sb, group, NULL);
3074 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3081 static int ext4_li_info_new(void)
3083 struct ext4_lazy_init *eli = NULL;
3085 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3089 INIT_LIST_HEAD(&eli->li_request_list);
3090 mutex_init(&eli->li_list_mtx);
3092 eli->li_state |= EXT4_LAZYINIT_QUIT;
3099 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3102 struct ext4_sb_info *sbi = EXT4_SB(sb);
3103 struct ext4_li_request *elr;
3105 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3111 elr->lr_next_group = start;
3114 * Randomize first schedule time of the request to
3115 * spread the inode table initialization requests
3118 elr->lr_next_sched = jiffies + (prandom_u32() %
3119 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3123 int ext4_register_li_request(struct super_block *sb,
3124 ext4_group_t first_not_zeroed)
3126 struct ext4_sb_info *sbi = EXT4_SB(sb);
3127 struct ext4_li_request *elr = NULL;
3128 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3131 mutex_lock(&ext4_li_mtx);
3132 if (sbi->s_li_request != NULL) {
3134 * Reset timeout so it can be computed again, because
3135 * s_li_wait_mult might have changed.
3137 sbi->s_li_request->lr_timeout = 0;
3141 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3142 !test_opt(sb, INIT_INODE_TABLE))
3145 elr = ext4_li_request_new(sb, first_not_zeroed);
3151 if (NULL == ext4_li_info) {
3152 ret = ext4_li_info_new();
3157 mutex_lock(&ext4_li_info->li_list_mtx);
3158 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3159 mutex_unlock(&ext4_li_info->li_list_mtx);
3161 sbi->s_li_request = elr;
3163 * set elr to NULL here since it has been inserted to
3164 * the request_list and the removal and free of it is
3165 * handled by ext4_clear_request_list from now on.
3169 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3170 ret = ext4_run_lazyinit_thread();
3175 mutex_unlock(&ext4_li_mtx);
3182 * We do not need to lock anything since this is called on
3185 static void ext4_destroy_lazyinit_thread(void)
3188 * If thread exited earlier
3189 * there's nothing to be done.
3191 if (!ext4_li_info || !ext4_lazyinit_task)
3194 kthread_stop(ext4_lazyinit_task);
3197 static int set_journal_csum_feature_set(struct super_block *sb)
3200 int compat, incompat;
3201 struct ext4_sb_info *sbi = EXT4_SB(sb);
3203 if (ext4_has_metadata_csum(sb)) {
3204 /* journal checksum v3 */
3206 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3208 /* journal checksum v1 */
3209 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3213 jbd2_journal_clear_features(sbi->s_journal,
3214 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3215 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3216 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3217 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3218 ret = jbd2_journal_set_features(sbi->s_journal,
3220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3222 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3223 ret = jbd2_journal_set_features(sbi->s_journal,
3226 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3227 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3229 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3230 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3237 * Note: calculating the overhead so we can be compatible with
3238 * historical BSD practice is quite difficult in the face of
3239 * clusters/bigalloc. This is because multiple metadata blocks from
3240 * different block group can end up in the same allocation cluster.
3241 * Calculating the exact overhead in the face of clustered allocation
3242 * requires either O(all block bitmaps) in memory or O(number of block
3243 * groups**2) in time. We will still calculate the superblock for
3244 * older file systems --- and if we come across with a bigalloc file
3245 * system with zero in s_overhead_clusters the estimate will be close to
3246 * correct especially for very large cluster sizes --- but for newer
3247 * file systems, it's better to calculate this figure once at mkfs
3248 * time, and store it in the superblock. If the superblock value is
3249 * present (even for non-bigalloc file systems), we will use it.
3251 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3254 struct ext4_sb_info *sbi = EXT4_SB(sb);
3255 struct ext4_group_desc *gdp;
3256 ext4_fsblk_t first_block, last_block, b;
3257 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3258 int s, j, count = 0;
3260 if (!ext4_has_feature_bigalloc(sb))
3261 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3262 sbi->s_itb_per_group + 2);
3264 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3265 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3266 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3267 for (i = 0; i < ngroups; i++) {
3268 gdp = ext4_get_group_desc(sb, i, NULL);
3269 b = ext4_block_bitmap(sb, gdp);
3270 if (b >= first_block && b <= last_block) {
3271 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3274 b = ext4_inode_bitmap(sb, gdp);
3275 if (b >= first_block && b <= last_block) {
3276 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3279 b = ext4_inode_table(sb, gdp);
3280 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3281 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3282 int c = EXT4_B2C(sbi, b - first_block);
3283 ext4_set_bit(c, buf);
3289 if (ext4_bg_has_super(sb, grp)) {
3290 ext4_set_bit(s++, buf);
3293 j = ext4_bg_num_gdb(sb, grp);
3294 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3295 ext4_error(sb, "Invalid number of block group "
3296 "descriptor blocks: %d", j);
3297 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3301 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3305 return EXT4_CLUSTERS_PER_GROUP(sb) -
3306 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3310 * Compute the overhead and stash it in sbi->s_overhead
3312 int ext4_calculate_overhead(struct super_block *sb)
3314 struct ext4_sb_info *sbi = EXT4_SB(sb);
3315 struct ext4_super_block *es = sbi->s_es;
3316 struct inode *j_inode;
3317 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3318 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3319 ext4_fsblk_t overhead = 0;
3320 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3326 * Compute the overhead (FS structures). This is constant
3327 * for a given filesystem unless the number of block groups
3328 * changes so we cache the previous value until it does.
3332 * All of the blocks before first_data_block are overhead
3334 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3337 * Add the overhead found in each block group
3339 for (i = 0; i < ngroups; i++) {
3342 blks = count_overhead(sb, i, buf);
3345 memset(buf, 0, PAGE_SIZE);
3350 * Add the internal journal blocks whether the journal has been
3353 if (sbi->s_journal && !sbi->journal_bdev)
3354 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3355 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3356 j_inode = ext4_get_journal_inode(sb, j_inum);
3358 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3359 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3362 ext4_msg(sb, KERN_ERR, "can't get journal size");
3365 sbi->s_overhead = overhead;
3367 free_page((unsigned long) buf);
3371 static void ext4_set_resv_clusters(struct super_block *sb)
3373 ext4_fsblk_t resv_clusters;
3374 struct ext4_sb_info *sbi = EXT4_SB(sb);
3377 * There's no need to reserve anything when we aren't using extents.
3378 * The space estimates are exact, there are no unwritten extents,
3379 * hole punching doesn't need new metadata... This is needed especially
3380 * to keep ext2/3 backward compatibility.
3382 if (!ext4_has_feature_extents(sb))
3385 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3386 * This should cover the situations where we can not afford to run
3387 * out of space like for example punch hole, or converting
3388 * unwritten extents in delalloc path. In most cases such
3389 * allocation would require 1, or 2 blocks, higher numbers are
3392 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3393 sbi->s_cluster_bits);
3395 do_div(resv_clusters, 50);
3396 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3398 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3401 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3403 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3404 char *orig_data = kstrdup(data, GFP_KERNEL);
3405 struct buffer_head *bh;
3406 struct ext4_super_block *es = NULL;
3407 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3409 ext4_fsblk_t sb_block = get_sb_block(&data);
3410 ext4_fsblk_t logical_sb_block;
3411 unsigned long offset = 0;
3412 unsigned long journal_devnum = 0;
3413 unsigned long def_mount_opts;
3417 int blocksize, clustersize;
3418 unsigned int db_count;
3420 int needs_recovery, has_huge_files, has_bigalloc;
3423 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3424 ext4_group_t first_not_zeroed;
3426 if ((data && !orig_data) || !sbi)
3429 sbi->s_daxdev = dax_dev;
3430 sbi->s_blockgroup_lock =
3431 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3432 if (!sbi->s_blockgroup_lock)
3435 sb->s_fs_info = sbi;
3437 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3438 sbi->s_sb_block = sb_block;
3439 if (sb->s_bdev->bd_part)
3440 sbi->s_sectors_written_start =
3441 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3443 /* Cleanup superblock name */
3444 strreplace(sb->s_id, '/', '!');
3446 /* -EINVAL is default */
3448 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3450 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3455 * The ext4 superblock will not be buffer aligned for other than 1kB
3456 * block sizes. We need to calculate the offset from buffer start.
3458 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3459 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3460 offset = do_div(logical_sb_block, blocksize);
3462 logical_sb_block = sb_block;
3465 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3466 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3470 * Note: s_es must be initialized as soon as possible because
3471 * some ext4 macro-instructions depend on its value
3473 es = (struct ext4_super_block *) (bh->b_data + offset);
3475 sb->s_magic = le16_to_cpu(es->s_magic);
3476 if (sb->s_magic != EXT4_SUPER_MAGIC)
3478 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3480 /* Warn if metadata_csum and gdt_csum are both set. */
3481 if (ext4_has_feature_metadata_csum(sb) &&
3482 ext4_has_feature_gdt_csum(sb))
3483 ext4_warning(sb, "metadata_csum and uninit_bg are "
3484 "redundant flags; please run fsck.");
3486 /* Check for a known checksum algorithm */
3487 if (!ext4_verify_csum_type(sb, es)) {
3488 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3489 "unknown checksum algorithm.");
3494 /* Load the checksum driver */
3495 if (ext4_has_feature_metadata_csum(sb) ||
3496 ext4_has_feature_ea_inode(sb)) {
3497 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3498 if (IS_ERR(sbi->s_chksum_driver)) {
3499 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3500 ret = PTR_ERR(sbi->s_chksum_driver);
3501 sbi->s_chksum_driver = NULL;
3506 /* Check superblock checksum */
3507 if (!ext4_superblock_csum_verify(sb, es)) {
3508 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3509 "invalid superblock checksum. Run e2fsck?");
3515 /* Precompute checksum seed for all metadata */
3516 if (ext4_has_feature_csum_seed(sb))
3517 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3518 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3519 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3520 sizeof(es->s_uuid));
3522 /* Set defaults before we parse the mount options */
3523 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3524 set_opt(sb, INIT_INODE_TABLE);
3525 if (def_mount_opts & EXT4_DEFM_DEBUG)
3527 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3529 if (def_mount_opts & EXT4_DEFM_UID16)
3530 set_opt(sb, NO_UID32);
3531 /* xattr user namespace & acls are now defaulted on */
3532 set_opt(sb, XATTR_USER);
3533 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3534 set_opt(sb, POSIX_ACL);
3536 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3537 if (ext4_has_metadata_csum(sb))
3538 set_opt(sb, JOURNAL_CHECKSUM);
3540 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3541 set_opt(sb, JOURNAL_DATA);
3542 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3543 set_opt(sb, ORDERED_DATA);
3544 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3545 set_opt(sb, WRITEBACK_DATA);
3547 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3548 set_opt(sb, ERRORS_PANIC);
3549 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3550 set_opt(sb, ERRORS_CONT);
3552 set_opt(sb, ERRORS_RO);
3553 /* block_validity enabled by default; disable with noblock_validity */
3554 set_opt(sb, BLOCK_VALIDITY);
3555 if (def_mount_opts & EXT4_DEFM_DISCARD)
3556 set_opt(sb, DISCARD);
3558 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3559 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3560 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3561 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3562 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3564 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3565 set_opt(sb, BARRIER);
3568 * enable delayed allocation by default
3569 * Use -o nodelalloc to turn it off
3571 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3572 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3573 set_opt(sb, DELALLOC);
3576 * set default s_li_wait_mult for lazyinit, for the case there is
3577 * no mount option specified.
3579 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3581 if (sbi->s_es->s_mount_opts[0]) {
3582 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3583 sizeof(sbi->s_es->s_mount_opts),
3587 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3588 &journal_ioprio, 0)) {
3589 ext4_msg(sb, KERN_WARNING,
3590 "failed to parse options in superblock: %s",
3593 kfree(s_mount_opts);
3595 sbi->s_def_mount_opt = sbi->s_mount_opt;
3596 if (!parse_options((char *) data, sb, &journal_devnum,
3597 &journal_ioprio, 0))
3600 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3601 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3602 "with data=journal disables delayed "
3603 "allocation and O_DIRECT support!\n");
3604 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3605 ext4_msg(sb, KERN_ERR, "can't mount with "
3606 "both data=journal and delalloc");
3609 if (test_opt(sb, DIOREAD_NOLOCK)) {
3610 ext4_msg(sb, KERN_ERR, "can't mount with "
3611 "both data=journal and dioread_nolock");
3614 if (test_opt(sb, DAX)) {
3615 ext4_msg(sb, KERN_ERR, "can't mount with "
3616 "both data=journal and dax");
3619 if (ext4_has_feature_encrypt(sb)) {
3620 ext4_msg(sb, KERN_WARNING,
3621 "encrypted files will use data=ordered "
3622 "instead of data journaling mode");
3624 if (test_opt(sb, DELALLOC))
3625 clear_opt(sb, DELALLOC);
3627 sb->s_iflags |= SB_I_CGROUPWB;
3630 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3631 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3633 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3634 (ext4_has_compat_features(sb) ||
3635 ext4_has_ro_compat_features(sb) ||
3636 ext4_has_incompat_features(sb)))
3637 ext4_msg(sb, KERN_WARNING,
3638 "feature flags set on rev 0 fs, "
3639 "running e2fsck is recommended");
3641 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3642 set_opt2(sb, HURD_COMPAT);
3643 if (ext4_has_feature_64bit(sb)) {
3644 ext4_msg(sb, KERN_ERR,
3645 "The Hurd can't support 64-bit file systems");
3650 * ea_inode feature uses l_i_version field which is not
3651 * available in HURD_COMPAT mode.
3653 if (ext4_has_feature_ea_inode(sb)) {
3654 ext4_msg(sb, KERN_ERR,
3655 "ea_inode feature is not supported for Hurd");
3660 if (IS_EXT2_SB(sb)) {
3661 if (ext2_feature_set_ok(sb))
3662 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3663 "using the ext4 subsystem");
3665 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3666 "to feature incompatibilities");
3671 if (IS_EXT3_SB(sb)) {
3672 if (ext3_feature_set_ok(sb))
3673 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3674 "using the ext4 subsystem");
3676 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3677 "to feature incompatibilities");
3683 * Check feature flags regardless of the revision level, since we
3684 * previously didn't change the revision level when setting the flags,
3685 * so there is a chance incompat flags are set on a rev 0 filesystem.
3687 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3690 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3691 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3692 blocksize > EXT4_MAX_BLOCK_SIZE) {
3693 ext4_msg(sb, KERN_ERR,
3694 "Unsupported filesystem blocksize %d (%d log_block_size)",
3695 blocksize, le32_to_cpu(es->s_log_block_size));
3698 if (le32_to_cpu(es->s_log_block_size) >
3699 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3700 ext4_msg(sb, KERN_ERR,
3701 "Invalid log block size: %u",
3702 le32_to_cpu(es->s_log_block_size));
3706 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3707 ext4_msg(sb, KERN_ERR,
3708 "Number of reserved GDT blocks insanely large: %d",
3709 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3713 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3714 if (ext4_has_feature_inline_data(sb)) {
3715 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3716 " that may contain inline data");
3719 err = bdev_dax_supported(sb, blocksize);
3724 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3725 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3726 es->s_encryption_level);
3730 if (sb->s_blocksize != blocksize) {
3731 /* Validate the filesystem blocksize */
3732 if (!sb_set_blocksize(sb, blocksize)) {
3733 ext4_msg(sb, KERN_ERR, "bad block size %d",
3739 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3740 offset = do_div(logical_sb_block, blocksize);
3741 bh = sb_bread_unmovable(sb, logical_sb_block);
3743 ext4_msg(sb, KERN_ERR,
3744 "Can't read superblock on 2nd try");
3747 es = (struct ext4_super_block *)(bh->b_data + offset);
3749 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3750 ext4_msg(sb, KERN_ERR,
3751 "Magic mismatch, very weird!");
3756 has_huge_files = ext4_has_feature_huge_file(sb);
3757 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3759 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3761 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3762 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3763 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3765 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3766 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3767 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3768 (!is_power_of_2(sbi->s_inode_size)) ||
3769 (sbi->s_inode_size > blocksize)) {
3770 ext4_msg(sb, KERN_ERR,
3771 "unsupported inode size: %d",
3775 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3776 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3779 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3780 if (ext4_has_feature_64bit(sb)) {
3781 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3782 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3783 !is_power_of_2(sbi->s_desc_size)) {
3784 ext4_msg(sb, KERN_ERR,
3785 "unsupported descriptor size %lu",
3790 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3792 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3793 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3795 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3796 if (sbi->s_inodes_per_block == 0)
3798 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3799 sbi->s_inodes_per_group > blocksize * 8) {
3800 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3801 sbi->s_blocks_per_group);
3804 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3805 sbi->s_inodes_per_block;
3806 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3808 sbi->s_mount_state = le16_to_cpu(es->s_state);
3809 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3810 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3812 for (i = 0; i < 4; i++)
3813 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3814 sbi->s_def_hash_version = es->s_def_hash_version;
3815 if (ext4_has_feature_dir_index(sb)) {
3816 i = le32_to_cpu(es->s_flags);
3817 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3818 sbi->s_hash_unsigned = 3;
3819 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3820 #ifdef __CHAR_UNSIGNED__
3823 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3824 sbi->s_hash_unsigned = 3;
3828 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3833 /* Handle clustersize */
3834 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3835 has_bigalloc = ext4_has_feature_bigalloc(sb);
3837 if (clustersize < blocksize) {
3838 ext4_msg(sb, KERN_ERR,
3839 "cluster size (%d) smaller than "
3840 "block size (%d)", clustersize, blocksize);
3843 if (le32_to_cpu(es->s_log_cluster_size) >
3844 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3845 ext4_msg(sb, KERN_ERR,
3846 "Invalid log cluster size: %u",
3847 le32_to_cpu(es->s_log_cluster_size));
3850 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3851 le32_to_cpu(es->s_log_block_size);
3852 sbi->s_clusters_per_group =
3853 le32_to_cpu(es->s_clusters_per_group);
3854 if (sbi->s_clusters_per_group > blocksize * 8) {
3855 ext4_msg(sb, KERN_ERR,
3856 "#clusters per group too big: %lu",
3857 sbi->s_clusters_per_group);
3860 if (sbi->s_blocks_per_group !=
3861 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3862 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3863 "clusters per group (%lu) inconsistent",
3864 sbi->s_blocks_per_group,
3865 sbi->s_clusters_per_group);
3869 if (clustersize != blocksize) {
3870 ext4_warning(sb, "fragment/cluster size (%d) != "
3871 "block size (%d)", clustersize,
3873 clustersize = blocksize;
3875 if (sbi->s_blocks_per_group > blocksize * 8) {
3876 ext4_msg(sb, KERN_ERR,
3877 "#blocks per group too big: %lu",
3878 sbi->s_blocks_per_group);
3881 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3882 sbi->s_cluster_bits = 0;
3884 sbi->s_cluster_ratio = clustersize / blocksize;
3886 /* Do we have standard group size of clustersize * 8 blocks ? */
3887 if (sbi->s_blocks_per_group == clustersize << 3)
3888 set_opt2(sb, STD_GROUP_SIZE);
3891 * Test whether we have more sectors than will fit in sector_t,
3892 * and whether the max offset is addressable by the page cache.
3894 err = generic_check_addressable(sb->s_blocksize_bits,
3895 ext4_blocks_count(es));
3897 ext4_msg(sb, KERN_ERR, "filesystem"
3898 " too large to mount safely on this system");
3899 if (sizeof(sector_t) < 8)
3900 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3904 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3907 /* check blocks count against device size */
3908 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3909 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3910 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3911 "exceeds size of device (%llu blocks)",
3912 ext4_blocks_count(es), blocks_count);
3917 * It makes no sense for the first data block to be beyond the end
3918 * of the filesystem.
3920 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3921 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3922 "block %u is beyond end of filesystem (%llu)",
3923 le32_to_cpu(es->s_first_data_block),
3924 ext4_blocks_count(es));
3927 blocks_count = (ext4_blocks_count(es) -
3928 le32_to_cpu(es->s_first_data_block) +
3929 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3930 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3931 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3932 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3933 "(block count %llu, first data block %u, "
3934 "blocks per group %lu)", sbi->s_groups_count,
3935 ext4_blocks_count(es),
3936 le32_to_cpu(es->s_first_data_block),
3937 EXT4_BLOCKS_PER_GROUP(sb));
3940 sbi->s_groups_count = blocks_count;
3941 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3942 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3943 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3944 EXT4_DESC_PER_BLOCK(sb);
3945 if (ext4_has_feature_meta_bg(sb)) {
3946 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3947 ext4_msg(sb, KERN_WARNING,
3948 "first meta block group too large: %u "
3949 "(group descriptor block count %u)",
3950 le32_to_cpu(es->s_first_meta_bg), db_count);
3954 sbi->s_group_desc = kvmalloc(db_count *
3955 sizeof(struct buffer_head *),
3957 if (sbi->s_group_desc == NULL) {
3958 ext4_msg(sb, KERN_ERR, "not enough memory");
3963 bgl_lock_init(sbi->s_blockgroup_lock);
3965 /* Pre-read the descriptors into the buffer cache */
3966 for (i = 0; i < db_count; i++) {
3967 block = descriptor_loc(sb, logical_sb_block, i);
3968 sb_breadahead(sb, block);
3971 for (i = 0; i < db_count; i++) {
3972 block = descriptor_loc(sb, logical_sb_block, i);
3973 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3974 if (!sbi->s_group_desc[i]) {
3975 ext4_msg(sb, KERN_ERR,
3976 "can't read group descriptor %d", i);
3981 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3982 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3983 ret = -EFSCORRUPTED;
3987 sbi->s_gdb_count = db_count;
3988 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3989 spin_lock_init(&sbi->s_next_gen_lock);
3991 setup_timer(&sbi->s_err_report, print_daily_error_info,
3992 (unsigned long) sb);
3994 /* Register extent status tree shrinker */
3995 if (ext4_es_register_shrinker(sbi))
3998 sbi->s_stripe = ext4_get_stripe_size(sbi);
3999 sbi->s_extent_max_zeroout_kb = 32;
4002 * set up enough so that it can read an inode
4004 sb->s_op = &ext4_sops;
4005 sb->s_export_op = &ext4_export_ops;
4006 sb->s_xattr = ext4_xattr_handlers;
4007 sb->s_cop = &ext4_cryptops;
4009 sb->dq_op = &ext4_quota_operations;
4010 if (ext4_has_feature_quota(sb))
4011 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4013 sb->s_qcop = &ext4_qctl_operations;
4014 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4016 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4018 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4019 mutex_init(&sbi->s_orphan_lock);
4023 needs_recovery = (es->s_last_orphan != 0 ||
4024 ext4_has_feature_journal_needs_recovery(sb));
4026 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4027 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4028 goto failed_mount3a;
4031 * The first inode we look at is the journal inode. Don't try
4032 * root first: it may be modified in the journal!
4034 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4035 err = ext4_load_journal(sb, es, journal_devnum);
4037 goto failed_mount3a;
4038 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4039 ext4_has_feature_journal_needs_recovery(sb)) {
4040 ext4_msg(sb, KERN_ERR, "required journal recovery "
4041 "suppressed and not mounted read-only");
4042 goto failed_mount_wq;
4044 /* Nojournal mode, all journal mount options are illegal */
4045 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4046 ext4_msg(sb, KERN_ERR, "can't mount with "
4047 "journal_checksum, fs mounted w/o journal");
4048 goto failed_mount_wq;
4050 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4051 ext4_msg(sb, KERN_ERR, "can't mount with "
4052 "journal_async_commit, fs mounted w/o journal");
4053 goto failed_mount_wq;
4055 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4056 ext4_msg(sb, KERN_ERR, "can't mount with "
4057 "commit=%lu, fs mounted w/o journal",
4058 sbi->s_commit_interval / HZ);
4059 goto failed_mount_wq;
4061 if (EXT4_MOUNT_DATA_FLAGS &
4062 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4063 ext4_msg(sb, KERN_ERR, "can't mount with "
4064 "data=, fs mounted w/o journal");
4065 goto failed_mount_wq;
4067 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4068 clear_opt(sb, JOURNAL_CHECKSUM);
4069 clear_opt(sb, DATA_FLAGS);
4070 sbi->s_journal = NULL;
4075 if (ext4_has_feature_64bit(sb) &&
4076 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4077 JBD2_FEATURE_INCOMPAT_64BIT)) {
4078 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4079 goto failed_mount_wq;
4082 if (!set_journal_csum_feature_set(sb)) {
4083 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4085 goto failed_mount_wq;
4088 /* We have now updated the journal if required, so we can
4089 * validate the data journaling mode. */
4090 switch (test_opt(sb, DATA_FLAGS)) {
4092 /* No mode set, assume a default based on the journal
4093 * capabilities: ORDERED_DATA if the journal can
4094 * cope, else JOURNAL_DATA
4096 if (jbd2_journal_check_available_features
4097 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4098 set_opt(sb, ORDERED_DATA);
4100 set_opt(sb, JOURNAL_DATA);
4103 case EXT4_MOUNT_ORDERED_DATA:
4104 case EXT4_MOUNT_WRITEBACK_DATA:
4105 if (!jbd2_journal_check_available_features
4106 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4107 ext4_msg(sb, KERN_ERR, "Journal does not support "
4108 "requested data journaling mode");
4109 goto failed_mount_wq;
4115 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4116 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4117 ext4_msg(sb, KERN_ERR, "can't mount with "
4118 "journal_async_commit in data=ordered mode");
4119 goto failed_mount_wq;
4122 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4124 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4127 if (!test_opt(sb, NO_MBCACHE)) {
4128 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4129 if (!sbi->s_ea_block_cache) {
4130 ext4_msg(sb, KERN_ERR,
4131 "Failed to create ea_block_cache");
4132 goto failed_mount_wq;
4135 if (ext4_has_feature_ea_inode(sb)) {
4136 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4137 if (!sbi->s_ea_inode_cache) {
4138 ext4_msg(sb, KERN_ERR,
4139 "Failed to create ea_inode_cache");
4140 goto failed_mount_wq;
4145 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4146 (blocksize != PAGE_SIZE)) {
4147 ext4_msg(sb, KERN_ERR,
4148 "Unsupported blocksize for fs encryption");
4149 goto failed_mount_wq;
4152 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4153 !ext4_has_feature_encrypt(sb)) {
4154 ext4_set_feature_encrypt(sb);
4155 ext4_commit_super(sb, 1);
4159 * Get the # of file system overhead blocks from the
4160 * superblock if present.
4162 if (es->s_overhead_clusters)
4163 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4165 err = ext4_calculate_overhead(sb);
4167 goto failed_mount_wq;
4171 * The maximum number of concurrent works can be high and
4172 * concurrency isn't really necessary. Limit it to 1.
4174 EXT4_SB(sb)->rsv_conversion_wq =
4175 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4176 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4177 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4183 * The jbd2_journal_load will have done any necessary log recovery,
4184 * so we can safely mount the rest of the filesystem now.
4187 root = ext4_iget(sb, EXT4_ROOT_INO);
4189 ext4_msg(sb, KERN_ERR, "get root inode failed");
4190 ret = PTR_ERR(root);
4194 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4195 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4199 sb->s_root = d_make_root(root);
4201 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4206 if (ext4_setup_super(sb, es, sb_rdonly(sb)))
4207 sb->s_flags |= MS_RDONLY;
4209 /* determine the minimum size of new large inodes, if present */
4210 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4211 sbi->s_want_extra_isize == 0) {
4212 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4213 EXT4_GOOD_OLD_INODE_SIZE;
4214 if (ext4_has_feature_extra_isize(sb)) {
4215 if (sbi->s_want_extra_isize <
4216 le16_to_cpu(es->s_want_extra_isize))
4217 sbi->s_want_extra_isize =
4218 le16_to_cpu(es->s_want_extra_isize);
4219 if (sbi->s_want_extra_isize <
4220 le16_to_cpu(es->s_min_extra_isize))
4221 sbi->s_want_extra_isize =
4222 le16_to_cpu(es->s_min_extra_isize);
4225 /* Check if enough inode space is available */
4226 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4227 sbi->s_inode_size) {
4228 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4229 EXT4_GOOD_OLD_INODE_SIZE;
4230 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4234 ext4_set_resv_clusters(sb);
4236 err = ext4_setup_system_zone(sb);
4238 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4240 goto failed_mount4a;
4244 err = ext4_mb_init(sb);
4246 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4251 block = ext4_count_free_clusters(sb);
4252 ext4_free_blocks_count_set(sbi->s_es,
4253 EXT4_C2B(sbi, block));
4254 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4257 unsigned long freei = ext4_count_free_inodes(sb);
4258 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4259 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4263 err = percpu_counter_init(&sbi->s_dirs_counter,
4264 ext4_count_dirs(sb), GFP_KERNEL);
4266 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4269 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4272 ext4_msg(sb, KERN_ERR, "insufficient memory");
4276 if (ext4_has_feature_flex_bg(sb))
4277 if (!ext4_fill_flex_info(sb)) {
4278 ext4_msg(sb, KERN_ERR,
4279 "unable to initialize "
4280 "flex_bg meta info!");
4284 err = ext4_register_li_request(sb, first_not_zeroed);
4288 err = ext4_register_sysfs(sb);
4293 /* Enable quota usage during mount. */
4294 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4295 err = ext4_enable_quotas(sb);
4299 #endif /* CONFIG_QUOTA */
4301 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4302 ext4_orphan_cleanup(sb, es);
4303 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4304 if (needs_recovery) {
4305 ext4_msg(sb, KERN_INFO, "recovery complete");
4306 ext4_mark_recovery_complete(sb, es);
4308 if (EXT4_SB(sb)->s_journal) {
4309 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4310 descr = " journalled data mode";
4311 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4312 descr = " ordered data mode";
4314 descr = " writeback data mode";
4316 descr = "out journal";
4318 if (test_opt(sb, DISCARD)) {
4319 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4320 if (!blk_queue_discard(q))
4321 ext4_msg(sb, KERN_WARNING,
4322 "mounting with \"discard\" option, but "
4323 "the device does not support discard");
4326 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4327 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4328 "Opts: %.*s%s%s", descr,
4329 (int) sizeof(sbi->s_es->s_mount_opts),
4330 sbi->s_es->s_mount_opts,
4331 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4333 if (es->s_error_count)
4334 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4336 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4337 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4338 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4339 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4346 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4351 ext4_unregister_sysfs(sb);
4354 ext4_unregister_li_request(sb);
4356 ext4_mb_release(sb);
4357 if (sbi->s_flex_groups)
4358 kvfree(sbi->s_flex_groups);
4359 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4360 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4361 percpu_counter_destroy(&sbi->s_dirs_counter);
4362 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4364 ext4_ext_release(sb);
4365 ext4_release_system_zone(sb);
4370 ext4_msg(sb, KERN_ERR, "mount failed");
4371 if (EXT4_SB(sb)->rsv_conversion_wq)
4372 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4374 if (sbi->s_ea_inode_cache) {
4375 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4376 sbi->s_ea_inode_cache = NULL;
4378 if (sbi->s_ea_block_cache) {
4379 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4380 sbi->s_ea_block_cache = NULL;
4382 if (sbi->s_journal) {
4383 jbd2_journal_destroy(sbi->s_journal);
4384 sbi->s_journal = NULL;
4387 ext4_es_unregister_shrinker(sbi);
4389 del_timer_sync(&sbi->s_err_report);
4391 kthread_stop(sbi->s_mmp_tsk);
4393 for (i = 0; i < db_count; i++)
4394 brelse(sbi->s_group_desc[i]);
4395 kvfree(sbi->s_group_desc);
4397 if (sbi->s_chksum_driver)
4398 crypto_free_shash(sbi->s_chksum_driver);
4400 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4401 kfree(sbi->s_qf_names[i]);
4403 ext4_blkdev_remove(sbi);
4406 sb->s_fs_info = NULL;
4407 kfree(sbi->s_blockgroup_lock);
4411 fs_put_dax(dax_dev);
4412 return err ? err : ret;
4416 * Setup any per-fs journal parameters now. We'll do this both on
4417 * initial mount, once the journal has been initialised but before we've
4418 * done any recovery; and again on any subsequent remount.
4420 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4422 struct ext4_sb_info *sbi = EXT4_SB(sb);
4424 journal->j_commit_interval = sbi->s_commit_interval;
4425 journal->j_min_batch_time = sbi->s_min_batch_time;
4426 journal->j_max_batch_time = sbi->s_max_batch_time;
4428 write_lock(&journal->j_state_lock);
4429 if (test_opt(sb, BARRIER))
4430 journal->j_flags |= JBD2_BARRIER;
4432 journal->j_flags &= ~JBD2_BARRIER;
4433 if (test_opt(sb, DATA_ERR_ABORT))
4434 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4436 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4437 write_unlock(&journal->j_state_lock);
4440 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4441 unsigned int journal_inum)
4443 struct inode *journal_inode;
4446 * Test for the existence of a valid inode on disk. Bad things
4447 * happen if we iget() an unused inode, as the subsequent iput()
4448 * will try to delete it.
4450 journal_inode = ext4_iget(sb, journal_inum);
4451 if (IS_ERR(journal_inode)) {
4452 ext4_msg(sb, KERN_ERR, "no journal found");
4455 if (!journal_inode->i_nlink) {
4456 make_bad_inode(journal_inode);
4457 iput(journal_inode);
4458 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4462 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4463 journal_inode, journal_inode->i_size);
4464 if (!S_ISREG(journal_inode->i_mode)) {
4465 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4466 iput(journal_inode);
4469 return journal_inode;
4472 static journal_t *ext4_get_journal(struct super_block *sb,
4473 unsigned int journal_inum)
4475 struct inode *journal_inode;
4478 BUG_ON(!ext4_has_feature_journal(sb));
4480 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4484 journal = jbd2_journal_init_inode(journal_inode);
4486 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4487 iput(journal_inode);
4490 journal->j_private = sb;
4491 ext4_init_journal_params(sb, journal);
4495 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4498 struct buffer_head *bh;
4502 int hblock, blocksize;
4503 ext4_fsblk_t sb_block;
4504 unsigned long offset;
4505 struct ext4_super_block *es;
4506 struct block_device *bdev;
4508 BUG_ON(!ext4_has_feature_journal(sb));
4510 bdev = ext4_blkdev_get(j_dev, sb);
4514 blocksize = sb->s_blocksize;
4515 hblock = bdev_logical_block_size(bdev);
4516 if (blocksize < hblock) {
4517 ext4_msg(sb, KERN_ERR,
4518 "blocksize too small for journal device");
4522 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4523 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4524 set_blocksize(bdev, blocksize);
4525 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4526 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4527 "external journal");
4531 es = (struct ext4_super_block *) (bh->b_data + offset);
4532 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4533 !(le32_to_cpu(es->s_feature_incompat) &
4534 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4535 ext4_msg(sb, KERN_ERR, "external journal has "
4541 if ((le32_to_cpu(es->s_feature_ro_compat) &
4542 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4543 es->s_checksum != ext4_superblock_csum(sb, es)) {
4544 ext4_msg(sb, KERN_ERR, "external journal has "
4545 "corrupt superblock");
4550 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4551 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4556 len = ext4_blocks_count(es);
4557 start = sb_block + 1;
4558 brelse(bh); /* we're done with the superblock */
4560 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4561 start, len, blocksize);
4563 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4566 journal->j_private = sb;
4567 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4568 wait_on_buffer(journal->j_sb_buffer);
4569 if (!buffer_uptodate(journal->j_sb_buffer)) {
4570 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4573 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4574 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4575 "user (unsupported) - %d",
4576 be32_to_cpu(journal->j_superblock->s_nr_users));
4579 EXT4_SB(sb)->journal_bdev = bdev;
4580 ext4_init_journal_params(sb, journal);
4584 jbd2_journal_destroy(journal);
4586 ext4_blkdev_put(bdev);
4590 static int ext4_load_journal(struct super_block *sb,
4591 struct ext4_super_block *es,
4592 unsigned long journal_devnum)
4595 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4598 int really_read_only;
4600 BUG_ON(!ext4_has_feature_journal(sb));
4602 if (journal_devnum &&
4603 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4604 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4605 "numbers have changed");
4606 journal_dev = new_decode_dev(journal_devnum);
4608 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4610 really_read_only = bdev_read_only(sb->s_bdev);
4613 * Are we loading a blank journal or performing recovery after a
4614 * crash? For recovery, we need to check in advance whether we
4615 * can get read-write access to the device.
4617 if (ext4_has_feature_journal_needs_recovery(sb)) {
4618 if (sb_rdonly(sb)) {
4619 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4620 "required on readonly filesystem");
4621 if (really_read_only) {
4622 ext4_msg(sb, KERN_ERR, "write access "
4623 "unavailable, cannot proceed");
4626 ext4_msg(sb, KERN_INFO, "write access will "
4627 "be enabled during recovery");
4631 if (journal_inum && journal_dev) {
4632 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4633 "and inode journals!");
4638 if (!(journal = ext4_get_journal(sb, journal_inum)))
4641 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4645 if (!(journal->j_flags & JBD2_BARRIER))
4646 ext4_msg(sb, KERN_INFO, "barriers disabled");
4648 if (!ext4_has_feature_journal_needs_recovery(sb))
4649 err = jbd2_journal_wipe(journal, !really_read_only);
4651 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4653 memcpy(save, ((char *) es) +
4654 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4655 err = jbd2_journal_load(journal);
4657 memcpy(((char *) es) + EXT4_S_ERR_START,
4658 save, EXT4_S_ERR_LEN);
4663 ext4_msg(sb, KERN_ERR, "error loading journal");
4664 jbd2_journal_destroy(journal);
4668 EXT4_SB(sb)->s_journal = journal;
4669 ext4_clear_journal_err(sb, es);
4671 if (!really_read_only && journal_devnum &&
4672 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4673 es->s_journal_dev = cpu_to_le32(journal_devnum);
4675 /* Make sure we flush the recovery flag to disk. */
4676 ext4_commit_super(sb, 1);
4682 static int ext4_commit_super(struct super_block *sb, int sync)
4684 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4685 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4688 if (!sbh || block_device_ejected(sb))
4691 * If the file system is mounted read-only, don't update the
4692 * superblock write time. This avoids updating the superblock
4693 * write time when we are mounting the root file system
4694 * read/only but we need to replay the journal; at that point,
4695 * for people who are east of GMT and who make their clock
4696 * tick in localtime for Windows bug-for-bug compatibility,
4697 * the clock is set in the future, and this will cause e2fsck
4698 * to complain and force a full file system check.
4700 if (!(sb->s_flags & MS_RDONLY))
4701 es->s_wtime = cpu_to_le32(get_seconds());
4702 if (sb->s_bdev->bd_part)
4703 es->s_kbytes_written =
4704 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4705 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4706 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4708 es->s_kbytes_written =
4709 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4710 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4711 ext4_free_blocks_count_set(es,
4712 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4713 &EXT4_SB(sb)->s_freeclusters_counter)));
4714 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4715 es->s_free_inodes_count =
4716 cpu_to_le32(percpu_counter_sum_positive(
4717 &EXT4_SB(sb)->s_freeinodes_counter));
4718 BUFFER_TRACE(sbh, "marking dirty");
4719 ext4_superblock_csum_set(sb);
4722 if (buffer_write_io_error(sbh)) {
4724 * Oh, dear. A previous attempt to write the
4725 * superblock failed. This could happen because the
4726 * USB device was yanked out. Or it could happen to
4727 * be a transient write error and maybe the block will
4728 * be remapped. Nothing we can do but to retry the
4729 * write and hope for the best.
4731 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4732 "superblock detected");
4733 clear_buffer_write_io_error(sbh);
4734 set_buffer_uptodate(sbh);
4736 mark_buffer_dirty(sbh);
4739 error = __sync_dirty_buffer(sbh,
4740 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4744 error = buffer_write_io_error(sbh);
4746 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4748 clear_buffer_write_io_error(sbh);
4749 set_buffer_uptodate(sbh);
4756 * Have we just finished recovery? If so, and if we are mounting (or
4757 * remounting) the filesystem readonly, then we will end up with a
4758 * consistent fs on disk. Record that fact.
4760 static void ext4_mark_recovery_complete(struct super_block *sb,
4761 struct ext4_super_block *es)
4763 journal_t *journal = EXT4_SB(sb)->s_journal;
4765 if (!ext4_has_feature_journal(sb)) {
4766 BUG_ON(journal != NULL);
4769 jbd2_journal_lock_updates(journal);
4770 if (jbd2_journal_flush(journal) < 0)
4773 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4774 ext4_clear_feature_journal_needs_recovery(sb);
4775 ext4_commit_super(sb, 1);
4779 jbd2_journal_unlock_updates(journal);
4783 * If we are mounting (or read-write remounting) a filesystem whose journal
4784 * has recorded an error from a previous lifetime, move that error to the
4785 * main filesystem now.
4787 static void ext4_clear_journal_err(struct super_block *sb,
4788 struct ext4_super_block *es)
4794 BUG_ON(!ext4_has_feature_journal(sb));
4796 journal = EXT4_SB(sb)->s_journal;
4799 * Now check for any error status which may have been recorded in the
4800 * journal by a prior ext4_error() or ext4_abort()
4803 j_errno = jbd2_journal_errno(journal);
4807 errstr = ext4_decode_error(sb, j_errno, nbuf);
4808 ext4_warning(sb, "Filesystem error recorded "
4809 "from previous mount: %s", errstr);
4810 ext4_warning(sb, "Marking fs in need of filesystem check.");
4812 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4813 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4814 ext4_commit_super(sb, 1);
4816 jbd2_journal_clear_err(journal);
4817 jbd2_journal_update_sb_errno(journal);
4822 * Force the running and committing transactions to commit,
4823 * and wait on the commit.
4825 int ext4_force_commit(struct super_block *sb)
4832 journal = EXT4_SB(sb)->s_journal;
4833 return ext4_journal_force_commit(journal);
4836 static int ext4_sync_fs(struct super_block *sb, int wait)
4840 bool needs_barrier = false;
4841 struct ext4_sb_info *sbi = EXT4_SB(sb);
4843 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4846 trace_ext4_sync_fs(sb, wait);
4847 flush_workqueue(sbi->rsv_conversion_wq);
4849 * Writeback quota in non-journalled quota case - journalled quota has
4852 dquot_writeback_dquots(sb, -1);
4854 * Data writeback is possible w/o journal transaction, so barrier must
4855 * being sent at the end of the function. But we can skip it if
4856 * transaction_commit will do it for us.
4858 if (sbi->s_journal) {
4859 target = jbd2_get_latest_transaction(sbi->s_journal);
4860 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4861 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4862 needs_barrier = true;
4864 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4866 ret = jbd2_log_wait_commit(sbi->s_journal,
4869 } else if (wait && test_opt(sb, BARRIER))
4870 needs_barrier = true;
4871 if (needs_barrier) {
4873 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4882 * LVM calls this function before a (read-only) snapshot is created. This
4883 * gives us a chance to flush the journal completely and mark the fs clean.
4885 * Note that only this function cannot bring a filesystem to be in a clean
4886 * state independently. It relies on upper layer to stop all data & metadata
4889 static int ext4_freeze(struct super_block *sb)
4897 journal = EXT4_SB(sb)->s_journal;
4900 /* Now we set up the journal barrier. */
4901 jbd2_journal_lock_updates(journal);
4904 * Don't clear the needs_recovery flag if we failed to
4905 * flush the journal.
4907 error = jbd2_journal_flush(journal);
4911 /* Journal blocked and flushed, clear needs_recovery flag. */
4912 ext4_clear_feature_journal_needs_recovery(sb);
4915 error = ext4_commit_super(sb, 1);
4918 /* we rely on upper layer to stop further updates */
4919 jbd2_journal_unlock_updates(journal);
4924 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4925 * flag here, even though the filesystem is not technically dirty yet.
4927 static int ext4_unfreeze(struct super_block *sb)
4929 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
4932 if (EXT4_SB(sb)->s_journal) {
4933 /* Reset the needs_recovery flag before the fs is unlocked. */
4934 ext4_set_feature_journal_needs_recovery(sb);
4937 ext4_commit_super(sb, 1);
4942 * Structure to save mount options for ext4_remount's benefit
4944 struct ext4_mount_options {
4945 unsigned long s_mount_opt;
4946 unsigned long s_mount_opt2;
4949 unsigned long s_commit_interval;
4950 u32 s_min_batch_time, s_max_batch_time;
4953 char *s_qf_names[EXT4_MAXQUOTAS];
4957 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4959 struct ext4_super_block *es;
4960 struct ext4_sb_info *sbi = EXT4_SB(sb);
4961 unsigned long old_sb_flags;
4962 struct ext4_mount_options old_opts;
4963 int enable_quota = 0;
4965 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4970 char *orig_data = kstrdup(data, GFP_KERNEL);
4972 /* Store the original options */
4973 old_sb_flags = sb->s_flags;
4974 old_opts.s_mount_opt = sbi->s_mount_opt;
4975 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4976 old_opts.s_resuid = sbi->s_resuid;
4977 old_opts.s_resgid = sbi->s_resgid;
4978 old_opts.s_commit_interval = sbi->s_commit_interval;
4979 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4980 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4982 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4983 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4984 if (sbi->s_qf_names[i]) {
4985 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4987 if (!old_opts.s_qf_names[i]) {
4988 for (j = 0; j < i; j++)
4989 kfree(old_opts.s_qf_names[j]);
4994 old_opts.s_qf_names[i] = NULL;
4996 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4997 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4999 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5004 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5005 test_opt(sb, JOURNAL_CHECKSUM)) {
5006 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5007 "during remount not supported; ignoring");
5008 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5011 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5012 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5013 ext4_msg(sb, KERN_ERR, "can't mount with "
5014 "both data=journal and delalloc");
5018 if (test_opt(sb, DIOREAD_NOLOCK)) {
5019 ext4_msg(sb, KERN_ERR, "can't mount with "
5020 "both data=journal and dioread_nolock");
5024 if (test_opt(sb, DAX)) {
5025 ext4_msg(sb, KERN_ERR, "can't mount with "
5026 "both data=journal and dax");
5030 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5031 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5032 ext4_msg(sb, KERN_ERR, "can't mount with "
5033 "journal_async_commit in data=ordered mode");
5039 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5040 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5045 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5046 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5047 "dax flag with busy inodes while remounting");
5048 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5051 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5052 ext4_abort(sb, "Abort forced by user");
5054 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5055 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5059 if (sbi->s_journal) {
5060 ext4_init_journal_params(sb, sbi->s_journal);
5061 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5064 if (*flags & MS_LAZYTIME)
5065 sb->s_flags |= MS_LAZYTIME;
5067 if ((bool)(*flags & MS_RDONLY) != sb_rdonly(sb)) {
5068 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5073 if (*flags & MS_RDONLY) {
5074 err = sync_filesystem(sb);
5077 err = dquot_suspend(sb, -1);
5082 * First of all, the unconditional stuff we have to do
5083 * to disable replay of the journal when we next remount
5085 sb->s_flags |= MS_RDONLY;
5088 * OK, test if we are remounting a valid rw partition
5089 * readonly, and if so set the rdonly flag and then
5090 * mark the partition as valid again.
5092 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5093 (sbi->s_mount_state & EXT4_VALID_FS))
5094 es->s_state = cpu_to_le16(sbi->s_mount_state);
5097 ext4_mark_recovery_complete(sb, es);
5099 /* Make sure we can mount this feature set readwrite */
5100 if (ext4_has_feature_readonly(sb) ||
5101 !ext4_feature_set_ok(sb, 0)) {
5106 * Make sure the group descriptor checksums
5107 * are sane. If they aren't, refuse to remount r/w.
5109 for (g = 0; g < sbi->s_groups_count; g++) {
5110 struct ext4_group_desc *gdp =
5111 ext4_get_group_desc(sb, g, NULL);
5113 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5114 ext4_msg(sb, KERN_ERR,
5115 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5116 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5117 le16_to_cpu(gdp->bg_checksum));
5124 * If we have an unprocessed orphan list hanging
5125 * around from a previously readonly bdev mount,
5126 * require a full umount/remount for now.
5128 if (es->s_last_orphan) {
5129 ext4_msg(sb, KERN_WARNING, "Couldn't "
5130 "remount RDWR because of unprocessed "
5131 "orphan inode list. Please "
5132 "umount/remount instead");
5138 * Mounting a RDONLY partition read-write, so reread
5139 * and store the current valid flag. (It may have
5140 * been changed by e2fsck since we originally mounted
5144 ext4_clear_journal_err(sb, es);
5145 sbi->s_mount_state = le16_to_cpu(es->s_state);
5146 if (!ext4_setup_super(sb, es, 0))
5147 sb->s_flags &= ~MS_RDONLY;
5148 if (ext4_has_feature_mmp(sb))
5149 if (ext4_multi_mount_protect(sb,
5150 le64_to_cpu(es->s_mmp_block))) {
5159 * Reinitialize lazy itable initialization thread based on
5162 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5163 ext4_unregister_li_request(sb);
5165 ext4_group_t first_not_zeroed;
5166 first_not_zeroed = ext4_has_uninit_itable(sb);
5167 ext4_register_li_request(sb, first_not_zeroed);
5170 ext4_setup_system_zone(sb);
5171 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5172 ext4_commit_super(sb, 1);
5175 /* Release old quota file names */
5176 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5177 kfree(old_opts.s_qf_names[i]);
5179 if (sb_any_quota_suspended(sb))
5180 dquot_resume(sb, -1);
5181 else if (ext4_has_feature_quota(sb)) {
5182 err = ext4_enable_quotas(sb);
5189 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5190 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5195 sb->s_flags = old_sb_flags;
5196 sbi->s_mount_opt = old_opts.s_mount_opt;
5197 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5198 sbi->s_resuid = old_opts.s_resuid;
5199 sbi->s_resgid = old_opts.s_resgid;
5200 sbi->s_commit_interval = old_opts.s_commit_interval;
5201 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5202 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5204 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5205 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5206 kfree(sbi->s_qf_names[i]);
5207 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5215 static int ext4_statfs_project(struct super_block *sb,
5216 kprojid_t projid, struct kstatfs *buf)
5219 struct dquot *dquot;
5223 qid = make_kqid_projid(projid);
5224 dquot = dqget(sb, qid);
5226 return PTR_ERR(dquot);
5227 spin_lock(&dquot->dq_dqb_lock);
5229 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5230 dquot->dq_dqb.dqb_bsoftlimit :
5231 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5232 if (limit && buf->f_blocks > limit) {
5233 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5234 buf->f_blocks = limit;
5235 buf->f_bfree = buf->f_bavail =
5236 (buf->f_blocks > curblock) ?
5237 (buf->f_blocks - curblock) : 0;
5240 limit = dquot->dq_dqb.dqb_isoftlimit ?
5241 dquot->dq_dqb.dqb_isoftlimit :
5242 dquot->dq_dqb.dqb_ihardlimit;
5243 if (limit && buf->f_files > limit) {
5244 buf->f_files = limit;
5246 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5247 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5250 spin_unlock(&dquot->dq_dqb_lock);
5256 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5258 struct super_block *sb = dentry->d_sb;
5259 struct ext4_sb_info *sbi = EXT4_SB(sb);
5260 struct ext4_super_block *es = sbi->s_es;
5261 ext4_fsblk_t overhead = 0, resv_blocks;
5264 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5266 if (!test_opt(sb, MINIX_DF))
5267 overhead = sbi->s_overhead;
5269 buf->f_type = EXT4_SUPER_MAGIC;
5270 buf->f_bsize = sb->s_blocksize;
5271 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5272 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5273 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5274 /* prevent underflow in case that few free space is available */
5275 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5276 buf->f_bavail = buf->f_bfree -
5277 (ext4_r_blocks_count(es) + resv_blocks);
5278 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5280 buf->f_files = le32_to_cpu(es->s_inodes_count);
5281 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5282 buf->f_namelen = EXT4_NAME_LEN;
5283 fsid = le64_to_cpup((void *)es->s_uuid) ^
5284 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5285 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5286 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5289 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5290 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5291 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5300 * Helper functions so that transaction is started before we acquire dqio_sem
5301 * to keep correct lock ordering of transaction > dqio_sem
5303 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5305 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5308 static int ext4_write_dquot(struct dquot *dquot)
5312 struct inode *inode;
5314 inode = dquot_to_inode(dquot);
5315 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5316 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5318 return PTR_ERR(handle);
5319 ret = dquot_commit(dquot);
5320 err = ext4_journal_stop(handle);
5326 static int ext4_acquire_dquot(struct dquot *dquot)
5331 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5332 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5334 return PTR_ERR(handle);
5335 ret = dquot_acquire(dquot);
5336 err = ext4_journal_stop(handle);
5342 static int ext4_release_dquot(struct dquot *dquot)
5347 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5348 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5349 if (IS_ERR(handle)) {
5350 /* Release dquot anyway to avoid endless cycle in dqput() */
5351 dquot_release(dquot);
5352 return PTR_ERR(handle);
5354 ret = dquot_release(dquot);
5355 err = ext4_journal_stop(handle);
5361 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5363 struct super_block *sb = dquot->dq_sb;
5364 struct ext4_sb_info *sbi = EXT4_SB(sb);
5366 /* Are we journaling quotas? */
5367 if (ext4_has_feature_quota(sb) ||
5368 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5369 dquot_mark_dquot_dirty(dquot);
5370 return ext4_write_dquot(dquot);
5372 return dquot_mark_dquot_dirty(dquot);
5376 static int ext4_write_info(struct super_block *sb, int type)
5381 /* Data block + inode block */
5382 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5384 return PTR_ERR(handle);
5385 ret = dquot_commit_info(sb, type);
5386 err = ext4_journal_stop(handle);
5393 * Turn on quotas during mount time - we need to find
5394 * the quota file and such...
5396 static int ext4_quota_on_mount(struct super_block *sb, int type)
5398 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5399 EXT4_SB(sb)->s_jquota_fmt, type);
5402 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5404 struct ext4_inode_info *ei = EXT4_I(inode);
5406 /* The first argument of lockdep_set_subclass has to be
5407 * *exactly* the same as the argument to init_rwsem() --- in
5408 * this case, in init_once() --- or lockdep gets unhappy
5409 * because the name of the lock is set using the
5410 * stringification of the argument to init_rwsem().
5412 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5413 lockdep_set_subclass(&ei->i_data_sem, subclass);
5417 * Standard function to be called on quota_on
5419 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5420 const struct path *path)
5424 if (!test_opt(sb, QUOTA))
5427 /* Quotafile not on the same filesystem? */
5428 if (path->dentry->d_sb != sb)
5430 /* Journaling quota? */
5431 if (EXT4_SB(sb)->s_qf_names[type]) {
5432 /* Quotafile not in fs root? */
5433 if (path->dentry->d_parent != sb->s_root)
5434 ext4_msg(sb, KERN_WARNING,
5435 "Quota file not on filesystem root. "
5436 "Journaled quota will not work");
5437 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5440 * Clear the flag just in case mount options changed since
5443 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5447 * When we journal data on quota file, we have to flush journal to see
5448 * all updates to the file when we bypass pagecache...
5450 if (EXT4_SB(sb)->s_journal &&
5451 ext4_should_journal_data(d_inode(path->dentry))) {
5453 * We don't need to lock updates but journal_flush() could
5454 * otherwise be livelocked...
5456 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5457 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5458 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5463 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5464 err = dquot_quota_on(sb, type, format_id, path);
5466 lockdep_set_quota_inode(path->dentry->d_inode,
5469 struct inode *inode = d_inode(path->dentry);
5473 * Set inode flags to prevent userspace from messing with quota
5474 * files. If this fails, we return success anyway since quotas
5475 * are already enabled and this is not a hard failure.
5478 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5481 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5482 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5483 S_NOATIME | S_IMMUTABLE);
5484 ext4_mark_inode_dirty(handle, inode);
5485 ext4_journal_stop(handle);
5487 inode_unlock(inode);
5492 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5496 struct inode *qf_inode;
5497 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5498 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5499 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5500 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5503 BUG_ON(!ext4_has_feature_quota(sb));
5505 if (!qf_inums[type])
5508 qf_inode = ext4_iget(sb, qf_inums[type]);
5509 if (IS_ERR(qf_inode)) {
5510 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5511 return PTR_ERR(qf_inode);
5514 /* Don't account quota for quota files to avoid recursion */
5515 qf_inode->i_flags |= S_NOQUOTA;
5516 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5517 err = dquot_enable(qf_inode, type, format_id, flags);
5520 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5525 /* Enable usage tracking for all quota types. */
5526 static int ext4_enable_quotas(struct super_block *sb)
5529 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5530 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5531 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5532 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5534 bool quota_mopt[EXT4_MAXQUOTAS] = {
5535 test_opt(sb, USRQUOTA),
5536 test_opt(sb, GRPQUOTA),
5537 test_opt(sb, PRJQUOTA),
5540 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5541 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5542 if (qf_inums[type]) {
5543 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5544 DQUOT_USAGE_ENABLED |
5545 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5547 for (type--; type >= 0; type--)
5548 dquot_quota_off(sb, type);
5551 "Failed to enable quota tracking "
5552 "(type=%d, err=%d). Please run "
5553 "e2fsck to fix.", type, err);
5561 static int ext4_quota_off(struct super_block *sb, int type)
5563 struct inode *inode = sb_dqopt(sb)->files[type];
5567 /* Force all delayed allocation blocks to be allocated.
5568 * Caller already holds s_umount sem */
5569 if (test_opt(sb, DELALLOC))
5570 sync_filesystem(sb);
5572 if (!inode || !igrab(inode))
5575 err = dquot_quota_off(sb, type);
5576 if (err || ext4_has_feature_quota(sb))
5581 * Update modification times of quota files when userspace can
5582 * start looking at them. If we fail, we return success anyway since
5583 * this is not a hard failure and quotas are already disabled.
5585 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5588 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5589 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5590 inode->i_mtime = inode->i_ctime = current_time(inode);
5591 ext4_mark_inode_dirty(handle, inode);
5592 ext4_journal_stop(handle);
5594 inode_unlock(inode);
5596 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5600 return dquot_quota_off(sb, type);
5603 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5604 * acquiring the locks... As quota files are never truncated and quota code
5605 * itself serializes the operations (and no one else should touch the files)
5606 * we don't have to be afraid of races */
5607 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5608 size_t len, loff_t off)
5610 struct inode *inode = sb_dqopt(sb)->files[type];
5611 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5612 int offset = off & (sb->s_blocksize - 1);
5615 struct buffer_head *bh;
5616 loff_t i_size = i_size_read(inode);
5620 if (off+len > i_size)
5623 while (toread > 0) {
5624 tocopy = sb->s_blocksize - offset < toread ?
5625 sb->s_blocksize - offset : toread;
5626 bh = ext4_bread(NULL, inode, blk, 0);
5629 if (!bh) /* A hole? */
5630 memset(data, 0, tocopy);
5632 memcpy(data, bh->b_data+offset, tocopy);
5642 /* Write to quotafile (we know the transaction is already started and has
5643 * enough credits) */
5644 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5645 const char *data, size_t len, loff_t off)
5647 struct inode *inode = sb_dqopt(sb)->files[type];
5648 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5649 int err, offset = off & (sb->s_blocksize - 1);
5651 struct buffer_head *bh;
5652 handle_t *handle = journal_current_handle();
5654 if (EXT4_SB(sb)->s_journal && !handle) {
5655 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5656 " cancelled because transaction is not started",
5657 (unsigned long long)off, (unsigned long long)len);
5661 * Since we account only one data block in transaction credits,
5662 * then it is impossible to cross a block boundary.
5664 if (sb->s_blocksize - offset < len) {
5665 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5666 " cancelled because not block aligned",
5667 (unsigned long long)off, (unsigned long long)len);
5672 bh = ext4_bread(handle, inode, blk,
5673 EXT4_GET_BLOCKS_CREATE |
5674 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5675 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5676 ext4_should_retry_alloc(inode->i_sb, &retries));
5681 BUFFER_TRACE(bh, "get write access");
5682 err = ext4_journal_get_write_access(handle, bh);
5688 memcpy(bh->b_data+offset, data, len);
5689 flush_dcache_page(bh->b_page);
5691 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5694 if (inode->i_size < off + len) {
5695 i_size_write(inode, off + len);
5696 EXT4_I(inode)->i_disksize = inode->i_size;
5697 ext4_mark_inode_dirty(handle, inode);
5702 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5704 const struct quota_format_ops *ops;
5706 if (!sb_has_quota_loaded(sb, qid->type))
5708 ops = sb_dqopt(sb)->ops[qid->type];
5709 if (!ops || !ops->get_next_id)
5711 return dquot_get_next_id(sb, qid);
5715 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5716 const char *dev_name, void *data)
5718 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5721 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5722 static inline void register_as_ext2(void)
5724 int err = register_filesystem(&ext2_fs_type);
5727 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5730 static inline void unregister_as_ext2(void)
5732 unregister_filesystem(&ext2_fs_type);
5735 static inline int ext2_feature_set_ok(struct super_block *sb)
5737 if (ext4_has_unknown_ext2_incompat_features(sb))
5741 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5746 static inline void register_as_ext2(void) { }
5747 static inline void unregister_as_ext2(void) { }
5748 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5751 static inline void register_as_ext3(void)
5753 int err = register_filesystem(&ext3_fs_type);
5756 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5759 static inline void unregister_as_ext3(void)
5761 unregister_filesystem(&ext3_fs_type);
5764 static inline int ext3_feature_set_ok(struct super_block *sb)
5766 if (ext4_has_unknown_ext3_incompat_features(sb))
5768 if (!ext4_has_feature_journal(sb))
5772 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5777 static struct file_system_type ext4_fs_type = {
5778 .owner = THIS_MODULE,
5780 .mount = ext4_mount,
5781 .kill_sb = kill_block_super,
5782 .fs_flags = FS_REQUIRES_DEV,
5784 MODULE_ALIAS_FS("ext4");
5786 /* Shared across all ext4 file systems */
5787 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5789 static int __init ext4_init_fs(void)
5793 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5794 ext4_li_info = NULL;
5795 mutex_init(&ext4_li_mtx);
5797 /* Build-time check for flags consistency */
5798 ext4_check_flag_values();
5800 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5801 init_waitqueue_head(&ext4__ioend_wq[i]);
5803 err = ext4_init_es();
5807 err = ext4_init_pageio();
5811 err = ext4_init_system_zone();
5815 err = ext4_init_sysfs();
5819 err = ext4_init_mballoc();
5822 err = init_inodecache();
5827 err = register_filesystem(&ext4_fs_type);
5833 unregister_as_ext2();
5834 unregister_as_ext3();
5835 destroy_inodecache();
5837 ext4_exit_mballoc();
5841 ext4_exit_system_zone();
5850 static void __exit ext4_exit_fs(void)
5852 ext4_destroy_lazyinit_thread();
5853 unregister_as_ext2();
5854 unregister_as_ext3();
5855 unregister_filesystem(&ext4_fs_type);
5856 destroy_inodecache();
5857 ext4_exit_mballoc();
5859 ext4_exit_system_zone();
5864 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5865 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5866 MODULE_LICENSE("GPL");
5867 module_init(ext4_init_fs)
5868 module_exit(ext4_exit_fs)